Patent Application
20050169506
1. Field of the Invention
The present invention relates to a system, apparatus, and method for the collection of friction ridge signatures from a subject. More particularly, the present invention relates to a low power consumption, small size, and low weight friction ridge capturing device and method for the collection of friction ridge signatures from a subject. Most particularly, the present invention relates to a low power consumption, compact, and portable digital friction ridge capturing apparatus, system and method for the collection of friction ridge signatures with a platen area of at least 3.0 square inches.
2. Description of the Related Art
Human beings have friction ridges on their hands and feet. Friction ridge impressions from a subject's fingers are commonly known as fingerprints. Animals also commonly have unique friction patterns on their footpads. In dogs and cats, for example, these patterns are called paw prints.
Digital scanning systems that capture friction ridge impressions, collectively termed herein as ‘fingerprints’, are old in the art. Many of these systems were designed to capture a smaller area of one or two fingerprints while others were designed to capture a much larger area. Such existing systems commonly use optical imaging, capacitance, infrared radiation, ultrasound, or other means to capture fingerprints.
Many optical imaging systems are designed to capture one or two fingerprints at a time. These systems are termed ‘single print devices’ herein. For example, the manufacturers listed in Table 1 provide optical devices that scan one or two fingerprints at a time.
Devices that capture a single fingerprint at a time are compact and draw minimal power during operation. One of the first issued patents that discloses how such a single digit device works is U.S. Pat. No. 3,200,701 to White, the entire contents of which is hereby incorporated by reference as if fully set forth herein. White teaches a device that uses a light source, a prismatic body, the principle of Total Internal Reflection “TIR”, and a scanning apparatus to capture a fingerprint image. A typical TIR scanning system that captures fingerprint images comprises a light source, a prism, a camera, and a host computer or other control device that is used to capture the image. It should be understood that the camera is any type of suitable sensor for the capture of image data. This includes Charge Coupled Devices (CCD) and Complimentary Metal Oxide Semiconductor (CMOS) cameras as well as sensor chips included in these cameras and both linear and area scan versions. The host computer or other control device is referred to as a host computer. There may be other components of a prior art system such as, e.g., polarizing filters, corrective optics, and holographic gratings.
Most commercially available large format optical systems today follow this single digit system configuration. That is, they use a light source, prismatic body, TIR, camera(s), and host computer to create fingerprint images. For prior art devices, “capable of capturing more than two fingerprints simultaneously” means optical devices having a surface capture area exceeding 3.0 square inches. This type of system is referred to as a large format fingerprint capture system. In addition, large format fingerprint capture systems include those that capture palm prints and writer's edge.
Large format fingerprint devices typically capture fingerprints from multiple fingers simultaneously and therefore, the area upon which the subjects place their fingers must be large enough to accommodate the maximum number of fingers to be captured simultaneously. Usually, this number is four, but Cross Match Technologies provides a system that is able to capture the fingerprints in two groups of two fingerprints apiece. In effect, this Cross Match Technologies system captures four fingerprints simultaneously.
Livescan systems, one common form of large format fingerprint system, typically use a glass or plastic surface, termed a platen, upon which the subject's fingers are rolled or pressed. Images of the fingers' ridges are typically captured from underneath the platen by one or multiple cameras and are then converted into digital files. Images of rolled fingers are called rolled prints and images of pressed fingers are called slaps. Livescan devices are available from the sources listed in TABLE 2.
A large body of patents exists for large format fingerprint scanning devices. U.S. Pat. No. 3,200,701 to White, discussed above, discloses one such system. Further, U.S. Pat. No. 4,933,976 to Fishbine et al., the entire contents of which are hereby incorporated by reference as if fully set forth herein, teaches a configuration comprising a prismatic-based TIR device for fingerprint capture. Fishbine et al. disclose a method that combines successively captured images into one image array.
U.S. Pat. Nos. 5,548,394, 5,629,764, 5,650,842, 6,178,255, and 6,407,804 all disclose variations of the TIR based prismatic platen device used to capture fingerprint images.
In U.S. Pat. No. 5,548,394 to Giles et al., the entire contents of which are hereby incorporated by reference as if fully set forth herein, teaches a TIR based system that uses a linear CCD camera (as opposed to an area based camera) and associated optics that is used to create rolled fingerprints. Prints are captured by changing the orientation of a mirror as a scan progresses.
U.S. Pat. No. 5,629,764 to Bahuguna et al., the entire contents of which are hereby incorporated by reference as if fully set forth herein, teaches correcting the aspect ratio of images using holographic gratings. Light incident on the fingerprint surface is totally internally reflected at the surface of the prism but immediately after being reflected back toward the prism the light enters a holographic grating that changes the light direction. Light, directed now at the hypotenuse of the prism at an angle greater than the critical angle, is total internally reflected toward the camera. In this patent, the aspect ratio is corrected before the light leaves the prism.
U.S. Pat. No. 5,650,842 to Maase et al., the entire contents of which are hereby incorporated by reference as if fully set forth herein, discloses a fingerprint capture system that optically corrects the aspect ratio of images after the image beam leaves the prism. Correction is achieved by including optical elements in the image path after the image exits the prism. The image may be generated using TIR or light dispersion from the friction ridges. A reference light source is an edge lit light panel manufactured by Hewlett-Packard that is illuminated by red Light Emitting Diodes (LEDs).
U.S. Pat. No. 6,178,255 Scott et al., the entire contents of which are hereby incorporated by reference as if fully set forth herein, discloses a method and apparatus in which a mechanism slides a prism over an imaged area of a camera. By using a linear encoding mechanism, the method and apparatus splices together complete fingerprint images from the smaller portions of the fingerprint that are captured when sliding the prism over the camera.
U.S. Pat. No. 6,407,804 to Hillman et al., the entire contents of which are hereby incorporated by reference as if fully set forth herein, discloses a fingerprint capture system with embedded reference targets in the optical path. Using these reference targets, the system can be calibrated at anytime using these internal reference targets. This patent also discloses use of a planar light source constructed from a two dimensional array of LEDs followed by a diffusor.
Several patented systems do not teach that TIR be used at the fingerprint imaging surface in order for a fingerprint to be captured. Such systems are disclosed in U.S. Pat. Nos. 5,621,516, 5,650,842 and 6,061,463.
U.S. Pat. No. 5,621,516 to Shinzaki et al., the entire contents of which are hereby incorporated by reference as if fully set forth herein, discloses a system that images randomly reflected light from the friction ridges. By providing means of minimizing the light content reflected by platen areas not in contact with the friction edges, the contrast of the resulting images are improved. U.S. Pat. No. 5,650,842 to Maase et al., discussed above, provides another mechanism for capture of friction ridges via light dispersion.
U.S. Pat. No. 6,061,463 to Metz et al., the entire contents of which are hereby incorporated by reference as if fully set forth herein, teaches using a slanted-fringed light diffractive grating to redirect light perpendicularly toward the platen surface. Light reflected from the surface is then captured by the camera. Light incident upon the friction ridges is dispersed and therefore this less intense reflection is imaged as dark. One disclosed embodiment uses volume holograms to redirect light. This patent contrasts with U.S. Pat. No. 5,629,764 to Bahuguna et al. in that this patent redirects the light before it reaches the platen surface and therefore image aspect correction is never required.
Based on the foregoing discussion, most, if not all, existing commercial large format fingerprint devices use the principle of TIR. In the majority of these devices, the object plane to be imaged (the fingerprint) and the image plane of the camera are not parallel, centered, and perpendicular to a common axis. To correct for optical perspective distortions introduced by the relative positions of the object and image planes optics within the device must correct the positions of the object and/or image planes before the camera captures an image or the system must employ an algorithmic solution for correcting the perspective distortion introduced into the image. In the first case, if additional optical components are added, the size and weight of the device increase. For example, see U.S. Pat. Nos. 5,650,842 and 6,407,804. In the later cases, to avoid an unfocused image the depth of field must be deep enough for the entire object area to be in focus. Since the image is not optically corrected for perspective distortion, the depth of field requirement is driven by the three dimensional geometry of the platen, the optics used between the platen surface and the camera, and the geometric relationship between the camera and the platen surface. Typically, lenses that allow larger depths of field have focal lengths such as 30 mm or greater with corresponding f-stops often greater than 4.0. Long focal length lenses often result in a distance from the object plane to the image plane that is too large to put the entire device into a physically compact solution. In addition, high f-stop lenses restrict the amount of light entering a camera and therefore more light, power, and/or exposure are needed for these systems, thus implying a larger power consumption.
Often aberrations in the periphery of the image such as barrel distortion and pincushion distortion distort the fingerprint to the point where the images are not truly usable. Barrel distortion occurs when the distance of the pixels along the image edges are farther away from the center of the image than the corresponding actual distance in the target. Pincushion distortion occurs when the pixels along the images edges are closer to the center of the image that the corresponding actual distance in the target. Barrel and pincushion distortions are introduced via lenses used in an optical system and both distortion types may be present at the same time.
The majority of existing optical fingerprint systems rely on LEDs for light sources. Light illuminating the platen may be either diffuse or collimated.
Electricity consumers in a fingerprint device include the light source(s), the camera(s), frame grabber electronics, magnetic stripe readers, barcode readers, radio frequency identification modules, proximity card readers, smartcard readers, displays, platen heaters, platen blowers, and servomotors, if present. In total, the power used by such systems is above 10 watts for all existing large format fingerprint systems. Therefore, prior art large format fingerprint devices are powered by external power sources connected via separate cabling since power provided over a single data/control/power cable will either be insufficient or the battery on the attached computer will be drained too quickly. In other words, prior art systems cannot be powered only from the computer to which the device is attached.
Most of the patented systems described above are either not compact or not lightweight and therefore they cannot be considered as portable. To be moved, these devices often require a protective case. In existing instances, the device and case weigh over 30 pounds. In addition, many devices must be re-calibrated once the device has been moved and reinstalled. Such re-calibration is required due to the presence of moving parts or the possibility that parts have moved relative to one another.
Such systems also commonly address the issue of condensation on the platen. Such condensation occurs when the dew point around the platen/finger is too high relative to the ambient temperature of the prism and therefore moisture from the finger condenses on the prism. In such cases, platen heaters and platen blowers have been used to minimize the condensation effects to the image.
No prior art large format fingerprint scanning device has the ability to pass all data, power, and control logic over a single physical connection to the device. In addition, image-processing means to identify the start and stop points of a fingerprint image capture session also do not currently exist. Rather, external controls such as foot pedals, touch screens, keyboard keys, and buttons located on the device are commonplace as means to identify start and stop points for capturing fingerprint images.
Thus, there is a need for a fingerprint capture device that combines the features of small size, low weight, low power consumption, and incorporates non-external fingerprint start-stop scanning control. The system, apparatus, and method of the present invention preferably provides these features by:
The present invention overcomes the deficiencies of prior art large format fingerprint devices by providing a fingerprint capture apparatus, system and method that is low power, compact, and lightweight and has a platen area greater than about 3.0 square inches, e.g., from about 3 to about 24 square inches. Further, the present invention is typically powered, controlled, and exchanges data over a single data/control/power connection to a host PC. In an alternative preferred embodiment the large format fingerprint device is directly connected to a completely disconnected (not plugged in to a wall power outlet or other external power source) portable PC, such as a laptop having a battery power source. In another preferred embodiment a wireless interface (e.g. infrared, 802.11b, Bluetooth, etc.) to the device exchanges data and accepts control functions from an attached computer processor while the device runs on a completely self-contained power subsystem such as a battery. In such an embodiment the device may have no physical connection ports. If desired, the device exchanges data and accepts control functions via internet and/or satellite connectivity to a computer processor.
In some embodiments the device may have an internal computational capability so it can direct fingerprint capture and fingerprint matching within the device using fingerprint templates that are delivered over a wired or wireless network attached to the device. Typical applications of such devices would be access control and tracking applications. The matching templates and algorithm could be updated over the electronic interface.
The primary device components of the present invention combine to minimize required power, size and weight. The device of the present invention comprises a light source, a prism, a camera (including the lens), a housing, and a host computer. Optional elements comprise holographic elements such as gratings and holographic optical elements (HOEs), a battery subsystem, an image processing subsystem, optical filters, magnetic stripe reader, RFID module, proximity card reader, smartcard reader, barcode reader, a platen heater, a platen blower, and mirrors used to divert the image beam.
To achieve minimal size and weight the number of components is minimized in the system, apparatus, and method of the present invention. Technology for aspect ratio changes, image aberration corrections, and perspective corrections may be used to minimize the depth of field, as taught by U.S. Pat. Nos. 5,629,764 and 6,061,463 incorporated herein by reference. Alternatively, aspect ratio, image aberration, and perspective corrections can be made algorithmically via hardware, firmware and software as provided for in the present invention. Ojanen, see Appendix A, teaches one way of removing image aberrations and perspective distortions via such algorithms. Also, technology may be used to maximize the amount of light generated per watt, as taught, by U.S. Pat. No. 5,359,691 to Tai et al., U.S. Pat. No. 5,390,276 to Tai et al., and U.S. Pat. No. 5,854,872 to Tai, which are hereby incorporated by reference as if fully set forth herein.
As disclosed in U.S. Pat. Nos. 5,629,764 and 6,061,463 incorporated herein by reference, technologies to reduce the required depth of field to near zero exist. These technologies typically work by redirecting the light in the device with holographic gratings or other suitable replacements. Any optical technology that allows the object plane and image plane to be properly aligned suffices but, the added advantage of the holographic elements as disclosed in U.S. Pat. Nos. 5,629,764 and 6,061,463 is that they take a small amount of space. Since the object plane and image plane are properly aligned, there is no need for heavier and more space consuming optics to perform optical image correction.
An alternative way to improve the depth of field without redirecting light or adding optical components is provided by this invention. Since the prisms used in these devices are formed from planar surfaces, a real object placed onto the planar platen surface can be described as a planar virtual object that appears closer to the camera than the real object. The plane orientation of the virtual object can be used in conjunction with the lateral magnification to generate an angle at which to orient the sensor chip in the camera so as to minimize the required depth of field.
More specifically, in referring to
Now referring to
αi=tan−1(m*tan(θ)/n).
If the image sensor plane is placed at angle αi in the ideal system setup, the depth of field will be zero. For example, if m=1/13, θ=45 degrees, and n=1.52, then αi=2.89 degrees.
Collimation or semi-collimation of the light source increases the efficiency at which the device uses light. Usage of more collimated light means that less overall light is needed for the camera to generate images. The higher the overall light efficiency, the lower the requirements for power and light size. U.S. Pat. Nos. 5,359,691, 5,390,276 and 5,854,872, all disclose compact light pipes that utilize microprismatic structures to guide light in a semi-collimated manner. These light pipes efficiently convert diffuse light from a Cold Cathode Fluorescent Light (CCFL) into semi-collimated light.
After an image has been captured, it may need to have perspective, barrel and pincushion distortions removed. Typically, barrel and pincushion distortions are removed first, then perspective distortion is removed. Optics have traditionally served this purpose. But, in this invention, barrel and pincushion aberrations can be removed with an algorithm developed by Ojanen. A pre-print describing this algorithm and software that implements the algorithm may be found in Appendix A and at http://www.math.rutgers.edu/˜ojanen/, the entire contents of which are hereby incorporated by reference as if fully set forth herein. The Ojanen algorithm relies on the capture of a high accuracy image of a fixed array of dots of a particular size. By knowing the size and locations of the dots within a grid, the systematic discovery of barrel and pincushion distortions can be characterized by running a least squares algorithm on the error between the observed dot locations and sizes and the known dot locations and sizes. Correction of the distortions is then completed using the returned least squares parameters. Typically, the present invention employs a perspective correction algorithm in addition to the Ojanen algorithm.
In the present invention, depth of field issues in some embodiments do not drive the lens requirement so a shorter focal distance lens can be used at the expense of potentially increasing aberrations. But, since such aberrations are predictable, they can be reasonably corrected using software that implements the Ojanen algorithm. Another important side effect of a shorter focal length lens is the f-stop setting. Since a low depth of field is required, a lower f-stop lens is used so that more efficient use is made of the light and thus size and power requirements are further reduced.
The system power requirements of the present invention are reduced to the point that the device can be run within the limits of the power supplied via common computer interfaces such as FireWire and USB, e.g., USB-2, (FireWire can provide up to 50 watts of power and USB up to 2.5 watts). The unanticipated and non-obvious innovation of the present invention is enabling the fingerprint device to be powered by a completely disconnected laptop computer while maximizing the battery life of the laptop and thereby extending the useful amount of time the system can be used. The invention provides for a device that consumes at most about 3.0 watts of power, preferably at most about 2.5 watts of power. Typically, the light uses at most about 1 watt, preferably at most about 0.7 watts and the camera uses at most about 2 watts, preferably at most about 1.8 watts.
In the device, apparatus, and method of the present invention, as one alternative to LED light sources, a CCFL is used. Other light sources include electroluminescent sources and lasers. The power source for this light is an electrical inverter that taps power off from the power originating from the host computer interface. A single CCFL is used to illuminate a light pipe that partially collimates the light before sending the light into a prism. This CCFL and light pipe construction, as described in U.S. Pat. Nos. 5,359,691, 5,390,276, and 5,854,872 generates enough light at a low power to serve as the system light source for the apparatus, system and method of the present invention. The rated lifetime of such CCFL's is about 10,000 hours of operation. As with LEDs, the rated lifetime is the amount of time the light is on at the rated power until the light output is one half of the original output. Not only does the light source emit enough light for the apparatus, system and method of the present invention, it also is delivered in a very compact size and thus contributes to the size of the apparatus, system and method of the present invention.
In the apparatus, system and method of the present invention, two electrical loads exist in every embodiment: the light source load and the camera load. In a preferred embodiment, the light source load comprises the electrical inverter and the CCFL. In alternative embodiments, there are at least several other loads: an optional electrical storage device, a magnetic stripe reader, RFID module, proximity card reader, smartcard reader, and a barcode reader. In alternative embodiments, a battery, solar cell or capacitor subsystem is used to supply electrical energy to system components. In particular, such a subsystem is needed in the case where more power is needed than the computer interface can provide. This can be the case, for instance, in an embodiment comprising at least one of a magnetic stripe reader, an RFID module, a proximity card reader and a smartcard reader to read demographic data from the back of a driver's license and an embodiment comprising a one dimensional or two dimensional barcode reader to read the demographic data from the bar code on the back of a driver's license.
In alternative embodiments, images are captured with either a one-dimensional (line scan) or two-dimensional (area scan) camera. Preferred embodiments comprise area scan cameras to increase system robustness by avoiding the use of moving parts. In these embodiments, the image generated must be large enough to capture the entire object area at a prescribed resolution. Two such cameras with USB 2.0 interfaces are the Silicon Imaging SI-3170-U and the Silicon Imaging SI-6600-U cameras.
In a preferred embodiment, the camera uses the light provided by the light pipe to capture images of rolled and slapped impressions of fingerprints. The electrical interface to the camera also comprises the control signals that operate the camera, operate the magnetic stripe reader, operate the barcode reader, operated the RFID module, operate the proximity card reader, operate the smart card reader, and turn the light source on and off. That is, in a preferred embodiment, all power, control, and data to be exchanged between the host computer and the fingerprint device are exchanged via the single connection between the computer and the device.
An important aspect of the control logic to capture fingerprints is how to determine when to start capturing a print and when to stop capturing a print. In preferred embodiments of the apparatus, system and method of the present invention, this control logic is implemented in software since the frame rate of the images delivered by the camera is high enough to allow processing on the host computer that algorithmically identifies a starting frame and an ending frame for each fingerprint capture. Typically, for a single finger roll the real frame rate is 20 frames or more per second, which results in a 12-13 or more processed frames per second rate. Typically, for a four finger slap image the real frame rate is at least 6 frames per second which results in an at least 4 processed frames per second rate. As a rolled fingerprint is being captured frame-by-frame, individual frames are analyzed and combined into a rolled print so that when the ending frame is identified, the combined rolled image is complete. Preferably, the frame rate for a rolled image is sufficient to obtain a real-time image.
In summary, in a preferred embodiment, the apparatus, system and method of the present invention provide a light source and camera combination that has a power and light efficiency that allows a large format fingerprint device to be powered, controlled, and to exchange data digital image frames over a single connection, such as a USB 2.0 cable connection. Alternative embodiments include FireWire 1.0, FireWire 2.0 and next generation peripheral interfaces. Alternative embodiments also include a power subsystem wherein a battery or capacitor is charged during periods of low power consumption and when more power is required more power is drawn from the power subsystem.
In order to ensure continued operation of the apparatus, system and method of the present invention, non-volatile memory is included inside the device so that statistical and diagnostic data can be collected and monitored. In a preferred embodiment, the number of times the light source switches on and off is maintained in non-volatile memory so that a predetermined maintenance schedule can be followed.
FIGS. 3A-B illustrate a large format fingerprint device in which a holographic grating has been incorporated according to an embodiment of the present invention;
FIGS. 8A-D are a flow diagram of the method of the present invention.
FIGS. 10A-B illustrate how changing the angle of the image sensor minimizes depth of field.
The present invention focuses on the use of optical imaging to provide an optical fingerprint scanning apparatus, system and method for capturing more than two fingerprints simultaneously.
A camera as used herein is minimally the combination of a lens and an image capture capability. For digital cameras, such image capture capability includes an image sensor and a frame grabber that converts the sensed image data to a digital format, e.g., an image frame. For film cameras, the image capture capability is provided by film. Digital cameras optionally have interfaces to transfer digitized images to a system for processing.
An image frame as used herein is data output by the camera that represents at least a portion of the scene, which the camera is capturing.
A capture sequence as used herein is a series of at least one image frame provided by the camera from which at least one image frame is selected for generating an output composite image.
A physical connection port as used herein is a connector whereby a cable or other physical electronic communication mechanism is operatively attached to the connector.
A roll capture sequence as used herein is a capture sequence used to generate a composite rolled fingerprint image.
A slap capture sequence as used herein is a capture sequence used to generate a slap fingerprint image.
Each of the preferred embodiments and alternatives thereof comprises a light source consisting of a light pipe providing semi-collimated light or a collimated light source.
Referring now to
As illustrated in
In each of the embodiments of the present invention, typically, an external stimulus causes a device according to the present invention to turn on the light source 201 before beginning fingerprint capture and turn off the light source 201 after the host computer is done interacting with the device. The preferred manner to control the light is via a software control from the host computer 206. This host computer 206 directs the device to change a switch that allows or disallows electricity to flow to the light thereby turning the light on and off.
In each of the embodiments of the present invention, a record of counts and other data is typically maintained in a non-volatile memory 203 that is located in the camera 204 electronics or elsewhere on the device. These counts are used to track the need for system maintenance. Such counts include, but are not limited to, amount of time the light source is on, the number of each type of fingerprint captured or rescanned, the number of times the light source was switched on, the number of times the light source was switched off, and the number of times the device detected that the light was off when it should have been on. Other data stored typically includes the device serial number, manufactured date, manufactured location, date and time latest used, and driver software version number. A diagnostic software or firmware component typically interrogates these counts in the non-volatile memory 203 to determine if the counts indicate device maintenance is needed. This diagnostic component, in an alternative embodiment, is also configured to perform tests to identify possible system errors. This diagnostic component outputs a diagnostic report based on the test results and the values of the counts. The diagnostic report can be viewed and browsed on a screen 210 of an attached host computer 206 or can be printed and, in any event, can be stored in a persistent storage (not shown) by the host computer 206.
The efficient light source 201 is preferably a CCFL using a light pipe (dimensionally flat, high and uniform output) or alternatively and LED or other source providing a semi-collimated light source as adapted from the teachings of U.S. Pat. Nos. 5,359,691, 5,390,276, and 5,854,872 and other collimated light sources. The patents teach light sources that inject light into the side of a light pipe. Microstructures within the light pipe and the light guide connecting the light source to the light pipe redirect the incident light into predefined directions. The arrangement and geometry of the microstructures enable the light pipe to output light from the light pipe surface in a substantially collimated fashion. The conical angles at which light leaves the surface of the light pipe are predetermined by the arrangement and geometry of the microstructures. The microstructures are typically microprisms. The light pipe configuration used for the present invention optionally also includes a filter to restrict light emanating from the filter surface to primarily semi-collimated light in a cone that diverges from the normal to the surface of the light pipe by approximately 30 degrees in each direction.
In an alternative embodiment an automatic feedback loop is used to control light source intensity 201 since light output varies over time. This feedback loop is implemented by at least one of an optoelectronic feedback loop for controlling input power to the light source 201 and a device-resident or host computer 206 resident software for adjusting the exposure time of the camera 204. For instance, Microsemi, 2381 Morse Avenue, Irvine, Calif. 92614, sells silicon chips that can easily be incorporated into an optoelectronic light feedback loop.
The apparatus, system and method of the present invention preferably captures fingerprints using image processing operations as trigger conditions. For example, placing fingers of the subject on the platen could start a capture sequence, and the subject removing contact with the platen could end the capture sequence or when a substantially similar image occurs more than a pre-determined number of times could end the capture sequence.
Alternative embodiments use foot pedals 212, buttons 211, keyboard keys (not shown), or touch screens 210. Image detection and capture processing operations are implemented in at least one of software, firmware, a dedicated circuit or a functionally dedicated chip. These operations are implemented on at least one of the host computer 206, a network computing resource 219, a frame grabber (not shown) that captures the frames, or within the device itself 200.
In preferred embodiments of the present invention, the interface 205 to the host computer 206 is a single tether 205 that handles data, control, and power for the light source 201, camera 204, platen heater, platen blower, and optional devices. Optional devices include a barcode reader 501 and a magnetic stripe reader 601. In preferred embodiments, the interface comprises at least one of USB (USB-2) connection or FireWire and their variants or other interface for exchanging data and conveying power to operate. If the device of the present invention is plugged into an external power source (not shown) such as a wall outlet or an internal or external battery power source (not shown), the interface may include Ethernet and its variants as well as optical fiber, suitable to enable high resolution images to be captured and transmitted for processing over the tether 205. Essentially, the interface can be any that provides the interconnectivity between the capture device of the present invention and a target system that performs at least one of receipt of captured images, and processing of received images.
A protective covering (not shown) is provided which comprises one of a coating placed directly onto the device, a coating placed directly onto the device combined with a removable cover, and a lightweight snap-on carrying case that the device easily slips into and out of.
For a higher resolution camera such that scanning can be performed at least at 500 dpi and 1000 dpi, the images captured can include at least one of a single digit, up to 8 digits simultaneously, a palm print, a writer's edge, and all the slaps and rolls and footprints and nose prints required of an apparatus, system and method according to the present invention.
In each of the embodiments, camera lenses 204 may introduce imaging defects such as barrel and pincushion distortion. The present invention may employ a suitable correction algorithm such as a correction algorithm substantially similar to that published by Ojanen, the entire contents of which are hereby incorporated by reference, or other suitable correction algorithm. A pre-print describing this Ojanen algorithm and software that implements this algorithm may be found at http://www.math.rutgers.edu/˜ojanen/ and is included in Appendix A. The first step in applying the algorithm is to scan a reference target on the device, which contains multiple geometric elements of known size and spacing with respect to one another. The size and spacing of these geometric elements is designed to capture barrel and pincushion distortions. For example, a rectangular array of circles measuring 0.5 mm in diameter and spaced 2.5 mm apart is used. Once the image of the target has been captured, the software attempts to automatically locate the geometric elements in the image and attempts to form a rectangular array of these elements. After a complete array of the geometric elements is identified, the software uses the known distance between elements, the size of the elements, and the center of the element array to measure differences (error terms) between the expected position of the elements and the actual position of the elements. These errors terms are used with a selected defect model to approximate the coefficients in the defect model using a least squares type of algorithm. The output coefficients and the defect model are then used to correct a captured image using the Ojanen method. Correction in the image occurs after the final roll or slap has been captured.
Preferably, perspective distortion, if present, would be compensated for optically, by firmware, by hardware or by software.
Referring now to FIGS. 3A-B, a second preferred embodiment of the present invention includes an efficient light source 201, a prism 202, a holographic grating 301 on an upper surface of the prism 202 and a light-transmitting substrate 302 on the holographic grating 301, a camera 204 having a lens, and an interface 205 to a host computer 206, and the host computer 206. The holographic grating 301, light transmitting substrate 302 and the upper surface 221 of the prism 202 together form a platen. Optionally, the holographic grating 301 is attached by an adhesive to the upper surface 221 of the prism 202 and to the lower surface of the light transmitting substrate 302 by an adhesive (not shown). Typically the holographic grating 301, light transmitting substrate 302 and the prism 202 are made of glass or acrylic polymer. In this second preferred embodiment, light emitted by the efficient light source 201 (light pipe) enters the prism 202 intersecting the prism surface 221 at an angle greater than the critical angle. Then, the light passes through a holographic grating 301 and the light transmitting substrate 302 on the surface of the prism 202, hits the finger ridges 220, and is scattered/absorbed or hits the surface of the substrate 302 and is reflected 208. Reflected light 208 passes back through the holographic grating 301 and is corrected on its way back through the holographic grating 301. The scattered and reflected light then is further reflected by a surface of the prism 202 and exits the prism 202 in the direction 208 of the camera 204. The camera 204 captures a frame and transmits the frame to the host computer 206.
In a preferred alternative of the second embodiment, the holographic grating 301 is adapted from the teaching of U.S. Pat. No. 5,629,764. The holographic grating 301 allows the size of the system to be reduced since the depth of field requirement is now near zero. This translates into a lower f-stop and lower focal length lens, which in turn translates into a shorter, required optical path length. In other words, the holographic grating 301 is significantly advantageous to the compactness (and portability) of the present invention.
An additional alternative to the second preferred embodiment and the above-mentioned alternative employs a holographic optical element (HOE) with a lensing function. Such an HOE serves a very similar function to the holographic grating 301 but, in addition, it has a focusing ability built in so that part of the function of the lens can be off-loaded onto the HOE. The net result in this additional alternative to both preferred second embodiments is that the optical path length can be made even shorter than with the holographic grating 301 since the lens 204 can be closer to the prism 202.
The embodiment illustrated in
In an alternative embodiment a capacitor 402 or a solar cell (not shown) replaces the Lithium ion battery 402. Such a construction can be used when a power shortage is temporary and can be served by the capacity of the capacitor being used.
The power for the barcode reader 501 is tapped off of the power subsystem 401. The 4100 unit from Hand Held Products, Inc. (HHP), 700 Visions Drive, Skaneateles Falls, N.Y., 13153-0208, is an example of a suitable reader. Since the barcode reader 501 and the fingerprint imager 204 are not operating at the same time, full power is only provided to one device at a time. The control logic that interfaces to the camera 204 and the barcode reader 401 is written as firmware, in a preferred embodiment. This firmware communicates with the camera 204 and the barcode reader 401 in their native formats. For instance, the HHP barcode reader mentioned above communicates with a serial interface that is implemented on a Silicon Imaging camera. The firmware within the camera 204 manages this serial interface to control the barcode reader 501. An external interface to the barcode reader 501 on the host computer 206 simply talks to the firmware, which talks with the barcode reader 501.
In a fifth preferred embodiment, illustrated in
In a sixth preferred embodiment, illustrated in
In a seventh preferred embodiment, the device illustrated in
The camera bracket 1122 has the lens threaded into the through hole 1123. A board level camera is secured in place onto camera bracket 1122 on the opposite side of the lens and the entire camera bracket assembly is mounted onto the camera mount 1121 by screwing the camera bracket to the camera mount through adjustment slot 1120. The inverter required for the light pipe can be mounted on the front side of camera mount 1119 as well. After the physical connection 1106 and the dichroic mirror 1107 have been mounted, the electrical connections within the system are completed and adjustments for the locations of the components are made so as to ensure the camera is capturing the platen area properly.
In operation, light originating at the light source 1104 enters the prism 1102 and intersects the platen surface. Light, which totally internally reflects is directed toward the camera around the optical axis 1105. The light intersects a dichroic mirror 1107 and reflects toward the camera lens 1108. The light is then capture by the image sensor 1109 on the camera unit 1110 and digitized as an image. Images captured by the camera 1110 are transmitted over the connection 1106 to an attached computer.
In an eighth preferred embodiment, the device illustrated in
Slip-Case
As illustrated in
Case
As illustrated in
Method
The capture fingerprint request 803 et seq. varies according to the type of fingerprint being captured. Several capture types exist but all such captures fit in either a roll or a slap format. In the following, a composite image is defined as an image formed from one or more image frames. Therefore, a composite image includes an output image created from a sequence of scans from a linear sensor, a rolled print generated from a series of images, or even a single slap image that was the result of one processed image frame. These formats determine the processing steps of the method to be performed. The different capture types translate to different camera configuration settings. Configuration settings include area of platen to be captured, clock rate at which pixels are captured, and camera exposure time.
When a capture fingerprint request is received 802, a new capture session is initialized if an existing session does not already exist. At the beginning of each session, the camera 204 is configured for obtaining a reference offset image 803 and a reference offset image is captured 804. Next, the light source 201 is turned on under software control 804. The system begins capturing frames from the camera 204 and a calibration process compares the current frame with the previous frame. The system calculates a metric that measures the change in the luminance of the light source 201 in the middle of each captured frame. When the change in luminance fall below a pre-set tolerance, i.e., levels-off, and the rate of increase drops below a pre-set threshold, the light 201 is deemed to be “on”. Once the light 201 is turned on 805, luminance levels may be measured again and adjustments to the exposure setting of the camera 204 are made until the luminance levels reach a pre-set threshold level. Exposure adjustment is necessary as light source 201 brightness decreases over time. If increasing the exposure cannot compensate for the lack of luminance from the light source 201 then the software reports that a maintenance check is required. If the session initialized properly, the settings of the camera 204 for exposure, frame time and viewing window (based on the fingerprint capture type) are set for a roll image 806 or a slap image 818. At this point, the system captures a blank frame from the camera 204 and keeps this frame as a reference gain image for a roll image 807 or a slap image 819. The process of capturing a roll or slap print now commences.
Roll Capture (
In one embodiment, the process for initializing a roll occurs 808. The subject then positions the center of the finger on the platen so that the subject sees the fingerprint centered in the viewing window of the client's user interface. The subject then rolls the finger in one direction to the nail bed of the finger and then completely rolls the finger in the opposite direction to the nail bed on the other side. The fingerprint roll is complete and the fingerprint system returns to the client software to let the subject know that the fingerprint roll is complete.
During this rolling process, the host computer 206 continuously captures frames 809 from the camera 304. For each frame, the image is preprocessed with offset and gain correction 810 before a fingerprint-locating algorithm is applied. The fingerprint-locating algorithm analyzes the each frame for fingerprint data and if the subject has placed a finger on the platen, then the system locates the fingerprint 811 in the frame and generates coordinates that describe a bounding box around the fingerprint.
To compute the fingerprint location in each frame, two histograms are generated. The histograms are based upon image variance and are calculated only on each row and column index that is evenly divisible by the estimated fingerprint ridge width. The variance of the grayscale values of the pixels is calculated over an area roughly equal to the width of two ridges on the current frame for every pixel whose row and column is evenly divisible by the estimated ridge width, and whose area resides entirely within the current frame. If the variance of the pixel is greater than a pre-set threshold, then the associated positions in each histogram are incremented. Once the two histograms have been generated, the first and last entries in each histogram that are above a pre-set tolerance provide a rectangle encompassing the location of the fingerprint.
The automatic trigger process employs the current and previous fingerprint bounding box locations to determine finger travel distance and direction between frames. The center column of each fingerprint is calculated as the middle of the corresponding bounding box determined by step 811. The centers of the current and previous locations are compared to determine if the fingerprint is moving and if so, which direction the finger is moving. If the Euclidian distance between the centers of the locations is less than or equal to a predetermined number of pixels, the fingerprint is determined to be stopped. If the current frame center is greater than a predetermined number of pixels right of the previous frame, the fingerprint is determined to be rolling right. If the current frame center is greater than a predetermined number of pixels left of the previous frame, the fingerprint is determined to be rolling left. The predetermined number of pixels is typically at least about 10.
A half roll in one direction is started with a frame whose roll direction is either left or right (the direction of the half roll). The half roll is composed of a sequence of frames that have a direction of either stopped or direction of the half roll. The half roll is completed when the current frame's roll direction is opposite the direction of the half roll. If the half roll has a sufficient number of frames with a roll direction equal to the half roll direction, the full roll is begun and the capture sequence is started. Otherwise, the software returns to waiting for a half roll. The full roll is composed of a sequence of frames with roll directions opposite the direction of the half roll direction, not including stopped. The full roll is completed when the roll direction of a frame is not equal to the direction of the full roll or a sufficient number of stationary frames have been captured. If a sufficient number of frames are reached between the beginning of the full roll and the end of the full roll, the software accepts the full roll as complete. If the number of frames is insufficient, the system cancels the full roll and returns to waiting for a half roll. If at any point during the rolls the finger is removed from the platen, the software returns to waiting for a half roll.
When a roll starts or is cancelled, the composite image that represents the fingerprint roll is initialized. As frames from the camera are captured, they are processed by applying offset and gain, fingerprint location, and trigger condition analysis. If the cancel condition is indicated then the current fingerprint roll is halted and the process returns to the beginning of the fingerprint roll process 808. If, instead, the end roll condition is set then the composite image is post-processed 814814A 815815A 815B. If there is no trigger condition set then the current frame is merged into the composite image 813813.1-813.5 to create a composite roll image from a sequence of frames. The process of grabbing and processing frames continues in this manner until the roll end trigger occurs. The roll end trigger signals the end of the capture sequence.
For a roll, merging 813 into a composite image is done in five steps: 1) identifying where the current composite image and new fingerprint image overlap, 2) calculating the direction of the roll, 3) computing an initial splice line by roughly aligning the fingerprint ridges between the composite image and new fingerprint image, 4) use a quality metric to refine the splice line from the top to the bottom of the image, 5) combine the new image frame into the merged composite image using morphing along the splice line.
Overlap area 813.1: The overlap area between the merged composite image and the new fingerprint image has been described above.
Roll direction 813.2: The direction of the roll can be determined by computing which side the new fingerprint image is located. For example, if the new fingerprint image is located on the left side of the composite image then the roll direction is to the left.
Initial splice line 813.3: Create an initial splice line based on the endpoints where the new fingerprint image and the composite fingerprint image intersect in the overlap area then compute the slope of this new splice line segment. Compute the center of this new splice line segment. Determine which two fingerprint ridges, near the center of the new image and the merged composite image, have the best alignment. A metric that can be used is the local gray scale average along the splice line. This center location of the splice line is updated so to this identified best match point so that splice line refinement can occur at this reliable anchor point. Copy this new splice line and its center location and call it the composite splice line.
Splice line refinement 813.4: Starting from the center of the new splice line segment iterate up the splice line segment a pre-determined number of pixels at a time. The starting position and the ending position on the splice line identify an area of the splice line that is being refined. In the top row of this refinement region, iterate from a threshold number of pixels left of the splice point to a threshold number of pixels to the right of the splice point. Form a splice line segment candidate from the iterated pixel to the starting position on the splice line. Compute a common similarity metric between all candidate splice line segments so formed. One similarity metric computes the pixel intensity average of the two areas and compares the averages. The result of the comparison is a score that represents how close these two areas match, which represents how well the ridges line up. Once all the comparisons are done, the best refined pixel location becomes the new point on the composite splice line for this particular row. This process iterates to the top of the overlap area and from the center of the splice line segment to the bottom of the overlap area. The result is a final composite splice line based on the initial splice line.
Morphing (merging) into composite image 813.5: The existing composite image and new fingerprint image form a new composite image. The initial splice line and composite splice line control what region of the composite image gets blended with the new fingerprint image region. Iterate from the bottom of the overlap region to the top of the overlap region along both splice lines simultaneously. For each row consider the pixels on that row between the splice lines and a predetermined threshold number of pixels outside of the splice lines. Iterate across this interval of pixels on the row and assign the value of the composite image at that location as a value weighted by distance between the two splice lines. Thus, data from the merged composite image is morphed into the data from the new fingerprint image.
The merging method to create the rolled fingerprint from a sequence of frames typically comprises the following steps. The first frame is copied to the composite image. The bounding boxes, located as described above for locate print 811, for the current and previous fingerprint locations are intersected to form an overlap area. The left of the overlap area is equal to the maximum of the left coordinates of the two bounding boxes. The top of the overlap area is equal to the maximum of the top coordinates of the two bounding boxes. The right of the overlap area is equal to the minimum of the right coordinates of the two bounding boxes. The bottom of the overlap area is equal to the minimum of the bottom coordinates of the two bounding boxes. The center columns of the overlap area in the current frame and composite image are examined to find where fingerprint ridges intersect the columns. These intersections are compared between the new frame and the composite image and they are used to perform a dynamic stretch on the current frame. If the current frame and composite image are too dissimilar then the merging is aborted and the subject is warned that the finger is moving too drastically. In this case, a new roll capture is automatically started. The current frame is stretched so that the ridges roughly intersect with the existing ridges from the composite image and the current image is morphed with the composite image to produce a new composite image. The final composite image becomes the fingerprint roll image.
Typically, adjacent opposed dark edges of two sequential bounding boxes are compared by taking a histogram to analyze each dark edge and a matching algorithm is used to match ridges to obtain an image such that the fingerprint ridges are continuous.
Image processing techniques remove image defects introduced by the camera and the lens. Six main processing steps occur: deviant pixel correction, offset and gain correction, high pass filtering, aberration correction, perspective correction, and noise filtering.
Offset and gain correction 810: Applying offset and gain is a pixel by pixel operation. For each pixel, a lowest and highest acceptable value are defined by the offset and gain images respectively. Offset and gain corrections stretch these lowest and highest values to the maximum range allowed in a pixel. Therefore, a mapping between the observed interval (from the offset and gain pixel) is made to the maximum range output. Using an observed pixel in an image as an input to this mapping yields an output pixel that has been offset and gain corrected. Input pixels below the lowest limit or above the highest limit are saturated at the respective values on output.
Deviant pixel correction 814: Most cameras contain imperfections on the sensory chip. These imperfections manifest themselves as intensity values that differ greatly from the normal expected intensity values. These deviant pixels are beyond a threshold away from the average intensity value of a neighborhood of pixels.
Deviant pixel correction involves two steps. The first step involves initializing the deviant pixel subsystem. The second step is the correction of individual frames.
Initialization, which occurs in the session initialization 801, requires an offset image acquired in camera initialization. The grayscale value of each pixel in the image below the second row and above the second to last row is compared to an average grayscale value of the two pixels above and two pixels below the current pixel. If the grayscale value of the current pixel significantly differs from the average of the other four pixels, the current pixel's location is added to a cached list for use in the second step.
Deviant pixel correction of the individual frames is relatively simple. For each pixel location cached in the first step, the system averages the grayscale values of the two pixels above and two pixels below and replaces the grayscale value of the current pixel with that average.
High pass filtering 814A: Edge details of the fingerprint ridges may be enhanced at the cost of increasing noise in the image frame. An approach to enhancing the edges is applying a high pass filter as is commonly known in the image processing field. Such filters convolve a high pass filter with the image. Typical high pass filters used may have a kernel size of 3, 5, or 7. The strength of the high-pass filter being used is driven by the application requirements.
Aberration correction 815: The camera lens also introduces image defects such as pincushion or barrel distortions. Parameters for correction of these defects are identified during calibration at the time of device manufacture and these parameters are stored in the non-volatile memory 203. In camera initialization, these parameters are read and the defect model is initialized.
The output coefficients and the defect model are used to correct a captured image using an algorithm such as the Ojanen algorithm. Correction amounts to local averaging in neighborhoods defined by the defect model and the output coefficients. Alternatively, an interpolation method such as bi-linear interpolation or nearest neighbor can be used to create an output pixel for each neighborhood.
Perspective correction 815A: Perspective correction of the image may also be performed. In using an algorithm such as the Ojanen algorithm, a mathematical description of the perspective model identified by the algorithm can be used in conjunction with an interpolation algorithm such as bi-linear interpolation to generate the final corrected composite image. Correction in the image occurs after the final roll or slap has been captured. Alternatively, if the device is made in a precision fashion, the perspective correction can be geometrically modeled as a three-dimensional relationship between planes. Once measured, the mathematical description of these planes can be used in conjunction with bi-linear interpolation to create a perspective corrected composite image. In both cases described here, the initialization step 801 preferentially precomputes the perspective correction parameters for each pixel so that in full operation extra time would not have to be spent on calculating needed weights repeatedly.
Noise filtering 815B: The noise filter algorithm convolves the image with an averaging convolution kernel. In the convolution operation, the variance within the neighborhood is used in conjunction with a fixed threshold. If the variance exceeds the threshold then the original pixel is left unchanged otherwise, the pixel is assigned the value of the convolution. The convolution kernel size is established experimentally according to an application's requirements.
When post-processing is completed the diagnostic values are updated 816. When all the statistical data has been written back into the non-volatile memory 203, control returns to the host program. The host program then saves the final image and returns to and signals completion 817. At this point, the host software requests the image just captured, requests another fingerprint acquisition, reads non-volatile memory, or closes its session.
In alternative embodiments of capturing the rolls, the image processing steps may have their order changed and merging of the composite rolled image may occur at any stage of processing the frame.
Slap Capture (
If a capture image trigger condition did not occur then the process of capturing and processing frames continues 821 to 824. The final captured frame is post processed 825825A 826826A 826B as described above for roll capture 814814A 815815A 815B. When the post processing is complete the software updates diagnostic data in non-volatile memory 827, saves the image and indicates the capture is complete 828 to the host program and processes a new request 802.
In alternative embodiments of capturing the slaps, the image processing steps may have their order changed and merging of the composite rolled image may occur at any stage of processing the frame.
The present invention applies to large format fingerprint/handprint/footprint scanning as well as pet imaging applications. Other extensions of this technology to other applications are also possible. The descriptions herein are not meant to be limiting to the applications described herein but were intended to be illustrative of the application of this invention. For instance, the same invention can be applied in newborn applications in which the footprints of newborns are digitally captured, in applicant processing applications for capture and submission of fingerprints for criminal background purposes, for arrestee or detainee applications, for verification of the person's identity when the collected prints are matched against a database of prints, for access control applications, and for applications to sampling pet paws or pet noses to maintain identities of animals with certified pedigrees. Many times demographic data must be collected in conjunction with the fingerprints so that the fingerprints can be associated with a name, address, identification number, etc. Such demographic data allows expedited matching in many database systems.
This is a non-provisional application claiming the benefit under 35 U.S.C. 119 of provisional application No. 60/534,394 filed on Jan. 7, 2004, the entire disclosure of which is hereby incorporated by reference.
| Number | Date | Country | |
|---|---|---|---|
| 60534394 | Jan 2004 | US |
