• 专利附录
  • 专利说明
  • 权利要求
  • 类似技术
  • 同族专利
  • 引用文献
  • 法律状态
  • 优先权
    • 专利摘要:
    The present invention relates to a system for determining the vividness of an image presented for authentication, a reference signal is processed on a display, and a reflection of the signal processed from a target is analyzed to determine its vividness. . The analysis includes spatial and / or temporal bandpass filtering of the reflected signal, and the determination of RGB values for each frame of the reflected signal and / or each pixel in one or more frames of the reflected signal. The correlations of frame level and / or pixel by pixel between the determined RGB values and the signal provided are calculated, and a determination as to whether a displayed image is live or false is made through one or both correlations.
    公开号:BR112017004427B1
    申请号:R112017004427-7
    申请日:2015-09-09
    公开日:2021-01-19
    发明作者:David Hirvonen
    申请人:EyeVerify Inc.；
    IPC主号:
    专利说明:

    CROSS REFERENCE TO RELATED REQUESTS
    [0001] This application is a continuation of, and claims priority for, US Patent Application No. 14 / 480,802 pending, filed September 9, 2014, entitled "Systems and Methods for Liveness Analysis", the description of which is here incorporated in its entirety as a reference. TECHNICAL FIELD
    [0002] The present disclosure relates, in general, to image analysis and, in particular, to image processing techniques to detect whether an individual shown in an image is alive. BACKGROUND
    [0003] It is often desirable to restrict access to goods or resources to specific individuals. Biometric systems can be used to authenticate an individual's identity to grant or deny access to a resource. For example, iris scanners can be used by a biometric security system to identify an individual based on unique structures in the individual's iris. Such a system can erroneously authorize an imposter, however, if the imposter presents to scan a pre-recorded image or video of an authorized person's face. This fake image or video can be displayed on a monitor, such as a cathode ray tube (CRT) or liquid crystal display (LCD), in bright photographs, etc., in front of a camera used for scanning. Some counterfeit detection systems can detect a false image when checking eye movement. But such a system may not be effective in identifying a fake video that includes the expected movement of the eye. Improved systems and methods are required for the efficient determination of fake images and videos from those provided live by authorized persons. SUMMARY
    [0004] In several implementations described here, differences in the reflective properties of real / authentic faces and false faces are used to distinguish live and authentic faces and / or eyes from fake images / videos. This is achieved, in part, by making a reference signal on a screen held in front of a target, which can be a real face or a false image, by recording a reflection of the reference signal by the target and when calculating one or more correlations between reflected and rendered signals.
    [0005] Therefore, in one aspect, a computer-implemented method of determining whether a real image is presented for authentication includes rendering, on a display device, a first time-varying signal that includes several respective different signals that are separated in phase from each other. The method also includes capturing, during rendering, numerous images of a target that is illuminated by the first rendered signal, and applying a bandpass filter temporarily to the images to generate a plurality of filtered images. The method also includes extracting a second signal from the filtered images and generating a first measurement based on at least a temporal correlation of the first signal and the second signal. In addition, the method includes, for each pixel location in several pixel locations, extracting a respective signal for the pixel location based on changes in the respective pixel location value over time in numerous filtered images and calculating a respective pixel location correlation score for each of the pixel locations based on a correlation of the respective signal extracted from the pixel location to the first signal. The method further includes generating a second measurement based on at least several of the pixel location correlation scores, and accepting or rejecting the target based on at least the first and second measurements.
    [0006] Each respective signal of the first signal can have a different color, and each color can be processed using the same frequency. In some implementations, each respective signal of the first signal is a different monochrome signal, and the first signal can be sinusoidal. Each respective signal of the first signal can be a distinct curvature, and the curvatures can be superimposed on the first signal. The respective signals from the first signal can be generated at random. In some implementations, a respective value specific to a pixel location can be a color.
    [0007] Each image can include the countless respective images that have undergone, each, a respective transformation, are in a respective different resolution or include a respective different spatial frequency band that corresponds to a selected lighting phenomenon. The first measurement can be based even more on whether a phase of the first signal corresponds to a phase of the second signal. Extracting a second signal from the filtered images may include extracting the second signal from a respective dominant color value for each of the filtered images. In some implementations, the method also includes stabilizing the target on several of the images captured and / or processed before applying the bandpass filter. The bandpass filter can be applied in the frequency domain or in the time domain.
    [0008] In some implementations, generating the second measurement based on at least a number of pixel location correlation results includes combining the pixel location correlation scores to generate the second measurement. The target can be a human face and the combined pixel location correlation scores can be for the pixel locations of a specific region of the face. The specific region of the face can be determined using one or more of: (i) dynamic image analysis to avoid, at least in part, one or more parts of the face that are occluded or overexposed in the plurality of images and (ii ) a mask or a weight map that represents knowledge about the characteristics of the face that are likely to reflect the first rendered signal.
    [0009] Each pixel location can represent several elements of image data, some or all of which may be in different resolutions. Each pixel location can be a weighted combination of the respective image data elements of the pixel location. The various images captured can represent a Gaussian pyramid or a Laplacian pyramid. A particular filtered image of the filtered images can represent a weighted combination of a number of pyramid levels. Other modalities of this aspect include systems, apparatus and corresponding computer programs.
    [0010] Particular implementations of the subject described in this specification can achieve one or more of the following advantages. For example, the detection technique depends on a multispectrum pattern signal that is processed during the detection. Images of a person's face and / or eye from any pre-recorded video / image provided for authentication are unlikely to correlate with the multispectrum pattern signal provided during the detection of vividness. In addition, any reflection of the multispectrum pattern signal from a screen rendering, such a video / image, is likely to be of a different nature than the reflection of the face and / or eye of a living person. Several implementations described here can detect these anomalies, as explained below and, as such, can be more robust to distinguish a live and authorized person from fake videos and / or images.
    [0011] The details of one or more modalities of the subject described in this specification are presented in the attached drawings and in the description below. Other characteristics, aspects and advantages of the subject will become evident from the description, drawings and claims. BRIEF DESCRIPTION OF THE DRAWINGS
    [0012] The patent or order file contains at least one password executed in color. Copies of that patent publication or patent application with color design (s) will be provided by the Office upon request and payment of the necessary fee; Figure 1 illustrates an example procedure for determining two measurements of liveliness; figures 2 (a) to 2 (c) represent an example of a captured video frame, a corresponding standardized and stabilized video frame and a corresponding time band filtered frame, respectively; figures 3 (a) to 3 (c) represent example recovered RGB signals that correspond to a sequence of captured video frames, the bandpass filter response of the recovered RGB signals and the corresponding rendered reference RGB signals, respectively; Figures 4 (a) to 4 (c) represent, from top to bottom, the Fast Fourier Transform (FFT) periodograms of the signals represented in Figures 3 (a) to 3 (c), respectively; figure 4 (d) represents a temporal Butterworth filter used to generate the bandpass filter response illustrated in figure 3 (b); figure 5 (a) illustrates a normalized and stabilized image frame in the example average; figures 5 (b) to 5 (d) represent a corresponding two-dimensional (2D) correlation image, a processed correlation image and a corresponding saturation image, respectively; figure 5 (e) represents an example of a face mask; figures 5 (f) to 5 (k) represent captured video frames of example, which correspond to a complete cycle of a reference RGB signal, with the use of which the 2D correlation image illustrated in figure 5 (b) is calculated; figure 6 represents an example configuration of an LCD monitor that renders a false image and a telephone that captures and analyzes the false image; figure 7 illustrates another example configuration of an LCD monitor that represents a false image and a telephone that captures and analyzes the false image; figures 8 (a) to 8 (k) represent a false average image frame captured from an LCD monitor configured as shown in figure 6, the corresponding 2D correlation image and video frames, which correspond to a complete cycle of the RGB signal, with the use of which, the 2D correlation image illustrated in figure 8 (b) is calculated; figures 9 (a) to 9 (k) represent a false average image frame captured from an LCD monitor configured as shown in figure 7, the corresponding 2D correlation image and video frames, which correspond to a complete cycle of the RGB signal, with the use of which, the 2D correlation image illustrated in figure 9 (b) is calculated; figures 10 (a) to 10 (c) represent the recovered RGB signals that correspond to the sequence of captured false video frames shown in figures 9 (f) to 9 (k), bandpass filter response of the recovered RGB signals and the corresponding rendered RGB reference signals, respectively; figures 11 (a) to 11 (c) represent, from top to bottom, rapid Fourier transformation (FFT) periodograms of the signals represented in figures 10 (a) to 10 (c), respectively; figure 11 (d) represents a temporal Butterworth filter used to generate the bandpass filter response illustrated in figure 10 (b); figure 12 represents Moiré patterns associated with a false image; figure 13 illustrates another example of a procedure for detecting the vividness of an eye; figures 14 (a) and 14 (b) describe an example of an eye that reflects a telephone that captures an image of the eye and the corresponding 2D correlation image, respectively; figure 15 (a) shows the false image represented in figure 12 with a higher resolution; figures 15 (b) and 15 (c) show a high-resolution cropped portion of the image represented in figure 15 (a) and the local 2D correlation, calculated according to the procedure shown in figure 13.
    [0013] Reference numbers and similar designations in the various drawings indicate similar elements. DETAILED DESCRIPTION
    [0014] Figure 1 illustrates a general framework for calculating two measurements that can be used to distinguish between images of an eye obtained from a real, living and a fake person (for example, images or video previously captured from a person alive). In step 102, a multispectral pattern is processed on the display device, such that a person's face (or "target") is illuminated by the pattern. In some implementations, the pattern is displayed for about a second, but other durations are possible. The display device can be the display device of a data processing device such as a smartphone, smart glasses, a smart watch, a tablet computer, a laptop, etc. The images of the target illuminated by the multispectral pattern are captured by a digital camera in step 104. In some implementations, the digital camera is a digital camera facing the front of the data processing apparatus. Other digital cameras can be used, including digital cameras on other devices.
    [0015] In several implementations, the multispectral pattern includes three overlapping sinusoidal signals. For example, red, green and blue (RGB) curvatures can be used to match the sensitivity of native filters for each color channel in standard Bayer digital cameras. Sinusoidal signals can be processed at substantially a single frequency, so that a single bandpass filter can be used for subsequent analysis (described below). In addition, the three sinusoidal signals can be uniformly separated in phase through the three color channels (for example, red = 0, green = (2 * pi) / 3 and blue = (2 * pi) * 2/3 ), to improve the separation capacity of the recovered signal and reduce lighting gaps that can exacerbate the blinking effects that may be uncomfortable for some users. In a deployment, a frequency of about 4 Hz is used, which is below the threshold for photosensitive epilepsy, however, it is fast enough to be easily separable from typical low-frequency lighting noise in a short period of time. Other multi-spectral patterns can be used, in addition to RGB curvatures, including patterns with less or more component signs, a red and blue curvature, for example.
    [0016] A video signal that includes the images captured by the digital camera is recorded in step 104. In some implementations, the video signal is a 0.75 second video clip at approximately 25 Hz, that is, 25 frames / second. Other durations and frame rates are possible. In step 106, each frame in the recorded video signal can be marked with the value (for example, the RGB value) of the pattern being rendered on the display device in step 102 at approximately the time the image was captured. Interchangeable image file (EXIF) metadata (or other metadata) can also be stored in step 106e, in general, to provide a measurement of ambient lighting for automatic threshold adjustment. Metadata can include ambient brightness, exposure time, ISO setting and / or aperture value.
    [0017] In some implementations, video stabilization (recording and deformation) can be performed on the video signal recorded in step 108 to map the points in the scene to a common reference coordinate system. After stabilization and deformation, the frames can be converted and, a normalized RGB color space to reduce sensitivity to shadows and other lighting artifacts in the environment, and thus a stabilized and normalized video signal is obtained in step 108.
    [0018] In step 110, the stabilized and normalized video is processed with the use of a temporary bandpass filter that is tuned to the frequency of the processed curvature, for example, 4 Hz in an example. As an illustration, the filter can be applied to Gau- sian pyramids that correspond to stabilized and standardized video frames. Temporal bandpass filtering can be performed to isolate from the normalized signal obtained in step 108, a response signal that corresponds to the multispectral standard processed in step 102. Finally, the filtered bandpass video signal is compared to the standard multispectral previously rendered, for example, at different scales, to obtain: (1) a global picture based on the temporal correlation in step 112 and / or (2) a correlation in the sense of the local pixel in step 114, as described below.
    [0019] In order to calculate a global time correlation measurement, each frame of the filtered response signal obtained in step 110 can be represented with a dominant RGB value in step 122. The dominant RGB value assigned in step 122 must match the color multispectral pattern of the rendered RGB value, as represented by the RGB values marked for the video signal recorded in step 106. As an illustration, the dominant RGB values can be calculated robustly from a chromaticity histogram or as a weighted average of pixel values for each frame. Other ways of determining the dominant RGB value are possible.
    [0020] An average saturation image is calculated from the filtered response signal (step 110) and can be used to provide the weights for the weighted average method (step 122). In some implementations, the average saturation image is the distance from a gray image that corresponds to the frame to be processed. The resulting two-dimensional (2D) saturation image is proportional to the intensity of the reflected RGB multispectral pattern. Then, in step 124, a linear trend withdrawal is performed independently on each of the estimated red, green and blue signals, in order to remove any ramp component from the data, making it more suitable for comparison with the reference RGB multispectral standard signal. The linear trend withdrawal can be calculated using a linear m-estimator, for example.
    [0021] Figure 3 (a) shows an example of a global RGB signal. The signal is called "global" because it represents the dominant RGB values corresponding to a frame and not any particular pixel in that frame. In step 126, this global signal is processed with a Butterworth temporal bandpass filter in the frequency domain to extract the appropriate frequency that corresponds to the recorded signal. Figures 2 (b) and 2 (c) show the filtered RGB signal and the rendered reference signal (that is, the RGB multispectral standard), respectively. These two signals are compared in step 128 using a normalized cross-correlation, and the resulting value, denoted nxcorr, indicates a first measurement of liveliness. In one implementation, a small one-dimensional (1D) temporal survey is performed in step 128 to compensate for the latency in the camera controller, which can cause a small change between the measured and processed RGB signals. The survey is a 1D survey, because each point in the combined waveform in figure 2 (a) represents an entire frame. Figures 4 (a) to 4 (c) describe fast Fourier transform (FFT) periodograms of the signals represented in figures 3 (a) to 3 (c), respectively, Correlation in the direction of Local Pixel
    [0022] In step 114, a local temporal normalized cross-correlation spatial mean is calculated at each pixel location in the filtered video response (ie, the signal obtained in step 110 by filtering the stabilized and normalized recorded signal through the temporal bandpass filter). The spatial average can produce a 2D correlation image (for example, in an interval [-1 ... +1]) which can indicate how accurately each pixel in the filtered response corresponds to the generated RGB signal. For example, figure 5 (b) shows a correlation image that corresponds to an example stabilized and normalized recorded image represented in figure 5 (a). Figure 5 (c) shows a processed 2D correlation image obtained, for example, by selecting the maximum left and right correlation images, as described below. To calculate a 2D correlation, a face mask can be applied at step 132, for example, to restrict processing to the skin part of the face and thereby remove dark characteristics of the low albedo face and / or to remove the movement noise independent of the eyes. Figure 5 (e) represents an example of a face mask. A local pixel-to-pixel correlation is then calculated in step 134, for example, for each of the images illustrated in figures 5 (f) to 5 (k). These images correspond to a complete cycle of the RGB multispectral standard and the respective pixel-by-pixel correlations can be calculated and processed to obtain the final 2D correlation image illustrated in figure 5 (c).
    [0023] In some implementations, when calculating the local pixel-to-pixel correlation, the phase delay recovered from the overall correlation above can be used in step 134 to avoid the need for an expensive correlation search on the volumetric data corresponding to the stabilized and normalized values obtained in step 110. In some implementations, the normalized mean values of spatial cross correlation are calculated separately, in steps 136, 138, respectively, for the left and right sides of the face mask. The maximum of the two spatial correlations can be selected in step 140. This can provide a more robust correlation measurement than a single average, since extreme lighting conditions are often limited to just one side of the face. Alternatively, the global average for all pixels of the face mask can be used if the ambient brightness value of the EXIF metadata is low enough to make saturation unlikely, as can be found in most indoor environments. Figure 5 (d) illustrates a saturation image that corresponds to the 2D correlation image illustrated in figure 5 (c). The final mean of the local correlation measurement, denoted nxcorr2, can be a second measurement of liveliness.
    [0024] Typically, the skin of a real face provides relatively diffuse reflection with high albedo and, as such, the correlation value in each pixel can be high. The correlation image tends to be quite uniform, with a relatively low spatial variance. In contrast, when a video monitor is used for impostor reproduction, the monitor tends to behave like a mirror and, depending on the angle of reflection of the light emitted from the display screen in which the multispectral RGB standard is processed , the light is either mainly reflected back locally in a small portion of the captured face image (as shown in figure 6) or reflected away from the display screen, as shown in figure 7.
    [0025] For example, figure 8 (a) represents a captured printer image that is displayed on an LCD screen in front of the device to which access must be authorized (for example, a phone), as shown in figure 6 Figures 8 (b) and 8 (c) show the corresponding 2D correlation images, figure 8 (d) shows the corresponding saturation image, figure 8 (e) shows the applied face mask, and figures 8 ( f) to 8 (k) describe several captured image images that correspond to a complete cycle of the RGB multispectral standard provided as shown in step 102 in figure 1. In this example, the second nxcorr2 measurement is high (about 0.63 ), because the LCD screen is kept parallel to the phone used to capture images and because the LCD screen works like a mirror. The first nxcorr measurement, that is, the overall correlation, is low, however, indicating that the captured images are probably not obtained from a live source. If the LCD screen displaying the imposter images is kept at an angle to the screen used to render the multispectral RGB standard, as shown in figure 7, for example, both the nxcorr2 and nxcorr values are expected to be low , that is, below a selected limit, such as 0.5, 0.4, 0.3, etc. A typical example that corresponds to this case, in which the light is reflected away from the camera, is illustrated in figures 9 (a) to 9 (k). In this case, neither the global correlation measurements nor the local averages correspond to the expected RGB signal, in general, causing both the nxcorr and nxcorr2 measurements to be low. As such, the filtered response signal obtained in step 124 can be very noisy, as the 1D RGB signal shown in figures 10 (a) to 10 (c) illustrates.
    [0026] In addition to exploring the mirror-like properties of many video playback screens, correlation measurements may reflect other anomalies in video reproduction, for example, artifact samples, such as vertical bands in the filtered output images. temporal bandpass, as can be seen in the last six frames in figure 9. In one implementation, a normalized FFT for each color signal represented in the filtered response signal is a strong indicator that the individual is an imposter, as can be seen in figure 11. The top three lines are the periodograms corresponding to the red, green and blue channels, obtained from the filtered response signal (obtained in step 110, figure 1). The final line is a temporary pass-through Butterworth filter set for the expected signal period in the recorded video. A low ratio of the filtered bandwidth signal to the total signal energy is another measurement that can be used to detect imposter cases.
    [0027] Reflection analysis from the LCD screen held in front of the image capture device (for example, a cell phone camera) can be used to help detect an impositor when, for example, nxcor2 is high , but nxcorr is low, as described with reference to figures 8 (a) to 8 (k). For example, figures 12 (a) to 12 (c) show a fake image displayed on an LCD screen held in front of a camera, a cropped image of the face region close to the eye and an edge image that corresponds to the cropped image , representing a reflection image from the phone that was used to capture the fake image displayed on the LCD screen. Another artifact is the monitor's texture patterns that are visible in the 2D correlation image, as can be seen in figure 12 (d). A 2D classifier, such as a Haar classifier, can be trained to identify the patterns in the correlation image that are unique to imposter cases. In general, in several implementations, an authentic classification is returned if, and only if, the global correlation (nxcorr) and the global correlation (nxcorr2) exceed a predetermined limit.
    [0028] Figure 13 illustrates another imposter detection technique that has the advantage of the reflective properties of a typical eye. Specifically, step 1302 of rendering an RGB multi-spectral pattern, step 1304 of capturing a video signal, step 1306 of marking each frame with an RGB value and step 1306 of stabilizing the recorded video signal and marked are performed in a similar manner as described above with reference to figure 1. Then, in step 1308, a space-time bandpass decomposition is performed to explore the convex reflective properties of the eye. It is observed that an eye typically has a convex reflective surface so that each image structure captured in step 1304 includes a reduced mirror image of the eye environment, which may include a compact image of the RGB standard processed on a display screen. in step 1302.
    [0029] In step 1310, the temporal bandpass filters are applied to a Laplacian pyramid that corresponds to the stabilized and marked signals. The Laplacian pyramid can provide a spatial bandpass decomposition of the input video to help isolate the primarily high spatial frequencies from the multi-spectral RGB pattern reflected from the eye.
    [0030] A 2D pixel-to-pixel correlation image is then produced using a temporal normalized cross-correlation between the reference signal and the video bandpass filtered output, in step 1312. A local average in a small neighborhood of the dominant peak can be used as a measure of additional liveliness. In general, this approach can detect the vividness of the eyes as opposed to detecting the vividness of the face using the first and second measurements described above. In a pixel-by-pixel local correlation for only the eye region of an authentic real eye, only a bright spot is expected which corresponds to the reflection of the RGB signal produced by the eye pupil, as can be seen in figures 14 (a) and 14 (b). If multiple points are observed or points are not detected, it is determined that the captured images are likely to be provided by an imposter.
    [0031] The systems and techniques described here can be implemented in a computing system that includes a final component (for example, as a data server) or that includes a middleware component (for example, an application server) or which includes an initial component (for example, a client computer with a graphical user interface or a Web browser through which a user can interact with an implementation of the systems and techniques described herein), or any combination of such final components, middleware or initials. The system components can be interconnected by any form or means of digital data communication (for example, a communication network). Examples of communication networks include a local area network ("LAN"), a wide area network ("WAN") and the Internet.
    [0032] The computing system can include clients and servers. A client and a server are, in general, remote from each other and can interact through a communication network. The client and server relationship arises because of computer programs running on the respective computers and which have a client-server relationship between them. Numerous modalities have been described. However, it must be understood that various modifications can be made without departing from the spirit and scope of the invention.
    [0033] The modalities of the subject and the operations described in this specification can be implemented in digital electronic circuits, or in computer software, firmware or hardware, including the structures revealed in this specification and their structural equivalents, or in combinations of one. or more of these. The modalities of the subject described in this specification can be implemented as one or more computer programs, that is, one or more computer program instruction modules, encoded in a computer storage medium for execution by, or to control the operation of, data processing apparatus. Alternatively or additionally, the program instructions can be encoded in an artificially generated propagated signal, for example, an electrical, optical or electromagnetic signal generated by the machine, which is generated to encode the information for transmission to a suitable receiver device for execution by a data processing device. A computer storage medium can be or be included in a computer-readable storage device, in a computer-readable storage substrate, in a random or serial access array or memory device, or in a combination of one or more of that. In addition, while a computer storage medium is not a propagated signal, a computer storage medium can be a source or destination of computer program instructions encoded in an artificially generated propagated signal. The computer's storage medium can also be or be included in one or more separate physical components or media (for example, multiple CDs, discs, or other storage devices).
    [0034] The operations described in this specification can be implemented as operations performed by a data processing device on data stored on one or more storage devices readable by computer or received from other sources.
    [0035] The term "data processing apparatus" encompasses all types of data processing apparatus, devices and machines, including, for example, a programmable processor, a computer, a system on a chip or multiples, or combinations of previous ones. The device may include special-purpose logic circuits, for example, an FPGA (field programmable port arrangement) or an ASIC (application-specific integrated circuit). The device may also include, in addition to the hardware, the code that creates an execution environment for the computer program in question, for example, a code that constitutes processor firmware, a protocol stack, a database management system data, an operating system, a multi-platform runtime environment, a virtual machine, or a combination of one or more of these. The device and the execution environment can carry out several different computing model infrastructures, such as web services, distributed computing and grid computing infrastructures.
    [0036] A computer program (also known as a program, software, software application, script or code) can be written in any form of programming language, including compiled or interpreted languages, declarative or procedural languages and can be implemented in any way. form, Including as a stand-alone program or as a module, component, subroutine, object or other unit suitable for use in a computing environment. A computer program can, but does not have to, correspond to a file in a file system. A program can be stored in a part of a file that contains other programs or data (for example, one or more scripts stored in a markup language resource), in a single file dedicated to the program in question or in several coordinated files, files that store one or more modules, subprograms, or pieces of code). A computer program can be implemented to run on one computer or on multiple computers that are located in one location or distributed in several locations and interconnected by a communication network.
    [0037] The modalities of the subject described in this specification can be implemented in a computer system that includes a final component, for example, as a data server, or that includes a middleware component, for example, an application server , or that includes an initial component, for example, a client computer that has a graphical user interface or a web browser through which a user can interact with an implementation of the subject described in this specification, or any combination of a or more such as final component, middleware, or initial. The system components can be linked together by any means or means of digital data communication, for example, a communication network. Examples of communication networks include a local area network ("LAN") and a wide area network ("WAN"), an inter-network (for example, the Internet) and point-to-point networks (for example, networks point to point for that purpose).
    [0038] The computing system can include clients and servers. A client and a server are, in general, remote from each other and can interact through a communication network. The client-server relationship arises because of the computer programs that run on the respective computers and that have a client-server relationship to each other. In some embodiments, a server transmits data (for example, an HTML page) to a client device (for example, for the purpose of presenting data to receive user input and from a user who interacts with the client's device) . The data generated on the client device (for example, a result of user interaction) can be received from the client device on the server.
    [0039] A system of one or more computers can be configured to perform specific operations or actions due to having software, firmware, hardware, or a combination of them installed in the system that, in operation, makes the system perform the actions. One or more computer programs can be configured to perform particular operations or actions by virtue of including instructions that, when executed by a data processing device, cause the device to perform the actions.
    [0040] Although this specification contains many specific details of implementation, these should not be interpreted as limitations on the scope of any inventions or what can be requested, but rather as descriptions of characteristics specific to specific modalities of particular inventions. Certain characteristics that are described in this specification, in the context of separate modalities, can also be implemented in combination in a single modality. On the other hand, several characteristics that are described in the context of a single modality can also be implemented in several modalities separately or in any suitable subcombination. In addition, although the characteristics can be described above as acting on certain combinations and still initially claimed as such, one or more characteristics of a claimed combination, in some cases, can be excised from the combination, and the claimed combination can be targeted to a subcombination or variation of a subcombination.
    [0041] Likewise, although the operations are represented in the drawings in a specific order, this should not be understood as requiring that these operations be carried out in the specific order shown or in sequential order, or that all illustrated operations are carried out for achieve the desired results. In certain circumstances, parallel and multitasking processing can be advantageous. In addition, the separation of the various components of the system in the modalities described above should not be understood as requiring such separation in all modalities, and it should be understood that the components and program systems described can, in general, be integrated together into a single software product or packaged in various software products.
    [0042] Thus, the specific modalities of the subject have been described. Other modalities are within the scope of the following claims. In some cases, the actions listed in the claims may be carried out in a different order and still achieve the desired results. In addition, the processes described in the attached figures do not necessarily require the specific order shown, or sequential order, to obtain the desired results. In certain implementations, multitasking and parallel processing can be advantageous.
    权利要求:
    Claims (21)
    [0001]
    1. Method implemented by computer, characterized by the fact that it comprises: rendering, in a display device, a first signal that varies over time, which comprises a plurality of different respective signals that are separated in phase with each other; during rendering, capture a plurality of images of a target that is illuminated by the first rendered signal; apply a bandpass filter in time to the images to generate a plurality of filtered images; extract a second signal from the filtered images; generating a first measurement based on at least one temporal correlation of the first signal and the second signal; for each pixel location in a plurality of pixel locations, extract a respective signal for the pixel location based on changes to a respective pixel location value over time in a plurality of the filtered images; calculating a respective pixel location correlation score for each of the pixel locations based on a correlation between the respective signal extracted from the pixel location to the first signal; generating a second measurement based on at least a plurality of the pixel location correlation scores; and accept or reject the target based on at least the first and second measurements.
    [0002]
    2. Method, according to claim 1, characterized by the fact that each respective sign of the first sign is of a different color.
    [0003]
    3. Method, according to claim 2, characterized by the fact that each color is processed using the same frequency.
    [0004]
    4. Method according to claim 1, characterized by the fact that each respective signal of the first signal is a different monochrome signal.
    [0005]
    5. Method according to claim 1, characterized by the fact that the first signal is sinusoidal.
    [0006]
    6. Method according to claim 1, characterized by the fact that each respective sign of the first sign is a distinct curvature and in which the curvatures are superimposed on the first sign.
    [0007]
    7. Method, according to claim 1, characterized by the fact that the respective signals of the first signal are generated randomly.
    [0008]
    8. Method according to claim 1, characterized by the fact that a respective specific value of a pixel location is a color.
    [0009]
    9. Method, according to claim 1, characterized by the fact that each image comprises a plurality of respective images which are each respectively transformed in a respective different resolution, or comprise a respective different spatial frequency band that corresponds to a selected lighting phenomenon.
    [0010]
    10. Method according to claim 1, characterized by the fact that the first measurement is still based on the fact that a phase of the first signal corresponds to a phase of the second signal.
    [0011]
    11. Method, according to claim 1, characterized by the fact that the extraction of a second signal from the filtered images comprises the extraction of the second signal from a respective value of the dominant color of each of the filtered images.
    [0012]
    12. Method according to claim 1, characterized by the fact that it additionally comprises the stabilization of the target in the plurality of images before applying the bandpass filter.
    [0013]
    13. Method, according to claim 1, characterized by the fact that the bandpass filter is applied in the frequency domain or in the time domain.
    [0014]
    14. Method according to claim 1, characterized in that the generation of the second measurement based on at least a plurality of the pixel location correlation scores comprises combining the pixel location correlation scores to generate the second measurement.
    [0015]
    15. Method according to claim 14, characterized by the fact that the target is a human face and in which the combined pixel location correlation scores are for the pixel locations of a specific region of the face.
    [0016]
    16. Method, according to claim 15, characterized by the fact that the specific region of the face is determined using at least one of: (i) dynamic image analysis in order to avoid, at least in part, a or more parts of the face that are occluded or overexposed in the plurality of images, and (ii) a mask or a weight map that represents the knowledge about the characteristics of the face that are likely to reflect the first processed signal.
    [0017]
    17. Method according to claim 1, characterized by the fact that each pixel location represents a respective plurality of image data elements.
    [0018]
    18. Method according to claim 17, characterized by the fact that a plurality of the image data elements are in different resolutions.
    [0019]
    19. Method according to claim 18, characterized by the fact that each pixel location is a weighted combination of the respective image data elements of the pixel location.
    [0020]
    20. Method, according to claim 1, characterized by the fact that the plurality of captured images represents a Gaussian pyramid or a Laplacian pyramid.
    [0021]
    21. Method according to claim 20, characterized in that a specific filtered image of the filtered images represents a weighted combination of a plurality of pyramid levels.
    类似技术:
    公开号 | 公开日 | 专利标题
    BR112017004427B1|2021-01-19|method for analyzing liveliness
    Boulkenafet et al.2017|OULU-NPU: A mobile face presentation attack database with real-world variations
    US10853625B2|2020-12-01|Facial signature methods, systems and software
    US10762334B2|2020-09-01|System and method for entity recognition
    US10380421B2|2019-08-13|Iris recognition via plenoptic imaging
    WO2018040307A1|2018-03-08|Vivo detection method and device based on infrared visible binocular image
    WO2016084072A1|2016-06-02|Anti-spoofing system and methods useful in conjunction therewith
    Nemoto et al.2015|Visual attention in LDR and HDR images
    TW202011260A|2020-03-16|Liveness detection method, apparatus and computer-readable storage medium
    CN109325933A|2019-02-12|A kind of reproduction image-recognizing method and device
    CN109670390A|2019-04-23|Living body face recognition method and system
    US9996743B2|2018-06-12|Methods, systems, and media for detecting gaze locking
    Farrukh et al.2020|FaceRevelio: a face liveness detection system for smartphones with a single front camera
    US20200302156A1|2020-09-24|System architecture and method of authenticating a 3-d object
    US20200314411A1|2020-10-01|Synchronizing an illumination sequence of illumination sources with image capture in rolling shutter mode
    KR20200034018A|2020-03-31|Face recognition method based on infrared image and learning method for the same
    CN112597466A|2021-04-02|User authentication method and system
    TRANSACTIONS2017|Face Spoofing and Counter-Spoofing: A Survey of State-of-the-art Algorithms
    EP2479722B1|2014-03-19|Motion-based image watermarking
    Wang et al.2013|Livrable D1. 3 of the PERSEE project-Perceptual Modelling: Softwares results and final report.
    同族专利:
    公开号 | 公开日
    US20160071275A1|2016-03-10|
    RU2671540C2|2018-11-01|
    CA2960397C|2018-05-01|
    ES2688399T3|2018-11-02|
    MX2017002987A|2017-07-28|
    JP6257840B2|2018-01-10|
    KR101902136B1|2018-09-27|
    EP3192008A2|2017-07-19|
    MX359546B|2018-10-02|
    US20170053406A1|2017-02-23|
    RU2017111779A3|2018-10-10|
    CN107077602A|2017-08-18|
    WO2016040487A2|2016-03-17|
    AU2018247216A1|2018-11-01|
    WO2016040487A3|2016-04-28|
    AU2018247216B2|2019-11-14|
    EP3192008B1|2018-06-27|
    US10237459B2|2019-03-19|
    AU2015315156A1|2017-03-23|
    CN107077602B|2018-11-02|
    RU2017111779A|2018-10-10|
    BR112017004427A2|2017-12-05|
    KR20170052636A|2017-05-12|
    SG11201701855RA|2017-04-27|
    US9396537B2|2016-07-19|
    JP2017534102A|2017-11-16|
    CA2960397A1|2016-03-17|
    引用文献:
    公开号 | 申请日 | 公开日 | 申请人 | 专利标题
    US5329379A|1992-10-22|1994-07-12|International Business Machines Corporation|System and method of measuring fidelity of decompressed video signals and images|
    US6075905A|1996-07-17|2000-06-13|Sarnoff Corporation|Method and apparatus for mosaic image construction|
    KR100320188B1|1999-03-23|2002-01-10|구자홍|Forgery judgment method for iris recognition system|
    KR100407319B1|2001-04-11|2003-11-28|학교법인 인하학원|Face feature tracking method using block-matching algorithm based on extended cost function|
    US7440590B1|2002-05-21|2008-10-21|University Of Kentucky Research Foundation|System and technique for retrieving depth information about a surface by projecting a composite image of modulated light patterns|
    RU2309672C2|2005-10-05|2007-11-10|Общество с ограниченной ответственностью "Научно-производственное предприятие "Лазерные системы"|Method for identifying the belonging of fingerprint to alive or dead person|
    US8260008B2|2005-11-11|2012-09-04|Eyelock, Inc.|Methods for performing biometric recognition of a human eye and corroboration of same|
    CN101493884B|2008-01-24|2012-05-23|中国科学院自动化研究所|Multi-optical spectrum image collecting device and method|
    KR100949801B1|2008-04-17|2010-03-30|한국전자통신연구원|Apparatus and method for polynomial reconstruction in fuzzy vault system|
    US7508960B1|2008-05-06|2009-03-24|International Business Machines Corporation|Projection of light patterns for liveness verification of biometrics|
    JP5814799B2|2012-01-06|2015-11-17|キヤノン株式会社|Image processing apparatus and image processing method|
    US8542879B1|2012-06-26|2013-09-24|Google Inc.|Facial recognition|
    US10452894B2|2012-06-26|2019-10-22|Qualcomm Incorporated|Systems and method for facial verification|
    US8457367B1|2012-06-26|2013-06-04|Google Inc.|Facial recognition|
    JP5955133B2|2012-06-29|2016-07-20|セコム株式会社|Face image authentication device|
    CN102831400B|2012-07-31|2015-01-28|西北工业大学|Multispectral face identification method, and system thereof|
    KR101417433B1|2012-11-27|2014-07-08|현대자동차주식회사|User identification apparatus using movement of pupil and method thereof|RU2737509C2|2014-04-07|2020-12-01|Айверифай Инк.|User authentication biometric binding|
    US9934443B2|2015-03-31|2018-04-03|Daon Holdings Limited|Methods and systems for detecting head motion during an authentication transaction|
    CN107851182B|2015-06-16|2019-04-19|眼验股份有限公司|For palming off the system and method for detection and liveness analysis|
    CN107135348A|2016-02-26|2017-09-05|阿里巴巴集团控股有限公司|Recognition methods, device, mobile terminal and the camera of reference object|
    US10111103B2|2016-03-02|2018-10-23|EyeVerify Inc.|Spoof detection using proximity sensors|
    FR3053499B1|2016-06-29|2018-06-29|Safran Identity & Security|METHOD AND DEVICE FOR DETECTING FRAUD BY EXAMINING TWO SEPARATE FOCUSES DURING AUTOMATIC FACE RECOGNITION|
    US10204262B2|2017-01-11|2019-02-12|Microsoft Technology Licensing, Llc|Infrared imaging recognition enhanced by 3D verification|
    US10372974B2|2017-01-11|2019-08-06|Microsoft Technology Licensing, Llc|3D imaging recognition by stereo matching of RGB and infrared images|
    US10679083B2|2017-03-27|2020-06-09|Samsung Electronics Co., Ltd.|Liveness test method and apparatus|
    CN108694357A|2017-04-10|2018-10-23|北京旷视科技有限公司|Method, apparatus and computer storage media for In vivo detection|
    US10657401B2|2017-06-06|2020-05-19|Microsoft Technology Licensing, Llc|Biometric object spoof detection based on image intensity variations|
    CN113095124A|2017-06-07|2021-07-09|创新先进技术有限公司|Face living body detection method and device and electronic equipment|
    GB2563953A|2017-06-28|2019-01-02|Cirrus Logic Int Semiconductor Ltd|Detection of replay attack|
    GB201713697D0|2017-06-28|2017-10-11|Cirrus Logic Int Semiconductor Ltd|Magnetic detection of replay attack|
    GB201801528D0|2017-07-07|2018-03-14|Cirrus Logic Int Semiconductor Ltd|Method, apparatus and systems for biometric processes|
    CN107506687B|2017-07-17|2020-01-21|Oppo广东移动通信有限公司|Living body detection method and related product|
    US20190108551A1|2017-10-09|2019-04-11|Hampen Technology Corporation Limited|Method and apparatus for customer identification and tracking system|
    GB2567503A|2017-10-13|2019-04-17|Cirrus Logic Int Semiconductor Ltd|Analysing speech signals|
    GB201801664D0|2017-10-13|2018-03-21|Cirrus Logic Int Semiconductor Ltd|Detection of liveness|
    GB201801874D0|2017-10-13|2018-03-21|Cirrus Logic Int Semiconductor Ltd|Improving robustness of speech processing system against ultrasound and dolphin attacks|
    GB201801661D0|2017-10-13|2018-03-21|Cirrus Logic International Uk Ltd|Detection of liveness|
    CN107832712A|2017-11-13|2018-03-23|深圳前海微众银行股份有限公司|Biopsy method, device and computer-readable recording medium|
    GB201801659D0|2017-11-14|2018-03-21|Cirrus Logic Int Semiconductor Ltd|Detection of loudspeaker playback|
    CN109376592B|2018-09-10|2021-04-27|创新先进技术有限公司|Living body detection method, living body detection device, and computer-readable storage medium|
    US11023756B2|2018-10-26|2021-06-01|Advanced New Technologies Co., Ltd.|Spoof detection using iris images|
    US11272156B2|2019-02-15|2022-03-08|Analog Devices International Unlimited Company|Spatial correlation sampling in time-of-flight imaging|
    CN112269975A|2020-03-31|2021-01-26|周亚琴|Internet of things artificial intelligence face verification method and system and Internet of things cloud server|
    DE102020123927A1|2020-09-15|2022-03-17|Audi Aktiengesellschaft|Method for recognizing a driver of a motor vehicle and motor vehicle as being alive|
    法律状态:
    2020-05-19| B06U| Preliminary requirement: requests with searches performed by other patent offices: procedure suspended [chapter 6.21 patent gazette]|
    2020-12-01| B09A| Decision: intention to grant [chapter 9.1 patent gazette]|
    2021-01-12| B09W| Correction of the decision to grant [chapter 9.1.4 patent gazette]|Free format text: RETIFICA-SE O PARECER DE DEFERIMENTO NOTIFICADO NA RPI NO 2605 DE 08-12-2020. |
    2021-01-19| B16A| Patent or certificate of addition of invention granted [chapter 16.1 patent gazette]|Free format text: PRAZO DE VALIDADE: 20 (VINTE) ANOS CONTADOS A PARTIR DE 09/09/2015, OBSERVADAS AS CONDICOES LEGAIS. |
    优先权:
    申请号 | 申请日 | 专利标题
    US14/480,802|2014-09-09|
    US14/480,802|US9396537B2|2014-09-09|2014-09-09|Systems and methods for liveness analysis|
    PCT/US2015/049195|WO2016040487A2|2014-09-09|2015-09-09|Systems and methods for liveness analysis|
    [返回顶部]