Systems and methods for picking objects using 3-D geometry and segmentation

Abstract

A method for controlling a robotic system includes: capturing, by an imaging system, one or more images of a scene; computing, by a processing circuit including a processor and memory, one or more instance segmentation masks based on the one or more images, the one or more instance segmentation masks detecting one or more objects in the scene; computing, by the processing circuit, one or more pickability scores for the one or more objects; selecting, by the processing circuit, an object among the one or more objects based on the one or more pickability scores; computing, by the processing circuit, an object picking plan for the selected object; and outputting, by the processing circuit, the object picking plan to a controller configured to control an end effector of a robotic arm to pick the selected object.

Description

FIELD

Aspects of embodiments of the present disclosure relate to computer vision, including systems and methods for picking objects using 3-D geometry and segmentation.

BACKGROUND

Robotic automation is an area of growing interest in increasing the efficiency of manufacturing and logistics operations through the picking and placing of objects in industrial settings as well as in home use, such as assisting infirm people in household activities.

Automating the process of picking disorganized objects from a bin is a very complex problem. While a human may be able to intuitively recognize which objects in a bin can be grasped and the best likely points at which to grip an object that the human wishes to take from the bin, replicating a human skill at this task is difficult for a variety of issues. One such issue is that while a human is able to generalize to different objects and different environments, current robotic systems are generally unreliable outside of carefully controlled environments, such as where objects are presented to the robotic systems in regular and orderly arrangements, such as where the objects are arranged in a grid and where all of the objects in the grid have the same orientation. In particular, such robotic systems may not generalize well to unknown or unseen objects or conditions in which the positions and orientations of the objects is irregular, such as the case of objects jumbled in a bin.

Generalizing to different objects and different environments is especially hard for vision systems that are often used to guide the bin-picking process. These vision systems typically capture images of a scene using cameras and analyze the captured images to estimate the location and orientation of the objects to pick. This is a challenging task, especially in the case of unknown environments (including varying lighting conditions) and irregular conditions and orientations of objects in the bin.

SUMMARY

Some aspects of embodiments of the present disclosure relate to systems and methods for picking objects using 3-D geometry and segmentation as computed from one or more images depicting a scene containing objects to be picked.

According to one embodiment of the present disclosure, a method for controlling a robotic system includes: capturing, by an imaging system, one or more images of a scene; computing, by a processing circuit including a processor and memory, one or more instance segmentation masks based on the one or more images, the one or more instance segmentation masks detecting one or more objects in the scene; computing, by the processing circuit, one or more pickability scores for the one or more objects; selecting, by the processing circuit, an object among the one or more objects based on the one or more pickability scores; computing, by the processing circuit, an object picking plan for the selected object; and outputting, by the processing circuit, the object picking plan to a controller configured to control an end effector of a robotic arm to pick the selected object.

The method may further include: computing one or more depth maps based on the one or more images of the scene; and segmenting the one or more depth maps into one or more point clouds using the one or more instance segmentation masks, the one or more point clouds corresponding to the one or more objects in the scene.

The computing the one or more pickability scores for the one or more objects may include computing a pickability score for an object of the one or more objects by: computing a measure of clutter in a neighborhood of the object based on the one or more point clouds; computing a measure of distance of the object from a current location of the end effector of the robotic arm based on the one or more point clouds; and computing a measure of how much an object protrudes from other objects of the one or more objects based on the one or more point clouds.

The object picking plan may include a picking position and a direction of approach, and the computing the object picking plan for the selected object may include computing the picking position based on a visible center of mass of a point cloud corresponding to the selected object.

The computing the picking position may further include computing an estimated center of mass based on adjusting the visible center of mass by a scalar value to be farther from a viewpoint corresponding to a camera used to capture the one or more images of the scene.

The computing the one or more pickability scores for the one or more objects may include supplying the one or more depth maps and the one or more instance segmentation masks to an object pickability neural network.

The computing the one or more pickability scores for the one or more objects may include computing a pickability score for an object of the one or more objects by: computing a measure of clutter in a neighborhood of the object based on the one or more instance segmentation masks; computing a measure of distance of the object from a current location of the end effector of the robotic arm based on the one or more instance segmentation masks; and computing a measure of how much an object protrudes from other objects of the one or more objects based on the one or more instance segmentation masks.

The computing the one or more pickability scores for the one or more objects may include supplying the one or more instance segmentation masks to an object pickability neural network.

The object picking plan may include a picking position and a direction of approach.

The one or more images may include images captured based on a plurality of different imaging modalities.

The one or more images may include polarization raw frames.

The one or more instance segmentation masks may be computed using a trained convolutional neural network.

The one or more images may include images captured from a plurality of different viewpoints, and the one or more instance segmentation masks may include a plurality of instance segmentation masks corresponding to the different viewpoints.

The pickability scores may be computed independently for different instance segmentation masks corresponding to the different viewpoints, and the selecting the object may be based on the one or more pickability scores aggregated across the objects visible from the different viewpoints.

The method may further include computing depth maps for the different viewpoints, and the pickability scores may be further computed based on the depth maps.

According to one embodiment of the present disclosure, a system includes: a processor; and memory storing instructions that, when executed by the processor, cause the processor to: receive one or more images of a scene from an imaging system; compute one or more instance segmentation masks based on the one or more images, the one or more instance segmentation masks detecting one or more objects in the scene; compute one or more pickability scores for the one or more objects; select an object among the one or more objects based on the one or more pickability scores; compute an object picking plan for the selected object; and output the object picking plan to a controller configured to control an end effector of a robotic arm to pick the selected object.

The memory may further store instructions that, when executed by the processor, cause the processor to: compute one or more depth maps based on the one or more images of the scene; and segment the one or more depth maps into one or more point clouds using the one or more instance segmentation masks, the one or more point clouds corresponding to the one or more objects in the scene.

The instructions to compute the one or more pickability scores for the one or more objects may further include instructions that, when executed by the processor, cause the processor to compute a pickability score for an object of the one or more objects by: computing a measure of clutter in a neighborhood of the object based on the one or more point clouds; computing a measure of distance of the object from a current location of the end effector of the robotic arm based on the one or more point clouds; and computing a measure of how much an object protrudes from other objects of the one or more objects based on the one or more point clouds.

The object picking plan may include a picking position and a direction of approach, the instructions to compute the object picking plan for the selected object may further include instructions that, when executed by the processor, cause the processor to compute the picking position based on a visible center of mass of a point cloud corresponding to the selected object.

The instructions to compute the picking position may further include instructions that, when executed by the processor, cause the processor to compute an estimated center of mass based on adjusting the visible center of mass by a scalar value to be farther from a viewpoint corresponding to a camera used to capture the one or more images of the scene.

The instructions to compute the one or more pickability scores for the one or more objects may further include instructions that, when executed by the processor, cause the processor to supply the one or more depth maps and the one or more instance segmentation masks to an object pickability neural network.

The instructions to compute the one or more pickability scores for the one or more objects may include instructions that, when executed by the processor, cause the processor to compute a pickability score for an object of the one or more objects by: computing a measure of clutter in a neighborhood of the object based on the one or more instance segmentation masks; computing a measure of distance of the object from a current location of the end effector of the robotic arm based on the one or more instance segmentation masks; and computing a measure of how much an object protrudes from other objects of the one or more objects based on the one or more instance segmentation masks.

The instructions to compute the one or more pickability scores for the one or more objects may further include instructions that, when executed by the processor, cause the processor to supply the one or more instance segmentation masks to an object pickability neural network.

The object picking plan may include a picking position and a direction of approach.

The one or more images may include images captured based on a plurality of different imaging modalities.

The one or more images may include polarization raw frames.

The one or more instance segmentation masks may be computed using a trained convolutional neural network.

The one or more images may include images captured from a plurality of different viewpoints, and the one or more instance segmentation masks may include a plurality of instance segmentation masks corresponding to the different viewpoints.

The pickability scores may be computed independently for different instance segmentation masks corresponding to the different viewpoints, and the memory may further store instructions that, when executed by the processor, cause the processor to select the object based on the one or more pickability scores aggregated across the one or more objects visible from the different viewpoints.

The memory may further store instructions that, when executed by the processor, cause the processor to compute depth maps for the different viewpoints, and the pickability scores may be further computed based on the depth maps.

BRIEF DESCRIPTION OF THE DRAWINGS

The accompanying drawings, together with the specification, illustrate exemplary embodiments of the present invention, and, together with the description, serve to explain the principles of the present invention.

FIG. 1A is a schematic diagram depicting a object pick estimator according to one embodiment of the present disclosure.

FIG. 1B is a high-level depiction of the interaction of light with transparent objects and non-transparent (e.g., diffuse and/or reflective) objects.

FIG. 2A is a perspective view of a camera array according to one embodiment of the present disclosure.

FIG. 2B is a cross sectional view of a portion of a camera array according to one embodiment of the present disclosure.

FIG. 2C is a perspective view of a stereo camera array system according to one embodiment of the present disclosure.

FIG. 3A is a block diagram of an object pick estimator according to one embodiment of the present disclosure.

FIG. 3B is a flowchart depicting a method for computing object pick estimates according to some embodiments of the present disclosure.

FIG. 4A is a flow diagram of a process for object level correspondence according to one embodiment.

FIG. 4B is a block diagram of an architecture for instance segmentation and mask generation according to one embodiment.

FIG. 4C is a more detailed flow diagram of a matching algorithm for identifying object-level correspondence for a particular object instance in a first segmentation mask according to one embodiment.

FIG. 5A is a block diagram of a feature extractor according to one embodiment of the present invention.

FIG. 5B is a flowchart depicting a method according to one embodiment of the present invention for extracting features from polarization raw frames.

FIG. 6 is a block diagram depicting a Polarized CNN architecture according to one embodiment of the present invention as applied to a Mask-Region-based convolutional neural network (Mask R-CNN) backbone.

FIG. 7 is a block diagram of an attention module that may be used with a polarized CNN according to one embodiment of the present invention.

DETAILED DESCRIPTION

In the following detailed description, only certain exemplary embodiments of the present invention are shown and described, by way of illustration. As those skilled in the art would recognize, the invention may be embodied in many different forms and should not be construed as being limited to the embodiments set forth herein.

Aspects of embodiments of the present disclosure relate to systems and methods for picking objects using 3-D geometry and segmentation as computed from one or more images depicting a scene containing objects to be picked.

Generally, in vision-guided robotics, an imaging system is used to capture images of a scene, which may include objects to be picked up by a robotic system such as a robot arm with a gripping end effector (e.g., fingers that can pinch or grasp an object from two or more directions and/or a suction gripper that forms a suction grip on a compatible surface of the object). A computing system or vision system applies various computer vision algorithms to the captured images to compute semantic information about objects in the scene, and the robotic system uses the semantic information to generate a motion path for picking an object from the scene.

Some comparative methods for picking known objects involves computing the six degree of freedom (6-DoF) poses of those objects (e.g., three-dimensional translation coordinates and three-dimensional rotational coordinates). As used herein, the term “known objects” will refer to objects that are pre-registered with a vision system that is used to analyze a scene to identify objects. For example, some vision systems store 3-D models or CAD models (e.g., 3-D mesh models) of the different known objects that the vision system will encounter and align the 3-D models to the objects as depicted in the images.

However, computing 6-DoF poses can be slow and may not be applicable in cases where the vision system does not have a 3-D model representing the rigid object in order to align the 3-D model to the observed object. The long processing times that are associated with computing 6-DoF poses may make such techniques unsuitable in cases where the desired cycle times (e.g., amount of time spent picking each object) are short. Computing 6-DoF poses may also be difficult in cases where the objects are flexible or deformable (e.g., where the objects are not rigid) because the 3-D model may not accurately match the shape of the object.

Some comparative approaches attempt to detect grasping points on objects purely using depth maps (e.g., 2.5D maps generated by a depth camera such as an RGBD camera or a time of flight depth camera, see, e.g., Guo, Jiaxin, et al. “Fast and Robust Bin-picking System for Densely Piled Industrial Objects.” arXiv preprint arXiv:2012.00316 (2020). and Nishina Yuki, et al. “Model-less Grasping Points Estimation for Bin-Picking of Non-Rigid Objects and Irregular-Shaped Objects.” Omron Technics. Vol. 52.012EN (2020).), and some approaches attempt to fit geometric primitives (e.g., hexahedrons, spheres, or cylinders) to the detected shapes of objects to select grasping points based on the geometric primitives (see, e.g., Torii, Takuya, and Manabu Hashimoto. “Model-less estimation method for robot grasping parameters using 3D shape primitive approximation.” 2018 IEEE 14th International Conference on Automation Science and Engineering (CASE). IEEE, 2018.).

Humans are able to intuitively pick objects from bins, even when cluttered and even when the objects may be new or novel (e.g., previously unseen objects).

Accordingly, some aspects of embodiments of the present disclosure relate to systems and methods for performing object picking analyses using instance segmentation to detect objects in a scene without performing an additional operation of estimating the 6-DoF pose of the object to be picked and/or without attempting to fit a 3-D model (e.g., a 3-D model of an idealized instance of the object or one or more geometric primitives) to the observed shape of the object (e.g., from a depth map). In addition, some aspects of embodiments of the present disclosure relate to further improving the pick using a depth map in conjunction with an instance segmentation mask computed by an instance segmentation module.

In general, an object pick estimation system according to embodiments of the present disclosure computes an object picking plan for a robotic arm to pick an object. The object pick estimation system may begin a process by using an imaging system to capture one or more images of a scene and then using the one or more images to compute an instance segmentation mask (or instance segmentation map) of the objects in the scene. The object pick estimation system may also estimate a depth map (e.g., based on capturing stereo images of a scene and computing depth from stereo). The instance segmentation mask identifies one or more instances of objects that appear in the images. The object pick estimation system computes pickability scores for the detected objects (using the instance segmentation mask and the depth map, if available) and an object is selected based on the pickability score. The system then computes a picking plan (e.g., coordinates of surfaces of the object that can be grasp and a direction along which to approach the object) for the robotic arm to pick the selected object. This picking plan may then be supplied to a robotic controller, which computes a motion plan to guide the robotic arm to this position and to pick up the object.

To provide some context, FIG. 1A is a schematic diagram depicting an object pick estimation system according to one embodiment of the present disclosure. As shown in FIG. 1A, a main camera 10 is arranged such that its field of view 12 captures images 40 of an arrangement 20 of objects 22 resting on a support platform in a scene 1. In the embodiment shown in FIG. 1A, the main camera 10 is located above the support platform (e.g., spaced apart from the objects 22 along the direction of gravity), but embodiments of the present disclosure are not limited thereto—for example, the main camera 10 can be arranged to have a downward angled view of the objects 22.

In some embodiments, one or more support cameras 30 are arranged at different poses or viewpoints around the scene containing the arrangement 20 of objects 22. Accordingly, each of the support cameras 30, e.g., first support camera 30a, second support camera 30b, and third support camera 30c, captures images 40 representing different views of the objects 22 from different viewpoints (e.g., a first viewpoint, a second viewpoint, and a third viewpoint, respectively) from one another and different viewpoints from the main camera 10. The viewpoints may be distinguished from one another in that they have substantially different optical axes, such as optical axes that are not parallel (non-parallel) to one another or that are spaced apart by a large distance if they are parallel to one another.

While FIG. 1A shows three support cameras 30, embodiments of the present disclosure are not limited thereto and may include, for example, at least one support camera 30 and may include more than three support cameras 30. In some embodiments, no support cameras are used and only a single main camera 10 is used from a single viewpoint.

In addition, while the main camera 10 is depicted in FIG. 1A as a stereo camera, embodiments of the present disclosure are not limited thereto, and may be used with, for example, a monocular main camera. As used herein, a stereo camera will be referred to as capturing images from a single viewpoint, as the camera modules of a stereo camera generally have optical axes that are substantially parallel to one another (and may be rectified to synthetically produce such parallel optical axes) and are generally spaced apart along a relatively short baseline to generate a depth map using stereo from a single viewpoint.

An object pick estimator 100 according to various embodiments of the present disclosure is configured to compute or estimate an object picking plan (e.g., grasping points and a direction of approach for picking or grasping) one or more the objects 22 based on information captured by the main camera 10 and the support cameras 30 (if present). The object picking plan computed by the object pick estimator 100 may then be supplied to a controller 28 for a robotic arm 24, which uses the object picking plan to generate a motion plan for moving the end effector 26 of the robotic arm to grasp or pick the selected object among the objects 22. According to various embodiments of the present disclosure, the object pick estimator 100 is implemented using one or more processing circuits or electronic circuits configured to perform various operations as described in more detail below. Types of electronic circuits may include a central processing unit (CPU), a graphics processing unit (GPU), an artificial intelligence (AI) accelerator (e.g., a vector processor, which may include vector arithmetic logic units configured efficiently perform operations common to neural networks, such dot products and softmax), a field programmable gate array (FPGA), an application specific integrated circuit (ASIC), a digital signal processor (DSP), or the like. For example, in some circumstances, aspects of embodiments of the present disclosure are implemented in program instructions that are stored in a non-volatile computer readable memory where, when executed by the electronic circuit (e.g., a CPU, a GPU, an AI accelerator, or combinations thereof), perform the operations described herein to compute a processing output, such as an instance segmentation mask and/or 6-DoF poses, from input images 40 (including, for example, polarization raw frames or the underlying images captured by polarization cameras or cameras with polarization filters in their optical paths). The operations performed by the object pick estimator 100 may be performed by a single electronic circuit (e.g., a single CPU, a single GPU, or the like) or may be allocated between multiple electronic circuits (e.g., multiple GPUs or a CPU in conjunction with a GPU). The multiple electronic circuits may be local to one another (e.g., located on a same die, located within a same package, or located within a same embedded device or computer system) and/or may be remote from one other (e.g., in communication over a network such as a local personal area network such as Bluetooth®, over a local area network such as a local wired and/or wireless network, and/or over wide area network such as the internet, such a case where some operations are performed locally and other operations are performed on a server hosted by a cloud computing service). One or more electronic circuits operating to implement the object pick estimator 100 may be referred to herein as a computer or a computer system, which may include memory storing instructions that, when executed by the one or more electronic circuits, implement the systems and methods described herein.

In more detail, the main camera 10 and the support cameras 30 are configured to capture images of scenes containing objects 22 within their fields of view 12 (while FIG. 1A illustrates a field of view 12 for the main camera 10 using dashed lines, the fields of view of the support cameras 30 are not explicitly shown). In the embodiment shown in FIG. 1A, the objects 22 are depicted abstractly as simple three-dimensional solids such as spheres, rectangular prisms, and cylinders. However, embodiments of the present disclosure are not limited thereto and object picking plans may be generated using any arbitrary object, including deformable objects mentioned above, such as flex circuits, bags or other pliable containers containing solids, liquids, and/or fluids, flexible tubing, and the like.

A “pose” refers to the position and orientation of an object with respect to a reference coordinate system. For example, a reference coordinate system may be defined with the main camera 10 at the origin, where the direction along the optical axis of the main camera 10 (e.g., a direction through the center of its field of view 12) is defined as the z-axis of the coordinate system, and the x and y axes are defined to be perpendicular to one another and perpendicular to the z-axis. (Embodiments of the present disclosure are not limited to this particular coordinate system, and a person having ordinary skill in the art would understand that poses can be mathematically transformed to equivalent representations in different coordinate systems.)

Each object 22 may also be associated with a corresponding coordinate system of its own, which is defined with respect to its particular shape. For example, a rectangular prism with sides of different lengths may have a canonical coordinate system defined where the x-axis is parallel to its shortest direction, z-axis is parallel to its longest direction, the y-axis is orthogonal to the x-axis and z-axis, and the origin is located at the centroid of the object 22.

Generally, in a three-dimensional coordinate system, objects 22 have six degrees of freedom—rotation around three axes (e.g., rotation around x-, y-, and z-axes) and translation along the three axes (e.g., translation along x-, y-, and z-axes). For the sake of clarity, symmetries of the objects 22 will not be discussed in detail herein, but may be addressed, for example, by identifying multiple possible poses with respect to different symmetries (e.g., in the case of selecting the positive versus negative directions of the z-axis of a right rectangular prism), or by ignoring some rotational components of the pose (e.g., a right cylinder is rotationally symmetric around its axis).

Based on a reference coordinate system (or camera space, e.g., defined with respect to a camera of the imaging system) and an object coordinate system (or object space, e.g., defined with respect to one of the objects), the pose of the object may be considered to be a rigid transform (rotation and translation) from object space to camera space. The pose of object 1 in camera space 1 may be denoted as P_c1¹, and the transform from object 1 space to camera space may be represented by the matrix:

$\left[\begin{array}{cccc}{R}_{11}& R_{12}& R_{13}& T_1\\ R_{21}& R_{22}& R_{23}& T_2\\ R_{31}& R_{32}& R_{33}& T_3\\ 0& 0& 0& 1\end{array}\right]$ where the rotation submatrix R:

$R=\left[\begin{array}{ccc}{R}_{11}& R_{12}& R_{13}\\ R_{21}& R_{22}& R_{23}\\ R_{31}& R_{32}& R_{33}\end{array}\right]$ represents rotations along the three axes from object space to camera space, and the translation submatrix T:

$T=\left[\begin{array}{c}{T}_1\\ T_2\\ T_3\end{array}\right]$ represents translations along the three axes from object space to camera space.

If two objects—Object A and Object B—are in the same camera C coordinate frame, then the notation P_CA is used to indicate the pose of Object A with respect to camera C and P_CB is used to indicate the pose of Object B with respect to camera C. For the sake of convenience, it is assumed herein that the poses of objects are represented based on the reference coordinate system, so the poses of objects A and B with respect to camera space C may be denoted P_A and P_B, respectively.

The pose of a particular object can be computed with respect to views from two different cameras. For example, images of Object A captured by a main camera C can be used to compute the pose P_CA of Object A with respect to main camera C. Likewise, images of Object A captured by a first support camera S₁ can be used to compute the pose P_S1A of object A with respect to the support camera S₁. If the relative poses of main camera C and support camera S₁ are known, then the pose P_S1A can be transformed to the coordinate system of the main camera C.

Some aspects of embodiments of the present disclosure also relate to providing information to assist in the control of a robotic arm 24 having an end effector 26 that may be used to grasp and manipulate objects 22 without computing the poses (e.g., 6-degree of freedom poses) of the objects to be picked or grasped. The robotic arm 24, including its end effector 26, may be controlled by a robotic arm controller 28, which, in some embodiments, receives the object picking plan computed by the object pick estimator 100.

Sensing Hardware

In the embodiment shown in FIG. 1A, the object pick estimating system includes an imaging system including a main camera 10 and one or more support cameras 30. In some embodiments of the present disclosure, the main camera 10 includes a stereo camera. Examples of stereo cameras include camera systems that have at least two monocular cameras spaced apart from each other along a baseline, where the monocular cameras have overlapping fields of view and optical axes that are substantially parallel to one another. While embodiments of the present disclosure will be presented herein in embodiments where the main camera 10 and the support cameras 30 are passive cameras (e.g., that are not connected to a dedicated light projector and that instead use ambient lighting or other light sources), embodiments of the present disclosure are not limited thereto and may also include circumstances where one or more active light projector are included in the camera system, thereby forming an active camera system, where the active light projector may be configured to project structured light or a pattern onto the scene. The support cameras 30 may be stereo cameras, monocular cameras, or combinations thereof (e.g., some stereo support cameras and some monocular support cameras). In some embodiments, the main camera 10 and/or one or more support cameras 30 may include one or more time-of-flight depth camera systems.

The main camera 10 and the support cameras 30 may use the same imaging modalities or different imaging modalities, and each of the main camera 10 and support cameras 30 may capture images using one or more different imaging modalities. Examples of imaging modalities include monochrome, color, infrared, ultraviolet, thermal, polarization, and combinations thereof (e.g., polarized color, polarized infrared, unpolarized ultraviolet, etc.).

The interaction between light and transparent objects is rich and complex, but the material of an object determines its transparency under visible light. For many transparent household objects, the majority of visible light passes straight through and a small portion (˜4% to ˜8%, depending on the refractive index) is reflected. This is because light in the visible portion of the spectrum has insufficient energy to excite atoms in the transparent object. As a result, the texture (e.g., appearance) of objects behind the transparent object (or visible through the transparent object) dominate the appearance of the transparent object. For example, when looking at a transparent glass cup or tumbler on a table, the appearance of the objects on the other side of the tumbler (e.g., the surface of the table) generally dominate what is seen through the cup. This property leads to some difficulties when attempting to detect surface characteristics of transparent objects such as glass windows and glossy, transparent layers of paint, based on intensity images alone:

FIG. 1B is a high-level depiction of the interaction of light with transparent objects and non-transparent (e.g., diffuse and/or reflective) objects. As shown in FIG. 1B, in some embodiments the main camera 10 includes a polarization camera 11 that captures polarization raw frames of a scene that includes a transparent object 41 in front of an opaque background object 42. A light ray 43 hitting the image sensor 14 of the polarization camera contains polarization information from both the transparent object 41 and the background object 42. The small fraction of reflected light 44 from the transparent object 41 is heavily polarized, and thus has a large impact on the polarization measurement, in contrast to the light 45 reflected off the background object 42 and passing through the transparent object 41.

Similarly, a light ray hitting the surface of an object may interact with the shape of the surface in various ways. For example, a surface with a glossy paint may behave substantially similarly to a transparent object in front of an opaque object as shown in FIG. 1B, where interactions between the light ray and a transparent or translucent layer (or clear coat layer) of the glossy paint causes the light reflecting off of the surface to be polarized based on the characteristics of the transparent or translucent layer (e.g., based on the thickness and surface normals of the layer), which are encoded in the light ray hitting the image sensor. Similarly, as discussed in more detail below with respect to shape from polarization (SfP) theory, variations in the shape of the surface (e.g., direction of the surface normals) may cause significant changes in the polarization of light reflected by the surface of the object. For example, smooth surfaces may generally exhibit the same polarization characteristics throughout, but a scratch or a dent in the surface changes the direction of the surface normals in those areas, and light hitting scratches or dents may be polarized, attenuated, or reflected in ways different than in other portions of the surface of the object. Models of the interactions between light and matter generally consider three fundamentals: geometry, lighting, and material. Geometry is based on the shape of the material. Lighting includes the direction and color of the lighting. Material can be parameterized by the refractive index or angular reflection/transmission of light. This angular reflection is known as a bi-directional reflectance distribution function (BRDF), although other functional forms may more accurately represent certain scenarios. For example, the bidirectional subsurface scattering distribution function (BSSRDF) would be more accurate in the context of materials that exhibit subsurface scattering (e.g. marble or wax).

A light ray 43 hitting the image sensor 14 of a polarization camera has three measurable components: the intensity of light (intensity image/I), the percentage or proportion of light that is linearly polarized (degree of linear polarization/DOLP/ρ), and the direction of that linear polarization (angle of linear polarization/AOLP/ϕ). These properties encode information about the surface curvature and material of the object being imaged, which can be used by the object pick estimator 100 to detect transparent objects, as described in more detail below. In some embodiments, by using one or more polarization cameras, the object pick estimator 100 can detect the shapes of optically challenging objects (e.g., that include surfaces made of materials having optically challenging properties such as transparency, reflectivity, or dark matte surfaces) based on similar polarization properties of light passing through translucent objects and/or light interacting with multipath inducing objects or by non-reflective objects (e.g., matte black objects).

In more detail, the polarization camera 11 may further includes a polarizer or polarizing filter or polarization mask 16 placed in the optical path between the scene 1 and the image sensor 14. According to various embodiments of the present disclosure, the polarizer or polarization mask 16 is configured to enable the polarization camera 11 to capture images of the scene 1 with the polarizer set at various specified angles (e.g., at 45° rotations or at 60° rotations or at non-uniformly spaced rotations).

As one example, FIG. 1B depicts an embodiment where the polarization mask 16 is a polarization mosaic aligned with the pixel grid of the image sensor 14 in a manner similar to a red-green-blue (RGB) color filter (e.g., a Bayer filter) of a color camera. In a manner similar to how a color filter mosaic filters incoming light based on wavelength such that each pixel in the image sensor 14 receives light in a particular portion of the spectrum (e.g., red, green, or blue) in accordance with the pattern of color filters of the mosaic, a polarization mask 16 using a polarization mosaic filters light based on linear polarization such that different pixels receive light at different angles of linear polarization (e.g., at 0°, 45°, 90°, and 135°, or at 0°, 60° degrees, and 120°). Accordingly, the polarization camera 11 using a polarization mask 16 such as that shown in FIG. 1B is capable of concurrently or simultaneously capturing light at four different linear polarizations. One example of a polarization camera is the Blackfly® S Polarization Camera produced by FLIR® Systems, Inc. of Wilsonville, Oregon

While the above description relates to some possible implementations of a polarization camera using a polarization mosaic, embodiments of the present disclosure are not limited thereto and encompass other types of polarization cameras that are capable of capturing images at multiple different polarizations. For example, the polarization mask 16 may have fewer than four polarizations or more than four different polarizations, or may have polarizations at different angles than those stated above (e.g., at angles of polarization of: 0°, 60°, and 120° or at angles of polarization of 0°, 30°, 60°, 90°, 120°, and 150°). As another example, the polarization mask 16 may be implemented using an electronically controlled polarization mask, such as an electro-optic modulator (e.g., may include a liquid crystal layer), where the polarization angles of the individual pixels of the mask may be independently controlled, such that different portions of the image sensor 14 receive light having different polarizations. As another example, the electro-optic modulator may be configured to transmit light of different linear polarizations when capturing different frames, e.g., so that the camera captures images with the entirety of the polarization mask set to, sequentially, to different linear polarizer angles (e.g., sequentially set to: 0 degrees; 45 degrees; 90 degrees; or 135 degrees). As another example, the polarization mask 16 may include a polarizing filter that rotates mechanically, such that different polarization raw frames are captured by the polarization camera 11 with the polarizing filter mechanically rotated with respect to the lens 18 to transmit light at different angles of polarization to image sensor 14. Furthermore, while the above examples relate to the use of a linear polarizing filter, embodiments of the present disclosure are not limited thereto and also include the use of polarization cameras that include circular polarizing filters (e.g., linear polarizing filters with a quarter wave plate). Accordingly, in various embodiments of the present disclosure, a polarization camera uses a polarizing filter to capture multiple polarization raw frames at different polarizations of light, such as different linear polarization angles and different circular polarizations (e.g., handedness).

As a result, the polarization camera 11 captures multiple input images (or polarization raw frames) of the scene including the surfaces of the objects 22. In some embodiments, each of the polarization raw frames corresponds to an image taken behind a polarization filter or polarizer at a different angle of polarization ϕ_pol (e.g., 0 degrees, 45 degrees, 90 degrees, or 135 degrees). Each of the polarization raw frames is captured from substantially the same pose with respect to the scene 1 (e.g., the images captured with the polarization filter at 0 degrees, 45 degrees, 90 degrees, or 135 degrees are all captured by a same polarization camera 11 located at a same location and orientation), as opposed to capturing the polarization raw frames from disparate locations and orientations with respect to the scene. The polarization camera 11 may be configured to detect light in a variety of different portions of the electromagnetic spectrum, such as the human-visible portion of the electromagnetic spectrum, red, green, and blue portions of the human-visible spectrum, as well as invisible portions of the electromagnetic spectrum such as infrared and ultraviolet.

FIG. 2A is a perspective view of a camera array 10′ according to one embodiment of the present disclosure. FIG. 2B is a cross sectional view of a portion of a camera array 10′ according to one embodiment of the present disclosure. Some aspects of embodiments of the present disclosure relate to a camera array in which multiple cameras (e.g., cameras having different imaging modalities and/or sensitivity to different spectra) are arranged adjacent to one another and in an array and may be controlled to capture images in a group (e.g., a single trigger may be used to control all of the cameras in the system to capture images concurrently or substantially simultaneously). In some embodiments, the individual cameras are arranged such that parallax shift between cameras is substantially negligible based on the designed operating distance of the camera system to objects 2 and 3 in the scene 1, where larger spacings between the cameras may be tolerated when the designed operating distance is large.

FIG. 2B shows a cross sectional view of two of the cameras or camera modules 10A′ and 10B′ of the camera array 10′ shown in FIG. 2A. As seen in FIG. 2B, each camera or camera module (10A′ and 10B′) includes a corresponding lens, a corresponding image sensor, and may include one or more corresponding filters. For example, in some embodiments, camera 10A′ is a visible light color camera that includes lens 12A′, image sensor 14A′, and color filter 16A′ (e.g., a Bayer filter). In the embodiment shown in FIG. 2B, the filter 16 is located behind the lens 12 (e.g., between the lens 12 and the image sensor 14), but embodiments of the present disclosure are not limited thereto. In some embodiments, the filter 16 is located in front of the lens 12, and in some embodiments, the filter 16 may include multiple separate components, where some components are located in front of the lens and other components are located behind the lens (e.g., a polarizing filter in front of the lens 12 and a color filter behind the lens 12). In some embodiments, camera 10B′ is a polarization camera that includes lens 12B′, image sensor 14B′, and polarizing filter 16B′ (a polarization camera may also include a visible light color filter or other filter for passing a particular portion of the electromagnetic spectrum, such as an infrared filter, ultraviolet filter, and the like). In some embodiments of the present disclosure, the image sensors four cameras 10A′, 10B′, 10C′, and 10D′ are monolithically formed on a same semiconductor die, and the four cameras are located in a same housing with separate apertures for the lenses 12 corresponding to the different image sensors. Similarly, the filters 16 may correspond to different portions of a single physical layer that has different optical filter functions (e.g., different linear polarizing angles or circular polarizers, color filters with corresponding spectral response functions, and the like) in different regions of the layer (corresponding to the different cameras). In some embodiments, a filter 16 of a polarization camera includes a polarization mask 16 similar to the Sony® IMX250MZR sensor, which includes a polarization mosaic aligned with the pixel grid of the image sensor 14 in a manner similar to a red-green-blue (RGB) color filter (e.g., a Bayer filter) of a color camera. In a manner similar to how a color filter mosaic filters incoming light based on wavelength such that each pixel in the image sensor 14 receives light in a particular portion of the spectrum (e.g., red, green, or blue) in accordance with the pattern of color filters of the mosaic, a polarization mask 16 using a polarization mosaic filters light based on linear polarization such that different pixels receive light at different angles of linear polarization (e.g., at 0°, 45°, 90°, and 135°, or at 0°, 60° degrees, and 120°). Accordingly, a camera of the camera array 10′ may use a polarization mask 16 to concurrently or simultaneously capture light at four different linear polarizations.

In some embodiments, a demosaicing process is used to compute separate red, green, and blue channels from the raw data. In some embodiments of the present disclosure, each polarization camera may be used without a color filter or with filters used to transmit or selectively transmit various other portions of the electromagnetic spectrum, such as infrared light.

As noted above, embodiments of the present disclosure relate to multi-modal and/or multi-spectral camera arrays. Accordingly, in various embodiments of the present disclosure, the cameras within a particular camera array include cameras configured to perform imaging in a plurality of different modalities and/or to capture information in a plurality of different spectra.

As one example, in some embodiments, the first camera 10A′ is a visible light camera that is configured to capture color images in a visible portion of the electromagnetic spectrum, such as by including a Bayer color filter 16A′ (and, in some cases, a filter to block infrared light), and the second camera 10B′, third camera 10C′, and fourth camera 10D′ are polarization cameras having different polarization filters, such filters having linear polarization angles of 0°, 60°, and 120°, respectively. The polarizing filters in the optical paths of each of the cameras in the array cause differently polarized light to reach the image sensors of the cameras. The individual polarization cameras in the camera array have optical axes that are substantially perpendicular to one another, are placed adjacent to one another, and have substantially the same field of view, such that the cameras in the camera array capture substantially the same view of a scene as the visible light camera 10A′, but with different polarizations. While the embodiment shown in FIG. 2A includes a 2×2 array of four cameras, three of which are polarization cameras, embodiments of the present disclosure are not limited thereto, and the camera array may more than three polarization cameras, each having a polarizing filter with a different polarization state (e.g., a camera array may have four polarization cameras along with the visible light color camera 10A′, where the polarization cameras may have polarization filters with angles of linear polarization, such as 0°, 45°, 90°, and 135°). In some embodiments, one or more of the cameras may include a circular polarizer.

As another example, one or more of the cameras in the camera array 10′ may operate in other imaging modalities and/or other imaging spectra, such as polarization, near infrared, far infrared, shortwave infrared (SWIR), longwave infrared (LWIR) or thermal, ultraviolet, and the like, by including appropriate filters 16 (e.g., filters that pass light having particular polarizations, near-infrared light, SWIR light, LWIR light, ultraviolet light, and the like) and/or image sensors 14 (e.g., image sensors optimized for particular wavelengths of electromagnetic radiation) for the particular modality and/or portion of the electromagnetic spectrum.

For example, in the embodiment of the camera array 10′ shown in FIG. 2A, four cameras 10A′, 10B′, 10C′, and 10D′ are arranged in a 2×2 grid to form a camera array, where the four cameras have substantially parallel optical axes. In addition, the optical axes of the camera modules of the camera array are arranged close together such that the camera modules capture images from substantially the same viewpoint with respect to the objects in the scene 1. One of skill in the art would understand that the acceptable spacing between the optical axes of the camera modules within an array in order to capture images of the scene from substantially the same viewpoint depends on the working distance to objects 22 in the scene, where longer working distances allow for larger spacing between the optical axes while shorter working distances may require closer or tighter spacing between the optical axes. The four cameras may be controlled together such that they capture images substantially simultaneously. In some embodiments, the four cameras are configured to capture images using the same exposure settings (e.g., same aperture, length of exposure, and gain or “ISO” settings). In some embodiments, the exposure settings for the different cameras can be controlled independently from one another (e.g., different settings for each camera), where the object pick estimator 100 jointly or holistically sets the exposure settings for the cameras based on the current conditions of the scene 1 and the characteristics of the imaging modalities and spectral responses of the cameras 10A′, 10B′, 10C′, and 10D′ of the camera array 10′.

In some embodiments, the various individual cameras of the camera array are registered with one another by determining their relative poses (or relative positions and orientations) by capturing multiple images of a calibration target, such as a checkerboard pattern, an ArUco target (see, e.g., Garrido-Jurado, Sergio, et al. “Automatic generation and detection of highly reliable fiducial markers under occlusion.” Pattern Recognition 47.6 (2014): 390-402.) or a ChArUco target (see, e.g., An, Gwon Hwan, et al. “Charuco board-based omnidirectional camera calibration method.” Electronics 7.12 (2018): 421.). In particular, the process of calibrating the targets may include computing intrinsic matrices characterizing the internal parameters of each camera (e.g., matrices characterizing the focal length, image sensor format, and principal point of the camera) and extrinsic matrices characterizing the pose of each camera with respect to world coordinates (e.g., matrices for performing transformations between camera coordinate space and world or scene coordinate space). Different cameras within a camera array may have image sensors with different sensor formats (e.g., aspect ratios) and/or different resolutions without limitation, and the computed intrinsic and extrinsic parameters of the individual cameras enable the object pick estimator 100 to map different portions of the different images to a same coordinate space (where possible, such as where the fields of view overlap).

FIG. 2C is a perspective view of a stereo camera array system 10 according to one embodiment of the present disclosure. For some applications, stereo vision techniques are used to capture multiple images of scene from different perspectives. As noted above, in some embodiments of the present disclosure, individual cameras (or camera modules) within a camera array 10′ are placed adjacent to one another such that parallax shifts between the cameras are small or substantially negligible based on the designed operating distance of the camera system to the subjects being imaged (e.g., where the parallax shifts between cameras of a same array are less than a pixel for objects at the operating distance). In addition, as noted above, in some embodiments, differences in the poses of the individual cameras within a camera array 10′ are corrected through image registration based on the calibrations (e.g., computed intrinsic and extrinsic parameters) of the cameras such that the images are aligned to a same coordinate system for the viewpoint of the camera array.

In stereo camera array systems according to some embodiments, the camera arrays are spaced apart from one another such that parallax shifts between the viewpoints corresponding to the camera arrays are detectable for objects in the designed operating distance of the camera system. This enables the distances to various surfaces in a scene (the “depth”) to be detected in accordance with a disparity measure or a magnitude of a parallax shift (e.g., larger parallax shifts in the locations of corresponding portions of the images indicate that those corresponding portions are on surfaces that are closer to the camera system and smaller parallax shifts indicate that the corresponding portions are on surfaces that are farther away from the camera system). These techniques for computing depth based on parallax shifts are sometimes referred to as Depth from Stereo

Accordingly, FIG. 2C depicts a stereo camera array system 10 having a first camera array 10-1′ and a second camera array 10-2′ having substantially parallel optical axes and spaced apart along a baseline 10-B. In the embodiments shown in FIG. 2C, the first camera array 10-1′ includes cameras 10A′, 10B′, 10C′, and 10D′ arranged in a 2×2 array similar to that shown in FIG. 2A and FIG. 2B. Likewise, the second camera array 10-2′ includes cameras 10E′, 10F′, 10G′, and 10H′ arranged in a 2×2 array, and the overall stereo camera array system 10 includes eight individual cameras (e.g., eight separate image sensors behind eight separate lenses). In some embodiments of the present disclosure, corresponding cameras of the camera arrays 10-1′ and 10-2′ are of the same type or, in other words, configured to capture raw frames or images using substantially the same imaging modalities or in substantially the same spectra. In the specific embodiment shown in FIG. 2C, cameras 10A′ and 10E′ may be of a same first type, cameras 10B′ and 10F′ may be of a same second type, cameras 10C′ and 10G′ may be of a same third type, and cameras 10D′ and 10H′ may be of a same fourth type. For example, cameras 10A′ and 10E′ may both have linear polarizing filters at a same angle of 0°, cameras 10B′ and 10F′ may both have linear polarizing filters at a same angle of 45°, cameras 10C′ and 10G′ may both be viewpoint-independent cameras having no polarization filter (NF), such as near-infrared cameras, and cameras 10D′ and 10H′ may both have linear polarizing filters at a same angle of 90°. As another example, cameras 10A′ and 10E′ may both be viewpoint-independent cameras such as visible light cameras without polarization filters, cameras 10B′ and 10F′ may both be thermal cameras, cameras 10C′ and 10G′ may both have polarization masks with a mosaic pattern polarization filters at different angles of polarization (e.g., a repeating pattern with polarization angles of 0°, 45°, 90°, and 135°), and cameras 10D′ and 10H′ may both be thermal (LWIR) cameras.

While some embodiments are described above wherein each array includes cameras of different types in a same arrangement, embodiments of the present disclosure are not limited thereto. For example, in some embodiments, the arrangements of cameras within a camera array are mirrored along an axis perpendicular to the baseline 10-B. For example, cameras 10A′ and 10F′ may be of a same first type, cameras 10B′ and 10E′ may be of a same second type, cameras 10C′ and 10H′ may be of a same third type, and cameras 10D′ and 10G′ may be of a same fourth type.

In a manner similar to that described for calibrating or registering cameras within a camera array, the various polarization camera arrays of a stereo camera array system may also be registered with one another by capturing multiple images of calibration targets and computing intrinsic and extrinsic parameters for the various camera arrays. The camera arrays of a stereo camera array system 10 may be rigidly attached to a common rigid support structure 10-S in order to keep their relative poses substantially fixed (e.g., to reduce the need for recalibration to recompute their extrinsic parameters). The baseline 10-B between camera arrays is configurable in the sense that the distance between the camera arrays may be tailored based on a desired or expected operating distance to objects in a scene—when the operating distance is large, the baseline 10-B or spacing between the camera arrays may be longer, whereas the baseline 10-B or spacing between the camera arrays may be shorter (thereby allowing a more compact stereo camera array system) when the operating distance is smaller.

As noted above with respect to FIG. 1B, a light ray 43 hitting the image sensor 14 of a polarization camera 10 has three measurable components: the intensity of light (intensity image/I), the percentage or proportion of light that is linearly polarized (degree of linear polarization/DOLP/ρ), and the direction of that linear polarization (angle of linear polarization/AOLP/ϕ).

Measuring intensity I, DOLP ρ, and AOLP ϕ at each pixel requires 3 or more polarization raw frames of a scene taken behind polarizing filters (or polarizers) at different angles, ϕ_pol (e.g., because there are three unknown values to be determined: intensity I, DOLP ρ, and AOLP ϕ. For example, a polarization camera such as those described above with respect to FIG. 1B captures polarization raw frames with four different polarization angles ϕ_pol, e.g., 0 degrees, 45 degrees, 90 degrees, or 135 degrees, thereby producing four polarization raw frames I_ϕpol, denoted herein as I₀, I₄₅, I₉₀, and I₁₃₅ and a camera module in accordance with some embodiments of FIGS. 2A, 2B, and 2C may capture polarization raw frames at three different polarization angles ϕ_pol, e.g., 0 degrees, 60 degrees, and 120 degrees, thereby producing three polarization raw frames I_ϕpol, denoted herein as I₀, I₆₀, and I₁₂₀.

The relationship between I_ϕpol and intensity I, DOLP ρ, and AOLP ϕ at each pixel can be expressed as:I_ϕpol=I(1+ρ cos(2(ϕ−ϕ_pol)))  (1)

Accordingly, with four different polarization raw frames I_ϕpol (I₀, I₄₅, I₉₀, and I₁₃₅), a system of four equations can be used to solve for the intensity I, DOLP ρ, and AOLP ϕ.

Shape from Polarization (SfP) theory (see, e.g., Gary A Atkinson and Edwin R Hancock. Recovery of surface orientation from diffuse polarization. IEEE transactions on image processing, 15(6):1653-1664, 2006.) states that the relationship between the refractive index (n), azimuth angle (θ_a) and zenith angle (θ_z) of the surface normal of an object and the ϕ and ρ components of the light ray coming from that object follow the following characteristics when diffuse reflection is dominant:

$\begin{array}{cc}\rho =\frac{{\left(n-\frac{1}{n}\right)}^2{\sin }^2\left({\theta }_z\right)}{2+2n^2-{\left(n+\frac{1}{n}\right)}^2{\sin }^2{\theta }_z+4\cos {\theta }_z\sqrt{n^2-{\sin }^2{\theta }_z}}& \left(2\right)\end{array}$ $\begin{array}{cc}\varphi ={\theta }_a& \left(3\right)\end{array}$ and when the specular reflection is dominant:

$\begin{array}{cc}\rho =\frac{2{\sin }^2{\theta }_z\cos {\theta }_z\sqrt{n^2-{\sin }^2{\theta }_z}}{n^2-{\sin }^2{\theta }_z-n^2{\sin }^2{\theta }_z+2{\sin }^4{\theta }_z}& \left(4\right)\end{array}$ $\begin{array}{cc}\varphi ={\theta }_a-\frac{\pi }{2}& \left(5\right)\end{array}$

Note that in both cases ρ increases exponentially as θ_z increases and if the refractive index is the same, specular reflection is much more polarized than diffuse reflection.

Accordingly, some aspects of embodiments of the present disclosure relate to applying SfP theory to detect or measure the gradients of surfaces (e.g., the orientation of surfaces or their surface normals or directions perpendicular to the surfaces) based on the raw polarization frames of the objects, as captured by the polarization cameras among the main camera 10 and the support cameras 30. Computing these gradients produces a gradient map (or slope map or surface normals map) identifying the slope of the surface depicted at each pixel in the gradient map. These gradient maps can then be used when estimating the shape and/or pose of the object by supplying these gradient maps or surface normals maps to a trained computer vision model (e.g., a convolutional neural network) and/or by aligning a pre-existing 3-D model (e.g., CAD model) of the object with the measured surface normals (gradients or slopes) of the object in based on the slopes of the surfaces of the 3-D model, as described in more detail below.

One example of an imaging system according to embodiments of the present disclosure includes a stereo pair of 2×2 camera arrays, in an arrangement similar to that shown in FIG. 2C, Each 2×2 camera array includes three color (RGB) cameras with polarization filters at different angles to capture a diverse range of polarization signatures of the scene in the spectral bands (red, green, and blue) and fourth near-IR camera without a polarization filter to capture the scene in the near-IR spectral band. This stereo pair of 2×2 camera arrays may be combined with other cameras located at different viewpoints with respect to the scene, thereby providing a multi-view imaging system. The other cameras may also be similar stereo camera arrays (e.g., similar stereo pairs of 2×2 camera arrays) or monocular camera arrays (e.g., single camera arrays of closely-spaced camera modules), and the camera arrays, in the stereo or monocular case, may have different arrangements and numbers of camera modules in the array (e.g., a 3×2 arrangement of 6 camera modules), and where the camera modules may operate in different modalities (e.g., thermal, ultraviolet, depth from time of flight, polarization, and the like).

FIG. 3A is a block diagram of an object pick estimator according to one embodiment of the present disclosure. FIG. 3B is a flowchart depicting a method for computing object pick estimates according to some embodiments of the present disclosure.

As shown in FIG. 3B, in operation 310, the object pick estimator 100 controls one or more cameras, such as the main camera 10 and the support cameras 30, to capture one or more images of the scene, which may be from multiple viewpoints in the case of multiple cameras. In embodiments using multiple cameras, the cameras are configured to capture images concurrently or substantially simultaneously. In the case of in imaging system that includes multiple cameras, each camera is arranged at a different pose with respect to the scene 1, such that each camera captures scene from its corresponding different viewpoint. Accordingly, the collection of images captured by multiple cameras represent a collection of multi-viewpoint images of the scene 1. (In some embodiments, the images are captured from multiple viewpoints using one or more cameras, such as by moving the one or more cameras between different viewpoints while keeping the scene fixed, and/or rigidly transforming the scene between captures by the one or more cameras.) The one or more images of the scene may be referred to herein as being “consistent” in that they are all pictures of the same consistent scene but providing different views of the scene from different viewpoints and/or different imaging modalities. This consistency between the images of the scene may be achieved by capturing all of the images substantially simultaneously or concurrently or by requiring that none of the objects of interest in the scene that are depicted in the image have moved (e.g., translated or rotated) between in the time between the capture of different images of the scene.

In some circumstances, one or more of the “cameras” are multi-modal cameras that capture multiple images from the same viewpoint, but having different modalities, such as different portions of the electromagnetic spectrum (e.g., red, green and blue portions of the visible light spectrum, near infrared light, far infrared light, ultraviolet light, etc.), different optical filters (e.g., linear polarization filters at different angles and/or circular polarization filters), and combinations thereof. Accordingly, a collection of multi-viewpoint images of a scene does not require that all images be captured from different viewpoints, but only that there are at least two images captured from different viewpoints. Such a collection of multi-viewpoint images therefore may include at least some images that are captured from the same viewpoint.

For the sake of exposition, the following discussion presents an embodiment where the imaging system includes only one camera, referred to as main camera 10, where the main camera 10 may include multiple camera modules (e.g., a multi-modal stereo camera such as that shown in FIG. 2C). Further embodiments of the present disclosure using multi-view capture (e.g., using a main camera 10 along with one or more support cameras 30) will be described in more detail following the discussion of embodiments using a single camera.

As shown in FIG. 3A, an object pick estimator 100 according to some embodiments of the present disclosure is configured to receive the one or more input images 40 captured by the imaging system including the main camera 10.

In operation 330, the object pick estimator 100 supplies the one or more images to an instance segmentation module 110 configured to compute instance segmentation masks based on input images. In more detail, aspects of embodiments of the present disclosure relate to the use of instance segmentation algorithms and trained instance segmentation networks that compute instance segmentation masks from supplied two-dimensional intensity images (e.g., color intensity images, such as red, green, blue (RGB) intensity images and/or monochrome intensity images). In some embodiments, the instance segmentation networks are existing, previously-trained instance segmentation convolutional neural networks that take 2-D intensity images as input and compute instance segmentation masks from the 2-D intensity images, where the instance segmentation masks label pixels of the images as being associated with a same instance of an object (e.g., assigning different instance identifiers to two different dogs appearing in an image), and may also label each instance with an object class (e.g., dogs, cats, people, cars, and the like).

The instance segmentation module 110 may include convolutional neural network backbones such as those described in Mask R-CNN (He, Kaiming, et al. “Mask R-CNN.” Proceedings of the IEEE International Conference on Computer Vision. 2017.), AlexNet (see, e.g., Krizhevsky, Alex, Ilya Sutskever, and Geoffrey E. Hinton. “ImageNet classification with deep convolutional neural networks.” Advances in neural information processing systems. 2012.), VGG (see, e.g., Simonyan, Karen, and Andrew Zisserman. “Very deep convolutional networks for large-scale image recognition.” arXiv preprint arXiv:1409.1556 (2014).), ResNet-101 (see, e.g., Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Deep residual learning for image recognition. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pages 770-778, 2016.), MobileNet (see, e.g., Howard, Andrew G., et al. “Mobilenets: Efficient convolutional neural networks for mobile vision applications.” arXiv preprint arXiv:1704.04861 (2017).), MobileNetV2 (see, e.g., Sandler, Mark, et al. “MobileNetV2: Inverted residuals and linear bottlenecks.” Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. 2018.), and MobileNetV3 (see, e.g., Howard, Andrew, et al. “Searching for MobileNetV3.” Proceedings of the IEEE International Conference on Computer Vision. 2019.), although embodiments of the present disclosure are not limited thereto. The feature pyramid network (FPN) 814 (see, e.g., Lin, Tsung-Yi, et al. “Feature pyramid networks for object detection.” Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. 2017.) enables the capture of multiscale features in the images (e.g., distinctive features of objects appearing at different sizes within the image), although embodiments of the present disclosure are not limited thereto, and other networks for capturing multiscale features may be used, such as a U-Net (see, e.g., Ronneberger, Olaf, Philipp Fischer, and Thomas Brox. “U-Net: Convolutional networks for biomedical image segmentation.” International Conference on Medical Image Computing and Computer-Assisted Intervention. Springer, Cham, 2015.) or a DeepMask (Pinheiro, Pedro O., et al. “Learning to Segment Object Candidates.” ArXiv:1506.06204 [Cs], September 2015. arXiv.org, http://arxiv.org/abs/1506.06204.).

In embodiments where the imaging system includes one or more polarization cameras, the images 40 include polarized images, and the instance segmentation module 110 may include a Polarized CNN backbone as described in PCT Patent Application No. PCT/US2020/048604, also filed as U.S. patent application Ser. No. 17/266,046, the content of which is incorporated herein by reference. Additional details regarding polarization CNN backbones is presented in a later section. The use of polarized images enables the detection and segmentation of optically-challenging objects that would otherwise be difficult (e.g., unreliable) to detect using standard color imaging techniques. Examples of optically challenging objects include objects that are transparent (e.g., glass or transparent plastic), highly reflective (e.g., shiny metallic surfaces, highly polished painted surfaces, mirrored surfaces, surfaces with specular reflections, and the like), dark matte surfaces (e.g., black rubber), and the like.

In some embodiments, the object pick estimator 100 includes a depth estimator 130 configured to compute one or more depth maps from the input images 40. These depth maps may be computed using any of a variety of techniques such as depth from stereo, depth from time of flight, and the like. For example, in the case of depth from stereo, the one or more images 40 includes a stereoscopic pair of images captured by a stereo pair (e.g., as shown in FIG. 2C), and a depth estimator 130 computes a disparity map from the stereoscopic pair of images, such as by performing block matching along epipolar lines or by using a trained disparity network. The disparity map generally indicates amount by which particular portions of the image have exhibited a parallax shift between the views of the two members of the stereo pair, where a larger parallax shift or larger disparity indicates that the corresponding imaged surfaces are closer to the depth camera and smaller disparities indicates that the corresponding imaged surfaces are farther from the depth camera. Accordingly, a depth camera can compute a 2.5D depth map where each pixel in the depth map indicates the estimated distance to the imaged surface, and where the depth map can be represented as a point cloud of 3-D points. As another example, in the case of time of flight, the depth estimator 130 computes the depth map based on differences between the arrival times of light in the various pixels in response to projecting coded light into the scene (e.g., based on a delay time of a pulse of light or based on a phase shift in projected light that varies sinusoidally in intensity over time).

In operation 350, in embodiments where the object pick estimator 100 has access to depth maps (e.g., as computed by a depth estimator 130), the object pick estimator 100 segments the depth map based on the instance segmentation mask computed in operation 330.

In more detail, continuing the above example where the main camera 10 is a stereo array camera, the instance segmentation mask and the depth map may be computed based on images (e.g., color images, polarization raw frames, and the like) captured by the same main camera 10, and therefore are captured from the same viewpoint. As such, the instance segmentation mask may be used to directly identify the pixels (or 3-D points) of the depth map that correspond to the same instances of objects. For example, an instance segmentation mask may identify three different boxes and assign arbitrary example labels ‘A’, ‘B’, and ‘C’ to the pixels of the instance segmentation mask that correspond to the three different boxes in a captured 2-D image (e.g., a 2-D color intensity image). Each pixel of a depth map encodes a distance between the camera and a depicted surface of scene 1, but does not encode information about which points of the depth map correspond to different objects (e.g., the depth map may encode the shapes of the different boxes in the scene, but does not encode semantic information such as whether the boxes are parts of a same object or multiple different objects). When the depth map and the instance segmentation mask depict the same scene from the same viewpoint, then each portion of the instance segmentation mask assigns an object instance to corresponding portions of the depth map, thereby labeling the points of the depth map with semantic information about the scene, such as the class of object and an instance identifier (e.g., labeling depth points corresponding to the shapes of three boxes in the scene with different labels such as ‘A’, ‘B’, and ‘C’).

In addition, in embodiments where the depth map and the instance segmentation mask are captured from different viewpoints, the object pick estimator 100 re-projects the points in the point cloud of the depth map to the viewpoint of the images used to compute the instance segmentation mask (in accordance with extrinsic camera parameters of the cameras used to capture the images used to capture the instance segmentation mask and the images used to generate the depth map), thereby enabling the instance segmentation mask to identify which points of the depth map correspond to the different detected instances of objects.

Accordingly, in embodiments where the object pick estimator 100 has access to a depth map of the scene, in operation 350 the object pick estimator 100 segments objects in the depth maps using instance segmentation masks computed directly from 2-D images (e.g., color intensity images, monochrome intensity images, polarization raw frames, polarization signatures, and the like) of the scene.

In operation 370, the object selector 150 of the object pick estimator 100 selects an object to be picked based on object pickability scores computed for the one or more objects detected through operation 330 and/or operation 350. In particular, the object pickability scores represent an estimated ease of picking up an object using a particular robotic arm 24 with a particular end effector 26 (e.g., an end effector with gripping fingers or an end effector with a suction gripper). In the discussion herein, a high pickability score indicates an object that is more easily pickable, while a low pickability score indicates an object that is more difficult to pick. For example, objects that are underneath or blocked by objects above them may be effectively unpickable in the current state of the bin and would have very low pickability scores, whereas objects that are on top of a pile and that are laterally spaced from other objects may have high pickability scores.

Aspects of embodiments of the present disclosure relate to computing pickability scores of objects using instance segmentation masks and, if available, depth maps of a scene containing objects, without additionally computing the poses (e.g., 6-DoF poses) of the objects to be picked.

For the sake of discussion, all of the objects x detected in a scene (e.g., detected by the instance segmentation module 110 in operation 330) may be represented by a set X. For each object x in the set, x_mask is the set of pixels that belong to object x, as identified by the instance segmentation mask. In addition, x_cloud is a point cloud of 3-D points (e.g., having x, y, and z coordinates) selected from the points in a depth map, as filtered or extracted from the depth map by the portion of the instance segmentation mask x_mask corresponding to that object x. The current location (e.g., 3-D coordinates) of the end effector 26 or gripper of the robot arm 24 may be represented by R.

In some embodiments of the present disclosure, the pickability score PS of an object x is defined as:PS(x)=k₁|x_mask|+k₂P_clutter(x)−k₃P_prox(x)+k₄P_outness(x)  (6)where the parameters k are parameters that are tuned for particular deployment environments (e.g., based on characteristics such as the shapes of the objects to be picked and the gripping methods used by the end effectors). |x_mask| represents the number of pixels of the instance segmentation mask that correspond to the current object x.

In the definition of pickability score PS above, P_clutter(x) is a measure of the clutter in the neighborhood of object x. In some embodiments, P_clutter(x) is calculated using the detected point clouds and/or depth maps by comparing the center of mass (a three-dimensional coordinate computed based on the mean coordinate of the point cloud of the object) of the current object x with the center of mass of each other detected object z:

$\begin{array}{cc}{P}_{\mathrm{clutter}}\left(x\right)=\sum _{z\in X}\max \left(d,\mathrm{mean}\left(x_{\mathrm{cloud}}\right)-\mathrm{mean}\left(z_{\mathrm{cloud}}\right)\right)& \left(7\right)\end{array}$ where d is a tunable parameter.

P_prox(x) in the definition above is a measure of the distance of the object from the current position R of the end effector 26 of the robot arm 24, predicted by the center of the detected point cloud x_cloud of the object x. In particular, this factor is used to favor closer objects, thereby reducing the cycle time by reducing the movements of the robot arm:P_prox(x)=∥mean(x_cloud)−R∥  (8)

P_outness(x) in the definition above is a measure of how much an object “sticks out” or protrudes from among the other objects in the bin. Intuitively, objects that protrude more from the other objects in a bin are easier to grasp than objects that are buried or otherwise blocked by other objects. In some embodiments, when computing a pickability score for an object x, the object selector 150 of the object pick estimator 100 defines a neighborhood of points N(x) around the object x but that are not a part of x. A plane p is then fit to these points, and P_outness computes a measure of how far x extends out of the plane p (and toward the position R of the end effector 26):

$\begin{array}{cc}{P}_{\mathrm{outness}}\left(x\right)=\mathrm{signum}\left(R-\underset{p}{\mathrm{proj}}\left(R\right)\right)\sum _{q\in x_{\mathrm{cloud}}}{\left(q-\underset{p}{\mathrm{proj}}\left(q\right)\right)}^3& \left(9\right)\end{array}$

Accordingly, an object x′ may be selected from the detected objects X by identifying the object having highest pickability score:

$\begin{array}{cc}{x}^{\prime }=\underset{x\in X}{\arg \max }\mathrm{PS}\left(x\right)& \left(10\right)\end{array}$

While the preceding discussion describes computing a pickability score PS using both an instance segmentation mask and a depth map according to some embodiments of the present disclosure, embodiments of the present disclosure are not limited thereto.

For example, in some embodiments, the pickability score PS includes a factor relating to the blocking of the object by other objects in the scene, which may be determined based on portions of the object being occluded from the view by other objects in the scene.

As another example, in some embodiments, the pickability score PS includes a score relating to the reachability of the object and/or whether the object is graspable using the end effector 26 installed on the robot arm 24 that is guided by the object pick estimator 100. For example, in the case of an end effector having a finger gripper, the pickability analysis may further include a determination of whether the fingers of the gripper can reach the lateral surfaces of the object x. As one example, this factor may include a finger insertion point analysis described in Nishina Yuki, et al. “Model-less Grasping Points Estimation for Bin-Picking of Non-Rigid Objects and Irregular-Shaped Objects.” Omron Technics. Vol. 52.012EN (2020).

As another example, in the case of a suction-based gripper, the analysis may include determining a shape of the object, such as from the depth map or from a surface normals map computing using shape from polarization based on captured polarization raw frames, to detect surfaces of the object x that are suitably-shaped for gripping by a suction gripper (e.g., sufficiently flat and smooth surfaces of the object).

As a further example, in some embodiments of the present disclosure, the pickability score is computed from an instance segmentation mask alone (e.g., without also using a depth map). In some such embodiments, the terms of the pickability score PS that require or use the depth map are modified to remove the reliance on the depth map.

For example, the P_clutter(x) factor may be modified to compute the distances to neighboring objects based solely on the distance between the objects in two-dimensional space (e.g., the centers of their respective masks):

$\begin{array}{cc}{P}_{\mathrm{clutter}}\left(x\right)=\sum _{z\in X}\max \left(d,\mathrm{mean}\left(x_{\mathrm{mask}}\right)-\mathrm{mean}\left(z_{\mathrm{mask}}\right)\right)& \left(11\right)\end{array}$ and, likewise, the P_mask factor may be computed based on distance within a plane parallel to the image plane p_mask of the instance segmentation mask:

$\begin{array}{cc}{P}_{\mathrm{prox}}\left(x\right)=\mathrm{mean}\left(x_{\mathrm{mask}}\right)-\underset{p_{\mathrm{mask}}}{\mathrm{proj}}\left(R\right)& \left(12\right)\end{array}$

In some embodiments where a pickability score is computed based on 2-D images (e.g., an instance segmentation mask) alone, the object selector 150 may further include an object pickability network (or object pickability neural network) that is trained to compute object pickability scores based on a supplied instance segmentation mask. For example, in some embodiments the object pickability network is trained to classify each instance of an object into one of a plurality of different pickability scores (e.g., “low pickability,” “moderate pickability,” and “high pickability”) or may be configured to compute a numerical pickability score (e.g., a normalized value from 0.0 to 1.0, where 0.0 represents a “difficult” pickability score and 1.0 represents an “easy” pickability score). In addition to an instance segmentation mask, the object pickability network may further take additional inputs such as the captured images 40

In some embodiments, the object pickability network is trained based on labeled training data, where the labels are pickability scores previously computed based on the pickability scores described above. For example, depth maps may be used along with instance segmentation masks for computing the pickability scores of objects when generating the training dataset, and the depth maps are omitted from the training data set, such that the instance segmentation masks (and, in some embodiments, other captured images) are included along with the calculated pickability scores when training the object pickability network. In some embodiments, the training dataset is further refined based on whether an actual robotic arm was able to pick the object (e.g., data points where the object pickability score was high but the robot arm 24 failed to pick the object may be removed or revised to indicate a low object pickability score).

In addition, in some embodiments of the present disclosure that include the use of depth maps, the object pickability network is trained to predict object pickability scores based on an instance segmentation mask and a depth map supplied as inputs.

In operation 390, the object picking planner 170 of the object pick estimator 100 computes an object picking plan for picking the selected object (e.g., the object selected by the object selector 150 in operation 370).

In one example embodiment, an object picking planner 170 of the object pick estimator 100 implements a grasp planner for a robotic arm based on the topography of the bin (as identified in a depth map) and identification of which object is associated with each depth estimate (through the instance segmentation mask).

In some embodiments, the object picking planner 170 computes the “visible center of mass” of an object x by computing the average or mean point within x_cloud:visible center of mass=mean(x_cloud)  (13)

Because some portions of the object may not be visible to the camera, the actual center of mass is likely to be farther from the camera than the visible center of mass. Accordingly, in some embodiments of the present disclosure, the estimated center of mass x is also adjusted by a scalar value k to be farther from the camera used to capture the images of the scene (assuming a coordinate system aligned with the camera):x=mean(x_cloud)+[k k k]^T  (14)where k is a tunable parameter.

As such, in some embodiments of the present disclosure, the object picking planner 170 determines a target picking position based on the estimated center of mass x, such as by directing the end effector (e.g., a center of the end effector between the multiple fingers of the end effector or a center of a suction end effector) toward the estimated center of mass x of the object x.

Some aspects of embodiments of the present disclosure also relate to computing a direction of approach for the end effector 26 of the robotic arm 24 toward the calculated picking position. In some embodiments, a plane p is fit to a neighborhood of points N(x) surrounding and including x_cloud. The direction normal (perpendicular or orthogonal) to the plane p may then be represented as {right arrow over (v)}, and the direction opposite to the normal vector may be used as the direction of approach (−{right arrow over (v)}).

In some embodiments of the present disclosure, the direction of approach also depends on the presence of other objects as detected in the scene, such as avoiding collisions with other objects in the scene. For example, in the case of a gripper using fingers, nearby objects may prevent the gripper from reaching some of the lateral sides of a particular object, while other lateral sides of the object may be unobstructed. Accordingly, finger insertion points (see, e.g., Nishina Yuki, et al. “Model-less Grasping Points Estimation for Bin-Picking of Non-Rigid Objects and Irregular-Shaped Objects.” Omron Technics. Vol. 52.012EN (2020).) may be determined from the depth map and/or based on the distances to neighboring objects to select an approach direction and orientation for the end effector to approach the object x.

In some embodiments of the present disclosure, the object picking planner 170 selects an approach direction based on other or additional criteria such as surface normals of the object x. For example, a finger gripper may be oriented such that the fingers of the gripper apply force along a direction that is substantially perpendicular to lateral surface of the object. As another example, a suction gripper may approach a surface of the object along a direction that is substantially perpendicular to the surface to improve contact.

Multi-Viewpoint Object Picking Estimation

Some aspects of embodiments of the present disclosure were described above in the context of computing an object picking plan for an object from a collection of objects, based on an instance segmentation mask viewing the objects from a single viewpoint. However, embodiments of the present disclosure are not limited thereto.

For example, in the case of an imaging system with multiple cameras (e.g., a main camera 10 and one or more support cameras 30) imaging a scene of objects from multiple different viewpoints, a separate instance segmentation mask can be calculated from each of the different viewpoints. Likewise, if applicable, a separate depth map may be calculated from each of these different viewpoints. In some embodiments, the object selector 150 computes pickability scores independently based on the instance segmentation masks (and, if applicable, the depth maps) captured for each of the viewpoints. This may result in multiple pickability scores being calculated for a same object that is visible from move than one viewpoint. However, the pickability scores may differ from the different viewpoints, based on the particular arrangement of the object with respect to its neighboring objects. The object selector 150 may then select an object to be picked based on the pickability scores aggregated across the objects visible in the different views.

According to some embodiments of the present disclosure, the object selector computes object-level correspondences between the objects depicted in different views. In other words, instances of the same object that are detected in multiple views are semantically connected, and a pickability score for the object is computed based on the combination of information from the different viewpoints. Some techniques for computing object-level correspondences are described in more detail below.

For example, in some embodiments, depth maps from multiple viewpoints are merged (e.g., based on known extrinsic parameters of the different cameras of the imaging system) and therefore more detailed and complete point clouds x_cloud may be computed for the objects in the scene. These point clouds may then be used to compute more accurate pickability scores PS for the objects.

The object pick estimator 100 outputs the object picking plan computed in operation 390 to a robot controller 28, which uses the object picking plan (including, for example, the picking position and the direction of approach) to control a robot arm 24 to move the end effector 26 to the object to pick the object up. The particular details of the motion plan is specific to characteristics of the robot arm 24 (e.g., the particular dimensions and arrangements of the mechanical linkages of the robot arm 24) as well as its current position R.

Computing Object-Level Correspondences

In the case of a sensing system using multi-viewpoint images or images of a scene from more than one viewpoint, in operation 330, the object pick estimator 100 computes object-level correspondences on the multi-viewpoint images of the scene. More specifically, instances of one or more types of objects are identified in the multi-viewpoint images of the scene, and corresponding instances of objects are identified between the multi-viewpoint images. For example, a scene 1 may include two cubes and three spheres, and various of the multi-viewpoint images may depict some or all of these five objects. A process of instance segmentation identifies the pixels in each of the images that depict the five objects, in addition to labeling them separately based on the type or class of object (e.g., a classification as a “sphere” or a “cube”) as well as instance labels (e.g., assigning a unique label to each of the objects, such as numerical labels “1,” “2,” “3,” “4,” and “5”). Computing object-level correspondences between the multi-viewpoint images further relates to computing consistent labels between the different viewpoints (for example, such that the same cube is labeled “1” from each of the viewpoint). Accordingly, the object pick estimator 100 generates collections of crops or patches of the multi-viewpoint images of the scene, where each collection of patches depicts the same instance from different viewpoints (cropped to the region containing the object and, in some cases, a small neighborhood or margin around the object).

In the case of a single image depicting a scene from a single viewpoint, in operation 330, the object pick estimator 100 may merely compute an instance segmentation mask, which similarly enables the generation of a crop or patch for each object instance detected in the image.

Systems and methods for computing object-level correspondences are described in International Patent Application No. PCT/US21/15926, titled “SYSTEMS AND METHODS FOR POSE DETECTION AND MEASUREMENT,” filed in the United States Patent and Trademark Office on Jan. 29, 2021, which is incorporated by reference herein in its entirety. For the sake of clarity, some techniques for computing object-level correspondences on images are described herein with reference to FIGS. 4A, 4B, and 4C.

In general terms, computing object-level correspondences reduces a search space for conducting image processing tasks such as, for example, pixel-level correspondence. In one embodiment, instance segmentation is performed to identify different instances of objects in images portraying a scene as viewed from different viewpoints, and instance segmentation maps/masks may be generated in response to the instance segmentation operation. The instance segmentation masks may then be employed for computing object level correspondences.

In one embodiment, object level correspondence allows the matching of a first instance of an object appearing in a first image that depicts a view of a scene from a first viewpoint, to a second instance of the same object appearing in a second image that depicts a view of a scene from a second viewpoint. Once object level correspondence is performed, the search space for performing, for example, pixel-level correspondence, may be limited to the regions of the image that correspond to the same object. Reducing the search space in this manner may result in faster processing of pixel-level correspondence and other similar tasks.

FIG. 4A is a flow diagram of a process for object level correspondence according to one embodiment. The process may be implemented by one or more processing circuits or electronic circuits that are components of the object pick estimator 100. It should be understood that the sequence of steps of the process is not fixed, but can be modified, changed in order, performed differently, performed sequentially, concurrently, or simultaneously, or altered into any desired sequence, as recognized by a person of skill in the art. The process described with respect to FIG. 4A may be used, in some embodiments of the present disclosure, to compute object level correspondences in operation 330 of FIG. 3B, but embodiments of the present disclosure are not limited thereto.

The process starts, and at block 400, the object pick estimator 100 receives multi-view images from the main and support cameras 10, 30. A first image captured by one of the cameras may depict one or more objects in a scene from a first viewpoint, and a second image captured by a second camera may depict the one or more objects in the scene from a second viewpoint different from the first viewpoint. The images captured by the cameras may be, for example, polarized images and/or images that have not undergone any polarization filtering.

At block 402 the object pick estimator 100 performs instance segmentation and mask generation based on the captured images. In this regard, the object pick estimator 100 classifies various regions (e.g. pixels) of an image captured by a particular camera 10, 30 as belonging to particular classes of objects. Each of the different instances of the objects in the image may also be identified, and unique labels be applied to each of the different instances of objects, such as by separately labeling each object in the image with a different identifier.

In one embodiment, instance segmentation masks delineating the various object instances are also be generated. Each instance segmentation mask may be a 2-D image having the same dimensions as the input image, where the value of each pixel may correspond to a label (e.g. a particular instance of the object depicted by the pixel). A different instance segmentation mask may be generated for different images depicting different viewpoints of the objects of interest. For example, a first segmentation mask may be generated to depict object instances in a first image captured by a first camera, and a second segmentation mask may be generated to depict object instances in a second image captured by a second camera. As convolutional neural network such as, for example, Mask R-CNN, may be employed for generating the instance segmentation masks.

At block 404, the object pick estimator 100 engages in identifying object-level correspondences of the objects identified in the instance segmentation masks. In this regard, the object pick estimator 100 may invoke a matching algorithm to identify a segmented instance of a particular object in one image as corresponding (or matching) a segmented instance of the same object in another image. The matching algorithm may be constrained to search for matching object instances along an epipolar line through an object instance in one image captured from a first viewpoint to find a corresponding object instance in a different image captured from a second viewpoint (where the epipolar line is defined in accordance with the extrinsic parameters of the cameras at the first and second viewpoints). In one embodiment, the matching algorithm compares different features of the regions corresponding to the segmented object instances to estimate the object correspondence. The matching of object instances from one image to another may narrow a search space for other image processing tasks such as, for example, evaluating a pickability of the object. The search space may be narrowed to the identified regions of the images that are identified as corresponding to the same object.

At block 406, the object pick estimator 100 generates an output based on the object-level correspondence. The output may be, for example, a pickability score of the object as depicted in the various images from the different viewpoints.

FIG. 4B is a block diagram of an architecture for instance segmentation and mask generation of step 402 according to one embodiment. Input images 410 captured by the various cameras 10, 30 are provided to a deep learning network 412 such as, for example, a CNN backbone. In the embodiments where the images include polarized images, the deep learning network may be implemented as a Polarized CNN backbone as described in PCT Patent Application No. PCT/US2020/048604, also filed as U.S. patent application Ser. No. 17/266,046, the content of which is incorporated herein by reference.

In one embodiment, the deep learning network 412 is configured to generate feature maps based on the input images 410, and employ a region proposal network (RPN) to propose regions of interest from the generated feature maps. The proposals by the CNN backbone may be provided to a box head 414 for performing classification and bounding box regression. In one embodiment, the classification outputs a class label 416 for each of the object instances in the input images 410, and the bounding box regression predicts bounding boxes 418 for the classified objects. In one embodiment, a different class label 416 is provided to each instance of an object.

The proposals by the CNN backbone may also be provided to a mask head 420 for generating instance segmentation masks. The mask head 416 may be implemented as a fully convolutional network (FCN). In one embodiment, the mask head 420 is configured to encode a binary mask for each of the object instances in the input images 410.

FIG. 4C is a more detailed flow diagram of a matching algorithm employed at step 404 (FIG. 4A) for identifying object-level correspondence for a particular object instance in a first segmentation mask according to one embodiment. The process may repeat for all object instances identified in the first segmentation mask. The sequence of steps of the process of FIG. 4C is not fixed, but can be modified, changed in order, performed differently, performed sequentially, concurrently, or simultaneously, or altered into any desired sequence, as recognized by a person of skill in the art.

At block 430, the matching algorithm identifies features of a first object instance in a first segmentation mask. The identified features for the first object instance may include a shape of the region of the object instance, a feature vector in the region, and/or keypoint predictions in the region. The shape of the region for the first object instance may be represented via a set of points sampled along the contours of the region. Where a feature vector in the region is used as the feature descriptor, the feature vector may be an average deep learning feature vector extracted via a convolutional neural network.

At block 432, the matching algorithm identifies an epipolar line through the first object instance in the first segmentation mask.

At block 434, the matching algorithm identifies one or more second object instances in a second segmentation mask that may correspond to the first object instance. A search for the second object instances may be constrained to the epipolar line between the first segmentation mask and the second segmentation mask that runs through the first object instance. In one embodiment, the matching algorithm searches approximately along the identified epiploar line to identify object instances in the second segmentation mask having a same class identifier as the first object instance. For example, if the first object instance belongs to a “dog” class, the matching algorithm evaluates object instances in the second segmentation mask that also belong to the “dog” class, and ignores objects that belong to a different class (e.g., a “cat” class).

At block 436, the matching algorithm identifies the features of the second object instances that belong the same class. As with the first object instance, the features of a particular second object instance may include a shape of the region of the second object instance, a feature vector representing the region, and/or keypoint predictions in the region.

At block 438, the matching algorithm compares the features of the first object instance to the features of second object instances for determining a match. In one embodiment, the matching algorithm identifies a fit between the features of the first object instance and features of the second object instances for selecting a best fit. In one embodiment, the best fit may be identified via a matching function such as the Hungarian matching function. In one embodiment, the features of the object instances are represented as probability distributions, and the matching function attempts to find a match of the probability distributions that minimizes a Kullback-Leibler (KL) divergence.

At block 440, a determination is made as to whether a match has been found. If the answer is YES, an output is generated at block 442. The output may include, for example, information (e.g. object ID) of the second object instance that matched the first object instance.

If the answer is NO, an output may be generate indicating a match failure at block 444.

Accordingly, object level correspondences can be computed from the multi-viewpoint images. These object level correspondences may be used to extract corresponding crops or patches from the multi-viewpoint images, where each of these crops or patches depicts a single instance of an object, and collections of corresponding crops or patches depict the same instance of an object from multiple viewpoints.

Polarized Convolutional Neural Networks

As discussed above, in some embodiments of the present disclosure, the captured images 40 may include polarization raw frames among the images 40 captured by polarization cameras of the imaging system, and the instance segmentation module 110 may include a polarized convolutional neural network (CNN) backbone to segment input images to detect instances of objects including optically challenging objects that might otherwise be difficult to detect in standard intensity images.

FIG. 5A is a block diagram of a feature extractor 500 according to one embodiment of the present invention. FIG. 5B is a flowchart depicting a method according to one embodiment of the present invention for extracting features from polarization raw frames. In the embodiment shown in FIG. 5A, the feature extractor 500 includes an intensity extractor 520 configured to extract an intensity image I 52 in an intensity representation space (e.g., in accordance with equation (7), as one example of a non-polarization representation space) and polarization feature extractors 530 configured to extract features in one or more polarization representation spaces. As shown in FIG. 5B, the extraction of polarization images in operation 410 may include extracting, in operation 411, a first tensor in a first polarization representation space from the polarization raw frames 48 (among the captured images 40) from a first Stokes vector. In operation 412, the feature extractor 500 further extracts a second tensor in a second polarization representation space from the polarization raw frames 48 (among the captured images 40) from a second Stokes vector. For example, the polarization feature extractors 530 may include a DOLP extractor 540 configured to extract a DOLP ρ image 54 (e.g., a first polarization image or a first tensor in accordance with equation (8) with DOLP as the first polarization representation space) and an AOLP extractor 560 configured to extract an AOLP ϕ image 56 (e.g., a second polarization image or a second tensor in accordance with equation (9), with AOLP as the second polarization representation space) from the supplied polarization raw frames 48. As another example, the polarization representation spaces may include combinations of polarization raw frames in accordance with Stokes vectors. As further examples, the polarization representations may include modifications or transformations of polarization raw frames in accordance with one or more image processing filters (e.g., a filter to increase image contrast or a denoising filter). The derived feature maps 52, 54, and 56 may then be supplied to a predictor 900 for further processing, such as performing inferences (e.g., generating instance segmentation masks, classifying the images, and generating textual descriptions of the images).

While FIG. 5B illustrates a case where two different tensors are extracted from the polarization raw frames 48 in two different representation spaces, embodiments of the present disclosure are not limited thereto. For example, in some embodiments of the present disclosure, exactly one tensor in a polarization representation space is extracted from the polarization raw frames 48. For example, one polarization representation space of raw frames is AOLP and another is DOLP (e.g., in some applications, AOLP may be sufficient for detecting transparent objects or other optically challenging objects such as translucent, non-Lambertian, multipath inducing, and/or non-reflective objects). In some embodiments of the present disclosure, more than two different tensors are extracted from the polarization raw frames 48 based on corresponding Stokes vectors. For example, as shown in FIG. 5B, n different tensors in n different representation spaces may be extracted by the feature extractor 500, where the n-th tensor is extracted in operation 414.

Accordingly, extracting features such as polarization feature maps or polarization images from polarization raw frames 48 produces first tensors 50 from which transparent objects or other optically challenging objects such as translucent objects, multipath inducing objects, non-Lambertian objects, and non-reflective objects are more easily detected or separated from other objects in a scene. In some embodiments, the first tensors extracted by the feature extractor 500 may be explicitly derived features (e.g., hand crafted by a human designer) that relate to underlying physical phenomena that may be exhibited in the polarization raw frames (e.g., the calculation of AOLP and DOLP images, as discussed above). In some additional embodiments of the present disclosure, the feature extractor 500 extracts other non-polarization feature maps or non-polarization images, such as intensity maps for different colors of light (e.g., red, green, and blue light) and transformations of the intensity maps (e.g., applying image processing filters to the intensity maps). In some embodiments of the present disclosure the feature extractor 500 may be configured to extract one or more features that are automatically learned (e.g., features that are not manually specified by a human) through an end-to-end supervised training process based on labeled training data.

Computing Predictions Such as Segmentation Masks Based on Polarization Features Computed from Polarization Raw Frames

As noted above, some aspects of embodiments of the present disclosure relate to providing first tensors in polarization representation space such as polarization images or polarization feature maps, such as the DOLP ρ and AOLP ϕ images extracted by the feature extractor 500, to a predictor such as a semantic segmentation algorithm to perform multi-modal fusion of the polarization images to generate learned features (or second tensors) and to compute predictions such as instance segmentation masks based on the learned features or second tensors. Specific embodiments relating to semantic segmentation or instance segmentation will be described in more detail below.

Generally, there are many approaches to semantic segmentation, including deep instance techniques. The various the deep instance techniques bay be classified as semantic segmentation-based techniques (such as those described in: Min Bai and Raquel Urtasun. Deep watershed transform for instance segmentation. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pages 5221-5229, 2017; Alexander Kirillov, Evgeny Levinkov, Bjoern Andres, Bogdan Savchynskyy, and Carsten Rother. Instancecut: from edges to instances with multicut. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pages 5008-5017, 2017; and Anurag Arnab and Philip HS Torr. Pixelwise instance segmentation with a dynamically instantiated network. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pages 441-450, 2017.), proposal-based techniques (such as those described in: Kaiming He, Georgia Gkioxari, Piotr Doll'ar, and Ross Girshick. Mask r-cnn. In Proceedings of the IEEE International Conference on Computer Vision, pages 2961-2969, 2017.) and recurrent neural network (RNN) based techniques (such as those described in: Bernardino Romera-Paredes and Philip Hilaire Sean Torr. Recurrent instance segmentation. In European Conference on Computer Vision, pages 312-329. Springer, 2016 and Mengye Ren and Richard S Zemel. End-to-end instance segmentation with recurrent attention. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pages 6656-6664, 2017.). Embodiments of the present disclosure may be applied to any of these semantic segmentation techniques.

While some comparative approaches supply concatenated polarization raw frames (e.g., images I₀, I₄₅, I₉₀, and I₁₃₅ as described above) directly into a deep network without extracting first tensors such as polarization images or polarization feature maps therefrom, models trained directly on these polarization raw frames as inputs generally struggle to learn the physical priors, which leads to poor performance, such as failing to detect instances of transparent objects or other optically challenging objects. Accordingly, aspects of embodiments of the present disclosure relate to the use of polarization images or polarization feature maps (in some embodiments in combination with other feature maps such as intensity feature maps) to perform instance segmentation on images of transparent objects in a scene.

One embodiment of the present disclosure using deep instance segmentation is based on a modification of a Mask Region-based Convolutional Neural Network (Mask R-CNN) architecture to form a Polarized Mask R-CNN architecture. Mask R-CNN works by taking an input image x, which is an H×W×3 tensor of image intensity values (e.g., height by width by color intensity in red, green, and blue channels), and running it through a backbone network: C=B(x). The backbone network B(x) is responsible for extracting useful learned features from the input image and can be any standard CNN architecture such as AlexNet (see, e.g., Krizhevsky, Alex, Ilya Sutskever, and Geoffrey E. Hinton. “ImageNet classification with deep convolutional neural networks.” Advances in neural information processing systems. 2012.), VGG (see, e.g., Simonyan, Karen, and Andrew Zisserman. “Very deep convolutional networks for large-scale image recognition.” arXiv preprint arXiv:1409.1556 (2014).), ResNet-101 (see, e.g., Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Deep residual learning for image recognition. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pages 770-778, 2016.), MobileNet (see, e.g., Howard, Andrew G., et al. “Mobilenets: Efficient convolutional neural networks for mobile vision applications.” arXiv preprint arXiv:1704.04861 (2017).), MobileNetV2 (see, e.g., Sandler, Mark, et al. “MobileNetV2: Inverted residuals and linear bottlenecks.” Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. 2018.), and MobileNetV3 (see, e.g., Howard, Andrew, et al. “Searching for MobileNetV3.” Proceedings of the IEEE International Conference on Computer Vision. 2019.)

The backbone network B (x) outputs a set of tensors, e.g., C={C₁, C₂, C₃, C₄, C₅}, where each tensor C_i represents a different resolution feature map. These feature maps are then combined in a feature pyramid network (FPN) (see, e.g., Tsung-Yi Lin, Piotr Doll'ar, Ross Girshick, Kaiming He, Bharath Hariharan, and Serge Belongie. Feature pyramid networks for object detection. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pages 2117-2125, 2017.), processed with a region proposal network (RPN) (see, e.g., Shaoqing Ren, Kaiming He, Ross Girshick, and Jian Sun. Faster r-cnn: Towards real-time object detection with region proposal networks. In Advances in Neural Information Processing Systems, pages 91-99, 2015.), and finally passed through an output subnetwork (see, e.g., Ren et al. and He et al., above) to produce classes, bounding boxes, and pixel-wise segmentations. These are merged with non-maximum suppression for instance segmentation.

Aspects of embodiments of the present invention relate to a framework for leveraging the additional information contained in polarized images using deep learning, where this additional information is not present in input images captured by comparative cameras (e.g., information not captured standard color or monochrome cameras without the use of polarizers or polarizing filters). Neural network architectures constructed in accordance with frameworks of embodiments of the present disclosure will be referred to herein as Polarized Convolutional Neural Networks (CNNs).

Applying this framework according to some embodiments of the present disclosure involves three changes to a CNN architecture:

(1) Input Image: Applying the physical equations of polarization to create the input polarization images to the CNN, such as by using a feature extractor 500 according to some embodiments of the present disclosure.

(2) Attention-fusion Polar Backbone: Treating the problem as a multi-modal fusion problem by fusing the learned features computed from the polarization images by a trained CNN backbone.

(3) Geometric Data Augmentations: augmenting the training data to represent the physics of polarization.

However, embodiments of the present disclosure are not limited thereto. Instead, any subset of the above three changes and/or changes other than the above three changes may be made to an existing CNN architecture to create a Polarized CNN architecture within embodiments of the present disclosure.

A Polarized CNN according to some embodiments of the present disclosure may be implemented using one or more electronic circuits configured to perform the operations described in more detail below. In the embodiment shown in FIG. 3A, a Polarized CNN is used as a component of the instance segmentation module 110 for computing an instance segmentation mask.

FIG. 6 is a block diagram depicting a Polarized CNN architecture according to one embodiment of the present invention as applied to a Mask-Region-based convolutional neural network (Mask R-CNN) backbone, where second tensors C (or output tensors such as learned feature maps) are used to compute an output prediction such as an instance segmentation mask.

While some embodiments of the present disclosure relate to a semantic segmentation or instance segmentation using a Polarized CNN architecture as applied to a Mask R-CNN backbone, embodiments of the present disclosure are not limited thereto, and other backbones such as AlexNet, VGG, MobileNet, MobileNetV2, MobileNetV3, and the like may be modified in a similar manner.

In the embodiment shown in FIG. 6, derived feature maps 50 (e.g., including input polarization images such as AOLP ϕ and DOLP ρ images) are supplied as inputs to a Polarized CNN backbone 610. In the embodiment shown in FIG. 6, the input feature maps 50 include three input images: the intensity image (I) 52, the AOLP (ϕ) 56, the DOLP (ρ) 54 from equation (1) as the input for detecting a transparent object and/or other optically challenging object. These images are computed from polarization raw frames 48 (e.g., images I₀, I₄₅, I₉₀, and I₁₃₅ as described above), normalized to be in a range (e.g., 8-bit values in the range [0-255]) and transformed into three-channel gray scale images to allow for easy transfer learning based on networks pre-trained on the MSCoCo dataset (see, e.g., Tsung-Yi Lin, Michael Maire, Serge Belongie, James Hays, Pietro Perona, Deva Ramanan, Piotr Doll'ar, and C Lawrence Zitnick. Microsoft coco: Common objects in context. In European Conference on Computer Vision, pages 740-755. Springer, 2014.).

In the embodiment shown in FIG. 6, each of the input derived feature maps 50 is supplied to a separate backbone: intensity B_I(I) 612, AOLP backbone B_ϕ(ϕ) 614, and DOLP backbone B_ρ(ρ) 616. The CNN backbones 612, 614, and 616 compute tensors for each mode, or “mode tensors” (e.g., feature maps computed based on parameters learned during training or transfer learning of the CNN backbone, discussed in more detail below) C_i,I, C_i,ρ, C_i,ϕ at different scales or resolutions i. While FIG. 6 illustrates an embodiment with five different scales i, embodiments of the present disclosure are not limited thereto and may also be applied to CNN backbones with different numbers of scales.

Some aspects of embodiments of the present disclosure relate to a spatially-aware attention-fusion mechanism to perform multi-modal fusion (e.g., fusion of the feature maps computed from each of the different modes or different types of input feature maps, such as the intensity feature map I, the AOLP feature map ϕ, and the DOLP feature map ρ).

For example, in the embodiment shown in FIG. 6, the mode tensors C_i,I, C_i,ρ, C_i,ϕ (tensors for each mode) computed from corresponding backbones B_I, B_ρ, B_ϕ at each scale i are fused using fusion layers 622, 623, 624, 625 (collectively, fusion layers 620) for corresponding scales. For example, fusion layer 622 is configured to fuse mode tensors C_2,I, C_2,ρ, C_2,ϕ computed at scale i=2 to compute a fused tensor C₂. Likewise, fusion layer 623 is configured to fuse mode tensors C_3,I, C_3,ρ, C_3,ϕ computed at scale i=3 to compute a fused tensor C₃, and similar computations may be performed by fusion layers 624 and 625 to compute fused feature maps C₄ and C₅, respectively, based on respective mode tensors for their scales. The fused tensors C_i (e.g., C₂, C₃, C₄, C₅), or second tensors, such as fused feature maps, computed by the fusion layers 620 are then supplied as input to a prediction module 650, which is configured to compute a prediction from the fused tensors, where the prediction may be an output such as an instance segmentation mask, a classification, a textual description, or the like.

FIG. 7 is a block diagram of an i-th fusion layer among the fusion layers 620 that may be used with a Polarized CNN according to one embodiment of the present invention. As shown in FIG. 7, in some embodiments of the present disclosure, a fusion layer (e.g., each of the fusion layer 620) is implemented using an attention module, in which the predictor 600 concatenates the supplied input tensors or input feature maps C_i,I, C_i,ρ, C_i,ϕ computed by the CNN backbones for the i-th scale and to generate concatenated tensor 710, where the concatenated tensor 710 is processed through a set of trained convolutional layers or attention subnetwork Ω_i for the i-th. scale. The attention subnetwork outputs a 3-channel image with the same height and width as the input tensors, and, in some embodiments, a softmax function is applied to each pixel of the 3-channel image to compute pixel-wise attention weights a for the i-th scale:[α_i,ϕ,α_i,ρ,α_i,I]=softmax(Ω_i([C_i,ϕ,C_i,ρ,C_i,I]))  (15)

These attention weights are used to perform a weighted average 720 per channel:C_i=α_i,ϕC_i,ϕ+α_i,ρC_i,ρ+α_i,IC_i,I  (16)

Accordingly, using an attention module allows a Polarized CNN according to some embodiments of the present disclosure to weight the different inputs at the scale i (e.g., the intensity I tensor or learned feature map C_i,I, the DOLP tensor or learned feature map C_i,ρ, and the AOLP tensor or learned feature map C_i,ϕ at scale i) based on how relevant they are to a given portion of the scene, where the relevance is determined by the trained attention subnetwork Ω_i in accordance with the labeled training data used to train the Polarized CNN backbone.

Accordingly, aspects of embodiments of the present disclosure relate to systems and methods for computing object picking plans for an object in a scene without explicitly computing of a six-degree of freedom pose of the object and/or other objects in the scene. In some circumstances, it is infeasible for a robotic vision system to estimate the pose of an object within the accuracy required by a robotic controller to perform an object pick. For example, pose estimation may be made difficult or impossible due to the deformation of the object, a small object size, and environmental constraints.

In addition, the computation of instance segmentation masks and depth maps in various embodiments of the present disclosure is relatively simple, faster, and less prone to error than some comparative pose estimation pipelines which perform pose refinement and other steps in addition to computing instance segmentation masks and depth maps. These additional steps performed by pose estimation pipelines may increase the processing time and costs, which may also make pose estimation infeasible in circumstances where short cycle times may make it difficult to compute an accurate pose within the time allotted. In circumstances where a pose estimation pipeline trades away accuracy in favor of speed, errors in the estimated object pose can drastically change the pickability analysis of the objects and may cause a robot to attempt to pick a more difficult object or worse make the robot attempt an entirely incorrect approach when trying to pick up the object.

In contrast, embodiments of the present disclosure are able to compute object picking plans based solely on instance segmentation masks and, in some embodiments instance segmentation masks in combination with depth maps. As such, the processing pipeline is shorter than pose estimation pipelines, and the avoided the extra computation time for computing the poses of objects in the scene results in lower latency computations.

While the present invention has been described in connection with certain exemplary embodiments, it is to be understood that the invention is not limited to the disclosed embodiments, but, on the contrary, is intended to cover various modifications and equivalent arrangements included within the spirit and scope of the appended claims, and equivalents thereof.

Claims

1. A method for controlling a robotic system, the method comprising: capturing, by an imaging system, one or more images of a scene;computing, by a processing circuit comprising a processor and memory, one or more instance segmentation masks based on the one or more images, the one or more instance segmentation masks detecting one or more objects in the scene;computing one or more depth maps based on the one or more images of the scene;segmenting the one or more depth maps into one or more point clouds using the one or more instance segmentation masks, the one or more point clouds corresponding to the one or more objects in the scene;computing, by the processing circuit, one or more pickability scores for the one or more objects, wherein the computing the one or more pickability scores for the one or more objects comprises computing a pickability score for an object of the one or more objects by: computing a measure of clutter in a neighborhood of the object based on the one or more point clouds,computing a measure of distance of the object from a current location of the end effector of the robotic arm based on the one or more point clouds, andcomputing a measure of how much an object protrudes from other objects of the one or more objects based on the one or more point clouds;selecting, by the processing circuit, an object among the one or more objects based on the one or more pickability scores;computing, by the processing circuit, an object picking plan for the selected object; andoutputting, by the processing circuit, the object picking plan to a controller configured to control an end effector of a robotic arm to pick the selected object.

2. The method of claim 1, wherein the object picking plan comprises a picking position and a direction of approach, wherein the computing the object picking plan for the selected object comprises computing the picking position based on a visible center of mass of a point cloud corresponding to the selected object.

3. The method of claim 2, wherein the computing the picking position further comprises computing an estimated center of mass based on adjusting the visible center of mass by a scalar value to be farther from a viewpoint corresponding to a camera used to capture the one or more images of the scene.

4. The method of claim 1, wherein the computing the one or more pickability scores for the one or more objects comprises supplying the one or more depth maps and the one or more instance segmentation masks to an object pickability neural network.

5. The method of claim 1, wherein the computing the one or more pickability scores for the one or more objects comprises computing a pickability score for an object of the one or more objects by: computing a measure of clutter in a neighborhood of the object based on the one or more instance segmentation masks;computing a measure of distance of the object from a current location of the end effector of the robotic arm based on the one or more instance segmentation masks; andcomputing a measure of how much an object protrudes from other objects of the one or more objects based on the one or more instance segmentation masks.

6. The method of claim 5, wherein the computing the one or more pickability scores for the one or more objects comprising supplying the one or more instance segmentation masks to an object pickability neural network.

7. The method of claim 1, wherein the object picking plan comprises a picking position and a direction of approach.

8. The method of claim 1, wherein the one or more images comprise polarization one or more respective raw frames.

9. The method of claim 1, wherein the one or more instance segmentation masks are computed using a trained convolutional neural network.

10. The method of claim 1, further comprising computing instance segmentation masks from images captured from a plurality of different viewpoints.

11. The method of claim 10, wherein the pickability scores for the one or more objects are computed independently for different viewpoints, and wherein the selecting the object is based on the one or more pickability scores of the object aggregated across different viewpoints.

12. The method of claim 11, further comprising computing depth maps for the different viewpoints, and wherein the pickability scores are further computed based on the depth maps.

13. A system comprising: a processor; andmemory storing instructions that, when executed by the processor, cause the processor to: receive one or more images of a scene from an imaging system;compute one or more instance segmentation masks based on the one or more images, the one or more instance segmentation masks detecting one or more objects in the scene;compute one or more depth maps based on the one or more images of the scene;segment the one or more depth maps into one or more point clouds using the one or more instance segmentation masks, the one or more point clouds corresponding to the one or more objects in the scene;compute one or more pickability scores for the one or more objects, wherein the instructions to compute the one or more pickability scores for the one or more objects further comprise instructions that, when executed by the processor, cause the processor to compute a pickability score for an object of the one or more objects by:computing a measure of clutter in a neighborhood of the object based on the one or more point clouds,computing a measure of distance of the object from a current location of the end effector of the robotic arm based on the one or more point clouds, andcomputing a measure of how much an object protrudes from other objects of the one or more objects based on the one or more point clouds;select an object among the one or more objects based on the one or more pickability scores;compute an object picking plan for the selected object; andoutput the object picking plan to a controller configured to control an end effector of a robotic arm to pick the selected object.

14. The system of claim 13, wherein the object picking plan comprises a picking position and a direction of approach, wherein the instructions to compute the object picking plan for the selected object further comprise instructions that, when executed by the processor, cause the processor to compute the picking position based on a visible center of mass of a point cloud corresponding to the selected object.

15. The system of claim 14, wherein the instructions to compute the picking position further comprise instructions that, when executed by the processor, cause the processor to compute an estimated center of mass based on adjusting the visible center of mass by a scalar value to be farther from a viewpoint corresponding to a camera used to capture the one or more images of the scene.

16. The system of claim 13, wherein the instructions to compute the one or more pickability scores for the one or more objects further comprise instructions that, when executed by the processor, cause the processor to supply the one or more depth maps and the one or more instance segmentation masks to an object pickability neural network.

17. The system of claim 13, wherein the instructions to compute the one or more pickability scores for the one or more objects comprise instructions that, when executed by the processor, cause the processor to compute a pickability score for an object of the one or more objects by: computing a measure of clutter in a neighborhood of the object based on the one or more instance segmentation masks;computing a measure of distance of the object from a current location of the end effector of the robotic arm based on the one or more instance segmentation masks; andcomputing a measure of how much an object protrudes from other objects of the one or more objects based on the one or more instance segmentation masks.

18. The system of claim 17, wherein the instructions to compute the one or more pickability scores for the one or more objects further comprise instructions that, when executed by the processor, cause the processor to supply the one or more instance segmentation masks to an object pickability neural network.

19. The system of claim 13, wherein the object picking plan comprises a picking position and a direction of approach.

20. The system of claim 13, wherein the one or more images comprise one or more respective polarization raw frames.

21. The system of claim 13, wherein the one or more instance segmentation masks are computed using a trained convolutional neural network.

22. The system of claim 13, wherein the memory further stores instructions that, when executed by the processor, cause the processor to compute instance segmentation masks from images captured from a plurality of different viewpoints.

23. The system of claim 22, wherein the pickability scores for the one or more objects are computed independently for different viewpoints, and wherein the memory further stores instructions that, when executed by the processor, cause the processor to select the object based on the one or more pickability scores of the object aggregated across different viewpoints.

24. The system of claim 22, wherein the memory further stores instructions that, when executed by the processor, cause the processor to compute depth maps for the different viewpoints, and wherein the pickability scores are further computed based on the depth maps.

25. A method for controlling a robotic system, the method comprising: capturing, by an imaging system, one or more images of a scene;computing, by a processing circuit comprising a processor and memory, one or more instance segmentation masks based on the one or more images, the one or more instance segmentation masks detecting one or more objects in the scene;computing one or more depth maps based on the one or more images of the scene;segmenting the one or more depth maps into one or more point clouds using the one or more instance segmentation masks, the one or more point clouds corresponding to the one or more objects in the scene;computing, by the processing circuit, one or more pickability scores for the one or more objects;selecting, by the processing circuit, an object among the one or more objects based on the one or more pickability scores;computing, by the processing circuit, an object picking plan for the selected object, wherein the object picking plan comprises a picking position and a direction of approach,wherein the computing the object picking plan for the selected object comprises computing the picking position based on a visible center of mass of a point cloud corresponding to the selected object; andoutputting, by the processing circuit, the object picking plan to a controller configured to control an end effector of a robotic arm to pick the selected object.

26. A method for controlling a robotic system, the method comprising: capturing, by an imaging system, one or more images of a scene;computing, by a processing circuit comprising a processor and memory, one or more instance segmentation masks based on the one or more images, the one or more instance segmentation masks detecting one or more objects in the scene;computing, by the processing circuit, one or more pickability scores for the one or more objects, wherein the computing the one or more pickability scores for the one or more objects comprises computing a pickability score for an object of the one or more objects by: computing a measure of clutter in a neighborhood of the object based on the one or more instance segmentation masks,computing a measure of distance of the object from a current location of the end effector of the robotic arm based on the one or more instance segmentation masks, andcomputing a measure of how much an object protrudes from other objects of the one or more objects based on the one or more instance segmentation masks;selecting, by the processing circuit, an object among the one or more objects based on the one or more pickability scores;computing, by the processing circuit, an object picking plan for the selected object; andoutputting, by the processing circuit, the object picking plan to a controller configured to control an end effector of a robotic arm to pick the selected object.

27. A method for controlling a robotic system, the method comprising: capturing, by an imaging system, one or more images of a scene from a plurality of different viewpoints;computing, by a processing circuit comprising a processor and memory, one or more instance segmentation masks based on the one or more images, the one or more instance segmentation masks detecting one or more objects in the scene;computing, by the processing circuit, one or more pickability scores for the one or more objects, wherein the pickability scores for the one or more objects are computed independently for different viewpoints;selecting, by the processing circuit, an object among the one or more objects based on the one or more pickability scores of the object aggregated across different viewpoints;computing, by the processing circuit, an object picking plan for the selected object; andoutputting, by the processing circuit, the object picking plan to a controller configured to control an end effector of a robotic arm to pick the selected object.

28. A system comprising: a processor; andmemory storing instructions that, when executed by the processor, cause the processor to: receive one or more images of a scene from an imaging system;compute one or more instance segmentation masks based on the one or more images, the one or more instance segmentation masks detecting one or more objects in the scene;compute one or more depth maps based on the one or more images of the scene; segment the one or more depth maps into one or more point clouds using the one or more instance segmentation masks, the one or more point clouds corresponding to the one or more objects in the scene;compute one or more pickability scores for the one or more objects;select an object among the one or more objects based on the one or more pickability scores;compute an object picking plan for the selected object, wherein the object picking plan comprises a picking position and a direction of approach,wherein the instructions to compute the object picking plan for the selected object further comprise instructions that, when executed by the processor, cause the processor to compute the picking position based on a visible center of mass of a point cloud corresponding to the selected object; andoutput the object picking plan to a controller configured to control an end effector of a robotic arm to pick the selected object.

29. A system comprising: a processor; andmemory storing instructions that, when executed by the processor, cause the processor to: receive one or more images of a scene from an imaging system;compute one or more instance segmentation masks based on the one or more images, the one or more instance segmentation masks detecting one or more objects in the scene;compute one or more pickability scores for the one or more objects, wherein the instructions to compute the one or more pickability scores for the one or more objects comprise instructions that, when executed by the processor, cause the processor to compute a pickability score for an object of the one or more objects by: computing a measure of clutter in a neighborhood of the object based on the one or more instance segmentation masks,computing a measure of distance of the object from a current location of the end effector of the robotic arm based on the one or more instance segmentation masks, andcomputing a measure of how much an object protrudes from other objects of the one or more objects based on the one or more instance segmentation masks;select an object among the one or more objects based on the one or more pickability scores;compute an object picking plan for the selected object; andoutput the object picking plan to a controller configured to control an end effector of a robotic arm to pick the selected object.

30. A system comprising: a processor; andmemory storing instructions that, when executed by the processor, cause the processor to: receive one or more images of a scene from a plurality of different viewpoints, from an imaging system;compute one or more instance segmentation masks based on the one or more images, the one or more instance segmentation masks detecting one or more objects in the scene;compute one or more pickability scores for the one or more objects, wherein the pickability scores for the one or more objects are computed independently for different viewpoints;select an object among the one or more objects based on the one or more pickability scores of the object aggregated across different viewpoints;compute an object picking plan for the selected object; andoutput the object picking plan to a controller configured to control an end effector of a robotic arm to pick the selected object.