Patent Application
20230349693
The present invention relates generally to determining a pose estimate of a pose assumed by a manipulated object in a real three-dimensional environment that has optical features by using light data obtained with a photodetector as well relative motion data indicative of changes in pose obtained with an auxiliary motion detection component, where the pose estimate is deployed in an application or interface.
An object's pose in a real three-dimensional environment can be expressed with respect to stationary references such as ground planes, reference surfaces, lines, solids, fixed points and other invariant features disposed in the real three-dimensional environment. It is convenient to parameterize the environment by a set of world coordinates with a chosen reference point. The reference point may be the origin of the world coordinates, the center of a particularly prominent invariant feature or the center of a distribution of two or more of these features. Once the locations and orientations of the invariant features distributed in the environment are known, then knowledge of the spatial relationship between the object and these invariant features enables one to compute the object's pose.
An object's pose information combines the three linear displacement coordinates (x,y,z) of any reference point on the object, as well as the three inclination angles, also called the Euler angles (ϕ,θ,ψ) that describe the pitch, yaw and roll of the object. Conveniently, all parameters (x,y,z,ϕ,θ,ψ) are expressed in world coordinates to yield an absolute pose. In some cases, alternative expressions for the inclination angles such as rotations defined by the four Cayley-Klein angles or quaternions are more appropriate.
Determination of a sequence of an object's absolute poses at different times allows one to compute and track the motion of the object in the real three-dimensional environment. Over time, many useful coordinate systems and method have been developed to track the pose of objects and to parametrize their equations of motion. For a theoretical background the reader is referred to textbooks on classical mechanics such as Goldstein et al., Classical Mechanics, 3rd Edition, Addison Wesley 2002.
Optical navigation is a particularly simple and precise way to track moving objects. The approach is also intuitive since our own human vision system computes locations and motion trajectories of objects in real three-dimensional environments.
The precision of optical navigation is due to the very short wavelength of electromagnetic radiation in comparison with typical object dimensions, negligible latency in short distance measurements due to the extremely large speed of light as well as relative immunity to interference. Thus, it is well known that the problem of determining an absolute pose or a motion trajectory of an object in almost any real three-dimensional environment may be effectively addressed by the application of optical apparatus and methods.
A particularly acute need for efficient, accurate and low-cost determination of the absolute pose of an object in a real three-dimensional environment is found in the field of hand-held objects used for interfacing with the digital world. This field encompasses myriads of manipulated objects such as pointers, wands, remote controls, gaming objects, smart phones, tablets, jotting implements, surgical implements, three-dimensional digitizers and various types of human utensils whose motion in real space is to be processed to derive a digital input for an application. In some realms, such application involves interactions that would greatly benefit from a rapid, low-cost method and apparatus for one-to-one motion mapping between real space and cyberspace.
Specific examples of cyberspace games played in three-dimensions (3-D) and requiring high-precision control object tracking involve scenarios where the manipulated control object is transported into or even mimicked in cyberspace. Exemplary gaming objects of this variety include a golfing club, a racket, a guitar, a gun, a ball, a steering wheel, a flying control or any other accoutrement that the player wishes to transport into and utilize in a cyberspace application. Indeed, event smart phones or tablets can serve the function of manipulated control objects. A very thorough summary of such 3-D interfacing needs for graphics are found in U.S. Pat. No. 6,811,489 to Shimizu, et al.
A major problem encountered by state of the art manipulated objects such as control wands and gaming implements is that they do not possess a sufficiently robust and rapid absolute pose determination system. In fact, many do not even provide for absolute pose determination. Rather, they function much like quasi three-dimensional mice. These solutions use motion detection components that rely on optical flow sensors, inertial sensing devices or other relative motion capture systems to derive the signals for interfacing with cyberspace. In particular, many such manipulated objects or interface devices try to obtain just a subset of the motion changes, e.g., inclination. An example of an inclination calculation apparatus is found in U.S. Pat. No. 7,379,841 to Ohta while a broader attempt at determining relative motion is taught in U.S. Pat. No. 7,424,388 to Sato and U.S. Application 2007/0049374 to Ikeda, et al.
Unfortunately, one-to-one motion mapping between space and cyberspace is not possible without the ability to digitize the absolute pose of the manipulated object with respect to a well-defined reference location in the real three-dimensional environment (real space). All prior art devices that do not solve the full motion problem, i.e., do not capture successive poses of the manipulated object with a method that accounts for all six degrees of freedom (namely, the very parameters (x,y,z,φ,θ,ψ) inherent in three-dimensional space) encounter limitations. Among many others, these include information loss, appearance of an offset, position aliasing, gradual drift and accumulating position error.
In general, the prior art has recognized the need for tracking all six degrees of freedom of manipulated objects that are rigid bodies translating and rotating in three-dimensions. Thus, optical navigation typically employs several cameras to determine the position or trajectory of an object in an environment by studying images of the object in the environment. Such optical capturing or tracking systems are commonly referred to as optical motion capture (MC) systems.
Typically, motion capture involves a tradeoff between computational cost and ergonomics. That is, the computational cost of image processing decreases with the degree of instrumentation of the environment or of the manipulated object. One particular system taught by McSheery et al. in U.S. Pat. No. 6,324,296 discloses a distributed-processing motion capture system that employs a number of point sources of light as markers, e.g., infrared LEDs, attached to the object whose motion is to be determined. The markers use unique sequences of light pulses to represent their unique identities and thus enable filtering out of information not belonging to the markers (i.e., background noise) by the imaging cameras located in the environment. Since McSheery's system permits a great deal of irrelevant information from the imaging sensors (e.g., CCDs) to be discarded before image processing, the system is less computationally expensive than more traditional motion capture systems.
Another three-dimensional position and orientation sensing system that employs markers on the object is taught by Kosaka et al. in U.S. Pat. No. 6,724,930. In this case the markers are uniquely identified based on color or a geometric characteristic of the markers in the extracted regions. The system uses an image acquisition unit or camera positioned in the environment and relies on image processing functions to remove texture and noise. Segmentation algorithms are used to extract markers from images and to determine the three-dimensional position and orientation of the object with respect to the image acquisition apparatus.
Still another way of employing markers in position and orientation detection is taught in U.S. Pat. No. 6,587,809 by Majoe. The object is tracked by providing it with markers that are activated one at a time and sensed by a number of individual sensors positioned in the environment. The position of the energized or active marker is determined by a control unit based on energy levels received by the individual sensors from that marker.
The above approaches using markers on objects and cameras in the environment to recover object position, orientation or trajectory are still resource-intensive for low-cost and low-bandwidth applications. This is due to the large bandwidth needed to transmit image data captured by cameras, the computational cost to the host computer associated with processing image data, and the data network complexity due to the spatially complicated distribution of equipment (i.e., placement and coordination of several cameras in the environment with the central processing unit and overall system synchronization).
Despite the above-mentioned limitations of general motion tracking systems, some aspects of these systems have been adopted in the field of manipulated objects used for interfacing with computers. Such objects are moved by users in three-dimensions to produce input for computer applications. Hence, they need to be tracked in all six degrees of freedom. Therefore, recent three-dimensional wands and controls do teach solving for all six degrees of freedom.
For example, U.S. Patent Application 2008/0167818 to Kimber et al. has a passive wand with no on-board devices or LEDs. The wand is viewed from multiple cameras finding the full 6 degrees of freedom to provide for more precise estimation of wand pose is expressly taught. Similarly, U.S. Pat. No. 6,982,697 to Wilson et al. teaches the use of external calibrated cameras to decode the orientation of the pointer used for control actions. U.S. Patent Application 2006/0109245 to Wilson, et al. further teaches how intelligent computing environments can take advantage of a device that provides orientation data in relative motion mode and absolute mode. Further teachings on systems that use external or not-on-board cameras to determine the pose and motion of a wand or control and use it as input into various types of applications can be found in U.S. Patent Applications: 2008/0192007, 2008/0192070, 2008/0204411, 2009/0164952 all by Wilson.
Still other notable teachings show as few as a single off-board camera for detecting three-dimensional motion of a controller employed for game control purposes. Such cameras may be depth sensing. Examples of corresponding teachings are found in U.S. Patent Application 2008/0096654 by Mondesir, et al., as well as U.S. Patent Applications 2008/0100825, 2009/0122146 both by Zalewski, et al.
Unfortunately, approaches in which multiple cameras are set up at different locations in the three-dimensional environment to enable stereo vision do not lead to a low-cost implementation. These solutions also require extensive calibration and synchronization of the cameras. Meanwhile, the use of expensive single cameras with depth sensing does not provide for robust systems. The sampling rate (i.e., frames per second) of such systems tends to be lower than desired, especially when the user is executing rapid and intricate movements with the manipulated object in a confined or close-range environment.
Another approach involves determining the position or attitude of a three-dimensional object in the absolute sense and using it for a graphical user interface. One example of this approach is taught in U.S. Pat. No. 6,727,885 to Ishino, et al. Here the sensor is on-board the manipulated object. A projected image viewed by the sensor and generated by a separate mechanism, i.e., a projection apparatus that imbues the projected image with characteristic image points is employed to perform the computation. Additional information about such apparatus and its application for games is found in U.S. Pat. No. 6,852,032 to Ishino and U.S. Pat. No. 6,993,206 to Ishino, et al.
The solution proposed by Ishino et al. is more versatile than the prior art solutions relying on hard-to-calibrate and synchronize multi-camera systems or expensive cameras with depth sensing capabilities. Unfortunately, the complexity of additional hardware for projecting images with characteristic image points is nontrivial. The same is true of consequent calibration and interaction problems, including knowledge of the exact location of the image in three-dimensional space. This solution is not applicable to close-range and/or confined environments, and especially environments with typical obstructions that interfere with line-of-sight conditions.
There are still other teachings attempting to improve on both the apparatus and method aspects of generating computer input with manipulated objects such as wands, pointers, remote controls (e.g., TV controls). A very illuminating overall review of state of the art technologies that can be used for interacting with virtual environments and their limitations are discussed by Richard Holloway in “Virtual Environments: A Survey of the Technology”, University of North Carolina at Chapel Hill, September 1993 (TR93-033). Still more recent teachings focusing on how absolute pose data can be used in specific contexts and for remote control applications is discussed in the following U.S. Patent Applications: 2007/0189737; 2008/0106517; 2008/0121782; 2008/0272272; 2008/0309511; 2009/0066647; 2009/0066648; 2009/0153389; 2009/0153475; 2009/0153478; 2009/0158203 and 2009/0158222.
In sum, despite considerable amount of work in the field, a clear and pressing need for low-cost, robust and accurate apparatus for absolute motion capture remains. Specifically, what is needed is an apparatus that permits one to obtain absolute pose data from the poses assumed by a manipulated object for purposes of interacting with the digital world. Such apparatus should not only be low-cost, robust and accurate, but it should also be convenient and easy to use at high frame rates in close-range and confined three-dimensional environments.
It is the object of the present invention to provide an efficient system and method to generate input data from pose estimates of poses assumed by a manipulated object in a real three-dimensional environment using on-board sensors. The on-board sensors include a photodetector or camera as well as an auxiliary motion detection component that generates relative motion data indicative of changes in pose. The system and method leverage the light data from the photodetector and the relative motion data indicative of changes in pose from the auxiliary motion detection component in an effective manner to provide pose estimates to an application that requests them. The system and method are well-suited for various applications including virtual reality applications (VR applications), augmented reality applications (AR applications), mixed reality applications (MR applications) and extended reality applications (XR applications). The system and method are robust and they are also well suited for convenient, low-cost, robust and accurate generation of input data while the manipulated object is operated in close-range and confined three-dimensional environments.
The objects and advantages of the present invention are accomplished by a system for generating input data. Input data covers various types of data produced by a user including input data, command data, control data and/or any other data to be used as input data in an application or in an interface. The input data is derived from or generated from a pose estimate for a pose (pose is position and orientation) that is assumed by a manipulated object in a real three-dimensional environment (as opposed to, e.g., a virtual environment that may be two-dimensional or three-dimensional). The real three-dimensional environment has optical features, sometimes called invariant features. The optical features can be active optical features, e.g., beacons that produce or emit light. The optical features can be passive optical features that reflect or scatter ambient light. Any combination of active optical features and/or passive optical features can be used.
The manipulated object is moved by hand, worn or otherwise handled by a human user. Because manipulated objects include those operated, gestured with or moved by hand they are referred to as manipulated herein. Such manipulated objects include wands, remote controls, three-dimensional mice, game controls, gaming objects, jotting implements, surgical implements, three dimensional digitizers, digitizing styli, smart phones, tablets, hand-held tools and the like. Other manipulated objects include wearable objects that are not necessarily manipulated by hand. Suitable manipulated objects of this type include Augmented Reality glasses (AR glasses), Virtual Reality glasses (VR glasses or VR Head-Mounted Displays (VR HMDs)), Mixed Reality glasses (MR glasses), Extended Reality glasses (XR glasses) and various other types of glasses, goggles, headgear and the like that is/are worn by a human user.
The system for generating the input data requires a photodetector mounted on-board the manipulated object. The photodetector is typically a camera, several cameras or still other photodetector, photosensor, optical sensor or one or more photosensitive devices. The photodetector accepts light from the optical features and generates light data corresponding to those optical features.
The system also requires an auxiliary motion detection component mounted on-board the manipulated object. The auxiliary motion detection component is typically made up of at least one relative motion sensor. Suitable motion sensors include accelerometers, gyroscopes, optical flow measuring units, electronic magnetic sensing components and the like. Equipped with one or more such relative motion sensors the auxiliary motion detection component is set up to generate relative motion data. The desired relative motion data is indicative of a change in the pose of the manipulated objects and specifically in any part of the object's pose. In other words, relative motion data is indicative of a change in an orientation, a change in a position or both.
The system is further equipped with a processor that is in communication with the photodetector and with the auxiliary motion component. The processor can be on-board the manipulated objects or it can be remote. For example, the processor can be in another device (e.g., a host device, a remote computer, a gaming console, a worn computer, a backpack computer, a tablet, a smart phone and the like).
The processor is configured, wired, coded, programmed or otherwise set up for performing a number of operations. These operations include receiving the light data from the photodetector. From the light data the processor determines an absolute pose estimate of the pose of the manipulated object. The pose estimate is called absolute because it is with respect to world coordinates describing the real three-dimensional environment. In other words, the absolute pose estimate is not relative, where relative means lacking ground truth or not with respect to world coordinates or with respect to a reference location in world coordinates. The light data is typically produced at certain times, e.g., when the photodetector captures an image frame of the three-dimensional environment. These times are referred to herein as times ti, where i stands for the i-th time or the i-th frame. For a photodetector embodied by a camera running at a certain frame rate (e.g., 100 Hz) the number of times ti obtained during each second will correspond to the number of frames per second (e.g., 100). Hence, the processor is able to determine absolute pose estimates of the pose assumed by the manipulated object at times ti (or in the case of 100 Hz, 100 times per second). Oversampling or subsampling techniques can be used to generate pose values at times other than ti.
The processor is also set up for receiving the relative motion data from the auxiliary motion detection component. From the relative motion data the processor determines a relative pose change. The relative pose change is indicative of at least one change, where the change can include zero or no change. The relative pose change can be a change in an orientation, i.e., a change in at least one of the orientation parameters describing orientation. The relative pose change can be a change in a position, i.e., a change in at least one of the position parameters describing position. Of course, the relative pose change can include changes in all orientation and position parameters.
Further, the processor is in communication with an application. The application is of any type that uses input data derived from or generated from the pose estimate obtained for the pose of the manipulated object (e.g., gesture information or digital ink). In some specific cases, the application can be a virtual reality (VR) application, an augmented reality (AR) application, a mixed reality (MR) application, an extended reality (XR) application or the like.
The application issues a request for the pose estimate to the processor at an application request time tr. The request can be continuous, periodic or sporadic. In response, the processor provides the application the pose estimate. The pose estimate is expressed in or with respect to world coordinates that describe the real three-dimensional environment.
The pose estimate generated by the processor is based on the absolute pose estimate from the light data and on the relative pose change from the relative motion data. In other words, the pose estimate is based on two types of information. It is based on estimate of absolute pose from light data that is obtained from the photodetector at time ti. It is also based on relative pose change provided by the auxiliary motion detection component typically at a much higher rate. In other words, relative pose change is sampled at a higher rate than light data. Still differently put, relative motion data is provided much more often than light data, or at several times between times ti and ti+1. Thus, relative motion data can be used to fill in between times ti and ti+1 (times when the photodetector captures successive image frames and can thus provide light data).
The application receives the pose estimate from the processor and uses is for generating input data. Note that the input data may be used by the application to render an augmented, virtual, mixed or extended scene to the user on a display. Input data can also be used as command data, e.g., to move virtual items or perform any operation within the application or even to turn the application on or off. Or the input data can be used to derive gestures or digital ink.
As already indicated above, the frame rate of the photodetector is slower and even significantly slower than the rate of the auxiliary motion detection component. In other words, between successive times ti and ti+1 when absolute pose estimates are obtained from light data captured by the photodetector the system sees many relative motion changes reported by the auxiliary motion detection component. This allows the processor to determine the pose estimate by combining the absolute pose estimates with relative pose changes between times ti and ti+1. The processor can use certain combining techniques to determine the pose estimates. These techniques include interpolation of the absolute pose estimate between times ti and ti+1 with relative pose changes. Another technique is numerical integration of the relative pose change between times ti and ti+1 while using the absolute pose estimate (e.g., at time ti) as an initial condition. Still another technique involved combining interpolation of the absolute pose with numerical integration of the relative pose change.
Alternatively, the processor can use more advanced techniques to estimate (or reconstruct) the absolute pose from the relative pose changes and the light data captured by the photodetector, where the latter (photodetector) is significantly slower than the former. In the present invention we will refer to this operation as “filling in”. A person of average skill in the art will be familiar with sensor fusion techniques where data from multiple sensors or sources is combined to create a more accurate and reliable estimate of a physical variable. “Filling in” can be understood as a type of sensor fusion operation where one source (the relative pose changes) generates measurements at a faster rate than the other (the absolute pose estimated from the light data). In fact, neither of the sources is required to be periodic for this operation to work.
In some embodiments the system uses active optical features that are modulated. Such optical features are sometimes referred to beacons or light sources. Such beacons can be modulated to produce an emission pattern that is temporally varied. In other words, the beacons can be turned on and off (e.g., flashed) at certain times or otherwise modulated. The beacons can also have emission patterns that change temporally in more complicated ways, e.g., by scanning the beacons to produce spatial and temporal (spatio-temporal) emission patterns. In most cases the active optical features are disposed in a certain pattern at known locations in world coordinates to support such temporal and spatio-temporal emission patterns.
Also provided is a method of generating input data from the pose estimate of the pose assumed by the manipulated object in the real three-dimensional environment that has optical features in it. The method calls for detecting by the photodetector mounted on-board the manipulated object light from the optical features. The photodetector generates light data corresponding to the optical features. The method further calls for generating by the auxiliary motion detection component mounted on-board the manipulated object relative motion data indicative of change in the pose. According to the method, the processor accepts the light data as well as the relative motion data indicative of change in the pose.
The application issues at an application request time tr a request to the processor for the pose estimate. The processor responds by providing to the application the pose estimate. The step of providing the pose estimate includes determining the absolute pose estimate of the pose from the light data at time ti, as well as determining the relative pose change from the relative motion data. The pose estimate is expressed with respect to world coordinates that describe the real three-dimensional environment. The pose estimate is based on the absolute pose estimate and the relative pose change. The application uses the pose estimate obtained from the processor to generate the input data.
The specifics of the invention and enabling details are described below with reference to the appended drawing figures.
To appreciate the basic aspects of the present invention, we initially turn to a simple version of an apparatus 10 in accordance with the invention, as shown in
Apparatus 10 processes absolute pose data 12 that describe the absolute pose of manipulated object 14 at a number of measurement times ti. Thus, successive pose data 12 collected at the chosen measurement times describe the motion that manipulated object 14 executes or is made to execute by a user 38.
Manipulated object 14 is any object that is moved either directly or indirectly by a user 38 and whose pose when object 14 is stationary or in motion yields useful absolute pose data 12. For example, manipulated object 14 is a pointer, a wand, a remote control, a three-dimensional mouse, a game control, a gaming object, a jotting implement, a surgical implement, a three-dimensional digitizer, a digitizing stylus a hand-held tool or any utensil. In fact, a person skilled in the art will realize that a manipulated object 14 can be even be an entire device such as a cell phone, a smart phone, a tablet or a smart object that is handled by user 38 to produce meaningful motion 40.
In the present case, manipulated object 14 is a pointer that executes motion 40 as a result of a movement performed by the hand of a user 38. Pointer 14 has a tip 16 that will be used as a reference point for describing its absolute pose in real three-dimensional environment 18. In general, however, any point on object 14 can be selected as reference point 16, as appropriate or convenient.
Pointer 14 has an on-board optical measuring arrangement 22 for optically inferring its absolute pose with the aid of one or more invariant features 32, 34, 36 disposed at different locations in real three-dimensional environment 18. Invariant features 32, 34, 36 are high optical contrast features such as edges of objects, special markings, or light sources. In the present embodiment, invariant feature 32 is an edge of an object such as a table (object not shown), invariant feature 34 is a special marking, namely a cross, and feature 36 is a light source. It is possible to use features 32, 34, 36 that are all located in a plane (coplanar) or else at arbitrary locations (non-coplanar) within real three-dimensional environment 18 as conveniently defined by global or world coordinates (Xo,Yo,Zo). The limitation is that, depending on the type of features 32, 34, 36, a sufficient number of them have to be visible to on-board optical measuring arrangement 22 at measurement times ti, as described in more detail below.
In the present embodiment the world coordinates (Xo,Yo,Zo) chosen to parameterize real three-dimensional environment 18 are Cartesian. A person skilled in the art will recognize that other choices including polar, cylindrical or still different coordinate systems can be employed. In addition, it will be appreciated that features 32, 34, 36 can be temporarily or permanently affixed at their spatial locations as required for measuring the pose of pointer 14. Indeed, the spatial locations of features 32, 34, 36 can be changed in an arbitrary manner, as long as on-board optical measuring arrangement 22 is appraised of their instantaneous spatial locations at times ti.
The spatial locations of features 32, 34, 36, whether temporary or permanent, are conveniently expressed in world coordinates (Xo,Yo,Zo). Furthermore, if possible, the spatial locations of features 32, 34 and 36 are preferably such that at least a subset of them is visible to on-board optical measuring arrangement 22 in all absolute poses that pointer 14 is expected to assume while undergoing motion 40. Invariant features 32, 34, 36 are used in deriving a relative or absolute position of tip 16 of pointer 14 in real three-dimensional environment 18. Features 32, 34, 36 are also used for optically inferring the remaining portion of the absolute pose, i.e., the orientation of pointer 14.
A number of optical measurement methods using optical measuring arrangement 22 to infer the relative or absolute pose of pointer 14 can be employed. In any of these methods, arrangement 22 uses one or more on-board components to obtain pose data 12 in accordance with any well-known absolute pose recovery technique including geometric invariance (among the geometric invariance approaches the ones using a collineation or a nomography matrix are especially suitable for this task), triangulation, ranging, path integration and motion analysis. In some embodiments arrangement 22 has a light-measuring component with a lens and an optical sensor that form an imaging system. In other embodiments arrangement 22 has an active illumination component that projects structured light or a scanning component that projects a scanning light beam into environment 18 and receives a scattered portion of the scanning light beam from features 32, 34. Specific examples of the various possible components will be explained in detail below.
Apparatus 10 has a processor 26 for preparing absolute pose data 12 corresponding to absolute pose of pointer 14 and for identifying a subset 48 of absolute pose data 12 required by an application 28. Specifically, application 28 uses subset 48 which may contain all or less than all of absolute pose data 12. Note that processor 26 can be located on pointer 14 or it can be remote, e.g., located in a remote host device, as is the case in this embodiment.
A communication link 24 is provided for sending absolute pose data 12 to application 28. Preferably, communication link 24 is a wireless communication link established with the aid of a wireless transmitter 30 mounted on pointer 14. In embodiments where processor 26 and application 28 are resident on pointer 14, communication link 24 can be a direct electrical connection. In still other embodiments, communication link 24 can be a wired remote link.
During operation user 38 holds pointer 14 in hand and executes a movement (i.e., user 38 manipulates pointer 14) such that pointer 14 executes motion 40 with respect to invariant features 32, 34, 36 in world coordinates (Xo,Yo,Zo) that parametrize real three-dimensional environment 18. For better visualization, motion 40 is indicated in dashed lines 42, 44 that mark the positions assumed by tip 16 and end 46 of pointer 14 during motion 40. For the purposes of this invention, line 42 is referred to as the trace of tip 16. In some specific applications of the present invention, trace 42 of tip 16 may be confined to a surface embedded in real three-dimensional environment 18. Such surface can be plane, e.g., a planar jotting surface, or it can be curved.
Motion 40 may produce no movement of end 46 or tip 16, i.e., no trace 42. In fact, motion 40 is not limited by any parameter other than those of standard mechanics of rigid body motion known form classical mechanics. Accordingly, changes in orientation of pointer 14 are considered to be motion 40. Likewise, changes in position of tip 16 (or any other reference point) in (x,y,z) coordinates conveniently expressed in world coordinates (Xo,Yo,Zo) are also considered to be motion 40. In the present case, orientation of pointer 14 is described by inclination angle θ, rotation angle φ and roll angle ψ referenced with respect to a center axis C.A. of pointer 14. A change in at least one of these angles constitutes motion 40. (Of course, zero change or no movement of pointer 14 is also a permitted motion 40).
In the present case, tip 16 moves along line 42 as pointer 14 is inclined with respect to a normal Z′ at inclination angle θ equal to θo. For simplicity, normal Z′ is selected to be parallel to the Zo axis of world coordinates (Xo,Yo,Zo). Furthermore, rotation and roll angles φ, ψ are equal to φo, ψo respectively. For convenience, in this embodiment angles θ, φ and ψ are Euler angles. Of course, other angles can be used to describe the orientation of pointer 14. In fact, a person skilled in the art will appreciate that any convention for describing the rotations of pointer 16 can be adapted for this description. For example, the four Cayley-Klein angles, the direction cosines, quaternions or still other descriptors of tilt, yaw and roll can be employed in such alternative conventions.
Now,
A third counterclockwise rotation by third Euler angle ψ is applied to twice rotated object coordinates (X′″,Y′″,Z′″) as shown in
Now, referring back to
For the purposes of the present invention, it is important to be able to optically infer, at least from time to time, the absolute pose of pointer 14. To do this, one relates Euler rotated object coordinates describing the orientation of pointer 14 to world coordinates (Xo,Yo,Zo). Note that the orientation of object axis Z′ in world coordinates (Xo,Yo,Zo) prior to the three Euler rotations is normal to plane (Xo,Yo). Second Euler angle θ defines the only counterclockwise rotation of object coordinates that is not about an object Z axis (this second rotation is about the X″=X′″ axis rather than axis Z′, Z″ or Z′″). Thus, Euler angle θ is an inclination angle θ between the completely Euler rotated object axis Z or axis C.A. and original object axis Z′, which is normal to plane (Xo,Yo).
Optical measuring arrangement 22 infers the absolute pose of pointer 14 during motion 40 at measurement times t1 and processor 26 prepares the corresponding absolute pose data 12. Note that absolute pose data 12 consist of inferred values of parameters (φ,θ,ψ,x,y,z) at measurement times ti. Invariant features 32, 34, 36 are located at positions that are defined in world coordinates (Xo,Yo,Zo). These positions stay fixed at least during measurement and usually permanently. Knowledge of the absolute positions of features 32, 34, 36 in world coordinates (Xo,Yo,Zo) allows the optical measuring arrangement 22 to describe the absolute pose of pointer 14 with absolute pose data 12 expressed in parameters (φ,θ,ψ,x,y,z) at measurement times ti in Euler rotated object coordinates within world coordinates (Xo,Yo,Zo). The expression of absolute pose data is preferably with respect to a reference location such as world origin (0,0,0) of world coordinates (Xo,Yo,Zo).
Of course, alternative locations within world coordinates can also be chosen as reference locations with respect to which the absolute pose of pointer 14 is expressed. For example, the center of invariant feature 34 may be chosen as the reference location and the locations of reference point 16 on pointer 14 at n measurement times ti can be denoted by corresponding n vectors Di, as shown in the drawing.
The frequency with which the absolute pose is inferred from the light data captured by on-board optical measuring arrangement 22, i.e., the times ti, depends on the use of absolute pose data 12 corresponding to that absolute pose and the desired performance, e.g., temporal resolution. It should be noted that periodic optical inference of absolute pose is not limited to any predetermined times ti or frequency schedule. In other words, the times between any two successive optical inferences or measurements of the absolute pose can be arbitrary. Preferably, however, arrangement 22 infers the absolute pose at a frequency that is high enough to obtain absolute pose data 12 that describe motion 40 at the temporal resolution required by application 28.
Wireless transmitter 30 of communication link 24 sends to processor 26 absolute pose data 12 here defined by parameters (φ,θ,ψ,x,y,z) and inferred from light data captured by on-board optical measuring arrangement 22 at measurement times ti. Absolute pose data 12 can be transmitted continuously, in bursts, in parts, at arbitrary or preset times or as otherwise desired. Processor 26 prepares a subset 48 of absolute pose data 12, for example the absolute position (x,y,z) of tip 16 and sends it to application 28. Application 28 uses absolute position (x,y,z) of tip 16 at measurement times ti to chart trace 42 of tip 16 as pointer 14 executes motion 40. In other words, unit 28 recovers trace 42 corresponding to the movement of tip 16. Note that the resolution of trace 42 in absolute space can be improved by increasing the sample of measurements of absolute trace points traversed in environment 18 by increasing the frequency of measurement times ti.
It should also be noted that pose data 12 should be formatted for appropriate communications between transmitter 30, processor 26 and application 28. Any suitable communication and formatting standards, e.g., IEEE interface standards, can be adapted for these purposes. For specific examples of formatting standards the reader is referred to Rick Poyner, LGC/Telegraphics, “Wintab™ Interface Specification: 16-bit and 32-bit API Reference”, revision of May 9, 1996; Universal Serial Bus (USB), “Device Class Definition for Human Interface Devices (HID)”, Firmware Specification, USB Implementers' Forum, Jun. 27, 2001 and six-degree of freedom interface by Ulrica Larsson and Johanna Pettersson, “Development and evaluation of a 6DOF interface to be used in a medical application”, Thesis, Linkopings University, Department of Science and Technology, Sweden, Jun. 5, 2002.
The orientation portion of absolute pose data 12, i.e., Euler angles (φ,θ,ψ) can also be used in the present embodiment. Specifically, processor 26 can prepare additional subsets or send all of the orientation parameters (φ,θ,ψ) of absolute pose data 12 as a single subset to application 28 or to a different application or device serving a different function. Any mix of orientation (φ,θ,ψ) and position (x,y,z) data derived from absolute pose data 12 can be used in subset 48. In fact, in some embodiments processor 26 keeps all absolute pose data 12 in subset 48 such that all of its parameters (φ,θ,ψ,x,y,z) can be used by application 28. This is done when application 28 has to reconstruct the entire motion 40 of pointer 14 and not just trace 42 of tip 16. For example, this is done when application 28 includes a motion-capture application. Once again, the temporal resolution of motion 40 can be improved by increasing the frequency of measurement times ti. Note that in this case parameters of pose data 12 that vary slowly are oversampled.
In
In step 58 all absolute pose data 12 are selected and passed to a subset formatting or preparing step 60A. In step 60A data 12 is prepared in the form of subset 48A as required by application 28. For example, data 12 is arranged in a particular order and provided with appropriate footer, header and redundancy bits (not shown), or as otherwise indicated by data porting standards such as those of Rick Poyner, LGC/Telegraphics (op. cit.).
In step 62, only a portion of data 12 is selected. Three exemplary cases of partial selection are shown. In the first case, only position data is required by application 28. Hence, in a step 59B only position data (x,y,z) are selected and the remaining data 12 are discarded. In a subsequent step 60B, position data (x,y,z) are prepared in the form of subset 48B as required by application 28 and/or as dictated by the porting standards.
In a second case, in a step 59C, only orientation data (ϕ,θ,ψ) are selected and the rest of data 12 are discarded. Then, in a step 60C, orientation data (ϕ,θ,ψ) are prepared in the form of a subset 48C for use by application 28.
In the third case, in a step 59D, a mix of data 12, including some position data and some orientation data are selected and processed correspondingly in a step 60D to prepare a subset 48D.
A person skilled in the art will appreciate that the functions described can be shared between processor 26 and application 28, e.g., as required by the system architecture and data porting standards. For example, some preparation of subset 48 can be performed by application 28 upon receipt. It should also be noted that in some embodiments data 12 can be pre-processed by transmitter 30 or post-processed at any point before or after preparation of the corresponding subset 48 in accordance with any suitable algorithm. For example, a statistical algorithm, such as a least squares fit can be applied to data 12 derived at different times ti or to successive subsets 48. Furthermore, quantities such as time derivatives of any or all parameters, i.e., (dx/dt, dy/dt, dz/dt, dφ/dt, dθ/dt, dψ/dt), can be computed. Also, various sampling techniques, e.g., oversampling can be used.
Subset 48 is transmitted to application 28 via a communication channel 72. Application 28 receives subset 48 as an input that is treated or routed according to its use. For example, in a step 64, subset 48 is used as control data. Thus, subset 48 is interpreted as an executable command 66 or as a part of an executable command. On the other hand, in a step 68, subset 48 is used as input data and saved to a data file 70.
In one embodiment, application 28 passes information to processor 26 to change the selection criteria for subset 48. Such information can be passed via communication channel 72 or over an alternative link, e.g., a feedback link 74. For example, application 28 requests subset 48A to be transmitted and uses subset 48A as input data for data file 70. At other times, application 28 requests subset 48C to be transmitted and uses subset 48C as command data for executable command 66. Alternatively, processor 26 can indicate a priori whether any subset 48 should be treated as input data or control data. In still another alternative, user 38 can indicate with the aid of a separate apparatus, e.g., a switch mounted on pointer 14 (not shown), whether subset 48 is intended as control data or input data. A person skilled in the art will recognize that there exist a large number of active and passive methods for determining the interpretation and handling of data being transmitted in subset 48 by both processor 26 and application 28.
In a specific application 28, subset 48 contains only position data (x,y,z) of reference point or tip 16 of pointer 14 collected at a number of measurement times ti. This subset corresponds to individual points along trace 42 and is an absolute trace expressed by points referenced with respect to origin (0,0,0) of world coordinates (Xo,Yo,Zo). For example, in a particular applications 28 trace 42 may be treated as a digital ink trace that is designed to be handled as input data or command data. Alternatively, the absolute points forming trace 42 can be expressed in world coordinates (Xo,Yo,Zo) with respect to a reference location other than world origin (0,0,0).
In practice, efficient inference of the absolute pose of pointer 14 in terms of absolute pose data expressed in parameters (φ,θ,ψ,x,y,z) representing Euler rotated object coordinates expressed in world coordinates (Xo,Yo,Zo) with respect to a reference location, such as world origin (0,0,0) imposes a number of important requirements. Since pointer 14 may be moving in a close-range environment 18 the field of view of on-board optical measuring arrangement 22 must be large. This is particularly crucial in situations where arrangement 22 has to tolerate frequent occlusions of one or more of invariant features 32, 34, 36. Such conditions arise when user 38 operates pointer 14 in a close-range home, gaming or work environment 18, i.e., in a room, a cubicle or other confined real space. Also, if full motion capture is desired, then the rate or frequency of measurement times ti has to be high in comparison to the rate of movement of the hand of user 38.
To learn how to address these and other practical considerations, we turn to another embodiment of an apparatus 100 according to the invention as shown in
Apparatus 100 is deployed within a real three-dimensional environment 108. In the case at hand, environment 108 is defined within a room 110 and it is parametrized by global or world coordinates (Xo,Yo,Zo) whose world origin (0,0,0) is posited in the lower left rear corner of room 110.
As in the previous embodiment, world origin (0,0,0) is selected as the reference location for expressing the measured values of parameters (φ,θ,ψ,x,y,z) that represent absolute pose data of manipulated object 102 in Euler rotated object coordinates (X,Y,Z). The three successive rotations by Euler angles (φ,θ,ψ) to obtain Euler rotated object coordinates (X,Y,Z) are also indicated in
A number of invariant features B1-B7 are placed at known locations in real three-dimensional environment 108 delimited by room 110.
Vectors R1-R7 define the locations of corresponding invariant features B1-B7. Following standard convention, vectors R1-R7 extend from world origin (0,0,0) to the centers of the corresponding invariant features B1-B7. All seven invariant features B1-B7 are high optical contrast features. More precisely, invariant features B1-B7 are light sources such as light-emitting diodes that emit electromagnetic radiation or light 112. Preferably, light 112 is in the infrared wavelength range of the electromagnetic spectrum. Light-emitting diodes in that range are typically referred to as infrared emitting diodes or just IR LEDs. For clarity, only four of the seven IR LEDs B1-B7 are shown simultaneously emitting light 112 in
Optical measuring arrangement 104 with light-measuring component 106 is mounted on-board, and more precisely on one of the sides of manipulated object 102. Component 106 is an absolute motion detection component equipped with a lens 114 and a photodetector or optical sensor 116 shown in detail in the cut-away view of manipulated object 102 depicted in
It should be noted, however, that the handling of manipulated object 102 does not need to be carried out directly by a user. In fact, object 102 can be a remotely controlled object or even an object that is cast or thrown by the user. Whether object 102 is manipulated directly or remotely and whatever its spatial trajectory in environment 108, it is crucial that light-measuring component 106 be optimally placed on object 102 to have a direct line-of-sight to most or all IR LEDs B1-B7 while object 102 is undergoing its intended motion. That is because component 106 needs to capture light 112 emitted by IR LEDs B1-B7 so that it can use these invariant features for optically inferring the values of parameters (φ,θ,ψ,x,y,z). Taken together, parameters (φ,θ,ψ,x,y,z) represent absolute pose data 118 that describes the absolute pose of manipulated object 102.
An appropriate choice of lens 114 will aid in addressing the above optics challenges. Obviously, lens 114 has to be small, robust and low-cost (e.g., moldable in acrylic or other plastic). Lens 114 should not require active focusing and it should have a low F-number (e.g., F #≈1.6 or less) to ensure high light gathering efficiency. At the same time, lens 114 should exhibit low levels of aberration and have a single viewpoint. In other words, lens 114 should exhibit quasi-pinhole optical characteristics. This last attribute is especially important when manipulated object 102 is expected to sometimes pass within a short distance of IR LEDs B1-B7. Under such conditions, the limited depth of field inherent in a normal refractive lens, especially one without active focal length adjustment, would cause a loss of optical information; a familiar problem in machine vision. U.S. Pat. Nos. 7,038,846 and 7,268,956, both to Mandella, teach a suitable design of a catadioptric lens that satisfies these stringent demands.
Apparatus 100 has a processor 120 for preparing pose data 118. In this exemplary embodiment, processor 120 is not on-board manipulated object 102 but is instead integrated in a computing device 122. For example, processor 120 may be a central processing unit (CPU), a graphics processing unit (GPU) or some other unit or combination of units resident on computing device 122. Computing device 122 is shown as a stationary device, but it is understood that it could be a portable device or an ultra-mobile device including a tablet, a PDA or a cell phone.
Besides preparing absolute pose data 118, processor 120 is entrusted with identifying a subset 118′ of pose data 118. As in the prior embodiment, the preparation of pose data 118 may include just collecting the inferred values of parameters (φ,θ,ψ,x,y,z) corresponding to the absolute pose of object 102. In more involved cases, the preparation of data 118 can include pre- and/or post-processing as well as computation of functions derived from measured values of one or more of parameters (φ,θ,ψ,x,y,z) (including the application of statistical algorithms to one or more these parameters). Meanwhile, identification of subset 118 has to do with the intended use of data 118 and the nature of its application.
Computing device 122 not only hosts processor 120, but also has a display 124 for displaying an output 126 to the user. Output 126 is generated by an application 128 that is running on computing device 122. Application 128 and its output 126 dictate what subset 118′ needs to be identified and supplied by processor 120. A simple case arises when application 128 is configured to produce as output 126 a visual element such as a token or even an image of object 102 and compute as well as show its absolute trajectory within room 110 in world coordinates (Xo,Yo,Zo) with respect to reference location (0,0,0). A person skilled in the art will easily discern, that under these constraints application 128 will require that all parameters (φ,θ,ψ,x,y,z) be included in subset 118′. This way, as time progresses, application 128 will be able to alter output 126 in response to the absolute pose of object 102 inferred from light data at different times ti and, if desired, display a replica of its full trajectory within room 110. Application 128 can display this information as output 126 on display 124 to the user as shown in
Computing device 122 employs its own internal communication link 130, e.g., a data bus, to transmit subset 118′ to application 128. Meanwhile, a wireless communication link 132 is provided for transmitting data 118 from manipulated object 102 to computing device 122. Wireless link 132 employs a transmitting unit 134A on object 102 and a receiving unit 134B on device 122.
When manipulated object 102 moves within room 110 on-board optical measuring arrangement 104 deploys absolute motion detection component 106. Here, component 106 is a light-measuring component that gathers light 112 emitted from IR LEDs B1-B7. Preferably, all IR LEDs B1-B7 are on at measurement times ti when the values of parameters (φ,θ,ψ,x,y,z) describing the absolute pose of object 102 are being measured.
As shown in more detail in
Electronics 142 can either fully process signals from sensor 116, or only pre-process them to obtain raw image data. The choice depends on whether fully processed or raw absolute pose data 118 is to be transmitted via wireless link 132 to computing device 122. When sufficient on-board power is available, performing most or all image processing functions on-board object 102 is desirable. In this case electronics 142 include all suitable image processing modules to obtain measured values of parameters (φ,θ,ψ,x,y,z) in their final numeric form. Data 118 being transmitted via link 132 to computing device 122 under these conditions is very compact. On the other hand, when on-board power is limited while the bandwidth of wireless communication link 132 is adequate, then electronics 142 include only the image processing modules that extract raw image data from sensor 116. In this case, raw absolute pose data 118 is transmitted to computing device 122 for further image processing to obtain the inferred or measured values of parameters (φ,θ,ψ,x,y,z) in their final numeric form.
In the present embodiment, sensor 116 is a CMOS sensor with a number of light-sensing pixels 144 arranged in an array 145, as shown in
The additional image processing modules depicted in
For example, extraction of invariant features such as edges, corners and markings will require the application of suitable image segmentation modules, contrast thresholds, line detection algorithms (e.g., Hough transformations) and many others. For more information on edge detection in images and edge detection algorithms the reader is referred to U.S. Pat. Nos. 6,023,291 and 6,408,109 and to Simon Baker and Shree K. Nayar, “Global Measures of Coherence for Edge Detector Evaluation”, Conference on Computer Vision and Pattern Recognition, June 1999, Vol. 2, pp. 373-379 and J. Canny, “A Computational Approach to Edge Detection”, IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 8, No. 6, November 1986 for basic edge detection all of which are herein incorporated by reference. Additional useful teachings can be found in U.S. Pat. No. 7,203,384 to Carl and U.S. Pat. No. 7,023,536 to Zhang et al. A person skilled in the art will find all the required modules in standard image processing libraries such as OpenCV (Open Source Computer Vision), a library of programming functions for real time computer vision. For more information on OpenCV the reader is referred to G. R. Bradski and A. Kaehler, “Learning OpenCV: Computer Vision with the OpenCV Library”, O'Reilly, 2008.
In the present embodiment, the absolute pose of object 102 including the physical location (x,y,z) of reference point 102′ (described by vector Go) and the Euler angles (φ,θ,ψ) are inferred with respect to world origin (0,0,0) with the aid of vectors R1-R7. To actually compute these parameters from on-board object 102 it is necessary to recover vectors R1-R7 from images 140 of IR LEDs B1-B7 contained in an image 160 of room 110 as shown on the surface of sensor 116 in
In practical situations, occlusion 162 as well as any other occlusions can be due to the user's body or other real entities or beings present in environment 108 obstructing the line-of-sight between lens 114 and IR LED B5. Also note, that if too few of IR LEDs B1-B7 are imaged, then inference of the absolute pose of object 102 may be impossible due to insufficient data. This problem becomes particularly acute if IR LEDs B1-B7 are not distinguishable from each other. Therefore, in a practical application it is important to always provide a sufficiently large number of IR LEDs that are suitably distributed within environment 108. Alternatively or in addition to these precautions, IR LEDs B1-B7 can be made distinguishable by setting them to emit light 112 at different wavelengths.
Referring again to
To properly perform its function, module 154 needs to calibrate the location of the center of image coordinates (Xi,Yi) with respect to reference point 102′. This calibration is preferably done prior to manipulating object 102, e.g., during first initialization and testing or whenever re-calibration of origin location becomes necessary due to mechanical reasons. The initialization can be performed with the aid of any suitable algorithm for fixing the center of an imaging system. For further information the reader is referred to Carlo Tomasi and John Zhang, “How to Rotate a Camera”, Computer Science Department Publication, Stanford University and Berthold K. P. Horn, “Tsai's Camera Calibration Method Revisited”, which are herein incorporated by reference.
Armed with the mapping provided by module 154, module 156 obtains the inferred values of parameters (φ,θ,ψ,x,y,z), which represent absolute pose data 118. Data 118 now properly represents the final numerical result that describes the absolute pose of object 102 inferred from light data. This description is made in terms of inferred values of parameters (φ,θ,ψ,x,y,z), which are the Euler rotated object coordinates expressed in world coordinates (Xo,Yo,Zo) with respect to world origin (0,0,0). In the last step, module 158 identifies a subset 118′ of parameters (φ,θ,ψ,x,y,z) to be sent to application 128.
In practice, due to certain optical effects including aberration associated with lens 114, the non-occluded portion of image 160 will exhibit a certain amount of rounding. This rounding can be compensated optically by additional lenses (not shown) and/or electronically during undistortion performed by module 154. Preferably, the rounding is accounted for by applying a transformation to the non-occluded and detected portion of image 160 by module 154. For example, module 154 has an image deformation transformer based on a plane projection to produce a perspective view. Alternatively, module 154 has an image deformation transformer based on a spherical projection to produce a spherical projection. Advantageously, such spherical projection can be transformed to a plane projection with the aid of well-known methods, e.g., as described by Christopher Geyer and Kostas Daniilidis, “A Unifying Theory for Central Panoramic Systems and Practical Implications”, www.cis.upenn.edu, Omid Shakernia, et al., “Infinitesimal Motion Estimation from Multiple Central Panoramic Views”, Department of EECS, University of California, Berkeley, and Adnan Ansar and Kostas Daniilidis, “Linear Pose Estimation from Points or Lines”, Jet Propulsion Laboratory, California Institute of Technology and GRASP Laboratory, University of Pennsylvania which are herein incorporated by reference.
It should also be remarked, that once image 160 is recognized and transformed, a part of the orientation, namely Euler angles (φ,θ) of object 102 can be inferred in several ways. For example, when working with the spherical projection, i.e., with the spherical projection of unobstructed portions of image 160, a direct three-dimensional rotation estimation can be applied to recover inclination angle θ and polar angle φ. For this purpose, a normal view of room 110 with IR LEDs B1-B7 is stored in a memory (not shown) such that it is available to module 154 for reference purposes. The transformation then yields the Euler angles (φ,θ) of object 102 with respect to IR LEDs B1-B7 and any other high optical contrast invariant features in room 110 by applying the generalized shift theorem. This theorem is related to the Euler theorem stating that any motion in three-dimensional space with one point fixed (in this case the reference point 102′ may be considered fixed for the duration of one measurement time ti) can be described by a rotation about some axis. For more information about the shift theorem the reader is referred to Ameesh Makadia and Kostas Daniilidis, “Direct 3D-Rotation Estimation from Spherical Images via a Generalized Shift Theorem”, Department of Computer and Information Science, University of Pennsylvania, which is herein incorporated by reference.
Alternatively, when working with a plane projection producing a perspective view of unobstructed portions of image 160 one can use standard rules of geometry to determine inclination angle θ and polar angle φ. Several well-known geometrical methods taking advantage of the rules of perspective views can be employed in this case, as mentioned above.
Referring back to
A person skilled in the art will realize that the embodiment of apparatus 100 shown in
Another embodiment of an apparatus 200 in accordance with the invention is shown in the three-dimensional diagrammatical view of
Environment 204 is an outdoor environment with ambient light 220 provided by the sun over the usual solar spectral range Δλamb. A certain number n of invariant features B1-Bn are affixed at known locations in environment 204. Vectors b1-bn are employed to describe the locations of corresponding invariant features B1-Bn in world coordinates (Xo,Yo,Zo). All invariant features B1-Bn are high optical contrast features, and, more specifically, they are IR LEDs for emitting electromagnetic radiation or light 222 in the infrared range of the electromagnetic spectrum.
When invariant features are embodied by light sources that are controlled they will be referred to as beacons. Beacons are preferably one-dimensional or point-like and they are implemented by light emitting diodes (LEDs), laser diodes, IR LEDs, optical fibers and the like. Of course, beacons can also be extended sources such as lamps, screens, displays and other light sources as well as any objects providing sufficiently highly levels of electromagnetic radiation that can be controlled. These include projected points and objects, as well as points and objects concentrating and reflecting radiation originating in environment 204 or active illumination from on-board manipulated object 202. The advantage of beacons over simple and uncontrolled light sources is that they are distinguishable.
It is the emission pattern of beacons B1-Bn that is controlled in the present embodiment. Hence, they are distinguishable and play the role of beacons. The emission pattern of beacons B1-Bn is dictated by locations b1-bn at which they are affixed in environment 204 and their on/off timing. In other words, the emission pattern is spatially set by placing beacons B1-Bn at certain locations and it is temporally varied by turning the beacons on and off at certain times. It will be appreciated that more complicated emission patterns can be practiced. For example, it is possible to configure beacons B1-Bn to produce spatial and temporal (spatio-temporal) emission patterns, such as scanning patterns.
Beacons B1, B2, . . . , Bn are controlled by corresponding controls C1, C2, . . . , Cn and a central unit 224 that communicates with the controls. The communications between unit 224 and controls C1, C2, . . . , Cn are carried by wireless up-link and down-link signals 226A, 226B. Of course, any method of communication, including wired or optical, can be implemented between central unit 224 and controls C1, C2, . . . , Cn. Different communication equipment will typically require different supporting circuitry, as will be appreciated by those skilled in the art. Taken together, controls C1, C2, . . . , Cn and unit 224 form an adjustment mechanism 228 for setting or adjusting a sequenced emission pattern of IR LEDs B1, B2, . . . , Bn. In other words, adjustment mechanism 228 is capable of modulating all IR LEDs B1-Bn in accordance with a pattern.
Object 202 has an on-board optical measuring arrangement 206 consisting of an absolute motion detection component 208. Component 208 is a light-measuring component with a lens 210 and a photodetector or an optical sensor 212. Light-measuring component 208 has an optical filter 216 positioned before sensor 212, as well as image processing electronics 218 connected to sensor 212. As in the prior embodiment, lens 210 is preferably a wide field of view lens with a substantially single viewpoint 214. Viewpoint 214 is selected as the reference point on manipulated object 202 for expressing the location parameters (x,y,z) of its absolute pose and its orientation parameters (φ,θ,ψ). Hence, vector Go in this embodiment extends from world origin (0,0,0) to viewpoint 214.
Once again, the absolute pose of object 202 in this embodiment is expressed in the Euler rotated object coordinates (X,Y,Z), whose origin is now attached to viewpoint 214. The manner in which rotations by Euler angles (φ,θ,ψ) are applied to object 202 to express the Euler rotated object coordinates (X,Y,Z) are analogous to the convention explained above and will therefore not be repeated.
The choice of viewpoint 214 of lens 210 as the reference point is very convenient for tracking object 202 and it does not limit the choice of object coordinates, as will be appreciated by those skilled in the art. As before, the absolute pose of object 202 is completely described by six parameters, namely the three components (x,y,z) of displacement vector Go from the origin of global coordinates (Xo,Yo,Zo) to the reference point, in this case viewpoint 214, and the three Euler angles (φ,θ,ψ). A trajectory 230 of object 202 is thus fully described by these six parameters and time t, i.e., (x,y,z,φ,θ,ψ,t).
Notice that lens 210, although shown as a single element in the previous and current embodiments can be compound. In other words, lens 210 can consist of several optical elements including various combinations of refractive and reflective elements. In any of these embodiments, the effective viewpoint 214 can be determined and chosen as reference point on object 202.
Optical sensor 212 of absolute motion detection component 208 is a photosensor designed for sensing light 222 from IR LEDs B1-Bn.
In fact, rather than being a sensor with an array of pixels, photosensor 212 in the present embodiment is a centroid sensing device that is a type of photo-diode often called a position-sensing device (PSD). PSD 212 determines a centroid of the flux of light 222 impinging on it.
Lens 210 has a field of view sufficiently large to capture electromagnetic radiation or light 222 emitted by most or all beacons B1-Bn and image it onto on-board centroid sensing device or PSD 212. Mathematically, it is known that to infer the absolute pose of object 202, i.e., to infer or measure the values of all parameters (x,y,z,φ,θ,ψ) of object 202 in environment 204, at least four among distinguishable beacons B1-Bn need to be in the field of view of lens 210.
Optical filter 216 placed before PSD 212 reduces the level of ambient light 220 impinging on PSD 212. Concurrently, the wavelengths of electromagnetic radiation or light 222 provided by LEDs B1-Bn are selected such that they are passed by filter 216. In the present case, ambient radiation 220 is produced by the sun and spans an emission spectrum Δλamb, whose intensity (I) peaks in the visible range and drops off in the infrared range as generally shown by graph 250 in
It is optional whether all wavelengths λ1, λ2, . . . , λn are different or equal. In some embodiments, different wavelengths can be used to further help differentiate between IR LEDs B1-Bn. In other embodiments different spatio-temporal or scan patterns can help distinguish between IR LEDs B1-Bn. The present embodiment, however, is simple for reasons of clarity. Here, all IR LEDs B1-Bn are emitting at the same emission wavelength λe equal to 950 nm. A transmittance (T) of filter 216 is selected as shown by graph 254 in
Returning to
Lens 210 produces an imaged distribution 232 of electromagnetic radiation 222 on PSD 212. PDS 212, in turn, generates electrical signals that represent the x-y position of a center-of-mass or centroid 234 of imaged distribution 232 in x-y plane of PSD 212. In the present case, IR LED B4 is a point-like source of electromagnetic radiation 222 and therefore lens 210 images it to a spot-type distribution 232. In general, it is desirable to keep spot 232 relatively small by appropriate design of lens 210, which is preferably a lens with good imaging properties including low aberration, single viewpoint imaging and high-performance modulation transfer function (MTF). In general, however, optic 210 can be refractive, reflective or catadioptric. It should also be noted that in the event of certain modulation patterns, e.g., scan patterns, it may even be unnecessary to use optic 210. This could be the case when the spatio-temporal modulation pattern is well controlled; e.g., it consists of well-timed scan-lines.
For a better understanding of PSD 212 we turn to the plan view diagram of its top surface 236 shown in
PSD 212 has two electrodes 238A, 238B for deriving signals corresponding to the x-position, namely xi+ and xi−, and two electrodes 238C, 238D for obtaining yi+ and yi− signals corresponding to the y-position. The manner in which these signals are generated and processed to obtain the location (xi,yi) of centroid 234 is well-known to those skilled in the art and will not be discussed herein. For more information on the subject the reader is referred to manufacturer-specific PSD literature, such as, e.g., “PSD (Position Sensitive Detector)” Selection Guide of Hamamatsu, Solid State Division, July 2003.
The intensities 232X, 232Y of imaged distribution 232, i.e., spot 232, along the Xi and Yi axes are visualized along the sides. Another imaged distribution 240 due to ambient radiation 220 is also indicated with a dashed line. Corresponding intensities 240X, 240Y along the Xi and Yi axes are also visualized along the sides. Because of the action of filter 216, intensities 240X, 240Y are low in comparison to 232X, 232Y and the corresponding centroid position thus includes a negligibly small shift error due to the background noise on the desired signal. Such background can be removed with any well-known electronic filtering technique, e.g., standard background subtraction. Corresponding electronics are known and will not be discussed herein.
PSD 212 is connected to image processing electronics 218 and delivers signals xi+, xi− and yi+, yi− to it. Electronics 218 are also in communication with central unit 224 by any suitable link so that it knows which beacon is active (here beacon B4) and thus responsible for centroid 234 at any given time. It is convenient to establish the link wirelessly with up-link and down-link signals 226A, 226B, as shown in
During operation, optical apparatus 200 uses the knowledge of which beacon produces centroid 234 described by image coordinates (xi,yi) and the beacon's location in environment 204 or world coordinates (Xo,Yo,Zo) to infer the absolute pose of object 202 in terms of measured values of parameters (x,y,z,φ,θ,ψ). Note that beacons B1-Bn need not be attached or affixed at any permanent location in environment 204, as long as their location at the time of emission of radiation 222 is known to apparatus 200. Moreover, any sequenced pattern of beacons B1-Bn can be used, even a pattern calling for all beacons B1-Bn to be on simultaneously. In the latter case, a constellation of n spots is imaged on PSD 212 and centroid 234 is the center of mass (C.O.M.) of the entire constellation of n spots 232, i.e., it is not associated with a single spot. Of course, in that case the ability to distinguish the beacons is removed and the performance of apparatus 200 will be negatively affected.
For better clarity of explanation, we first consider a modulation or sequenced pattern with only one beacon on at a time. Following such pattern, beacon B4 is turned off and beacon Bm is turned on to emit radiation 222. Note that an intensity distribution 242 of radiation 222 has a wide cone angle such that lens 210 can image radiation 222 even at steep angles of incidence. Alternatively, given knowledge of all possible relative positions between object 202 and beacon Bm, a mechanism can be provided to optimize angular distribution 242 for capture by lens 210.
To commence motion capture, controls C1-Cn and unit 224, i.e., adjustment mechanism 228 implements an initial sequenced pattern of IR LEDs B1-Bn. The initial pattern can be provided by image processing electronics 218 to unit 224 of adjustment mechanism 228 via up-link signals 226A. The initial pattern can be based on any parameter of the last known or inferred absolute pose or any other tracking information. Alternatively, initial sequenced pattern is standard.
A flow diagram in
Motion capture starts in step 276. In step 278 signals xi+, xi− and yi+, yi− encoding centroid 234 of activated beacon are sent from PSD 212 to electronics 218 for processing. In step 280 signals are tested for presence (sufficient power level for further processing) and are then filtered in step 282 to obtain filtered data corresponding to centroid 234. Filtering includes background subtraction, signal gain control including lock-in amplification and/or other typical signal processing functions. Absence of signals xi+, xi− and yi+, yi− is used to flag the corresponding beacon in step 284.
After filtering, the data is normalized in step 286. This step involves time-stamping, removing effects of known optical aberrations due to lens 210 and preparing the data for processing by either absolute or relative tracking or navigation algorithms. Normalization also formats data points from each cycle and may include buffering the data, if necessary, while centroid 234 from the next beacon in the pattern is queued up or buffering until a sufficient number of centroids 234 have been captured to perform reliable normalization. In a preferred embodiment, beacons B1, B2, . . . , Bn are amplitude modulated with a series of pulses. In this embodiment, normalization further includes selection of the pulse with most suitable amplitude characteristics (e.g., full dynamic range but no saturation) and discarding signals from other pulses.
In step 288 normalized data of centroid 234 is sent to a tracking or navigation algorithm 290. Contemporaneously, or earlier depending on timing and buffering requirements, absolute pose and motion capture program 270 submits a query 292 whether the first cycle of initial sequenced pattern in complete. The answer is used by navigation algorithm 290 in determining at least one parameter of the pose of object 202 and to prepare for capturing the next centroid in step 294.
Navigation algorithm 290 preferably determines all parameters (x,y,z,φ,θ,ψ) at initialization time tinit. in global coordinates (Xo,Yo,Zo) based on known locations of beacons B1, B2, . . . , Bn, i.e., known vectors b1, b2, . . . , bn. Only centroids 234 that are available (i.e., no drop out of corresponding beacon or other failure) and yield reliable centroid data are used. At least four centroids 234 need to be captured from the initial sequenced pattern to measure the values of parameters (x,y,z,φ,θ,ψ) in world coordinates (Xo,Yo,Zo). The pose is called absolute when all parameters are known in global coordinates (Xo,Yo,Zo) at a given time, e.g., at tinit. Navigation using absolute pose or at least one parameter of absolute pose is referred to as absolute tracking or absolute navigation.
In a particular embodiment, beacons B1, B2, . . . , Bn are positioned on a plane in a rectangular grid pattern and parameters (x,y,z,φ,θ,ψ) are inferred or measured based on projective, i.e., perspective geometry. In this approach the rules of perspective geometry using the concept of vanishing points lying on a horizon line are applied to determine the location of point of view 214. Specifically, given the locations of at least four coplanar beacons lying on at least three straight intersecting lines framing a rectangular grid in the field of view F.O.V. of lens 210, absolute navigation algorithm 290 defines a horizon and finds conjugate vanishing points from which point of view 214 is determined. Once point of view 214 is known, parameters (x,y,z,φ,θ,ψ) of object 202 are inferred or measured. Initially, point of view 214 is the origin or reference point at (x,y,z). As mentioned above, any other point on object 202 can be used as a reference point based on a coordinate transformation. The perspective geometry and vector algebra necessary to perform absolute navigation are known to skilled artisans of optical image processing and will not be discussed herein. For more details, the reader is referred to K. Kanatani, “Geometric Computation for Machine Vision”, Oxford Science Publications; Clarendon Press, Oxford; 1993, Chapters 2-3 and to U.S. Pat. No. 7,203,384 to Carl.
In embodiments where a large number of beacons are used and are available (low drop out), the rules of perspective geometry can be employed to filter beacons that are non-conformant therewith. In other words, the perspective geometry constraint can be used as an additional filter for high-precision absolute tracking or navigation.
Absolute pose expressed with inferred or measured values of parameters (x,y,z,φ,θ,ψ) computed by image processing electronics 218 at initial time tinit. in step 290 is used to update trajectory 230 during pose update step 296. Depending on the motion of object 202 and required resolution or accuracy for trajectory 230, the centroid capture rate and time between determinations of absolute pose should be adjusted. At high-speed capture rates absolute navigation algorithm 290 can keep updating parameters (x,y,z,φ,θ,ψ) in a continuous fashion based on at least four most recently captured centroids or even as each successive centroid is obtained. This can be accomplished by substituting the most recently captured centroid for the oldest centroid. Computed trajectory 230, expressed with absolute pose parameters and time (x,y,z,φ,θ,ψ,t), is output in step 298 to an application in the form of a subset. The subset may contain all or fewer than all of the parameters (x,y,z,φ,θ,ψ,t), depending on the requirements of the application.
The application requires knowledge of object's 202 movements for operation, feedback, input, control or other functions. The application has a control mechanism that initiates and terminates operation of motion capture program via control command 300. In several advantageous applications object 202 is a hand-held object that is manipulated directly by the user and trajectory 230 is used as input for the application, as will be addressed in more detail below.
Preferably, upon completion of one cycle of initial sequenced pattern a re-evaluation is performed in step 302. During re-evaluation beacons flagged during step 284 are removed from the data set or the optimized sequenced pattern to speed up operation. Beacons that fail in filtering or normalization steps 282, 286 may be adjusted or left out as well. Finally, any high quality beacons as determined by tracking or navigation algorithm 290 can be used for benchmarking or weighting. Of course, these decisions can be periodically re-checked to ensure that beacons yielding high quality data at a different pose are not turned off permanently. Additionally, intermittent background measurements are made with all beacons off at regular intervals or on an as-needed basis for background subtraction.
Alternatively, optimization and re-evaluation of the sequenced pattern is performed on-the-fly. In this case the initial cycle does not need to be completed and information from some beacons, e.g., the latter portion of the cycle may be disregarded altogether.
In a preferred embodiment of the method, the sequenced pattern of emission of radiation 222 by the beacons is controlled based on the one or more absolute pose parameters determined by tracking or navigation algorithm 290. The control can be a temporal control as in when the beacons are on, or spatial control of which beacons should be used and/or which beacons should be relocated and affixed at new locations in the environment. To this effect, in step 304 an optimized sequenced pattern is prepared based on the re-evaluation from step 302. If the application issues request 306 for further output from motion capture program 270, then the optimized sequenced pattern is activated in step 308 and the cycle of centroid capture re-starts at step 278. Otherwise, motion capture program is terminated in step 310.
In an alternative embodiment, motion capture program 270 employs an absolute navigation algorithm 290 that only determines a subset of absolute pose parameters (x,y,z,φ,θ,ψ). In one example, only (x,y,z) parameters defining the position of point of view 214 (vector Go) or some other reference point on object 202 are determined. These parameters can be used when orientation parameters (φ,θ,ψ) are not required by the application. An example of such application is a three-dimensional digitizer. In another example, only orientation parameters (φ,θ,ψ) of the pose of object 202 are determined. These can be used by an application that requires only orientation or angle information for its input or control functions, e.g., when object 202 is a remote pointer, joystick, three-dimensional controller, pointer, other hand-held object or indeed any object in need of angular tracking or navigation only.
In still another alternative embodiment, motion capture program 270 employs a relative navigation algorithm 290′ that only determines changes in some or all parameters (Δx,Δy,Δz,Δφ,Δθ,Δψ). For example, navigation algorithm 290′ determines linear velocities (dx/dt, dy/dt, dz/dt) and/or angular velocities (dφ/dt, dθ/dt, dψ/dt), accelerations (d2x/dt2, d2y/dt2, d2z/dt2, d2φ/dt2, d2θ/dt2, d2ψ/dt2) or higher order rates of change, such as jerk, of any absolute pose parameter or combinations thereof. It should be noted that some applications may not require absolute pose. Thus, absolute pose may not be inferred or measured at all and only relative navigation algorithm 290′ is deployed. Thus, the rates of change may be the results of variations of unknown combinations of absolute pose parameters. Relative navigation algorithm 290′ is advantageous for applications that do not require knowledge of trajectory 230 but just rates of change. Such applications include navigation of relative hand-held devices such as two-dimensional mice, three-dimensional mice, relative mouse-pens and other low-accuracy controls or relative input devices.
Apparatus 200 is inherently low-bandwidth, since PSD 212 reports just four values, namely (xi+, xi−, yi+, yi−) corresponding to the location of centroid 234 produced by one or more known beacons. The intrinsically high signal-to-noise ratio (SNR) of centroid 234 due to low background noise allows apparatus 200 to operate at high capture rates, e.g., up to 10 kHz and higher, rendering it ideal for tracking fast moving objects. In fact, apparatus 200 is sufficiently robust to navigate even rapidly moving hand-held objects, including pointers, controllers, mice, high-precision gamer instruments, jotting implements and the like in close-range environments or constrained areas such as desks, hand-held notepads, point-of-sale environments and various game- and work-spaces.
Optical navigation apparatus 200 admits of many more specific embodiments. First and foremost, centroid sensing device 212 can use various physical principles to obtain the centroid of imaged distribution 232 of electromagnetic radiation 222 (and ambient radiation 220). A person skilled in the art will recognize that even a regular full field sensor, e.g., a digital CMOS sensor, can act as centroid sensing device 212. In general, however, the use of a standard full-frame capture CMOS sensor with a large number of individual pixels will not be very efficient. That is due to the large computational burden associated with processing large numbers of image pixels and lack of intrinsic facility in centroid sensing. In addition, fast motion capture and high frame rates required for navigating hand-held objects with on-board optical measuring arrangement are not compatible with the high-power and large bandwidth requirements of digital CMOS sensors.
Optical apparatus 200 for processing pose data can employ many other types of centroid sensing devices as PSD 212. Some examples of such devices can be found in U.S. Patent Application 2007/0211239 to Mandella et al. A particularly convenient centroid sensing device has circular and planar geometry conformant to the naturally circular F.O.V. of lens 210.
The circular geometry of PSD 250 enables operation in polar coordinates (R,θ). In this convention each of four spots 232A, 232B, 232C and 232D has a centroid 234A, 234B, 234C and 234D described by polar coordinates (R1,θ1), (R2,θ2), (R3,θ3) and (R4,θ4). However, due to its principles of operation PSD 350 reports to electronics 218 only polar coordinates (Rc,θc), of the C.O.M.
A set of dashed arrows show the movement of centroids 234A, 234B, 234C and 234D and C.O.M. as a function of time. Note that applying optical flow without inferring or measuring the absolute pose of object 202 indicates an overall rotation and can be used as input for any relative motion device, e.g., an optical mouse. In such functional mode, absolute motion component 208 operates as an auxiliary motion component and more precisely an optical flow measuring unit that determines relative motion. Relative motion information obtained from optical flow can be very valuable and it can supplement absolute pose data in certain cases. For example, it can be used to estimate motion of object 202 between times ti when absolute pose is inferred or measured from light data.
In the last step, absolute pose data 248 consisting of all absolute pose parameters (x,y,z,φ,θ,ψ) are transmitted to an application running on control unit 224 via a wireless communication link 244 using a transceiver 246A on-board object 202 and a transceiver 246B on unit 224. In this embodiment unit 224 is running a monitoring application to supervise manipulated object 202 without displaying any output.
Note that in this embodiment, electronics 218 can pick the subset that is needed for the monitoring application running on unit 224. An uplink exists from unit 224 back to electronics 218 (as indicated) to communicate changes in the required subset or subsets for the application as they may arise. Thus, if manipulated object 202 is not experiencing any linear displacements, i.e., the coordinates (x,y,z) of its viewpoint 214 are static, then the subset of orientation parameters (φ,θ,ψ) is not relevant and does not need to be requested by unit 224.
A series of IR LEDs B1-Bn (not all shown) are located in environment 402 at known locations in world coordinates (Xo,Yo,Zo). IR LEDs B1-Bn are distinguishable since they are modulated as beacons in a sequenced pattern that is remotely controlled by a computing device 414. Beacons B1-Bn emit light 416 at a fixed wavelength in the infrared range. Each beacon has a large cone angle 418, as exemplified by beacon B2.
In a manner similar to the previous embodiment, component 410 infers the absolute pose of manipulated object 406 it terms of measured values of parameters (x,y,z,φ,θ,ψ) from observing sequentially flashing beacons B1-Bn. The reference location is the world origin and the reference point on object 406 is its tip 406′.
Absolute pose of object 406 is determined at a rate of 100 Hz or more and is processed by an on-board processor 407. Processor 407 may be a part of absolute motion measuring component 410 or it can be a separate processor. Processor 407 separates absolute pose data 420 into two subsets P and O. Subset P contains only position parameters (x,y,z) of tip 406′, or equivalently, the components of vector Go. Subset O contains only orientation parameters (φ,θ,ψ) of object 406. A trajectory of tip 406′ is designated by P(t), which is the collection of subsets P at measurement times ti, or P(t)=(x,y,z,ti). Meanwhile a history of orientations of object 406 is designated by O(t), which is the collection of subsets O at measurement times ti, or O(t)=(φ,θ,ψ,ti).
Both trajectory P(t) and a representation of orientations O(t) are indicated in dashed lines in
A wireless communication link 422 employing a transmitter 424 on object 406 and a receiver 426 on computing device 414 is used to transmit pose data 420 to computing device 414. In the present case absolute pose data 420 is broken up into time-synchronized subsets P and O. These subsets are transmitted via link 422 to an application 428 running on computing device 414. More specifically, subsets (P,O) captured at times t1, t2, . . . ti are transmitted sequentially to application 428 at a rate of about 100 Hz or higher.
Subsets 420a, 420b, . . . 420i arriving sequentially via communication link 422 provide the input for interacting with output 432 of application 428. Application 428 is programmed in such a manner that prior and newly arrived subsets O and P are represented graphically in the form of trajectories O(t)′ and P(t)′. In addition, manipulating object 406 in real three-dimensional space 402 of room 404 such that image 406″ lands on icon 438 turns application 428 on and off. Furthermore, placing image 406″ over commands and options 434 selects them. Finally, trajectory P(t)′ can be converted into a digital ink trace and converted into text using standard conversion algorithms analogous to those used in tablet PCs and known to those skilled in the art. The converted text can be displayed along text 436 already present on display screen 430. In this manner, subsets P and O are employed by application 428 as input for interacting with its output 432.
Computing device 414 also has a speaker 440 mounted to the side of display screen 430. Application 428 can thus also take advantage of audio elements 442 to supplement output 432 consisting of only visual elements. For example, audio elements 442 can be constituted by tones, e.g., warning tones when image 406″ of object 406 is moving off screen. Another audio element 442 can be a tune, e.g., to announce the launch or termination of application 428. Still another audio element 442 may be a musical composition that is selected or adjusted in volume or other auditory parameter by data from subsets P and O. For example, the location of tip 406′ as communicated by P(t) can control the volume. Finally, audio element 442 may simply be an alert signal when either subset P or O exhibit certain type of data. For example, when trajectory P(t) changes too rapidly and the user manipulating object 406 in real three-dimensional space 402 should slow down in moving object 406.
On-board optical measuring arrangement 508 has a lens and a PSD in its absolute motion detection component. Their arrangement and operation is analogous to those described in the previous two embodiments. Meanwhile, beacons B1-B4 are IR LEDs mounted on a reference object 510 that is positioned at a known location and in a known spatial relationship to world origin (0,0,0). In other words, the pose (position and orientation) of reference object 510, itself parametrized by coordinates (X1,Y1,Z1), as embedded in world coordinates (Xo,Yo,Zo) is known.
The angular motion or change in orientation parameters of manipulated object 502 in environment 504 is expressed with the aid of Euler angles (φ,θ,ψ). The reference point for describing the Euler rotated object coordinates is a tool tip 512 of object 502. Position of tool tip 512 is expressed in Cartesian coordinates (x,y,z). The successive positions of tool tip 512 are defined with the aid of vectors Go obtained at different times ti; i.e., by vectors Go(ti). The actual trajectory of tool tip 512 is expressed by vectors Di connecting the tips of successive vectors Go(ti). The trajectory of a distal end 514 of object 502 is indicated by reference 516.
IR LEDs B1-B4 emit infrared light 518 according to a modulation scheme imposed by a suitable control mechanism (not shown) integrated into reference object 510. The modulation scheme renders IR LEDs B1-B4 distinguishable, as required of light sources serving as beacons. The number of IR LEDs should be increased from the minimum of 4 to at least 16 and preferably 32 or more if sub-millimeter accuracy on the absolute pose and absolute motion of object 502 is required. Furthermore, they should be spaced as far apart as possible given the dimensions of reference object 510. For example, a two- or three-dimensional grid pattern is a good spatial arrangement for IR LEDs. Additionally, it is advantageous if IR LEDs are placed in a grid structure that subtends a portion of environment 504 designated as work space 520 in which tool 502 will be operated. For planar arrangements of IR LEDs integrated into reference object 510, it is also advantageous to operate tool tip 512 as close as possible to the centroid of the smallest convex set containing the IR LEDs (i.e., the distribution's convex hull).
When the spatial arrangement and number of IR LEDs is sufficiently optimized to yield sub-millimeter accuracy on the location of tool tip 512, and sub-degree accuracy on orientation parameters (φ,θ,ψ) within work space 520 then object 502 can be a precision tool. For example, in this embodiment manipulated object 502 can be a jotting implement, a surgical implement, a three-dimensional digitizer, a digitizing stylus, a hand-held tool such as a cutting implement or a utensil. More specifically, in the present embodiment tool 502 is a scalpel, work space 520 is an operating area (patient and incision not shown) and tool tip 512 is a blade tip.
The absolute motion tracking method of the invention with scalpel 502 is implemented by transmitting absolute pose data 506 inferred from light data via a communication link 522 to processor 524 at times ti. Processor 524 picks out as subset 526 orientation parameters (φ,θ,ψ) and position parameters of tool tip 512 described by vectors Di at times ti. In order to keep good track of the sequence of absolute poses, each subset 526 is appended with its corresponding measurement time ti. Thus, subsets 526 are expressed as (φ,θ,ψ,Di,ti). Note that vectors Di could alternatively be expressed in coordinates (X1,Y1,Z1) of reference object 510, since the full spatial relationship between world coordinates (Xo,Yo,Zo) and reference object 510 is known.
After preparation of absolute pose data 506 and identification of subsets 526, processor 524 forwards them to an application 528. Application 528 is preferably implemented on a physician's computer (not shown). Application 528 can be a reality simulation that allows an intern to follow an actual surgery in real time or perform their own mock surgery with scalpel 502. Application 528 can also be a remote control application, in which a physician performs a surgery with a mock version of tool 502. Then, a communication link such as the world wide web 530 relays subsets 526 to another module of remote surgery application 528 that is implemented on a remote device 532 that duplicates the motion encoded in subsets 526 to perform an actual surgery on an actual patient at the remote location with an actual scalpel (not shown).
In an alternative embodiment, tool 502 is a hand-held utensil whose working tip 512 is used for performing some useful function, e.g., stamping or marking an object located in work space 520. In this case application 228 is a general motion-capture application and the frequency of measurement times ti is on the order of 75 Hz. In some motion-capture applications such as biometric applications requiring precise knowledge of the motion of utensil 502, e.g., to derive a biometric aspect of hand 503, more frequent measurement times ti, e.g., in excess of 100 Hz or even in excess of 200 Hz can be used. In particular, such precise knowledge can be required when the biometric application is a user verification application.
In a first step 540, subset 526 is received by either a local host or a network via communication link 530. If subset 526 is intended for a remote host, then it is forwarded to the remote host in a step 542. In a second step 544, a processor in the intended host (local host or remote host, as the case may be) determines the requirements for subset 526. This selection can be made based on an intended final application 546. For example, when final application 546 only requires the parameters already contained in subset 526, then subset 526 is forwarded to step 548 for preparation and direct use. Alternatively, when application 546 requires additional parameters, subset 526 is forwarded to step 550 for derivation of these additional parameters.
For example, the additional parameters are derivatives of one or more of the parameters in subset 526. Thus, subset 526 is sent to a differentiation module 552 and then to a preparation module 554 for supplementing subset 526 with the derivatives. In the example shown, time derivatives of Euler angles φ and θ are required and thus, supplemented and prepared subset 526′ contains these time derivatives. Alternatively, statistical information about one or more of the parameters in subset 526 are required. Thus, subset 526 is sent to a statistics module 556 and then to a preparation module 558 for supplementing subset 526 with the statistical information. In the present example, the statistical information is a standard deviation of second Euler angle θ. Thus, supplemented and prepared subset 526″ contains the parameters of subset 526 and standard deviation σ(θ) of angle θ.
A person skilled in the art will appreciate that the functions described can be shared between local and remote hosts as well as application 546, e.g., as required by the system architecture and data porting standards. For example, some preparation and supplementing of subset 526 can be performed by application 546 upon receipt.
Subset 526 is transmitted to application 546 for use as an input that is treated or routed according to its use. For example, in a step 560, subset 526′ is used as control data. Thus, subset 526′ is interpreted as an executable command 562 or as a part of an executable command and used in an executable file 564. On the other hand, in a step 566, subset 526″ is used as input data and saved to a data file 568.
In general, application 546 has an output that is presented to one or more users. Meanwhile, the handling of tool 502 generates subsets 526 that are used as input; either in the form of control data or input data. There is a feedback loop between motion of tool 502 in real three-dimensional environment 504 and the output of application 546. Subsets 526 produced from motion of tool 502 by hand 503 in real space serve as input for interacting with the output of application 546 that runs on a computer, e.g., tablet PC 532. This relationship between input derived from motion of tool 502 in real space and output of computer-implemented application 528 renders the method of invention ideal for interfaces that require a more direct and kinesthetically intuitive interaction with applications in the digital world. This is particularly true of applications that include simulations of real-world events or applications that try to render cyberspace more accessible to human users.
Object 602 has an on-board optical measuring arrangement 608 for absolute pose tracking. Unlike in the prior embodiments, arrangement 608 does not rely only on ambient light. Instead, it has an active illumination component 610. Component 610 includes a source 612 for generating a light 614 and optics 616A, 616B for conditioning light 614 and projecting it into environment 604. Specifically, optic 616A is a beam splitter and optic 616B is a mirror. Additional optics, such as lenses may be included as well (not shown) for condition and projecting light 614.
Active illumination component 610 is simultaneously designed to receive a scattered portion 614′ of light 614 coming from one or more invariant features 618A, 618B located in environment 604. In the present embodiment, features 618A, 618B are markings deposited on the surface of tablet 606. It is particularly advantageous in this embodiment, if markings 618A, 618B are high optical contrast features under projected light 614 by virtue of being highly reflective to light 614. In fact, preferably markings 618A, 618B are retro-reflectors or made of a retro-reflective material.
Arrangement 608 employs scattered portion 614′ of light 614 for optically inferring or measuring the absolute pose of wand 602. The absolute pose 620 inferred from light data is again reported with parameters (φ,θ,ψ,Di,ti), which include the values of vector Do at times ti, herein again denoted as Di. In order to provide the requisite information in its scattered portion 614′, projected light 614 needs to carry spatial information. One way to imbue light 614 with such information is to provide it with structure. For example, light 614 can be a structured light projected in some pattern 622. Pattern 622 can be a time-invariant grid pattern or it can be a time-varying pattern. These options are well known to those skilled in the art of optical scanners with constant and time-varying scan patterns.
In the present embodiment, pattern 622 is a time-varying scanned pattern. To accomplish this, active illumination component 610 has a scanning unit 624. Unit 624 drives and controls mirror 616B, which is a scanning mirror in this case. When correctly driven, scanning mirror 616B executes an appropriate movement to trace out pattern 622.
In
Scanning mirror 616B may be a tiltable or rotatable mirror, depending on scan pattern 622 desired. In the event mirror 616B is tiltable, it can be uniaxial for executing a one-dimensional scan pattern 622, or biaxial for executing a two-dimensional scan pattern 622. A scan point Po of scan pattern 622 produced with projected light 614 intersecting tablet 606 and shown in
In the present embodiment, scanning mirror 616B is a tiltable biaxial mirror that executes a two-dimensional scan pattern 622 parametrized by scan angle σ referenced to mirror axis M.A. Additionally, projected light 614 is collimated into a scanning light beam 626. Angle δ denotes the angle of incidence of scanning light beam 626 on tablet 606 at scan point Po. Angle λ is the inclination angle of wand 602 with respect to the surface of tablet 606. Since invariant features 618A, 618B are retro-reflecting, angle δ is also the angle at which scattered portion 614′ returns from them to arrangement 608. A photodetector 628 is provided on-board wand 602 for receiving scattered portion 614′. Mirror 616B and beam splitter 616A guide scattered portion 614′ to photodetector 628 in this embodiment.
Preferably, the scan of an entire scan pattern 622 is executed rapidly, e.g., at kHz rates. Such rapid scanning is required to generate many scattered portions 614′ of light 614 coming from retro-reflecting invariant features 618A, 618B during each second. This ensures that there is sufficient data for arrangement 608 to infer absolute pose 620. In addition, scan pattern 622 should cover enough real space to ensure that scanning light beam 626 intersects features 618A, 618B from any of the absolute poses that wand 602 is expected to assume during regular operation. This can be accomplished by choosing a dense scan pattern 622 and a large scan angle σ. One possible two-dimensional scan pattern that satisfies these constraints is a Lissajous figure projected over a scan angle σ extending from −35° to +35°.
The times during the scan pattern 622 when scattered portions 614′ are detected by photodetector 628 indicate where, with respect to wand 602, invariant features 618A, 618B are located at those times. It should be noted that employing scan pattern 622 is also very useful in recognizing invariant features such as bar codes and other markings extensively used in commerce. Therefore, wand 602 with active illumination component 610 can be particularly useful in applications having to locate and simultaneously identify bar-code bearing objects that are present in environment 604 and may or may not be placed on tablet 606.
Note that this approach can be used in embodiments that use optical features in the environment that are active, e.g., beacons. In other words, the beacons can be configured to provide structured light in a time-invariant or a time-varying pattern (e.g., a certain repeating scan pattern) in a manner similar to that of active illumination component 610. The patterns may include two-dimensional scan patterns such as one or more scan lines.
Active illumination component 702 has a light source, in this case consisting of two laser diodes that produce two laser beams. Component 702 has two rotating scanning mirrors that produce two planes 708, 710 of projected light 712, 714 respectively. Each of these projected planes of light 708, 710 is produced by a respective laser beam, which is scanned within its respective plane by a respective rotating scanning mirror. These types of rotating scanning mirrors are well known to those skilled in the art. Preferably, the laser diodes emit in the infrared so that light 712, 714 is not visible or disruptive to a human user of tablet computer 706. Planes 708, 710 are at right angles to each other and are perpendicular to a central axis C.A. of stylus 700.
Four reflective elements 716A, 716B, 716C, 716D are mounted on the four sides of a display screen 718 belonging to tablet PC 706. Elements 716 have different numbers of retro-reflecting strips 720 that scatter light 712, 714 back along the direction from which it arrived. Specifically, element 716A, has two retro-reflecting strips 720, element 716B has three, element 716C has one and element 716D has four.
Component 702 is one part of an on-board optical measuring arrangement 722 of stylus 700. Above component 702, arrangement 722 includes a lens 724 and a sensor (not shown) for receiving light portions 712′ and 714′ that are back-scattered towards component 702 from environment 704. A suitable beam splitter, as in the prior embodiment, can be provided in order to separate back-scattered portions 712′, 714′ of light 712, 714 that is being projected into environment 704 in the form of planes 708, 710. It is known how to position such a beam splitter such that it directs back-scattered portions 712′, 714′ to the sensor. Lens 724 has its field of view (F.O.V.) chosen such that it can receive back-scattered portions 712′ and 714′, after they have been directed by the beam splitter and thus image them onto the sensor.
Alternatively, lens 724 can be designed to have a wide-angle panoramic F.O.V. such that it can directly view back-scattered portions 712′, 714′ emanating from retro-reflecting strips 720. This alternative design eliminates the need for a beam splitter. In either case, back-scattered portions 712′, 714′ received at the sensor will comprise a time-sequence of four back-scattered optical signals as they arrive in the same order that the beams are scanned over each of retro-reflecting strips 720. The timing of these optical signals can be processed infer the absolute pose of manipulated object 700 in Euler rotated coordinates (x,y,z,φ,θ,ψ) relative to the reference location (0,0,0) of tablet PC 706.
During operation, as the two scanning mirrors rotate at a suitable angular velocity, light 712, 714 of planes 708, 710 generates either one, two, three or four back scattered portions 712′, 714′. The number of these back scattered portions 712′, 714′ depends on which of the four reflective elements 716 is being intersected by planes 708, 710 respectively. At the instant shown in
Back-scattered portions 712′, 714′ are rapidly collected by lens 724 and projected onto the optical sensor. The optical sensor then converts this rapid sequence of optical signals into electrical signals for further processing into absolute pose data (x,y,z,φ,θ,ψ). In other words, lens 724 images all scattered portions 712′, 714′ onto the sensor to generate raw image signals. From these signals and their angular distribution, arrangement 722 can infer the absolute pose of stylus 700 and prepare it in the form of a suitable subset to serve as input for tablet computer 706 in a manner analogous to that explained above.
A person skilled in the art will realize that a large variety of active illumination components can be implemented in the apparatus of invention. However, whether any given optical measuring arrangement has an absolute motion detection component with a lens and an optical sensor or with an active illumination component or even with both, it is often advantageous to supplement it with an auxiliary motion detection component. Preferably, such auxiliary motion detection component tracks a relative position or movement and is used for estimation of absolute pose between measurement times ti.
Implement 750 has an on-board optical measuring component 760 with a lens that images its field of view onto a photosensor (not shown). Component 760 uses pixels 756 as beacons. For this reason, the processor of reader 752 modulates pixels 756 in a known pattern. At the time shown, only pixel 756′ is emitting light 758.
With the aid of pixels 756 acting as distinguishable light sources or beacons, the absolute pose of implement 750 is optically inferred by component 760. Nib 750′ of implement 750 is selected as the reference point. The absolute pose is expressed as absolute pose data in world coordinates (Xo,Yo,Zo) with respect to world origin (0,0,0). As before, the absolute pose data are in the form of Euler rotated object coordinates (x,y,z,φ,θ,ψ) or their equivalent. Depending on the application, the processor of reader 752 identifies among parameters (x,y,z,φ,θ,ψ) the subset that will serve as input to the application running on reader 752. For example, only (x,y) parameters in the plane of display screen 754 are employed if the input is to represent digital ink.
Implement 750 also has an auxiliary component 762 mounted on-board. Component 762 is an inertial sensing device such as a gyroscope or accelerometer. The principle of operation of these relative motion devices relies on detecting or integrating changes in motion. While undergoing these changes, such devices may take into account the constant presence of the gravitational field g in the Earth's frame of reference (Xi,Yi,Zi). In addition, may be subject to spurious measurements in accelerating frames of reference, such as in a car or on an airplane. For this reason, inertial devices are not suitable for determining the absolute pose of implement 750. However, over short periods of time, e.g., between times ti when absolute pose is inferred optically by component 760, these devices can detect relative changes in pose.
In cases where it may be required to minimize the computational load of the on-board absolute motion detection component 760 by collecting absolute pose data (x,y,z,φ,θ,ψ) at a slower rate, then it may be advantageous to use such inertial devices for estimating the motion between times ti. This is a variant of sensor fusion where a sparse set of absolute measurements from light data is combined with a dense set of relative measurements to produce a dense set of estimates. In one embodiment, the dense set of estimates is computed by numerically integrating the output of the inertial device from time ti to time ti+1 using the strapdown inertial navigation equations with the absolute pose inferred from light data at time ti set as initial condition. For more information about strapdown inertial navigation the reader is referred to Oliver J. Woodman, “An Introduction to Inertial Navigation”, University of Cambridge Laboratory Technical Report Number 696.
In an alternative apparatus 800 shown in
Arrangement 804 infers the absolute pose of object 802 from light data at measurement times ti. It sends the corresponding absolute pose data (x,y,z,φ,θ,ψ) via a communication link 803 to a processor 805. For better visualization, times ti when absolute pose is inferred correspond to tip 802′ locations indicated by points 801. Then, as in the prior embodiments, processor 805 identifies the necessary subset or subsets and provides them to an application 807 for use as input.
Object 802 has an auxiliary motion detection component 808 in the form of an optical flow measuring unit. Unit 808 has an emitter 810 for emitting a light 812 and a detector 814 for measuring scattered light 812′. During operation, scattered light 812′ returning from a scattering point 816 on a surface, or else from miniature scattering centers provides a relative measure of change in pose.
Unit 808 will be familiar to those skilled in the art and is analogous to those used by an optical flying mouse or a regular optical mouse, if tip 802′ is maintained near a scattering surface. In the case of an optical flying mouse, the image flow data is derived from the moving images of distant microscopic 3-D objects that are imaged onto a CCD camera sensor playing the function of detector 814. The information gained by this type of motion is used to track primarily only the relative angular motion of the mouse with respect to the 3-D environment containing the distant objects. In the case where component 808 is that of an ordinary optical mouse, the image flow data is derived from the moving images of microscopic features 811 on a surface 813 that object 802 is moving over, as shown in the present embodiment. Features 811 are imaged up close and magnified onto CCD camera 814, and the information gained by this method allows relative tracking of primarily only the translational motion of the mouse with respect to surface 813 containing features 811.
In both cases, the relative tracking data can be in the form of angular or linear velocities. These data can be integrated to give points along a relative path of motion between times ti when absolute pose data is found. Thus, as absolute data is used to define an absolute motion of hand-held manipulated object 802 at a certain resolution dictated by times ti, relative data is used to fill in relative motion information between times ti.
A person skilled in the art will realize that the absolute motion detection arrangements of the invention can itself be operated in a relative capture mode in addition to operating in the absolute motion capture or tracking mode. In other words, they can also double as auxiliary motion detection modules that provide relative motion information in some embodiments.
Environment 842 is of the kind in which there exists a stationary magnetic field 3, here indicated by a corresponding vector. This type of environment 842 is found, for example, on the surface of the Earth. Apparatus 840 has an auxiliary motion detection component 852 that is represented by an electronic magnetic sensing component. Component 852 is located in the body of manipulated object 844 for sensing changes in rotation of object 844 with respect to the magnetic field lines established by field B. Such changes produce a signal that represents the relative rotational velocity of manipulated object 844. These relative rotational velocities can be used to estimate pose between times ti, or when absolute pose is not being measured by arrangement 846.
The various embodiments of apparatus and methods of the invention for optically inferring absolute pose from on-board a manipulated object and reporting absolute pose data in a priori established world coordinates is useful for many applications. In particular, any application for which actions or movements of the manipulated object in real three-dimensional environment yields useful input stands to benefit from the apparatus and method. Such application may involve a simulation in which real environments are reproduced in a cyberspace or in a virtual space used by the application as part of its output.
Conveniently, environment 886 is parametrized by a Cartesian coordinate system (Xo,Yo,Zo) whose origin (0,0,0) is set at the lower right corner of frame 892. This Cartesian coordinate system serves as the world coordinates for application 880 and for arrangement 888. In addition, origin (0,0,0) is selected as the reference location with respect to which absolute poses of racket 884 will be optically inferred.
A computing device 894 that runs game 880 employs screen 890 for presenting an output 896 to user 882. Computing device 894 can be a personal computer, a dedicated gaming computer, a portable computer, a television system, any general computing device, hosting network or computing platform with sufficient resources to run game 880 on screen 890. In the present case, game 880 is a cyber game of tennis, and thus output 896 includes visual elements 898 necessary to represent a tennis court and a tennis match. Elements 898 include a tennis net 898A, a tennis ball 898B, an adversary with a tennis racket 898C, a court 898D and a replica or image 884′ of racket 884 held by user 882 playing game 880. In addition, an avatar 900 representing user 882 is added to output 896. It is avatar 900 that is shown holding a token of the racket; in this particular case it is just replica 884′ of racket 884.
Output 896 is in fact a cyberspace in which tennis game 880 unfolds and in which its elements 898, racket replica 884′ and avatar 900 are represented. Cyberspace 896 does not need to be parametrized like real three-dimensional environment 886. However, to provide user 882 with a realistic game experience, it is preferable that cyberspace 896 bear a high degree of correspondence to real space. For that reason, cyberspace 896 is parameterized with three-dimensional Cartesian coordinates (X1,X2,X3) that are at least loosely related to world coordinates (Xo,Yo,Zo). In the most realistic scenarios, game 880 can even use a one-to-one mapping of cyberspace 896 to real space 886.
Racket 884 has a reference point 902, which is in the center of its face and corresponds to the “sweet spot” of a normal tennis racket. Unlike the previous embodiments, reference point 902 is not an actual point on manipulated object 884 but a point that is defined in a clear relation thereto. Nonetheless, reference point 902 is used for reporting absolute pose data (x,y,z,φ,θ,ψ) inferred at measurement times ti by arrangement 888.
Racket 884 is also provided with an auxiliary motion detection component 904. In this embodiment, component 904 is an inertial sensing device. This specific device has a three-axis accelerometer 906 and a three-axis gyroscope 908. Between measurement times ti, gyroscope 908 provides information about changes in the orientation. This information can be represented by some or all Euler angles (φ,θ,ψ), any subset or combination thereof, some other angular description of orientation changes including concepts such as pan angles and changes therein. Meanwhile, also between measurement times ti, accelerometer 906 provides information about linear displacements that can be expressed in parameters (x,y,z), their subset, some combination thereof or still another description of linear displacement.
The combination of the subset or subsets from absolute pose data (x,y,z,φ,θ,ψ) and relative motion data are used by tennis game 880 as input for interacting with output 896. Specifically, the visual elements 898B, 898C as well as avatar 900 and replica 884′ of racket 884 are modified and re-arranged as a function of the input in accordance with the rules of the game of tennis implemented by the software programming of game 880. Thus, visual element 898B representing the ball bounces from replica 884′ as the latter is “swung” in cyberspace 896 to hit gall element 898B. When “hit” correctly, ball element 898B flies to the side of court 898D of adversary 898C. Meanwhile, avatar 900 follows the presumed motion of player 882 in real three-dimensional environment 886. The input does not re-arrange or modify court element 898D, since that part of the game is a stationary part of cyberspace 896.
A person skilled in the art will recognize that with minor modifications to cyberspace 896, game 880 could be a squash match where game object 884 is a squash racket. Game 880 could also be a golf game in which game object 884 is a golf club, or a baseball game in which game object 884 is a bat. Similar modifications can be made to implement games in cyberspace 896 in which game object 884 is a club, a bowling ball, a knife, a sword, a spear, a joystick, a steering wheel or a flying control. It should also be noted, that replica 884′ could be a different visual element or a token that does not even correspond in appearance to the physical appearance of game object 884. In this manner, a generally elongate game object 884 could be represented by suitable token 884′ within game 880. Such token would not be an image or a replica of game object 884 but, rather, the appropriate game object required by game 880. It is especially useful, when implementing game 880 to perform to make gamer 882 feel like they are performing moves with game objects 884 better than in real life, as this type of ego stroking will promote more usage.
Invariant features in this embodiment are two sets of temporally modulated IR LEDs acting as beacons, namely: 926A-D and 928A-D. Beacons 926A-D are mounted on a remote control 930, and more precisely on a flying control. Beacons 928A-D are mounted around a landing strip 932. Beacons 928A-C may emit light 929 at a different wavelength λ than that of light 927 emitted by beacons 926A-D. This makes it easier to differentiate beacons that are stationary in environment 922 from those that are moving (on flying control 930).
A computer 934 remotely controls the modulations of all beacons 926A-D, 928A-D and also receives absolute pose data 936 from arrangement 924 via a wireless communication link 938. The processor of computer 934 determines which of absolute pose data 936 to include the subsets to be used by a flying application 940 running on computer 934.
Flying application 940 requires one-to-one mapping between real three-dimensional environment 922 and its cyberspace. For this reason, world coordinates (Xo,Yo,Zo) with a reference location at their origin that is coincident with a corner of landing strip 932 are chosen as global coordinates. The reference point on aircraft 920 for reporting absolute pose data 936 in Euler rotated object coordinates (X,Y,Z)—shown with all three rotations in the upper right corner for easy reference—is its center of mass (C.O.M).
Meanwhile, flying control 930 defines an auxiliary reference coordinate system (Xr,Yr,Zr) with its origin at the lower right-hand corner of control 930. At each measurement time ti, computer 934 computes the relative pose of control 930 in global coordinates (Xo,Yo,Zo). This relative information is made available to arrangement 924 via link 938. Thus, arrangement 924 has all the requisite information about the instantaneous locations of all beacons 926, 928. This enables it to optically infer its absolute pose at measurement times ti. In addition, the pose of flying control 930 can be used to remotely control the flying behavior of aircraft 920. For example, the pose in which flying control 930 is held, corresponds to the pose that the user is instructing aircraft 920 to assume next. The mechanisms for aircraft control to implement such command are well known and will not be discussed herein.
Application 940 may keep track of the orientation O(t) and position P(t) of the center or mass (C.O.M.) of aircraft 920. It may further display this information in a visual form to the user on its display 942. For example, it may display O(t) and P(t) at the various times during flight in the form of a view from the cockpit. Such display may serve for flight simulation programs, training purposes or military drills. In addition, audio output, such as danger signals or tones can be emitted when O(t) and P(t) indicate an impending stall situation based on the application of standard avionics algorithms.
Yet another apparatus 950 supporting two manipulated objects 952A, 952B in a real three-dimensional environment 954 according to the invention is illustrated in
Computer 964 defines world coordinates (Xo,Yo,Zo) having an origin coinciding with its lower left corner. These are the global coordinates for reporting absolute pose data of both objects 952A, 952B. Computer 964 also controls the modulation pattern of beacons 958 via link 962. Furthermore, it sends corresponding information about the full location (absolute pose) of object 960 with its beacons 958 in world coordinates (Xo,Yo,Zo) to arrangements 956A, 956B via corresponding wireless communication links 966A, 966B. Thus, arrangements 956A, 956B are appraised of the location and modulation of beacons 958 at all measurement times ti to permit absolute motion capture or tracking of objects 952A, 952B.
Object 952A is a gun, a laser shooter, a general projectile launcher or another war object or implement. War object 952A is handled by a military trainee 968 in the conventional manner.
The reference point of war object 952A corresponds to the center of the outlet of its projectile launching nozzle. The coordinates defining the Euler rotated object coordinates (X1,Y1,Z1) of object 952A are shown on the nozzle with direction X1 being collinear with a projectile direction PD. The origin of these object coordinates (X1,Y1,Z1) is described by vector Gl in world coordinates (Xo,Yo,Zo).
Object 952B is a wearable article, in this case a pair of glasses worn by military trainee 968. The reference point of object 952B is not a point on object 952B, but rather an estimated position of the center of the trainee's head. Thus, the orientation portion (φ,θ,ψ) of the absolute pose of object 952B as optically inferred by arrangement 956B is also an indication of the attitude of the trainee's head. Specifically, trainee's looking direction LD can thus be automatically inferred and tracked. The Euler rotated object coordinates (X2,Y2,Z2) of object 952B are thus drawn centered on the trainees head and described by vector G2 in world coordinates (Xo,Yo,Zo).
A virtual reality simulation program 970, which is a military drill runs on computer 964. Program 970 displays the combat scenario in a virtual reality 972 on a projected display 974 to help monitor the progress of trainee 968. Scenario is constructed in cyberspace with output that includes visual elements 976, 978, 980. Elements 976, 978, 980 correspond to two virtual enemy combatants and a virtual projectile. Also, the projectile direction PD′ and looking direction LD′ are visualized. An avatar 968′ corresponding to trainee 968 is located in virtual reality 972 and is displayed on projected display 974 for monitoring purposes.
Preferably, trainee 968 is provided with the same visual elements of virtual reality 972 as shown on display 974 via a virtual retinal display or a display integrated with glasses 952B. This way, trainee can test his war skills on enemy combatants 976, 978. However, for pedagogical reasons, avatar 968′ is not displayed to trainee 968. Direct display technologies are well known to those skilled in the art of virtual reality or augmented reality.
During operation, arrangements 956A, 956B infer their absolute poses in environment 954 and transmit the corresponding absolute pose data to computer 964. The computer uses a subset of the data to enact the war exercise. Note that because objects 952A, 952B report their absolute pose data separately, they can be decoupled in virtual reality program 970. This is advantageous, because it allows to simulate a more realistic scenario in which trainee 968 can point and shoot gun 952A in a direction PD that is different from where he or she is looking, i.e., direction LD. In fact, in the present situation this behavior is required in order to deal with two virtual combatants 976, 978 simultaneously.
A person skilled in the art will realize that the application will be important in dictating the appropriate selection of manipulated object or objects. In principle, however, there is no limitation on what kind of object can be outfitted with an on-board optical arrangement for inferring its absolute pose with respect to a reference location in global coordinates parametrizing any given real three-dimensional environment. Of course, many applications that simulate the real world and many gaming applications, virtual reality simulations and augmented reality in particular, may request subsets that include all absolute pose data (φ,θ,ψ,x,y,z). This request may be necessary to perform one-to-one mapping between space and the cyberspace or virtual space employed by the application.
Whether fully virtual or not, applications typically provide the user with output of some variety. Normally, a rather small subset of absolute pose data can allow the user to interact with the output. For example, the supported interaction may include text input, which only requires a trace or re-arrangement of the output. In another case, it may only require a subset of one translational parameter to move or re-arrange some visual elements of the output. Given that the output may include audio elements and visual elements, the interaction applies to either or both of these types of output elements at the same time or sequentially. Since in many cases not all of the absolute pose data is necessary to interact with the output, the remainder of the absolute pose data can be used for still other purposes. For example, a certain absolute motion sequence executed with the manipulated object can be reserved for commands outside the application itself, such as dimming the display, adjusting display brightness, rotating or touching-up visual elements or even turning the computer running the application on and off.
Some augmented reality applications may further superpose one or more virtual elements onto the real three-dimensional environment. The virtual element or elements can be then rendered interactive with the manipulated object by the application.
This situation is illustrated in
Mobile device 994 is simultaneously a manipulated object in the sense of the present invention. Thus, device 994 has an on-board optical measurement arrangement 998 for inferring its absolute pose at times ti with respect to environment 996. The coordinate systems, reference location and reference point on object 994 are not shown in this drawing for reasons of clarity. Also, in this case the invariant features used by arrangement 998 are not light sources but, rather, are known objects in environment 996, including house 1000, road 1002 and other features that preferably have a high optical contrast and are easy for arrangement 998 to detect.
Augmented reality application 990 displays not only an image of environment 998, but also has a virtual element 1004. In the present case, element 1004 is a description of services provided in house 1000 at which device 994 is pointed. Element 1004 is superposed on the image of environment 996 at an appropriate position to make it easily legible to the user.
A person skilled in the art will appreciate that the Euler convention used to report absolute pose data is merely a matter of mathematical convention. In fact, many alternative parametrization conventions that are reducible to the Euler parameters or subsets of the Euler parameters can be employed.
It should further be noted that the manipulated object can be any type of device whose absolute pose can yield useful data. Thus, although the above examples indicate a number of possible manipulated objects other types of objects can be used. Also, the subset identified from the absolute pose data can be supplemented with various additional data that may be derived from other devices that are or are not on-board the manipulated object. For example, pressure sensors can indicate contact of the manipulated device with entities in the real three-dimensional environment. Other sensors can be used to indicate proximity or certain relative position of the manipulated object with respect to these entities. Furthermore, the absolute pose data and/or supplemental data in the subset can be encrypted for user protection or other reasons, as necessary.
Remote control 1012 moves in a real three-dimensional environment 1018. For example, remote control 1012 is a device that is designed for handling by a user (not shown) and is associated with or coupled to a screen or display 1020. In the present embodiment remote control 1012 is a wand. Environment 1018 is a volume in front of and around display 1020.
System 1010 has a number of invariant features 1022. In this embodiment, features 1022 are high optical contrast features instantiated by light sources. Preferably, light sources 1022 are infrared diodes or other point sources that output light 1024 in the infrared range of the electromagnetic spectrum into environment 1018.
IR LEDs 1022 are grouped into four groups. A first group 1022A is aligned along a first edge 1020A of display 1020. A second group 1022B is aligned along a second edge 1020B, a third group 1022C along a third edge 1020C and a fourth group 1022D along a fourth edge 1020D. Edges 1020A-D are the right, top, left and bottom edges of display 1020 in this embodiment. A frame 1023 girds display 1020 and supports all IR LEDs 1022. Note that any circuitry required to modulate IR LEDs 1022 in accordance with any suitable modulation pattern that makes them distinguishable (beacons) can be integrated into frame 1023. This is especially useful in cases where frame 1023 is provided separately from display 1020 and/or is expected to work with many different display types (e.g., touch-sensitive displays).
System 1010 has a photodetector 1026 provided on-board wand 1012 for detecting light 1024. Photodetector 1026 outputs data 1028 indicative of detected light 1024. In fact, data 1028 in this case is just raw image data. Preferably, photodetector 1026 is a position-sensing two-dimensional diode or a PSD. More precisely, photodetector 1026 is analogous to optical sensor 212 of absolute motion detection component 208 designed for sensing light 222 from IR LEDs B1-Bn in the embodiment described in reference to
Photodetector 1026 is located on-board wand 1012 for receiving light 1024 emitted by IR LEDs of the four groups 1022A-D. As described in the above embodiment, suitable optics (not shown) for imaging, guiding and conditioning ensure that light 1024 is properly imaged from environment 1018 onto PSD 1026.
Further, system 1010 has a controller 1030 configured to determine an absolute position of remote control 1012 based on data 1016 output by relative motion sensor 1014 and data 1028 output from photodetector 1026. Controller 1030 is not on-board wand 1012, but is instead resident in an electronic device 1032 that contains further circuitry 1034 for executing one or more applications. Both relative motion data 1016 and data 1028 from photodetector 1026 are communicated to controller 1030 with the aid of communications circuitry 1038. Only communications circuitry 1038 of electronic device 1032 is shown for reasons of clarity. Corresponding circuitry is also present on-board wand 1012. Communications circuitry 1038 provides an up-link 1040 for transmitting data 1016, 1028 to controller 1030 from wand 1012, and a down-link 1042 for controller 1030 requests, e.g., changes in subset data or operation parameters of wand 1012.
The absolute position of wand 1012 is determined with respect to a reference location, which is the lower right corner of display 1020 set to be world origin (0,0,0) of world coordinates (Xo,Yo,Zo). These coordinates are Cartesian and they parametrize environment 1018. World coordinates (Xo,Yo,Zo) are posited in a certain relationship to an image 1044 that is produced on display 1020. More specifically, a first axis or the Xo world axis is co-extensive with edge 1020D of display 1020, while a second axis or the Yo axis is co-extensive with edge 1020A.
Image 1044 is thus substantially defined or parametrized by two orthogonal axes Xo, Yo. The location of any part of image 1044, e.g., visual elements that constitute the output of any application running on circuitry 1034, is thus immediately defined along the Xo and Yo axes. In other words, all such visual elements are displayed on display 1020 in the (Xo,Yo) plane. No further coordinate transformations are required from the (Xo,Yo) plane of image 1044 to world coordinates (Xo,Yo,Zo).
Of course, choices in which image 1044 is not co-planar with a plane in world coordinates (Xo,Yo,Zo) can be made. In those cases, coordinate transformations from image coordinates to world coordinates will need to be performed to express the absolute position of wand 1012 with respect to image 1044 and any of its visual elements. These transformations are well understood and can be made in the Euler rotation convention explained above. Also note, the location of world origin (0,0,0) in the (Xo,Yo) plane can be re-set from time to time, as necessary (e.g., during calibration of image 1044 on display 1020).
Now, electronic device 1032 that hosts controller 1030 and circuitry 1034 that runs an application whose output produces image 1044 on display 1020 can be any type of device. In practice, device 1032 will most often be a television box, a game console or a stand-alone computing device. However, device 1032 can also be an application-specific computer or a mobile device that communicates with display 1020 via a wireless link (not shown). For example, device 1032 can be a cell phone or a personal digital assistant. In the present embodiment, device 1032 is a stand-alone computing device that can perform the functions of a television box and is in direct communication with display 1020.
A reference point 1012′ is selected on wand 1012 for expressing its absolute position in world coordinates (Xo,Yo,Zo). In the present case, reference point 1012′ is in the middle of the front face of wand 1012. Thus, absolute pose of wand 1012 is expressed by absolute pose data (x,y,z,φ,θ,ψ) in Euler rotated object coordinates using reference point 1012′ as their origin. Absolute pose data (x,y,z,φ,θ,ψ) is inferred optically or measured from on-board wand 1012 using output data 1028 which is the raw image data output by PSD 1026. All the necessary operations, including the application of the rules of perspective geometry, image warping etc. (see teachings above, especially in reference to
Controller 1030 is configured to generate signals for rendering display 1020. For this purpose, controller 1030 identifies a subset of absolute pose data (x,y,z,φ,θ,ψ) that will be used in the signals that render display 1020. In the present embodiment, that subset contains only one of the three absolute position parameters (x,y,z), namely (z) which is the absolute position of remote control or wand 1012 in or along a third axis that is orthogonal to the Xo, Yo axes defining image 1044. Because of advantageous parametrization, this third orthogonal axis is simply the Zo axis of world coordinates (Xo,Yo,Zo). The subset also contains requisite orientation parameters (φ,θ,ψ) to express the roll of wand 1012 around center axis C.A. In particular, orientation parameters (φ,ψ) are required to completely express that roll. Therefore, the subset is just (z,φ,ψ). In some cases, a single orientation parameter derived from (φ,ψ) can be employed to express the roll, as will be appreciated by those skilled in the art.
During operation, IR LEDs 1022 are modulated and emit infrared radiation or light 1024. In this embodiment of the method, the four groups 1022A-D of IR LEDs 1022 are modulated in a sequential pattern. Thus, only one IR LED 1022 emits light 1024 at any measurement time ti. For better understanding,
Now, PSD 1026 outputs data 1028 which is the raw image data corresponding to the centroid of the flux of light 1024 emitted by the IR LED 1022 that is on at time ti. Data 1028 is transmitted to controller 1030 via up-link 1040 of communications circuitry 1038. From data 1028 collected from a number of IR LEDs 1022 at different times ti, controller 1030 infers the absolute pose of wand 1012 in terms of absolute pose data (x,y,z,φ,θ,ψ). This part of the method of invention has been described in detail in the above embodiments (see, e.g.,
In addition to data 1028, controller 1030 receives relative motion data 1016 from relative motion sensor 1014. Controller 1030 uses data 1016 for interpolating the position of wand 1012 between times ti. Specifically, in the present embodiment, controller 1030 uses relative motion data 1016 to determine the change in pose parameters (z,φ,ψ). Once again, the use of relative motion data for estimating pose between times ti has been described above (see, e.g.,
Supplied with absolute pose parameters (z,φ,ψ) of the subset identified from among absolute pose data (x,y,z,φ,θ,ψ) and changes in pose parameters (z,φ,ψ) of the subset obtained from data 1016, controller 1030 is ready to generate signals that render display 1020. Specifically, controller 1030 uses the change in parameter (z) for generating signals for zooming in on or zooming out of at least a portion 1044A of image 1044 shown on display 1020. Additionally, controller 1030 uses parameters (φ,ψ) and changes therein to generate signals for rotating at least a portion 1044A or visual elements contained in portion 1044A of image 1044 on display 1020.
These actions will now be explained in more detail. First, controller 1030 uses all parameters (x,y,z,φ,θ,ψ) as the subset in rendering and displaying a visual element or cursor 1046 at the location where a center axis C.A. of wand 1012 intersects display 1020 or, equivalently, image 1044. In doing so it uses absolute data 1028 as well as relative motion data 1016, in accordance with any suitable combination or data fusion technique that is efficient. Such sensor fusion and corresponding data fusion techniques are well known in the art.
The computation and displaying of cursor 1046 is performed periodically at a sufficiently high rate (e.g., 60 Hz or higher) to be acceptable to a human viewer of display 1020. Note that cursor 1046 is a visual element that forms a part of the output of the application running on circuitry 1034 of device 1032. In addition, cursor 1046 defines a center of rotation for a visual element 1048. Element 1048 is also a part of the output of the application running on circuitry 1034. In this case element 1048 is an icon originally located at the lower left corner of display 1020.
A user moves wand 1012 in environment 1018 and by doing so interacts with visual elements 1046, 1048 of the output of the application displayed as image 1044 on display 1020. First, user holds wand 1012 such that its center axis C.A. intersects image 1044 at the original location of icon 1048. Thus, cursor 1046 is displayed on top of icon 1048 at that time. By subsequently depressing a button 1050, user informs controller 1030 that he or she wishes to select icon 1048 produced by the application. The corresponding button depressed signal (not shown) can be communicated to controller 1030 and then the application by using up-link 1040. The operations required to implement such selection are well known in the art.
Once icon 1048 is selected in the application, the user moves wand 1012 diagonally and up such that the motion of cursor 1046, which traces the point of intersection between center axis C.A. and display 1020, executes movement M1. At the end of movement M1, icon 1048 is within image portion 1044A. Now, the user depresses button 1050 again to instruct the application running on device 1032 to leave or stop dragging icon 1048. At this point, user executes a motion S1 with wand 1012 during which only cursor 1046 is displaced to the point of intersection between center axis C.A. and display 1020.
The user now depresses button 1050 twice to inform the application that he or she wishes to fix the location of cursor 1046 on display 1020. This fixed location will be the center of rotation for visual elements in image portion 1044A. Presently, only icon 1048 has been placed in portion 1044A.
At this point, the user rotates icon 1048 about the center of rotation defined by the location of cursor 1046. In particular, the user simply twists wand 1012 clockwise around its central axis C.A. as shown in the figure. Correspondingly, icon 1048 undergoes clockwise rotation. This rotation is broken down into two stages M2 and M3 for better understanding.
While rotating icon 1048 by turning wand 1012 clockwise, the user also moves wand 1012 in or along the Zo axis. Of course, this axis is orthogonal to axes Xo, Yo that define the plane (Xo,Yo) of image 1044. Specifically, at the start of stage M2 wand 1012 is at absolute position z1 along the Zo world coordinate axis. At the end of stage M2 it is at z2, and finally it is at absolute position z3 at the end of stage M3. It should be noticed that reference point 1012′ is instrumental in expressing the absolute positions. In fact, the absolute positions in Zo correspond to the absolute positions z1, z2, z3 of reference point 1012′.
Controller 1030 generates signals corresponding to absolute positions z1, z2, z3 of wand 1012 in the third axis Zo for zooming.
Specifically, since these values are increasing, the user is moving away. Hence, the application zooms in on portion 1044A of image 1044 shown on display 1020 to enlarge it. As a result, icon 1048 grows in size. When the absolute position values in Zo decrease, the application zooms out of portion 1044A. Of course, this convention could be inverted or otherwise changed depending on the application.
To simplify and reduce the processing required, controller 1030 can be configured to first determine the absolute position of wand 1012 in third axis Zo. Then, controller 1030 can determine a change in a position of wand 1012 in Zo by combining the initial absolute position with relative motion data 1016 that encode the change in position. This represents an efficient and wise usage of relative motion data under the assumption that the user does not appreciably change the orientation part (i.e., the inclination angles) of the absolute pose of wand 1012. In particular, if the user changes one or more of the orientation parameters, then more frequent reliance on absolute pose data obtained from raw image data 1028 will be necessary.
The above embodiment can be further enhanced by addition of more controllers and wands. In addition, other subsets of absolute and relative orientation and position data can be used to produce useful input for the application of system 1010.
The application supported by device 1032 is a gallery and painting touch-up application. Hence, the output of the application includes visual elements 1052A, 1052B, 1052C displayed on display 1020. Elements 1052 represent a gallery in cyberspace. Specifically, element 1052A is a gallery wall, element 1052B is a re-touching station, and element 1052C is a specific painting taken off wall 1052A. As before, cursor 1046 is located at the instantaneous intersection of center axis C.A. of wand 1012 and image 1044 presented on display 1020. Note that the instantaneous pose (position and orientation) of wand 1012 is drawn in solid lines, while prior and later poses are drawn in dashed lines.
To alert the user that the gallery application is running, an icon 1054 is enlarged and displayed on display 1020. Other icons, representing non-active applications are posted in the lower left corner of display 1020 for user reference.
During operation, controller 1030 uses all absolute pose data (x,y,z,φ,θ,ψ) in the subset for generating signals. It also uses all relative motion data 1016 for estimating pose between measurement times ti.
At time t1, when cursor was at location 1046′, the user depressed button 1050. This informed controller 1030 to generate input for interacting with the gallery application. Specifically, motion RM during the time interval t1 to tn, while button 1050 remains depressed is used to drag painting 1052C from gallery wall 1052A to re-touching station 1052B. At the instant shown, i.e., at time ti, painting 1052C is being moved and rotated into position on re-touching station 1052B. Note that all six absolute pose parameters (x,y,z,φ,θ,ψ) can be used by controller 1030 to generate signals for this operation.
Gallery application indicates motion RM by a corresponding motion RM′ in cyberspace of the gallery. In other words, motion RM in real three-dimensional environment 1018 is being mapped to motion RM′ in cyberspace of the gallery application. The mapping can be one-to-one when all parameters (x,y,z,φ,θ,ψ) are employed, or it can be simplified. Simplified mapping allows the user to drag painting 1052C without having to appreciably move wand 1012 in the Zo axis or pay attention to changes in orientation of painting 1052C while it is being dragged. Simplified mapping is performed by controller 1030 identifying a sufficient subset of parameters (x,y,z,φ,θ,ψ) to translate motion RM from environment 1018 to requisite motion RM′ in cyberspace.
In the simplest mapping, any rotation of wand 1012 is detected. Then, the selected portion of the image, namely painting 1052C is rotated in response to the detecting step. As painting 1052C is rotated, it is also brought closer in and undergoes a zooming operation, too. In practice, the detecting step is broken down into receiving a transmission from wand 1012 that communicates the output of at least one of motion detection components 1014, 1026 that are incorporated in wand 1012 and detecting that wand 1012 was rotated based on the received transmission.
Painting 1052C is placed on re-touching station 1052B at time tn. At this time the user depresses button 1050 again to inform controller 1030 that subsequent motion DI is to be interpreted as digital ink. Motion DI takes place between times tn and tq. Digital ink DI′ thus generated on painting 1052C is shown in more detail in
Referring back to system 1010 of
In addition, in this same variant, relative motion data 1016 is accepted by controller 1030 from relative motion sensor 1014 at times ti. As pointed out above, data 1016 is not absolute. Instead, it is indicative of a change in the pose (orientation and position of reference point 1012′) of wand 1012. However, if relative motion data 1016 does not exhibit a large amount of drift (usually due to senor drift and noise), then data 1016 can be used together with absolute pose data (x,y,z,φ,θ,ψ) derived from light data 1028 to track the absolute pose of wand 1012 with respect to reference location (0,0,0). In particular, if the orientation portion of the pose is not important for a given application, then the absolute position of reference point 1012′ can be tracked by combining absolute and relative data in this manner until relative drift becomes unacceptably large. A similar approach can be employed to track absolute orientation only, or any combination of position and orientation parameters, including the full set of parameters (x,y,z,φ,θ,ψ) and/or their mathematical equivalents.
The method of invention is adapted for entering text in a media system 1060 as illustrated in
With the aid of the pose data, electronic device 1062 determines where center axis C.A. of wand 1064 intersects the plane of an image 1072 displayed on a display 1074. System 1060 places a cursor 1076 at that location. In the event of mis-calibration or offset, a cursor centering routine can be provided prior to launching any applications. For example, the user points wand 1064 at the four corners of display 1074 attempting to hit suitable displayed fiducials. Electronic device 1062 computes the necessary adjustment and employs it to compensate for any offset or mis-calibration. Such routines are well known to those skilled in the art and will not be described further herein.
The application running on device 1062 is a search. It uses display 1074 of system 1060 to display to a user a number of selectable characters 1078. In this case characters 1078 are the letters of the alphabet. Of course, they could also be numerals found on a conventional QWERTY alphanumeric keyboard or other lettering or signage that is capable of conveying information.
The search application has a box 1080 for text entry. The text entered represents search terms as conventionally understood. To enter text in box 1080, user navigates cursor 1076 to a particular selectable character among characters 1078 by moving wand 1064. In other words, the output of a motion detection component, e.g., 1070 and/or 1068 is used for navigating cursor 1076 on display 1074. The selection of the particular selectable character, in the case shown the letter “H” on which cursor 1076 has come to rest, is received by depressing button 1066. This action informs device 1062 to accept the selection.
In the embodiment shown, a user has employed this method to type in the search term “Find my McIntosh” into box 1080. Upon accepting this search term, system 1060 launches the corresponding search via its device 1062 and its computational and search resources. Such resources may include access to networks (e.g., the world wide web), as is well known to those skilled in the art. The result of the search, namely McIntosh apple 1082 the user was searching for additional visual information in the form of text 1083 are displayed above box 1080.
The user can also use cursor 1076 to launch other applications and interact with other data structures. For example, in
The apple selection made by the user is displayed on screen 1074 in
It will be apparent to a person skilled in the art, that navigating cursor 1076 can be used with virtually any input modality in which visual elements are manipulated, altered, entered, removed or otherwise interacted with. These include conventional interfaces as well as three-dimensional interfaces, e.g., in cyberspace, as enabled by the present invention.
Media system 1100 has a display screen 1108, which is preferably high-resolution or high-definition and also touch sensitive. In addition, system 1100 has a remote control or wand in the shape of a game object 1110. The operation of object 1110 is equivalent to the wand. Object 1110 has a button 1112 and at least one absolute motion detection component 1114 with a photodetector such as a PSD. Component 1114 faces media system 1100 so as to receive light 1116 from light sources 1118. Light sources 1118 are modulated IR LEDs mounted in a frame 1120 that girds display 1108. An auxiliary motion detection component 1122, such as a relative motion detection component with a gyroscope and/or an accelerometer, is provided on board object 1110.
Object 1110 is operated by a user in a real three-dimensional environment 1124 in front of media system 1100 where component 1114 receives sufficient light 1116 from IR LEDs 1118. During operation object 1110 provides optical data to a controller residing in electronic device 1102 or even on-board. The controller determines the absolute pose of object 1110 and uses any subset of the absolute pose parameters to generate input for application 1084. As described above, the controller may also use relative motion data from relative motion detection component 1122. For example, controller tracks the absolute position of a reference point on object 1110, or the orientation of object 1110.
Controller may also compute and keep track of derived quantities, such as the intersection of the center axis C.A. of object 1110 with screen 1108.
During application 1084, an image 1126 is displayed on screen 1108. Image 1126 contains visual elements 1182, 1128, 1130 and a sight 1132. A cursor having the image of a reticle sight 1132 is placed at the intersection of C.A. and screen 1108. The path of sight 1132 as object 1110 is moved by the user is visualized by trajectory ST. Element 1082 is the McIntosh apple found by the user in a previous search application. Element 1128 is an apple tree, and element 1130 is a visible branch of another apple tree on which McIntosh 1082 is maturing.
Application 1084 allows the user to pick apple 1082 by skillfully detaching its stem from branch 1130. This is done by aiming and shooting with object 1110. First, sight 1132 is centered on the stem, and then button 1112 is depressed to execute the shot.
The result of a successful execution is shown in
We now return to system 1010 as illustrated in
System 1010 has a photodetector 1026 provided on-board wand 1012 for detecting light 1024. Photodetector 1026 outputs data 1028 indicative of detected light 1024.
In this embodiment, controller 1030 of system 1010 is configured to identify a derivative pattern of light sources 1022 from photodetector data 1028. The derivative pattern is indicative of the asymmetric and generally linear patterns of groups 1022A-D of IR LEDs 1022 along edges 1020A-D. As the absolute pose of photodetector 1026 in wand 1012 changes, the asymmetric and generally linear patterns undergo a well-understood transformation. Such transformation is described by perspective distortion plus any optical aberrations introduced by imaging lenses and/or other optics elements cooperating with photodetector 1026. Knowledge of this transformation enables one to correlate the asymmetric and generally linear pattern to the derivative pattern and obtain information about the pose of photodetector 1026 and hence of wand 1012.
It should be noted that in another alternative embodiment, light sources 1022 can simply reflect light. For example, they can reflect light projected from on-board a wand, as described above in conjunction with
More generally, first group 1022A, of light sources can be disposed proximate any edge of display 1020, at another location, or else on, near, or even beneath display 1020. In this latter case, display 1020 has to be transparent to light 1024. In fact, even certain pixels of display 1020, especially in the case of an OLED display, can serve as light sources 1022 (see embodiment described in conjunction with
In the preferred embodiment of system 1010, the system is coupled to display 1020 that has first and second edges 1020A, 1020B. System 1010 also has first and second groups of light sources 1022A, 1022B. In this preferred embodiment, the first group of light sources 1022A are disposed proximate first edge 1020A of display 1020 and second group of light sources 1022B are disposed proximate second edge 1020B of display 1020. This arrangement is preferred because of the orthogonal arrangement of groups 1022A and 1022B.
Light sources 1022 can be identified or processed in triads or larger tuples, depending on the specific tracking or navigation algorithms that are employed to determine the absolute pose or position of wand 1012. It should be noted that for determination of the complete absolute pose it is preferable to consider at least four light sources 1022 in each tuple that is positioned proximate the corresponding edge of display 1020.
The apparatus and method of invention are particularly useful in ubiquitous computing environments, as well as applications that run virtual realities, augmented realities and other complex and multi-dimensional representational spaces including three-dimensional cyberspaces. Furthermore, it should be noted that the apparatus supports multiple manipulated objects such as wands or game objects cooperating in the overall system, e.g., media system, simultaneously. This enables collaboration as well as multi-player games. Further, the addition of touch-sensitive screens with multi-touch support expands the modalities in which the user can interact with the application.
Room 1206 has invariant features or optical features OPF of which only a few are indicated in areas that are encircled and only some of those indicated are expressly designated. Optical features OPF in the present example are passive optical features that offer sufficiently high optical contrast to be processed as required by system 1200. As mentioned above, optical features OPF can include any visible yet invariant features of room 1206. Invariant features stay in place within room 1206; this makes them useful for optical pose recovery. Invariant features or optical features OPF in the present case include edges of objects, corners of objects, planes (e.g., walls of room 1206), any special markings that can include, without limitation, designations and codes (e.g., bar codes, QR codes) as well as any optically detectable parts of doors, windows, walls, furniture and so on. Optical features OPF in the present example are all passive, in other words they only reflect or scatter ambient light. It will be understood that any active features, i.e., features that emit light can be used as well, as shown in several previous embodiments. Also, the choice of optical features OPF should be such that they are not all along the same line (non-collinear) in order to be able to support absolute pose recovery or absolute pose estimation.
System 1200 has a manipulated object 1208 embodied by a smart phone. Smart phone 1208 is manipulated or moved by hand or otherwise handled by a human user (not shown). A photodetector 1210 embodied by a camera and shown schematically is mounted on-board smart phone 1208. Camera 1210 faces the back (away from user) and has a viewpoint P. Although only one camera is shown, it will be understood that smart phone 1208 may have several cameras including one or more front-facing cameras. Camera is offset by an offset ob from the origin 1212 of body coordinates 1214 parameterizing or describing smart phone 1208. Body coordinates 1214 use Cartesian axes (Xb, Yb, Zb). In this case origin 1212 is chosen at the center of mass (C.O.M.) of smart phone 1208, but another convenient point on smart phone 1208 can be used as well.
Smart phone 1208 also has an auxiliary motion detection component 1216 mounted on-board. Auxiliary motion detection component 1216 is offset by an offset oi from the origin 1212 of body coordinates 1214. Auxiliary motion detection component 1216 has its own auxiliary coordinates 1218 with Cartesian axes (Xi, Yi, Zi) chosen for clarity of explanation.
Auxiliary motion detection component 1216 has at least one relative motion sensor. Suitable relative motion sensors include accelerometers, gyroscopes, optical flow measuring units, electronic magnetic sensing components and the like. In the present case auxiliary motion detection component 1216 has a three-axis accelerometer 1220 and a three-axis gyroscope 1222. Both are only indicated schematically for reasons of clarity.
System 1200 further requires a processor 1224 that is in communication with camera 1210 and with auxiliary motion detection component 1216. In general, processor 1224 can be remote or in a remote device such as in a host device, a remote computer, a gaming console, a worn computer, a backpack computer, a tablet and the like. In the present embodiment, however, processor 1224 is local and on-board smart phone 1208.
Conveniently, room 1206 is described by global or world coordinates 1226 that use Cartesian axes (Xo, Yo, Zo). For reasons of clarity and convenience, world coordinates 1226 are shown with their origin in the upper left corner of room 1206. A person skilled in the art will recognize that world coordinates 1226 can be determined in various ways and positioned at various convenient places in room 1206.
The remaining aspects of system 1200 and its operation will be explained by reference to the diagram of
Processor 1224 is in communication with application 1228. Application 1228 is of any type that uses input data derived from or generated from the pose estimate obtained for the pose of manipulated object or smart phone 1208. In some specific cases, application 1228 can be a virtual reality (VR) application, an augmented reality (AR) application, a mixed reality (MR) application, an extended reality (XR) application or the like. Such applications can include a game, a virtual try-on application, a digital education application and generally any application that relies on virtual objects or input (e.g., gesture information or digital ink).
Processor 1224 is connected to camera 1210 and to auxiliary motion detection component 1216. Thus, processor 1224 is configured to receive light data 1230 from camera 1210. Light data 1230 can be pre-processed in cases where camera 1210 has such capabilities. In the present case, camera 1210 does not preprocess light data 1230 beyond the necessary operations required to produce a digital image. Consequently, light data 1230 contains raw image data.
Processor 1224 receives raw image data 1230 in an image processing module 1232. Image processing module 1232 performs the necessary pre-processing and processing functions, as mentioned above, to obtain a processed image 1234 suitable for submitting to a navigation or pose recovery algorithm. For example, extraction of invariant features or optical features OPF such as edges, corners and markings will require the application of suitable image segmentation modules, contrast thresholds, line detection algorithms (e.g., Hough transformations) and many others. For more information on edge detection in images and edge detection algorithms the reader is referred to U.S. Pat. Nos. 6,023,291 and 6,408,109 and to Simon Baker and Shree K. Nayar, “Global Measures of Coherence for Edge Detector Evaluation”, Conference on Computer Vision and Pattern Recognition, June 1999, Vol. 2, pp. 373-379 and J. Canny, “A Computational Approach to Edge Detection”, IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 8, No. 6, November 1986 for basic edge detection all of which are herein incorporated by reference. Additional useful teachings can be found in U.S. Pat. No. 7,203,384 to Carl and U.S. Pat. No. 7,023,536 to Zhang et al. A person skilled in the art will find all the required modules in standard image processing libraries such as OpenCV (Open Source Computer Vision), a library of programming functions for real time computer vision. For more information on OpenCV the reader is referred to G. R. Bradski and A. Kaehler, “Learning OpenCV: Computer Vision with the OpenCV Library”, O'Reilly, 2008.
Processed image 1234 is forwarded to absolute tracking or absolute navigation algorithm module 1236 to obtain an absolute pose estimate. As mentioned above, absolute navigation algorithm module 1236 can use any well-known absolute pose recovery technique including geometric invariance, triangulation, ranging, path integration and motion analysis. Among the geometric invariance approaches the ones using a collineation or nomography matrix are especially suitable for this task. Under some circumstances reduced nomography approaches deploying a reduced representation can be used by absolute navigation algorithm module 1236. The reader is referred to U.S. Pat. No. 11,577,159 to Mandella et al. as well as to U.S. Pat. No. 9,852,512 to Gonzalez-Banos et al. for teachings on reduced representations and reduced homographies in optical navigation.
As shown in
Absolute pose estimates 1238A and 1238B are called absolute because they are with respect to world coordinates 1226 describing room 1206. In other words, absolute pose estimates 1238A and 1238B are not relative, where relative means lacking ground truth or not with respect to world coordinates 1226. For this reason, absolute pose estimates 1238A and 1238B are expressed in terms of parameters that describe the six degrees of freedom (6 DOFs). These six degrees of freedom include all six degrees of freedom, namely parameters (φ,θ,ψ,x,y,z) that fully describe rigid body absolute pose in three-dimensional space.
The parameters (φ,θ,ψ,x,y,z) are with respect to world coordinates 1226. Of course, since light data 1230 is collected from viewpoint P and processed image 1234 is also from viewpoint P, absolute pose estimates 1238A and 1238B express the position and orientation of viewpoint P in world coordinates 1226. However, since offset ob is known, a standard coordinate transformation can be used to express these estimates 1238A and 1238B in body coordinates 1214 that describe smart phone 1208.
Further, the six degrees of freedom here expressed by parameters (φ,θ,ψ,x,y,z) obtained at successive times ti describe rigid body motion in three-dimensional space. As mentioned above, expressing the 6 DOFs using Euler Angles (φ,θ,ψ) is done to keep the explanation easier to follow. Other mathematical conventions for expressing orientation or rotation angles, e.g., quaternions, Cayley-Klein angles or still other conventions can be used instead, as will be appreciated by those skilled in the art.
Light data 1230 is received by image processing module 1232 at certain times, e.g., when the camera 1210 captures an image frame of room 1206. These times are referred to herein as times where i stands for the i-th time or the i-th frame. For this reason, absolute pose estimates 1238A and 1238B inferred from light data and expressed in terms of the six parameters (φ,θ,ψ,x,y,z) are supplemented with times ti and ti+1 to keep track of when they were obtained. For example, with camera 1210 running at 100 Hz or frames per second the number of times ti during which light data 1230 is obtained and forwarded to image processing module 1232 to yield processed image 1234 for absolute pose recovery or estimation by absolute navigation module 1236 during each second is 100. Hence, processor 1224, and more specifically its absolute navigation algorithm 1236 is able to determine absolute pose estimates of the poses assumed by smart phone 1208 one hundred (100) times per second. Differently put, absolute pose estimates 1238A and 1238B expressed as (φi,θi,ψi,xi,yi,zi,ti) and (φi+1,θi+1,ψi+1,xi+1,yi+1,zi+1,ti+1) are 1/100 of a second apart. Of course, camera 1210 can run at a different frame rate including a variable frame rate. Additionally, there may be an occasional frame drop-out or other situation (e.g., occlusion), that may render some frames unusable for determining absolute pose estimates.
Absolute pose estimates 1238A and 1238B expressed by absolute navigation algorithm module 1236 as (φi,θi,ψi,xi,yi,zi,ti) and (φi+1,θi+1,ψi+1,xi+1,yi+1,zi+1,ti+1) are delivered to a pose estimation algorithm module 1240. Pose estimation algorithm module 1240 uses absolute pose estimates 1238A and 1238B in conjunction with other data, as described below.
Processor 1224 is also set up for receiving relative motion data from auxiliary motion detection component 1216. In the present example, auxiliary motion detection component 1216 has a three-axis accelerometer 1220 and a three-axis gyroscope 1222. Of course, as indicated above, other relative motion sensors including optical flow measuring units, electronic magnetic sensing components and the like can also be used in auxiliary motion detection component 1216. Such units and their operation have been described in several of the preceding embodiments.
In the present case, relative motion data is broken down into relative orientation data 1242 from gyroscope 1222 and relative position data 1244 from accelerometer 1220. As in the case of camera 1210, accelerometer 1220 and gyroscope 1222 do not pre-process their output data beyond the necessary operations required to produce acceleration and angular rate measurements. Thus, relative orientation data 1242 is raw relative orientation change data and relative position data 1244 is raw relative position change data.
Relative motion data is by definition relative (not with respect to world coordinates 1226) and indicative of a change in at least one aspect or parameter of the pose of smart phone 1208. In the present embodiment, relative orientation data 1242 is indicative of a change in at least one angle that describes the orientation of smart phone 1208. Meanwhile, relative position data 1244 is indicative of a change in at least one linear parameter that describes the position of smart phone 1208. Relative motion data 1242, 1244 can include changes in all orientation and position parameters (change in all parameters of pose). A zero change in any parameter of orientation and/or position is also a valid indication of change, namely no change. Note that sometimes such relative data is simply referred to as motion data by those skilled in the art.
Auxiliary motion detection component 1216 delivers relative motion data made up of relative orientation data 1242 and relative position data 1244 to processor 1224. More specifically, relative motion data is delivered to a relative data processing module 1246 belonging to processor 1224. Relative data processing module 1246 performs the necessary pre-processing and processing functions to obtain processed motion data 1248. These pre-processing and processing steps can include removal of the signal associated with the gravitational field ag (also previously referred to as just g) of Earth 1204 using sensor fusion techniques such as indirect Kalman filter methods fusing accelerometer and gyroscope data to estimate device orientation and linear acceleration. Further, noise compensation based on models that can include Allan Variance, systematic bias errors, known drift characteristics as well as various filtering techniques and outlier rejection can be applied by relative data processing module 1246 in obtaining processed motion data 1248.
Processed motion data 1248 are delivered to a relative navigation algorithm module or relative pose change algorithm module 1250. Relative pose change algorithm module 1250 determines a relative pose change. In the present embodiment relative pose change has two parts. It has a relative orientation change 1252 and a relative position change 1254.
Just as light data 1230 is collected from viewpoint P that is offset from body coordinates 1214, relative pose change is collected at offset oi from body coordinates 1214. Thus, relative orientation changes 1252 and relative position changes 1254 are reported in auxiliary coordinates 1218 and not in reference to world coordinates 1226. Relative orientation changes 1252 are reported as changes in terms of (Δφ,Δθ,Δψ) angles reported in auxiliary coordinates 1218. Similarly, relative position changes 1254 are reported as changes in terms of (Δ2x, Δ2y, Δ2z) with the second-order displacements being reported in auxiliary coordinates 1218.
The times at which the changes are reported are indicated separately for gyroscope 1222 and for accelerometer 1220 in the present embodiment. Gyroscope 1222 provides its relative orientation data 1242 at times tjro, where the superscript “ro” indicates relative orientation and index j is the time increment.
Here, relative orientation data 1242 is provided from time increment j=1 through j=n. In other words, gyroscope 1222 provides relative orientation data 1242 for n time increments. Relative orientation changes 1252 are therefore reported along with time increment as (Δφj,Δθj,Δψj,tj)ro. Similarly, accelerometer 1220 provides its relative position data 1244 at times tkrp (where superscript “rp” indicates relative position and index k is the time increment). There are a number of m time increments for relative position data 1244. Relative second-order position changes 1254 with time increments are hence reported as (Δ2xk,Δ2yk, Δ2zk,tk)rp with k=1 through k=m.
As mentioned above, absolute pose estimates 1238A and 1238B inferred from light data and expressed by absolute navigation algorithm module 1236 as (φi,θi,ψi,xi,yi,zi,ti) and (φi+1,θi+1,ψi+1,xi+1,yi+1,zi+1,ti+1) are delivered to pose estimation algorithm module 1240 in conjunction with other data. These other data consist of time-stamped relative orientation changes 1252 expressed as (Δφj,Δθj,Δψj,tj)ro and time-stamped relative second-order position changes 1254 expressed as (Δ2xk,Δ2yk,Δ2zk,tk)rp.
It should be noted that gyroscope 1222 and accelerometer 1220 typically run at rates that exceed the frame rate of camera 1210. For example, gyroscope 1222 and accelerometer 1220 operate at rates at or in excess of 200 Hz while smartphone cameras rarely operate at framerates above 60 Hz (60 fps). Therefore, there will usually be several relative orientation changes 1252 and several relative position changes 1254 generated by relative pose change algorithm module 1250 between two successive absolute pose estimates 1238 inferred from light data, e.g., between absolute pose estimates 1238A and 1238B. Differently put, the frame rate of camera 1210 is slower and even significantly slower than the rate of auxiliary motion detection component 1216. Hence, between successive times ti when absolute pose estimates are obtained from light data 1230 captured by camera 1210 system 1200 sees many relative motion changes reported by auxiliary motion detection component 1216.
Pose estimation algorithm module 1240 receives both absolute data and relative data. All data are time-stamped. It is important that time-stamps be properly synchronized, e.g., by using a master or global clock (not shown). Synchronization issues over different data streams and implementation are well-understood by those skilled in the art and will not be addressed herein.
Pose estimation algorithm module 1240 uses absolute pose estimates 1238 inferred from light data to establish ground truth in room 1206 with respect to world coordinates 1226. Thus, at time ti absolute pose estimate 1238A is used to recover ground truth. Then, at time ti+1 absolute pose estimate 1238B is used for that purpose. Ground truth recovered at time ti remains valid until next ground truth at time ti+1 is established. In the present embodiment, ground truth is used to ensure that the transformation that expresses the pose of smart phone 1208 in world coordinates 1226 based on body coordinates 1214 is correct.
Given the ground truth valid from time ti until time ti+1 from absolute pose estimate 1238A, pose estimation algorithm module 1240 is in a position to combine absolute and relative data in a useful manner. First, pose estimation algorithm module 1240 reconciles relative pose changes composed of relative orientation changes 1252 and relative position changes 1254 with ground truth. Specifically, from time ti until time ti+1 absolute pose estimate 1238A for ground truth is reconciled with time-stamped relative orientation changes 1252 expressed as (Δφj,Δθj,Δψj,tj)ro and time-stamped relative second-order position changes 1254 expressed as (Δ2xk,Δ2yk,Δ2zk,tk)rp. Reconciliation uses a coordinate transformation to express relative orientation changes and relative second-order position changes in body coordinates 1214, rather than in auxiliary coordinates 1218 in which these data are initially collected and reported by gyroscope 1222 and accelerometer 1220.
The next step in combining absolute and relative data involves a filling in, interpolation or other manner of combining absolute and relative data expressed in the same reference or coordinate system (body coordinates 1214 with respect to world coordinates 1226). In other words, processor 1224 can use data combining techniques in its pose estimation algorithm module 1240 to determine pose estimates that are useful to application 1228.
The easiest data combining techniques include simple interpolation between times ti and ti+1 where relative motion data is altogether ignored. This may be done when relative orientation changes (Δφj,Δθj,Δωj,tj)ro and relative second-order position changes (Δ2xk,Δ2yk,Δ2zk,tk)rp are unreliable due to drift, noise or other issues. It should be noted that in general second-order position changes are more likely to be subject to drift and noise and relative orientation changes. Therefore, in some embodiments the relative second-order position changes (Δ2xk,Δ2yk,Δ2zk,tk)rp are ignored between times ti and ti+1 (not included by the data combining technique) while the usually more reliable relative orientation changes are included. Standard measures for determining unacceptable levels of noise known to those skilled in the art can be used in making the determination that relative motion data or a subset of it, e.g., relative second-order position changes (Δ2xk,Δ2yk,Δ2zk,tk)rp are unreliable. In the below explanation all relative motion is included in data combining for clarity, but it will be clear to a skilled artisan that the combining techniques presented can be implemented with less input data, e.g., only with relative orientation changes (Δφj,Δθj,Δψj,tj)ro.
In general, the processor uses combining techniques to combine relative pose change with the absolute pose estimate. One such technique numerically integrates relative orientation changes (Δφj,Δθj,Δψj,tj)ro and relative second-order position changes (Δ2xk,Δ2yk,Δ2zk,tk)rp from time ti to time ti+1 using the strapdown inertial navigation equations. The strapdown inertial navigation equations are differential equations that relate the outputs of auxiliary motion detection component 1216 to the position, velocity and orientation of smart phone 1208 in world coordinates 1226. The strapdown inertial navigation equations can be integrated numerically from time ti to time ti+1 to obtain the trajectory of smart phone 1208 between times ti and ti+1, given some initial condition. In one embodiment, the absolute pose inferred from the light data at time ti is used as the required initial condition. In another embodiment, the strapdown inertial navigation equations are numerically integrated under the constraint that the trajectory at time ti+1 matches the absolute pose inferred from light data at time ti+1. In yet another embodiment, the trajectory computed using the strapdown inertial navigation equations is combined with simple interpolation of the absolute pose between times ti and ti+1. For more information about strapdown inertial navigation the reader is referred to Oliver J. Woodman, “An Introduction to Inertial Navigation”, University of Cambridge Computer Laboratory Technical Report Number 696.
Processor 1224 forwards request 1256 to pose estimation algorithm module 1240. Module 1240 uses absolute pose estimate 1238A at time ti expressed by parameters (φi,θi,ψi,xi,yi,zi,ti) as the reference for executing the filling in technique. The specific example in
By request time tr three relative pose changes 1252 expressed by parameters (Δφ1,Δθ1,Δψ1,t1)ro, (Δφ2,Δθ2,Δψ2,t2)ro, (Δφ3,Δθ3,Δψ3,t3)ro have been received by module 1240 since time ti. Similarly, by request time tr three relative second-order position changes 1254 expressed by parameters (Δ2x1,Δ2y1,Δ2z1,t1)rp, (Δ2x2,Δ2y2,Δ2z2,t2)rp, (Δ2x3,Δ2y3,Δ2z3,t3)rp have been received by module 1240 since time ti. (It should be noted that in general gyroscope 1222 and accelerometer 1220 do not need to provide their data 1242 and 1244 at the same rate.)
In executing the filling in technique module 1240 adds the collective relative pose changes to the absolute pose estimate for time ti for each parameter that is used to express pose in the convention chosen herein. Thus, module 1240 obtains current pose estimate or pose estimate 1258 at request time tr by aggregating changes to the ground truth value. In other words, the absolute pose (φi,θi,ψi,xi,yi,zi,ti) inferred from light data at time ti is updated using the three orientation changes (Δφ,Δθ,Δψ) and the three second-order position changes (Δ2x,Δ2y,Δ2z) to produce a timely pose estimate for request time tr. As explained above, the changes in parameter values for orientation and the second-order changes for position can be numerically integrated using the strapdown navigation equations.
Further, the time parameter is carried for reasons of clarity. The final times for orientation parameters and position parameters can be different and do not need to sum to request time tr (or current time). Note that pose estimate 1258 of the pose of smart phone 1208 thus determined is expressed with phone 1208 being described in body coordinates 1214 and with respect to world coordinates 1226.
As shown in
Input data 1260 can be used by application 1228 to render an augmented, virtual, mixed or extended scene to the user on the display of smart phone 1208. Input data 1260 can also be used as control or command data. For example, control data 1260 can be converted to executable commands. These commands can include instructions (e.g., executable files) to move virtual items in application 1228, turn application 1228 on/off, change modes, input gesture commands that affect the way application 1228 is running or how it is being displayed.
Input data 1260 can be used to derive gestures or digital ink. In fact, application 1228 that receives pose estimate 1258 from processor 1224 and uses input data 1260 thus derived from absolute pose estimate and relative pose changes can be an application that renders an augmented, virtual, mixed or extended scene to the user on a display. Indeed, input data 1260 can be handled and deployed in any manner already described above in conjunction with preceding embodiments.
Input data 1260 is used by an augmented reality application that adjusts the position and orientation of virtual objects 1312A, 1312B displayed on screen 1302. In this case the subset of data used is the entire set of input data 1260. The subset thus contains the six absolute pose parameters (φr,θr,ψr,xr,yr,zr,tr) obtained at request time tr. All are used by the augmented reality operation of rendering and displaying virtual objects 1312A, 1312B.
Input data 1260 is used by an augmented reality application that adjusts the position and orientation of a virtual objects 1350 displayed on screen 1342. In this case the subset is again the entire set of input data 1260, namely the six absolute pose parameters (φr,θr,ψr,xr,yr,zr,tr) obtained at request time tr are used for by the augmented reality operation of rendering and displaying virtual object 1350.
An application for stylus 1400 requires as output digital ink that is either real (actually deposited) or virtual (traced) by a nib 1408 of stylus 1400 while in contact with surface 1402. In this case the subset of data used is only a part of the set of input data 1260. The subset contains five absolute pose parameters (φr,θr,ψr,xr,yr,tr) obtained at request time tr. The z parameter is absent because it is not required when only tracking motion in (x,y). Conveniently, in this example the world coordinate axes (Xo,Yo) are chosen to contain surface 1402 for simpler coordinate transformations. Should the application require 3D gesture input, then all six absolute pose parameters are used in the subset (subset is the full set of input data 1260). Also, in some applications even fewer than five absolute pose parameters (φr,θr,ψr,xr,yr,tr) are needed. For example, some digital ink applications only require smaller subsets such as just pose parameters (x,y) or (θ,x,y).
An application for marker 1420 requires as output digital ink 1434 produced by marker 1420 while in contact with writing surface 1432. Thus, the subset of data needed is again only a part of the set of input data 1260. The subset contains five absolute pose parameters (φr,θr,ψr,xr,yr,tr) obtained at request time tr. The z parameter is absent because it is not required when only tracking motion in (x,y). Conveniently, in this example the world coordinate axes (Xo,Yo) are chosen to be coplanar with writing surface 1432 for simpler coordinate transformations. Should the application require 3D gesture input, then all six absolute pose parameters are used in the subset (subset is the full set of input data 1260).
This embodiment uses active optical features 1448 such as IR LEDs or other emitters that are part of writing tablet 1442 and passive optical features 1450 (not expressly indicated). Passive optical features 1450 belong to any suitable objects in environment 1404 as discussed above and including table 1446. In addition, passible optical features 1450 include a Manhattan object 1450A placed on the surface of table 1446. Given the many optical features in the present embodiment, one can set up many suitable world coordinates for parameterizing/expressing the six degrees of freedom. For example, world coordinates 1452 can be associated with Manhattan object 1450A. Alternatively, world coordinates 1454 can be associated with writing tablet 1442.
An application for digital stylus 1440 requires as output digital ink 1456 produced by digital stylus 1440 while in contact with writing surface of writing tablet 1442. Thus, the subset of data needed is again only a part of the set of input data 1260. The subset contains five absolute pose parameters (φr,θr,ψr,xr,yr,tr) obtained at request time tr. Again, should the application require 3D gesture input, then all six absolute pose parameters are used in the subset (subset is the full set of input data 1260).
Furthermore, it should be noted that in this system with digital stylus 1440 and writing tablet 1442 active optical features 1448 can be modulated. Such modulated optical features 1448 are sometimes referred to as beacons or light sources. They can be modulated to produce specific emission patterns that are temporally varied. In other words, beacons 1448 can be turned on and off (e.g., flashed) at certain times or otherwise modulated. Beacons 1448 can also have emission patterns that change temporally in more complicated ways, e.g., by scanning beacons 1448 using an appropriate device (e.g., a scanning mirror) to produce spatial and temporal (spatio-temporal) emission patterns. In any event active optical features 1448 are disposed in a certain pattern at known locations, since they are affixed to writing tablet 1442 and/or to its holder (not expressly referenced) to support such temporal and spatio-temporal emission patterns.
It will be apparent to one skilled in the art that the approach of combining absolute pose estimates with relative pose changes can be deployed in many different types of systems with manipulated objects that are handled, worn or otherwise deployed by a user to generate input data. In fact, the above examples are to be treated as merely a few of the many possible applications. Further, the manner of combining the absolute pose estimates with relative pose changes can use techniques more advanced than simple filling in.
A person skilled in the art will recognize that in any of the above embodiments the reference location at which the origin of world coordinates is placed need not be permanent. Depending on the apparatus and changes in the real three-dimensional environment the reference location can be redefined. This may happen as a part of a re-calibration process or continuously while the application is running. In still another alternative embodiment, the reference coordinates in world coordinates could be made to travel along with the location of the cursor in cyberspace. Skilled artisans understanding the nature of coordinate transformations in three-dimensional space will understand how to implement these kinds of transformations.
It will be evident to a person skilled in the art that the present invention admits of various other embodiments. Therefore, its scope should be judged by the claims and their legal equivalents.
This application is a Continuation-in-part application of U.S. patent application Ser. No. 16/653,084 filed on Oct. 15, 2019 which is a reissue of U.S. patent application Ser. No. 13/134,006 filed on May 25, 2011 which is a Division of U.S. patent application Ser. No. 12/586,226 filed on Sep. 18, 2009 and a Continuation-in-part of U.S. application Ser. No. 12/584,402 filed on Sep. 3, 2009.
| Number | Date | Country | |
|---|---|---|---|
| Parent | 12586226 | Sep 2009 | US |
| Child | 13134006 | US |
| Number | Date | Country | |
|---|---|---|---|
| Parent | 16653084 | Oct 2019 | US |
| Child | 18198223 | US | |
| Parent | 12584402 | Sep 2009 | US |
| Child | 12586226 | US |
| Number | Date | Country | |
|---|---|---|---|
| Parent | 13134006 | May 2011 | US |
| Child | 16653084 | US |
