JP5076563B2 - Face matching device - Google Patents

Description

  The present invention relates to a face matching device, and more particularly to a technique for preventing spoofing using a photograph.

  A face collation device applying face recognition technology has both “convenience” that does not require a special operation from the user and “reliability” by using biometric information (face). It is used for user authentication and building entrance security systems.

  In this type of face matching device, countermeasures against “spoofing” using photographs have been a problem. Spoofing is an act in which another person tries to authenticate using his / her face photo or the like.

  For example, if you drop a bag containing a mobile device or if it is stolen, there is a high possibility that a document with a photo of your face, such as a driver's license or student ID card, will be included there. If impersonation can be easily performed using those face photographs, there is a risk that the damage of unauthorized use will increase.

  As an idea for preventing impersonation, there is known a method for preventing impersonation by a photograph by requesting a user to perform a predetermined action and checking whether or not the user has performed a movement as requested (Patent Document 1). reference). However, this method imposes an operational burden on the user, so that “convenience” that is an advantage of the face recognition technology is lost.

  In addition, as other ideas, a method of photographing a plurality of times while changing the illumination conditions to check how the face is shaded (see Patent Document 2), or examining whether it is a three-dimensional by photographing from various angles The method is known. However, these methods require a plurality of illumination devices and a plurality of cameras or a camera moving mechanism, so that the apparatus configuration becomes large. In addition, it is necessary to change the lighting conditions and the shooting angle, so that many photos must be taken.

Although it does not prevent spoofing, as a related technology, in an automatic transaction apparatus that captures an image of a customer, the visual field of the camera is blocked when an area of the same color in the image is larger than a predetermined area. A method for determining that the peeling has occurred is known (see Patent Document 3).
JP 2004-259255 A Japanese Patent Laid-Open No. 11-339048 JP 2004-139447 A

  The present invention has been made in view of the above circumstances, and an object of the present invention is to provide a technique capable of preventing impersonation by photographs as much as possible without causing an increase in the size of the apparatus and a decrease in convenience. There is to do.

  In order to achieve the above object, the present invention adopts the following configuration.

The face collation apparatus according to the present invention includes an imaging unit, a face collation unit, and an impersonation inspection unit. At the time of face collation, the image capturing unit photographs the subject, and the face collation unit performs face collation using an image captured by the image capturing unit. However, in the present invention, the spoofing inspection means is a “spoofed image” obtained by taking a picture of the image (inspection target image) before, in parallel, or after face matching with the image. A process for checking whether there is an impersonation (spoofing process) is executed.

  The inspection result of the spoofing inspection means may be a simple determination result that “the inspection target image is a spoofed image / not a spoofed image” or “the possibility that the inspection target image is a spoofed image (or not a spoofed image)”. It may be a score (estimated value) shown. The former determination result can also be obtained by comparing the latter score with a predetermined determination condition. The inspection result of the impersonation inspection means is used for the purpose of preventing impersonation using a photograph.

  Specifically, when the image is determined to be an impersonation image by the impersonation inspection unit, face matching using the image may be prohibited. Further, if the image is determined to be a spoofed image by the spoofing inspection means after the face matching or in parallel, the result of the face matching by the image can be invalidated. This can be expected to prevent spoofing using photographs.

  Alternatively, it is also preferable that the face matching unit uses the inspection result of the spoofing inspection unit as a face matching parameter. For example, the face matching unit may use the score as a parameter so that the higher the possibility of being a spoofed image, the lower the success rate of face matching. This method can also be expected to prevent impersonation using photographs.

The spoofing inspection means inspects whether the inspection target image is a spoofed image based on the feature amount extracted from the inspection target image. As the feature amount, for example, “uniformity of color or / and pattern in the first inspection region set in a part other than the face in the image” can be preferably adopted. This feature value focuses on the fact that the background of the ID photo is often a solid color or a ground pattern. As the feature amount, may be employed to "presence or absence of characters in the second inspection area set in a portion other than the face in the image." This feature value is focused on the fact that the edges of a photograph appear as a figure in a spoofed image, and that figures (ruled lines, marks), characters, etc. existing in the vicinity of the photograph often appear. Further, as the feature amount, “a blur amount of the third inspection region set in the face portion in the image” may be employed. This feature amount is focused on the fact that a spoofed image tends to have a significantly larger amount of blur than a legitimate image (an image obtained by photographing a person). By using these feature amounts, it is possible to perform an impersonation inspection very easily.

  In the present invention, a spoofing inspection may be performed based on any one of the feature amounts, or a spoofing inspection may be performed based on a combination of two or more feature amounts. If a plurality of feature amounts are combined, the reliability of the inspection result is improved.

  When the face matching device includes a display unit for allowing a user to check an image used for face matching, an image excluding the first inspection region and the second inspection region is displayed on the display unit. It is preferable to do so. By hiding the first inspection region and the second inspection region, it is possible to prevent the user from noticing that the uniformity of the background portion and the presence or absence of figures and characters are being checked, and improve the reliability of the impersonation inspection. be able to.

The present invention can be understood as a face collation apparatus having at least a part of the above means. The present invention can also be understood as a face matching method including at least a part of the above processing, a program for realizing the method, or a recording medium recording the program. Each of the above means and processes can be combined with each other as much as possible to constitute the present invention.

  According to the present invention, since impersonation can be inspected from an image for verification, the apparatus is not increased in size and the convenience is not lowered. In addition, since it can be realized with a simple algorithm, it can be expected to speed up the spoofing inspection and face matching.

  Exemplary embodiments of the present invention will be described in detail below with reference to the drawings.

<Device configuration>
FIG. 1 is a block diagram showing a functional configuration of a face collation apparatus according to an embodiment of the present invention. This face collation device is a device that performs personal authentication or personal identification of a person to be collated using a face image. For example, a security device in a computer with a camera or a mobile phone, a monitoring device that performs intruder detection, entry / exit management It can be applied to various uses such as a device that controls a door or a door.

  The face collation apparatus includes an imaging unit 10, an image storage unit 11, a face detection unit 12, an impersonation inspection unit 13, a face collation unit 14, a display unit 15, and a face definition information storage unit 16. In this embodiment, these functional elements are realized by an arithmetic processing unit executing software (program) and controlling hardware resources such as a camera, a memory, and a display as necessary. However, these functional elements may be configured by a dedicated chip.

<Face matching function>
The function and processing flow of the face matching device will be described with reference to the flowchart of FIG.

  When the face collation function is activated, the imaging unit 10 performs photographing, and an image for use in face collation is captured (step S10). The input image is stored in the image storage unit 11 and displayed on the display unit 15. The user can confirm the verification image by looking at the display on the display unit 15.

  As the imaging unit 10, a digital camera including an optical system and an imaging device (CCD, CMOS sensor, etc.) can be preferably applied. The image storage unit 11 is a storage device that temporarily stores an image to be processed. As the storage device, any specific technique such as a volatile memory or a nonvolatile memory may be applied. As the display unit 15, any specific technique such as a liquid crystal display may be applied.

  Next, the face detection unit 12 detects a human face from the image and identifies the position and size of the face (step S11).

  Any existing technique may be applied to the face detection processing by the face detection unit 12. For example, (1) a method for detecting a face by template matching using a reference template corresponding to the contour of the entire face, (2) a face by template matching based on a facial organ (eg, eyes, nose, ears). (3) A technique for detecting a vertex such as a head by chroma key processing and detecting a face based on the vertex, and (4) detecting an area close to the skin color and detecting the area as a face. And (5) learning using a teacher signal using a neural network and detecting a face-like area as a face. Instead of automatically detecting a face, a person may specify the position and size of the face. When a plurality of human faces are detected from the image, the face to be processed may be determined based on predetermined criteria such as the size, orientation, and position of the face. Therefore, the face matching process may be interrupted.

  Next, the impersonation inspection unit 13 inspects whether the image obtained by photographing is a “spoofed image” (step S12). In the present embodiment, as an inspection result of the impersonation inspection unit 13, a determination result of “Is an impersonation image” or “Not an impersonation image” is obtained. A specific process of the spoofing inspection will be described later.

  If the impersonation inspection unit 13 determines that the image is an impersonation image (step S13; YES), the process ends immediately without performing face matching using the image. Thereby, the impersonation using the photograph can be prevented.

  On the other hand, when it is determined that the image is not a spoofed image (step S13; NO), face collation using the image is started.

  The face collation unit 14 first detects a plurality of feature points from the face detected by the face detection unit 12 (step S14). A feature point is a point that serves as an extraction criterion for a feature amount, and which point in the face is selected as a feature point is determined in advance. For example, a plurality of locations such as the center of the eye, the apex of the nose, and the end of the mouth are selected as the feature points.

  Any existing technique may be applied to the feature point detection process. For example, (1) a method for detecting a feature point by learning a pattern indicating the position of a feature point in advance and performing matching using the learned data; and (2) an edge on the inside of the detected face. There is a method of detecting end points of facial organs by performing detection and pattern matching, and detecting feature points based on the end points.

  Subsequently, the face matching unit 14 extracts a feature amount based on the detected feature point (step S15). Any kind of feature amount may be adopted. For example, the gray value near the feature point, its periodicity / direction, the positional relationship of the feature point, etc. can be adopted as the feature quantity. The number of feature quantities can be arbitrarily set according to the expected matching accuracy. In general, tens to hundreds of feature quantities are extracted. A set of feature quantities extracted from the collation image in this way is called a face feature quantity vector. It can be said that the face feature vector is a numerical expression of the facial feature. Just as the facial features differ from person to person, the tendency (direction, size) of the facial feature vector varies from person to person.

  The face matching unit 14 performs face matching by comparing the face feature vector obtained from the image with the face definition information of the registrant registered in the face definition information storage unit 16, and outputs the determination result. (Step S16). In the face definition information storage unit 16, face definition information of one or more registrants is registered in advance. It can be said that the face definition information defines the registrant's face (that is, the tendency of the face feature vector extracted from the registrant's face).

  Any existing technique may be applied to the comparison process between the face feature vector and the face definition information. For example, in the case of identity authentication, the similarity between the face feature vector of the image and the registrant's (person's) face definition information is calculated, and if the similarity is greater than a predetermined threshold, the identity is determined. Good. In the case of individual identification, the degree of similarity may be calculated for each of a plurality of registrants, and it may be determined which registrant the matching target person corresponds to based on the magnitude of the degree of similarity. The determination result of the face collation device is used for permission of operation of a computer or a mobile phone, detection of an intruder, entrance / exit management, door lock control, and the like.

<Impersonation inspection process>
Next, the spoofing inspection process in step S12 will be described in detail.

The face matching device of the present embodiment uses a face photo (certification photo) attached to a document (license, passport, student card, membership card, cash card, credit card, etc.) that requires personal authentication. The main purpose is to prevent it. Therefore, the present inventors have found the following characteristics by closely observing this type of document and an image obtained by photographing the document.

  ・ In many cases, the background of the ID photo is plain or ground. In an image obtained by taking such a photograph, the color and pattern of the background portion are almost uniform.

  ・ The ID photo is cropped on multiple sides (usually four sides) surrounding the face. Therefore, the edge of the photograph appears as a straight line in the image. In addition, graphics (ruled lines, marks, etc.) and characters printed on the document may be shown together in the image.

  ・ The size of the face in the photo is small. Therefore, it is difficult to focus on the face in the photograph when photographing it, and so-called out-of-focus image is often obtained. In addition, for impersonation purposes, enlargement must be performed so that the size is the same as that of a real face, which tends to result in a blurred image.

  ・ In the ID photo, the distance between the face and the background is relatively short, so the face and the background are often in focus. In an image obtained by taking such a photograph, the degree of blur between the face portion and the background portion is about the same. On the other hand, when a living body is photographed with a mobile phone or an entrance / exit monitoring device, the background is usually far away, so only the face portion is focused and the background portion is often blurred.

  In the face collation apparatus according to the present embodiment, an impersonation inspection is performed by evaluating the above-described features by image processing. The following five methods are illustrated as specific evaluation methods. In the following description, an image 2 as shown in FIG. 3 is assumed as the inspection target image. FIG. 3 is an example of an image obtained by photographing a license.

(1) Method for Evaluating Color Uniformity First, an inspection region 21 is set in a part other than the face in the image (a part corresponding to the background of the ID photo). The position, size, range, etc. of the examination area 21 may be determined based on the relative relationship with the face area 20 obtained in step S11. Then, the color uniformity in the inspection region 21 is evaluated. Examples of the feature amount corresponding to the color uniformity include a color or density value range, variance, and standard deviation of all pixels in the inspection area 21. The higher the color uniformity, the higher the possibility that the image 2 is a spoofed image.

  A specific example is shown in FIG. FIG. 9 is a histogram of the B component of the pixels in the examination area 21, (a) is an example of an image (spoofed image) obtained by photographing an identification photograph, and (b) is an image (biological image) obtained by photographing a living body. It is an example. In a spoofed image with a uniform background color, the range and dispersion (standard deviation) values are extremely small compared to a biological image. Therefore, if the value range, variance, standard deviation, etc. of the pixels in the inspection area 21 are smaller than a predetermined threshold value, it can be regarded as a “spoofed image”.

(2) Method for Evaluating Pattern Uniformity Similar to the method (1), the inspection area 21 is set, and the uniformity of the pattern in the inspection area 21 is evaluated. For example, by evaluating the power concentration in the low frequency region in the frequency space, it is possible to check whether the inspection region 21 includes a uniform (regular) pattern. It can be said that the higher the uniformity of the pattern, the higher the possibility that the image 2 is a spoofed image.

Also, pattern uniformity can be evaluated using a pattern matching technique. For example, as shown in FIG. 10, the inspection area 24 is set in a part other than the face in the image, and a part of the inspection area 24 is cut out as a master pattern 25. Then, the matching with the master pattern 25 is checked for a plurality of locations (target pattern 26) in the inspection region 24. If a predetermined number or more of matching locations are found, it is determined that the pattern is uniform and is “spoofed image”. As a pattern matching method, any algorithm such as a normalized correlation method may be used. Further, matching may be performed between images, and it is also preferable to perform matching between frequency distributions after frequency conversion of the master pattern 25 and the target pattern 26, respectively.

(3) Method for evaluating presence / absence of figure As shown in FIG. 3, an inspection region 22 is set in a portion other than a face in an image. The position / size / range of the inspection area 22 may be determined based on the relative relationship with the face area 20 obtained in step S11. Then, it is evaluated whether or not a graphic exists in the inspection area 22. Geometric figures such as edges and ruled lines of photographs can be detected by Hough transform. A figure having a known shape can also be detected by template matching. It is also preferable to evaluate the number, size, shape, and clarity of the figure. If there is a figure in the inspection area 22, it can be said that there is a high possibility that the image 2 is a spoofed image.

(4) Method for Evaluating Presence / Absence of Characters Similar to the method (3), the inspection area 22 is set, and whether or not there is a character in the inspection area 22 is evaluated. A character detection method such as template matching may be used, or a known character recognition method may be used. It is also preferable to evaluate the number, size, content, and clarity of characters. If there are characters in the examination area 22, it can be said that the image 2 is likely to be a spoofed image.

(5) Method for Evaluating Blur Amount As shown in FIG. 3, an inspection region 23 is set in the face portion in the image. The position / size / range of the inspection area 23 may be determined based on the relative relationship with the face area 20 obtained in step S11. Then, the amount of blur in the inspection area 23 is evaluated. For example, the deviation of the frequency component of the inspection region 23 toward the low region may be evaluated as a blur amount. It can be said that the larger the blur amount, the higher the possibility that the image 2 is a spoofed image. Specifically, a low frequency component and a high frequency component in a predetermined region are obtained from the frequency distribution (power spectrum) of the inspection region 23, and if the value of “low frequency component ÷ high frequency component” is larger than a predetermined threshold, “spoofing” It may be determined as “image”. Alternatively, a low frequency component threshold value LTh and a high frequency component threshold value HTh are set in advance, and the low frequency component value is larger than the threshold value LTh and the high frequency component value is set. If the frequency component is smaller than the threshold value HTh, it may be determined that the frequency component is biased to a low frequency range and determined as a “spoofed image”.

Other methods may be used for evaluating the blur amount of the image. For example, the blur amount can be evaluated by estimating a PSF (point spread function) parameter of the image. Further, a contrast detection method or a phase difference detection method used for autofocus of a digital camera may be applied to the evaluation of the blur amount.

  In the contrast detection method, the contrast of the image (typically, the dispersion of luminance values) can be evaluated as a blur amount. For example, if the contrast value of the inspection area 23 is smaller than a predetermined threshold value (if the variance value is larger than the predetermined threshold value), it can be considered that the face portion is blurred and can be determined as an “spoofed image”. .

The phase difference detection method is a method in which two images are generated by a separator lens from image information of a subject input through a lens, and the image interval is larger or smaller than an appropriate value (image interval in a focused state). This is a method for detecting the amount of deviation. When this phase difference detection method is applied, the difference amount between the image interval and the appropriate value can be evaluated as a blur amount. For example, in the case of a face matching device mounted on a mobile phone, an entrance / exit monitoring device, etc., the distance between the imaging unit and the living body (subject) is known in advance (about 50 to 60 cm). In general, the focal length is set as described above (the image interval at this distance is stored as an appropriate value). On the other hand, in the case of impersonation using a photograph, the photograph (subject) must be brought close to the imaging unit (for example, 10 cm or less), and the image interval at this time greatly deviates from an appropriate value. Therefore, if the amount of difference between the image interval of the inspection region 23 (face portion) and the appropriate value is larger than a predetermined threshold value, the face portion is considered to be blurred and can be determined as a “spoofed image”.

  In the above method, the “possibility of being an impersonation image” is evaluated by the amount of blur of the face portion, but it can also be evaluated by the difference between the amount of blur of the face portion and the amount of blur of the background portion. In the spoofed image using the ID photo, the blur amount of the face portion and the background portion is about the same (the difference is small), but in the image obtained by photographing the living body, the background amount is larger than the blur amount of the face portion. This is because there is a high probability that the amount of blur in the portion becomes very large (the difference is large). Specifically, the blur amount of the face portion (inspection region 23) and the blur amount of the background portion (inspection region 21) are calculated, and the difference between them is compared with a threshold value to determine whether the image is a “spoofed image”. What is necessary is just to determine.

  The impersonation inspection unit 13 performs an impersonation inspection process using at least one of the five evaluation methods. 4 to 7 show variations of the spoofing inspection process. In these figures, evaluation methods A, B, and C correspond to any of the above five evaluation methods.

  FIG. 4 is an example in which inspection is performed by a single evaluation method A. The impersonation inspection unit 13 evaluates the inspection target image using the evaluation method A, and calculates a score representing the possibility of being an impersonation image. When the score satisfies a predetermined determination condition, it is determined that the inspection target image is a spoofed image, and the result is output. According to this processing, since it is sufficient to implement a single evaluation method, the program size can be reduced and the required resources can be reduced.

  FIG. 5 is an example in which a plurality of evaluation methods A to C are combined in series for inspection. The impersonation inspection unit 13 first performs evaluation using the evaluation method A. The evaluation by the evaluation method B is executed only when it is determined that the image is not a spoofed image. If it is determined that the evaluation method B is not a spoofed image, the evaluation by the evaluation method C is executed. According to this process, the spoofing inspection process ends when it is determined to be a spoofed image by any of the evaluation methods, so that high-speed processing can be expected.

  FIG. 6 is an example in which inspection is performed by combining a plurality of evaluation methods A to C in parallel. The impersonation inspection unit 13 performs evaluation with each of the evaluation methods A to C, and combines the determination results to output a final result indicating whether the image is an impersonation image. For example, if it is determined that the image is a spoofed image by an evaluation method of 1/2 or more (in this example, two evaluation methods), a final result indicating that the image is a spoofed image may be output. Even if the reliability of each evaluation method is low, it is possible to perform a spoofing inspection with high accuracy by comprehensively judging the evaluation results. In particular, it is effective in suppressing overdetection (misidentifying an image that is not a spoofed image as a spoofed image).

FIG. 7 is also an example in which inspection is performed by combining a plurality of evaluation methods A to C in parallel. However, in the inspection of FIG. 6, the judgment results of the respective evaluation methods are combined, whereas in the inspection of FIG. 7, the scores of the respective evaluation methods are combined. For example, the score calculated by each evaluation method is substituted into one evaluation formula, the final score is calculated, and whether or not the final score satisfies a predetermined determination condition is determined as a spoofed image. This process can also improve reliability. In addition, high inspection accuracy can be expected by optimizing the evaluation formula.

  4 to 7, three evaluation methods are combined, but two or four or more evaluation methods may be combined. Moreover, you may employ | adopt combinations other than FIGS. 4-7.

  As described above, the face collation apparatus according to the present embodiment can inspect impersonation from an image photographed for collation by skillfully using the features found in the impersonation image. Since this method does not require special illumination and photographing as in the prior art, it does not cause an increase in the size of the apparatus or a decrease in convenience. In addition, since it can be realized by the very simple algorithm as described above, it is possible to expect an increase in the speed of spoofing inspection and face matching.

<Modification>
The above embodiment is merely an example of the present invention. The scope of the present invention is not limited to the above embodiment, and various modifications can be made within the scope of the technical idea. Hereinafter, preferable modifications will be described.

  (1) Instead of displaying the entire image for verification on the display unit 15, only a part of the image for verification may be displayed on the display unit 15. Specifically, as shown in FIG. 8, it is preferable that only the image 30 excluding the portion of the inspection area 22 for evaluating the presence or absence of graphics or characters is enlarged and displayed. In this way, by hiding the inspection area 22 from the user's eyes, it is possible to prevent the user from noticing the presence or absence of graphics / characters in the vicinity of the photo or in the vicinity of the photo, and improve the reliability of the impersonation test Can be improved. Further, the portion of the inspection area 21 for evaluating the uniformity of the color and pattern may be hidden.

  (2) The spoofing inspection unit 13 may pass the score of the evaluation method as the inspection result to the face matching unit 14, and the face matching unit 14 may use the score as a parameter for face matching. For example, the reliability parameter which normalizes a score and takes the value of 0-1 is produced | generated. The reliability parameter takes a value closer to 0 as the possibility of being a spoofed image is higher, and takes a value closer to 1 as the possibility of being a spoofed image is lower. Then, the degree of similarity between the face of the matching image and the registrant's face is multiplied by the reliability parameter so that the higher the possibility of being a spoofed image, the lower the success rate of face matching. Also by this method, it is possible to effectively prevent spoofing using a photograph, as in the above embodiment.

  (3) In the above embodiment, as shown in the flowchart of FIG. 2, the spoofing inspection process (steps S12 and S13) is executed before the face matching process (steps S14 to S16), but the order of these processes is changed. It may be performed in parallel. When face matching is executed first or when two processes are executed in parallel, it may be determined whether to adopt or invalidate the result of face matching based on the result of the spoofing inspection process.

FIG. 1 is a block diagram illustrating a functional configuration of the face matching device. FIG. 2 is a flowchart showing the flow of processing of the face matching device. FIG. 3 is a diagram showing an inspection area in the spoofing inspection process. FIG. 4 is an example of a determination chart of impersonation inspection processing. FIG. 5 is an example of a determination chart of impersonation inspection processing. FIG. 6 is an example of a determination chart of impersonation inspection processing. FIG. 7 is an example of a determination chart of impersonation inspection processing. FIG. 8 is a diagram illustrating an example of a display image excluding the portion of the inspection area. FIG. 9 is a diagram illustrating an example of a histogram of a spoofed image and a biological image. FIG. 10 is a diagram illustrating an example of pattern uniformity evaluation using a pattern matching method.

Explanation of symbols

2 images 10 imaging unit 11 image storage unit 12 face detection unit 13 impersonation inspection unit 14 face collation unit 15 display unit 16 face definition information storage unit 20 face area 21, 22, 23, 24 inspection area 25 master pattern 26 target pattern 30 display For images

Claims (17)

Imaging means;
Face collation means for performing face collation using an image photographed by the imaging means;
Spoofing that checks whether the image is a spoofed image obtained by taking a picture based on the uniformity of the color or / and pattern in the first inspection region set in a part other than the face in the image. An inspection means ,
The impersonation test means, said first color and / or face matching device the image higher uniformity Assess and likely a spoof image pattern in the examination region. Imaging means;
Face collation means for performing face collation using an image photographed by the imaging means;
Spoofing inspection means for inspecting whether the image is a spoofed image obtained by taking a picture based on the presence or absence of characters in a second inspection region set in a part other than the face in the image; Prepared ,
The spoofing inspection means, face matching device Assess and are likely to be the image spoofing image when there is a character in the second inspection area. Imaging means;
Face collation means for performing face collation using an image photographed by the imaging means;
Based on the amount of blur in the third inspection area set in the face portion in the image used in the face matching, it is checked whether the image used in the face matching is a spoofed image obtained by taking a picture. Spoofing inspection means ,
The impersonation test unit, the third inspection region face matching device as the image is larger blurring amount Assess and likely a spoof image. Imaging means;
Face collation means for performing face collation using an image photographed by the imaging means;
Color uniformity in the first inspection area set for a portion other than the face in the image, pattern uniformity in the first inspection region, and first portion set for a portion other than the face in the image. A combination of two or more of the presence / absence of a graphic in the second inspection area, the presence / absence of characters in the second inspection area, and the blur amount of the third inspection area set in the face portion in the image A spoofing inspection means for inspecting whether the image is a spoofed image obtained by taking a picture ,
The spoofing inspection means evaluates that the higher the color uniformity in the first inspection region, the higher the possibility that the image is a spoofed image, and the higher the uniformity of the pattern in the first inspection region. The third inspection is evaluated as having a high possibility that the image is a spoofed image, and when the character is present in the second inspection region, the image is highly likely to be a spoofed image. face matching device Assess and are likely to be the image spoofing image larger the amount of blurring of the region. It further comprises display means for allowing the user to check the image used for face matching,
The face collation apparatus according to claim 1, wherein an image excluding the portion of the first inspection area is displayed on the display unit. It further comprises display means for allowing the user to check the image used for face matching,
The face collation apparatus according to claim 2 or 4, wherein an image excluding the portion of the second inspection area is displayed on the display means.   The face collation apparatus according to claim 1, wherein face collation by the image is prohibited when the impersonation inspection unit determines that the image is a spoofed image.   The face matching apparatus according to claim 1, wherein when the image is determined to be a spoofed image by the spoofing inspection unit, a face matching result based on the image is invalidated.   The face matching apparatus according to claim 1, wherein the face matching unit uses an inspection result of the spoofing checking unit as a parameter for face matching. Based on the uniformity of the color or / and pattern in the first inspection region set in the portion other than the face in the image, the face matching device that performs face matching using the image taken by the imaging means, A face matching method including a spoofing inspection step for inspecting whether the image is a spoofed image obtained by taking a photograph ,
In the spoofing inspection step, a face matching method for evaluating that the higher the uniformity of the color or / and pattern in the first inspection region, the higher the possibility that the image is a spoofed image . A face collation apparatus that performs face collation using an image captured by an imaging unit, based on the presence or absence of characters in a second inspection region set in a portion other than the face in the image, the image is a photograph. A face matching method including a spoofing inspection step for inspecting whether the image is a spoofed image obtained by photographing ,
In the spoofing inspection step, a face matching method that evaluates that the image is likely to be a spoofed image when a character is present in the second inspection area . A face matching device that performs face matching using an image photographed by an imaging unit is configured to perform face matching based on a blur amount of a third inspection region set in a face portion in an image used in the face matching. A face matching method including a spoofing inspection step for inspecting whether an image used in the step is a spoofed image obtained by taking a picture ,
In the spoofing inspection step, a face matching method that evaluates that a possibility that the image is a spoofed image is higher as a blur amount of the third inspection region is larger . A face collation device that performs face collation using an image photographed by an imaging means, color uniformity in a first inspection region set in a portion other than the face in the image, and in the first inspection region Pattern uniformity, presence / absence of a graphic in the second inspection area set in a part other than the face in the image, presence / absence of characters in the second inspection area, and face part in the image in a third blur amount of the inspection area, based on two or more combinations of the face collation method wherein the image comprises a test step spoofing examining whether the resulting spoofing image obtained by photographing a photograph is set to There,
In the spoofing inspection step, the higher the color uniformity in the first inspection area, the higher the possibility that the image is a spoofed image, and the higher the uniformity of the pattern in the first inspection area. The third inspection is evaluated as having a high possibility that the image is a spoofed image, and when the character is present in the second inspection region, the image is highly likely to be a spoofed image. A face matching method for evaluating that the larger the amount of blur in a region, the higher the possibility that the image is a spoofed image . Based on the uniformity of the color or / and pattern in the first inspection area set in the portion other than the face in the image, in the face matching device that performs face matching using the image taken by the imaging means, A program for executing an impersonation inspection process for inspecting whether the image is an impersonation image obtained by taking a photograph ,
In the spoofing inspection process, a program for evaluating that the higher the uniformity of the color or / and pattern in the first inspection area, the higher the possibility that the image is a spoofed image . Based on the presence or absence of characters in the second inspection region set in a portion other than the face in the image, a face verification device that performs face verification using the image captured by the imaging unit captures a photograph. A program for executing an impersonation inspection process for inspecting whether an image is an impersonation image obtained by photographing ,
In the spoofing inspection process, a program that evaluates that the image is likely to be a spoofed image when a character exists in the second inspection area . In a face matching device that performs face matching using an image captured by an imaging unit, the face matching is performed based on a blur amount of a third inspection region set in a face portion in an image used in the face matching. A program for executing an impersonation inspection process for inspecting whether an image used in is an impersonation image obtained by taking a photograph ,
In the spoofing inspection process, a program that evaluates that the possibility that the image is a spoofed image is higher as the blur amount of the third inspection region is larger . In a face collation apparatus that performs face collation using an image photographed by an imaging means, color uniformity in a first inspection region set in a portion other than the face in the image, and in the first inspection region Pattern uniformity, presence / absence of a graphic in the second inspection area set in a part other than the face in the image, presence / absence of characters in the second inspection area, and face part in the image based on the third blur amount of the inspection area, two or more combinations of the set in, met the program to execute the inspection process spoofing the image is examined whether the obtained spoofing image obtained by photographing a photograph And
In the spoofing inspection process, the higher the color uniformity in the first inspection region, the higher the possibility that the image is a spoofed image, and the higher the uniformity of the pattern in the first inspection region. The third inspection is evaluated as having a high possibility that the image is a spoofed image, and when the character is present in the second inspection region, the image is highly likely to be a spoofed image. A program for evaluating that the larger the amount of blur in a region, the higher the possibility that the image is a spoofed image .

Images