JP5871764B2 - Face recognition device - Google Patents

Description

  The present invention relates to a face authentication device that determines whether or not to verify a person to be verified by comparing face information of a person to be collated extracted from a captured image with face information of a registered user registered in advance.

  2. Description of the Related Art A face authentication apparatus that authenticates a person to be collated by collating a face image obtained by photographing the face of the person to be collated with a face image of a user registered in advance is used. The face authentication device evaluates the similarity between the face information, which is facial feature information obtained from the face image at the time of collation, and the face information obtained from the face image at the time of registration, and determines whether or not the person to be collated is a user. judge. This face information may be affected by the wearing object when the user or the person to be verified wears the wearing object such as glasses or accessories on the face.

  For example, if a user who was not wearing glasses at the time of registration is wearing glasses at the time of collation, the similarity between the registered face information and the face information at the time of collation will be low, and even though he is the person himself / herself. Authentication may fail (so-called principal rejection). In order to cope with this, for example, in Patent Document 1, a plurality of face image data is registered by photographing a face image of a person to be authenticated under a plurality of illumination conditions, photographing angles, and subject states (such as the presence or absence of glasses). Then, the degree of similarity between the registered data and the collation data representing the face image at the time of authentication of the person to be authenticated is calculated, and if there is a combination of the registration data and the collation data whose similarity is equal to or greater than a predetermined threshold, authentication is performed. A face authentication system for determining success has been proposed.

JP 2006-85289 A

  For each user, face information when the attachment is attached to the face and face information when the attachment is not attached are registered in advance, and the person to be collated is extracted from the captured image at the time of collation If the face information is similar to any of the face information, the success of the authentication can suppress the rejection of the person due to the difference in the wearing state of the wearing object at the time of registration and the matching. However, for example, if a person to be collated wears glasses at the time of collation, and a user other than the person to be collated wears glasses at the time of registration, both face information is similar due to the influence of the glasses There is. As a result, there is a possibility that authentication may succeed even though the person to be verified is not the user himself (so-called acceptance of others). In a face authentication device like patent document 1, since it will be attestation success also when face information of others is similar for a wearing thing, acceptance of others cannot be controlled.

  Therefore, the present invention reduces the possibility that the user himself / herself fails to authenticate due to the wearing object, or that the unregistered person to be verified is any user and that the authentication is erroneously successful. An object is to provide an authentication device.

  In order to solve such a problem, the present invention provides, for each registered user registered in advance, face information of a wearing state in which a wearing object is worn on a registered user's face and non-wearing in which no wearing object is worn. Storage means for storing the face information of the state as registered face information, the face information of the person to be collated extracted from the photographed image is collated with each of the plurality of registered face information, and the face of the person to be collated among the plurality of registered face information A collation means for identifying candidate face information similar to the face information, and when the candidate face information is the registered face information in the wearing state, the face information of the person to be collated is a non-registered user related to the registered face information in the wearing state. There is provided a face authentication device comprising supplementary determination means for performing supplementary determination for determining that a person to be collated is a registered user on condition that the registered face information in the wearing state is similar.

  When the candidate face information is registered face information in a non-wearing state, the supplementary determination unit preferably determines that the person to be collated is a registered user corresponding to the candidate face information without performing supplementary determination. .

  It is preferable that the similarity determination criterion used for supplementary determination by the supplemental determination unit is looser than the similarity determination criterion used for specifying candidate face information by the collating unit.

  The collating means calculates the similarity between the face information of the person to be collated and each of the plurality of registered face information, and specifies the registered face information that has the maximum similarity and satisfies the first criterion as candidate face information. When the candidate face information is the registered face information of the wearing state, the supplementary determination unit is similar to the registered face information of the non-wearing state related to the registered user corresponding to the registered face information of the wearing state and the face information of the person to be verified. It is preferable to determine that the person to be verified is a registered user when the degree satisfies the second standard that is looser than the first standard.

  When the candidate face information is the registered face information of the wearing state, the supplementary determining means performs the supplementary determination on the condition that the registered face information of the wearing state and the face information of the person to be collated are similar in a region where there is no wearing object. It is preferable to do.

  The face authentication apparatus according to the present invention reduces the possibility that the user himself / herself fails to authenticate due to the wearing object, or that the unregistered person to be verified is one of the users, and that the authentication is erroneously successful. be able to.

It is a schematic block diagram of the face authentication apparatus which is one Embodiment of this invention. It is a figure for demonstrating the registration data memorize | stored in a memory | storage part. It is the flowchart which showed the operation example of the process part of the face authentication apparatus. It is the flowchart which showed another operation example of the process part of the face authentication apparatus.

  Hereinafter, a face authentication apparatus according to an embodiment of the present invention will be described with reference to the drawings. In the face authentication apparatus according to the present embodiment, face information in a state where glasses are worn as an example of a wearing object (hereinafter referred to as glasses face) and face information in a state where glasses are not worn (hereinafter referred to as naked eye face). Are registered in advance for each user. The face authentication device collates the face information of the person to be collated with each of registered user face information (hereinafter referred to as registered face information), evaluates the similarity, and registers face information (candidate face) of the individual candidate. Information). The user corresponding to the registered face information is a candidate who may be the same person as the person to be verified. Further, the face authentication device, when the candidate face information is registered face information of a spectacle face, that is, the face information of the person to be collated is registered face information of the candidate's glasses face rather than the registered face information of the candidate's naked eye face. If they are similar, it is determined whether or not the registered face information of the naked eye face is similar to the face information of the person to be collated, and if the face information of the candidate is similar, the person to be collated is determined to be a user. By this supplementary determination, the face authentication apparatus of the present embodiment suppresses the acceptance of others due to the glasses.

  Further, in the face authentication apparatus according to the present embodiment, when the candidate face information is registered face information of the naked eye face, that is, the face information of the person to be collated is more than the registered face information of the candidate's glasses face. If it is similar to the registered face information of the face, it is determined that the person to be collated is a candidate without referring to the registered face information of the spectacle face. In this way, supplementary determination is not performed when a naked eye face with high reliability of matching is an individual candidate, and only when a relatively unreliable spectacle face is an individual candidate, A supplementary decision is made to confirm the similarity. Thereby, the acceptance of others when wearing glasses can be suppressed without increasing the person's rejection when the person to be verified is not wearing glasses.

  FIG. 1 is a schematic configuration diagram of a face authentication apparatus 1 according to an embodiment of the present invention. The face authentication device 1 includes a storage unit 20, a processing unit 30, and an output unit 40. In the face authentication device 1, an image captured by the imaging unit 10 is input to the processing unit 30. Although only one imaging unit 10 is shown in FIG. 1, the face authentication device 1 is connected to a plurality of imaging units and external devices such as automatic doors and electric locks, and these devices are connected according to the authentication result. You may control.

  The imaging unit 10 is installed so that the face of the person to be collated can be photographed according to the environment in which the face authentication device 1 is operated. For example, it is installed so that a predetermined range in front of an automatic door that restricts traffic is taken as an object to be photographed. The imaging unit 10 generates an image showing the face of the person to be verified as an input image. For this purpose, the imaging unit 10 includes a camera including an optical system that forms an image of the face of the person to be collated on a two-dimensional array of solid-state imaging elements such as a CCD or a CMOS. The imaging unit 10 outputs the generated input image to the processing unit 30.

  Note that the imaging unit 10 may create a color multi-tone image as an input image, or may be a camera that has a sensitivity in the near infrared region and creates a gray image. Moreover, it is preferable that the image pick-up element array which the image pick-up part 10 has has the pixel number of the grade which can distinguish the features, such as the eyes of the face, nose, and mouth which are reflected on the input image.

  The storage unit 20 includes at least one of a semiconductor memory, a magnetic recording medium, and an optical recording medium. The storage unit 20 stores a computer program, various parameters, data, and the like for controlling the face authentication device 1. Further, the storage unit 20 stores user identification information and face information that is feature information of the user's face extracted from the user's face image or face image. In the face authentication device 1, user identification information and face information are registered in advance in the storage unit 20 for each user. Hereinafter, the registered user face information is referred to as registered face information.

  For users who may wear glasses, the storage unit 20 stores the face information of the glasses face and the face information of the naked eye face in association with each user. These pieces of face information are an example of face information in a wearing state in which an attachment is attached to the user's face and face information in a non-wearing state in which no attachment is attached. For a user who does not wear glasses, the storage unit 20 stores face information of the naked eye face.

  FIG. 2 is a diagram for explaining registration data stored in the storage unit 20. The registered data in FIG. 2 has an ID number as identification information and a face image as registered face information for each user. In this example, there are four users. ID01, ID03 and ID04 users wear glasses, but ID02 users do not wear glasses. For this reason, for the users of ID01, ID03, and ID04, face images of both the spectacle face and the naked eye face are registered. On the other hand, only the face image of the naked eye face is registered for the user of ID02.

  In the example of FIG. 2, the glasses of the glasses face registered for the users ID01 and ID04 are similar. In this case, for example, a user with ID01 wearing glasses is a person to be collated, and when collated with each registered facial image, the user's glasses face with ID04 is more than the face of his own glasses. The similarity may be high. Even if the registered spectacle faces are not similar, for example, if the user with ID01 wears glasses different from those at the time of registration and becomes a person to be collated, the glasses will be used by other users. In the same manner, there is a possibility that the degree of similarity with the glasses face of other users is higher than the glasses face of oneself.

  The user identification information may be anything that can uniquely identify the user, and may be, for example, a user name. Further, FIG. 2 shows an example in which a face image is registered for the purpose of explanation. However, according to the method of collation by the processing unit 30, the face image or the feature information of the user's face extracted from the face image is registered. That's fine. The registered face information may be acquired from the outside, or may be extracted by the feature extraction unit 32 of the processing unit 30 described later.

  The processing unit 30 has one or a plurality of processors and their peripheral circuits. Then, the processing unit 30 extracts face information of the person to be collated from the face image input by the imaging unit 10, collates the face information with the face information of the user registered in advance in the storage unit 20, and It is determined whether or not the user is a registered user. For this purpose, the processing unit 30 includes a face detection unit 31, a feature extraction unit 32, a collation unit 33, a supplement determination unit 34, and an authorization unit 35. These units of the processing unit 30 are implemented as a functional module of a program that operates on the microprocessor unit, for example. These means may be constituted by independent integrated circuits, firmware, a microprocessor, and the like.

  The output unit 40 includes, for example, a communication interface connected to an automatic door, a passage gate, or an external device that controls an electric lock, and a control circuit thereof. And the output part 40 will output the signal which requests | requires the unlocking | release of an electric lock, for example to the connected apparatus, if the signal which shows the authentication success about a collation subject is received from the process part 30. FIG. Or the output part 40 may output the signal which shows the authentication result by the process part 30 to the display etc. which are installed in the vicinity of the imaging part 10, and notify an authentication result to a collation subject by a display or an audio | voice.

  Next, functions of each unit included in the processing unit 30 will be described.

  The face detection unit 31 detects a face area, which is an area where the face of the person to be collated is shown, from the input image. Therefore, for example, the face detection unit 31 calculates one or a plurality of feature amounts from a partial region in the input image, and if the feature amount satisfies a predetermined condition considered to correspond to a human face, A partial area is detected as a face area. Specifically, the face detection unit 31 performs edge pixel extraction using, for example, a Sobel filter, and calculates a feature quantity such as a direction distribution of edge pixels in the partial region or a luminance distribution near the edge. . Alternatively, the predetermined conditions described above may be determined in advance by obtaining the values of these feature values in advance from a plurality of images obtained by photographing a person's face and determining the range thereof.

  Alternatively, the face detection means 31 performs template matching with the input image using a face area template prepared in advance, or detects the edge distribution of the contour shape of the face or head area from the edge image of the input image. By doing so, the face area may be detected. For example, the face detection unit 31 passes an image obtained by extracting only pixels included in the face area from the input image to the feature extraction unit 32.

  The feature extraction unit 32 extracts face information (hereinafter also referred to as input face information) that is feature information of the face of the person to be collated from the image representing the face area. For this purpose, the feature extraction unit 32 extracts facial feature points such as the center of the eye region, the nose tip, and the mouth corner point from the image obtained by the face detection unit 31. Further, the feature extraction unit 32 extracts the extracted feature point type and position information on the input image (for example, a two-dimensional coordinate value with the upper left corner of the face image as the origin) as face feature information. The feature extraction unit 32 stores the input face information in the storage unit 20.

  The feature extraction means 32 can use various known methods for extracting feature points from a face image. For example, the feature extraction unit 32 performs edge extraction processing on the face image to extract edge pixels having a large luminance difference from surrounding pixels. Then, the feature extraction unit 32 checks whether or not the feature amount obtained based on the position and pattern of the edge pixel satisfies a predetermined condition for the part such as the eyes, the nose, and the mouth, and determines the position of each part. By specifying, each feature point can be extracted.

  The feature extraction unit 32 extracts face information when each user is registered by the same method as extracting the input face information, and stores them in the storage unit 20 as registered face information. .

  The collation means 33 collates the input face information of the person to be collated with each registered face information stored in the storage unit 20, and collates the user corresponding to the registered face information similar to the face information of the person to be collated. The candidate is identified as a candidate who may be the same person as the target person. For this purpose, the matching unit 33 first compares the input face information with each registered face information to calculate a similarity indicating the degree of similarity between the user's face and the face of the person to be verified for each user. To do. For a user who has registered both face information of a spectacle face and face information of a naked eye face, the collating unit 33 calculates both the similarity degree with the spectacle face and the similarity degree with the naked eye face. The matching unit 33 stores the calculated similarity in the storage unit 20 for each registered face information.

  For example, the collation unit 33 obtains individual similarity between the input face information and the registered face information by using the peripheral image features for each feature point extracted by the feature extraction unit 32, and calculates the sum of the total similarity. Calculate as degrees. In this case, the smaller the difference between the input face information and the registered face information, that is, the greater the similarity between the face of the person to be collated and the face of the user, the greater the similarity value.

  Alternatively, the matching unit 33 may obtain a position difference (distance) for each feature point of the same type between the input face information and the registered face information, and calculate the sum of squares as the similarity. In this case, the degree of similarity decreases as the difference between the input face information and the registered face information decreases, that is, as the face of the person to be collated and the face of the user are similar. In the following description, it is assumed that the higher the degree of similarity, the higher the degree of similarity between two faces to be compared.

  Alternatively, the matching unit 33 may calculate the similarity by pattern matching between the input image and a registered user face image (hereinafter referred to as a registered face image). In this case, the matching unit 33 calculates the normalized cross-correlation value between the input image and the registered face image while changing the relative position of the registered face image with respect to the input image, and the maximum value of the normalized cross-correlation value May be the similarity between the input image and its registered face image. In this case, since the matching unit 33 does not use the feature information extracted by the feature extraction unit 32 for calculating the similarity, the feature extraction unit 32 may be omitted.

  The collation means 33 obtains the maximum value of the similarity calculated for each registered face information, and when the maximum value is equal to or greater than a threshold value Th1 (an example of the first criterion), the registered face information having the maximum value is collated. The face information (candidate face information) of the person candidate similar to the face information of the subject person is determined. Thereafter, the user corresponding to the registered face information is treated as a candidate.

  The supplementary determination unit 34 determines whether or not supplementary determination is necessary based on whether the candidate face information specified as the principal candidate is the registered face information of the glasses face or the registered face information of the naked eye face. In the case of the registered face information of the spectacle face, since the reliability of the similarity determination by the matching unit 33 is low, as a supplementary determination, the registered face information of the naked eye face registered corresponding to the user is also checked. It is further evaluated whether or not it is similar to the person's face information, and it is determined whether or not the person to be collated is the user. When the most similar registered face information is a spectacle face, the similarity between the face of the person to be collated and the face of the candidate may have increased due to the similar characteristics of the spectacle image There is. For this reason, when the face information of the person candidate is the registered face information of the spectacle face, the supplement determination unit 34 performs the supplement determination to determine whether or not the person to be verified is really the same person as the candidate. Do.

  Since the matching unit 33 calculates the similarity for both the glasses face and the naked eye face, the supplementary determination unit 34 stores the similarity calculated by the matching unit 33 for the naked eye face corresponding to the glasses face of the person candidate. Get from and reference. When the similarity with respect to the naked-eye face is equal to or greater than a threshold Th2 (an example of the second criterion), the supplementary determination unit 34 determines that the collation target person is a user who is a candidate. On the other hand, when the similarity with respect to the naked-eye face is less than the threshold Th2, the supplementary determination unit 34 determines that the collation target person is not a user who is a candidate.

  On the other hand, when the face information of the person to be collated is similar to the registered face information of the naked eye face, the supplementary determination unit 34 considers that the authentication has been performed with high accuracy. In this case, the supplementary determination unit 34 determines that the person to be collated is a user who is a candidate without performing supplementary determination.

  The authorization unit 35 outputs a signal for unlocking the electric lock, for example, as a signal indicating successful authentication via the output unit 40 when the supplementary determination unit 34 determines that the person to be verified is a registered user. Or a display device installed in the vicinity of the imaging unit 10 is notified of the successful authentication. On the other hand, when the verification unit 33 determines that there is no candidate, or when the supplemental determination unit 34 determines that the verification target person is not a user, the authorization unit 35 performs an authentication failure via the output unit 40. A signal may be output.

  Here, the threshold value Th2 that is the similarity determination criterion used for the supplementary determination of the supplementary determination unit 34 is set lower (that is, the criterion is looser) than the threshold value Th1 that is the similarity determination criterion used by the matching unit 33. This is because the supplementary determination is an auxiliary to suppress the acceptance of others due to the similarity of the glasses. That is, in the supplementary determination, when it is determined that the person to be collated is similar to the spectacle face of a certain user, the auxiliary person is determined to be the same person on the condition that the person to be collated is somewhat similar to the naked eye face of the same user. It is because it is typical. Further, if the threshold value Th2 is set to be equal to or greater than the threshold value Th1, there is a high possibility that the person will be rejected due to the supplementary determination. For this reason, it is preferable to set threshold value Th2 lower than threshold value Th1. When the difference is used instead of the similarity, the threshold of the difference determined to be similar in the supplementary determination by the supplementary determination unit 34 is higher than the threshold of the difference determined to be similar by the matching unit 33. Is set.

  FIG. 3 is a flowchart illustrating an operation example of the processing unit 30 of the face authentication apparatus 1. First, the face detection unit 31 acquires an input image showing the face of the person to be collated from the imaging unit 10 (step S10). Then, the face detection unit 31 detects an input face area, which is an area where the face of the person to be collated is shown, from the input image (step S15). For example, the face detection unit 31 passes an image obtained by extracting only pixels included in the input face area from the input image to the feature extraction unit 32.

  The feature extraction unit 32 extracts input face information that is feature information of the face of the person to be collated from the image of the input face area (step S20). The feature extraction unit 32 passes the obtained input face information to the collation unit 33 and stores it in the storage unit 20. Note that when the matching unit 33 calculates the similarity by pattern matching between the image of the input face area and the registered face image, the process of step S20 may be omitted.

  The collation unit 33 calculates the similarity between the input face information and each registered face information for each user (step S25). For a user who has registered both the face information of the spectacle face and the face information of the naked eye face, the matching unit 33 determines the similarity between the input face information and the registered face information of the spectacle face, and the input face information and the naked eye face information. The degree of similarity with the registered face information is calculated. The matching unit 33 stores the calculated similarity in the storage unit 20.

  And the collation means 33 determines whether the maximum value of the calculated similarity degree is more than threshold value Th1 (step S30). If the maximum value is less than the threshold value Th1 (No in step S30), the collation unit 33 determines that there is no candidate, that is, the collation target person is not any user, and the authentication fails (step S55). .

  On the other hand, when the maximum value of the similarity is equal to or greater than the threshold value Th1 (Yes in step S30), the matching unit 33 uses the registered face information obtained as the maximum value as candidate face information, and the user corresponding to the registered face information Is a candidate who may be the same person as the person to be verified. Then, the supplementary determination unit 34 determines whether the candidate face information is registered face information of the candidate's naked eye face or registered face information of the spectacle face (step S35).

  When the face information of the person to be collated is similar to the registered face information of the naked eye face (No in step S35), the supplementary determination unit 34 regards that the authentication has been performed with high accuracy and does not perform supplementary determination. It is determined that the person is a candidate user. At this time, the authorization unit 35 outputs a signal indicating successful authentication via the output unit 40 (step S50).

  On the other hand, when the face information of the person to be collated is similar to the registered face information of the spectacle face (Yes in step S35), the supplementary determination unit 34 determines that supplementary determination is necessary, and determines the naked-eye face related to the candidate. The similarity calculated by the matching unit 33 for the registered face information is acquired from the storage unit 20 and referenced (step S40).

  Then, the supplementary determination unit 34 determines whether or not the similarity with respect to the naked eye face is equal to or greater than a threshold value Th2 (step S45). If the naked eye face similarity is equal to or greater than the threshold Th2 (Yes in step S45), the supplementary determination unit 34 determines that the person to be collated is a candidate user. At this time, the authorization unit 35 outputs a signal indicating successful authentication via the output unit 40 (step S50). On the other hand, if the degree of similarity of the naked eye face is less than the threshold Th2 (No in step S45), the supplementary determination unit 34 determines that the person to be verified is not a candidate user, and determines that the authentication has failed (step S55). Thus, the operation of the processing unit 30 ends.

  As another example of operation, when the candidate face information is a spectacle face, the supplementary determination unit 34 inputs the registered face information of the spectacle face and the input of the person to be collated regarding areas other than the peripheral area of the eye wearing the spectacles. Supplementary determination may be executed on the condition that the degree of similarity with face information is high to some extent. Below, the process of the process part 30 in this example is demonstrated.

  In this example, the storage unit 20 has a positional relationship between a peripheral region (referred to as a mounting region) of an eye that is worn with spectacles and is affected by the spectacles, and a region other than the peripheral region of the eye (referred to as a non-wearing region) in the face. (For example, the range of the face area image and the ratio in the vertical direction) are stored. Since each feature point constituting the face image is associated with position information on the image, it is possible to specify whether each feature point belongs to the mounting area or the non-wearing area.

  When the face information of the person to be collated is similar to the registered face information of the spectacle face, the supplementary determination unit 34 calculates the degree of similarity by increasing the weight of the portion of the non-wearing area of each face information. At this time, for example, the supplementary determination unit 34 determines whether the feature point is the wearing region or the non-wearing region according to the individual similarity between the feature point of the input face region and the feature point of the candidate's glasses face. The corresponding weight α is multiplied, and the total sum for each feature point is calculated as the overall weighted similarity.

  Further, the weight α is set such that the weight of the non-wearing area is set higher than the weight of the wearing area, for example, the weight of the non-wearing area is 1 and the weight of the wearing area is 0.3. The similarity of the area having a small influence is easily expressed in the weighted similarity. Note that the supplementary determination unit 34 may set the weight for the attachment area to 0 so that it is determined whether or not the non-attachment area is similar.

  The supplementary determination unit 34 performs the above-described supplemental determination when the weighted similarity is equal to or greater than the threshold Th3. That is, the supplementary determination unit 34 refers to the degree of similarity with the registered face information of the corresponding user's naked eye face, and determines whether the similarity with the naked eye face is equal to or greater than the threshold value Th2. It is determined whether or not is a user. On the other hand, when the weighted similarity is less than the threshold Th3, the supplementary determination unit 34 determines that the person to be collated is not the user without performing the above-described supplemental determination.

  FIG. 4 is a flowchart illustrating another operation example of the processing unit 30 of the face authentication apparatus 1. In the process shown in FIG. 4, it is determined whether the face information of the person to be collated in FIG. 3 is more similar to the registered face information of the candidate's glasses face than the registered face information of the candidate's naked eye face. A process using weighted similarity is added between step S35 and step S40 for referring to the face information of the naked eye face. Since the other processes are the same as those in FIG. 3, only the parts different from the processes shown in FIG. 3 will be described.

  If the candidate face information specified in the process of step S35 is a spectacle face, that is, if the registered face information that is most similar to the face information of the person to be collated is a spectacle face (Yes in step S35), supplementary determination means 34 calculates a weighted similarity between the input face information focusing on the similarity of the non-wearing area and the registered face information of the spectacle face (step S36). Then, the supplementary determination unit 34 determines whether or not the calculated weighted similarity is greater than or equal to the threshold Th3 (step S38).

  If the weighted similarity is equal to or greater than the threshold Th3 (Yes in step S38), the supplementary determination means 34 performs supplementary determination with reference to the similarity of the corresponding user's naked eye (after step S40). On the other hand, when the weighted similarity is less than the threshold value Th3 (No in step S38), the supplementary determination unit 34 determines that the person to be verified is not the user and determines that the authentication has failed (step S55). Since other processes are the same as those in FIG. 3, the description thereof is omitted.

  Note that the mounting area and the non-mounting area may not be set according to a predetermined positional relationship, but may be set dynamically for each feature point. For example, since the surroundings of the eyes are affected by the glasses, when the feature extraction unit 32 extracts the feature points of the eyes, a certain range centered on the feature points (for example, average spectacles in the face image) A horizontal belt-like range having a length in the vertical direction as a width may be set as the mounting area.

  As described above, the face authentication device according to the above-described embodiment is configured such that when face information of a registered user's glasses face is extracted as a candidate for the input face information of the person to be collated, the user's naked eye After confirming that the face information is similar to some extent, it is determined whether or not the person to be collated is the user. Accordingly, it is possible to suppress the acceptance of others due to the similarity of the glasses. Further, when face information of a registered user's naked eye face is extracted as a candidate for the person, it is determined that the person to be collated is the user without performing supplementary determination. That is, supplementary determination is made only when the input face information of the person to be collated is similar to the face information of the spectacle face. Thereby, the person's rejection when the verification target person is not wearing glasses can also be suppressed.

  Further, in the above description, the matching unit 33 identifies one user corresponding to the registered face information having the maximum similarity and the threshold value Th1 or more as a candidate. However, when there are a plurality of registered face information of the spectacle face whose similarity is equal to or greater than the threshold Th1, the supplementary determination means 34 may perform supplementary determination on these. In the supplementary determination in that case, the candidate having the maximum score obtained by combining both the similarity to the spectacle face and the sum of the similarity to the naked eye face is selected from the candidates, and the score is predetermined. When the above criteria are satisfied, the verification target person is determined as the candidate (user). Alternatively, the candidate having the maximum similarity of the naked eye face that is referred to at the time of supplementary determination may be selected as a determination target candidate.

  Moreover, the collation means 33 does not necessarily need to collate with all the users and collation target person's input face information. For example, only a group of users who are authorized to enter the restricted area may be collated. Alternatively, a priority is given to each user, the information is associated with each user and stored in the storage unit 20, and the collating unit 33 collates preferentially for a user with a high priority. Good.

  Further, the face authentication device 1 is not limited to authenticating a person to be collated with one input image, but may be one that finally authenticates a person to be collated using a plurality of input images having different shooting times. For example, the face authentication apparatus 1 extracts images of a person's face area from a plurality of input images taken at a plurality of times of a person walking in the monitoring area, and applies the known tracking technique to the same person. For the face areas that are determined to include the face of the face, the similarity between the face area image and each of a plurality of registered face images registered in advance is calculated in chronological order, and the similarity is used. It may be a face authentication device (so-called walk-through face authentication device) that authenticates a person walking in the monitoring area.

  In that case, for example, the face authentication apparatus 1 adds the number of first hits for the user whose similarity with the face information of the person to be collated at each point of time is the maximum, and ranks the first hit within the determination target period. Count the number for each user. Then, the registered face information obtained with the maximum similarity within a predetermined period is set as candidate face information, and when this is a spectacle face, the maximum value of the similarities with the naked eye face of the same user is used as supplementary determination. Check if it is above a predetermined threshold. When it is determined that the degree of similarity is equal to or greater than the threshold and is similar to the naked eye face to some extent, a first-rank hit ratio that is a ratio of the first-number hit number of the user to the number of input images in the period is calculated. If the first hit rate of the determination target user is greater than or equal to a threshold (for example, 70%), the face authentication device 1 determines that the verification target user is the user. Alternatively, when the number of first hits of a certain user is equal to or greater than the number of times (for example, seven times) that is a threshold, the face authentication device 1 may determine that the person to be verified is the user.

  Moreover, although the example in the case of wearing spectacles as an attachment has been described above, the attachment is not limited to spectacles. The wearing items include items worn by the user on the face such as accessories including piercings. For users wearing accessories such as pierced earrings, facial images or facial information with and without accessories are stored in the storage unit 20, and the processing unit 30 is described above for glasses. What is necessary is just to authenticate the person to be verified by performing the same process as the above.

  Further, the number of attachments is not limited to one, and there may be a plurality or a plurality of types. In that case, for example, as the registered face information, the face information in a state where one or a plurality of attachments are attached and the face information in a state where none of the attachments are attached are registered in the storage unit 20. Keep it. Further, in the above example in which the supplementary determination unit 34 calculates the weighted similarity, the positional relationship between the mounting region and the non-mounting region may be stored in the storage unit 20 for each type of mounting object.

  In the case where the problem of accepting another person is more important than the person's rejection, the supplementary determination may be performed even when the face information of the candidate person is determined to be similar to the face information of the person to be verified.

DESCRIPTION OF SYMBOLS 1 Face authentication apparatus 10 Imaging part 20 Memory | storage part 30 Processing part 31 Face detection means 32 Feature extraction means 33 Collation means 34 Supplementary determination means 35 Authorization means 40 Output part

Claims (5)

For each registered user registered in advance, the face information of a wearing state in which an attachment is attached to the face of the registered user and the face information in a non-wearing state in which the attachment is not attached are stored as registered face information. Storage means for
A collation that collates face information of a person to be collated extracted from a captured image with each of the plurality of registered face information, and identifies candidate face information similar to the face information of the person to be collated among the plurality of registered face information Means,
When the candidate face information is the registered face information in the wearing state, the face information of the person to be collated is similar to the registered face information in the non-wearing state related to the registered user corresponding to the registered face information in the wearing state. Supplementary determination means for performing supplementary determination to determine that the person to be collated is the registered user under the condition
A face authentication apparatus comprising:   When the candidate face information is the registered face information in the non-wearing state, the supplementary determination unit is a registered user corresponding to the candidate face information without performing the supplementary determination. The face authentication apparatus according to claim 1, wherein:   3. The face according to claim 1, wherein a similarity criterion used for the supplementary determination by the supplementary determination unit is looser than a similarity criterion used for specifying the candidate face information by the matching unit. 4. Authentication device. The collation means calculates the similarity between the face information of the person to be collated and each of the plurality of registered face information, and sets the registered face information that has the maximum similarity and satisfies the first criterion as the candidate face Identified as information,
When the candidate face information is the registered face information of the wearing state, the supplementary determination unit determines the registered face information of the non-wearing state and the matching target person's information regarding the registered user corresponding to the registered face information of the wearing state. 4. The face according to claim 3, wherein when the degree of similarity with face information satisfies a second criterion that is looser than the first criterion, the person to be collated is determined to be the registered user. 5. Authentication device.   When the candidate face information is the registered face information in the wearing state, the supplementary determination unit determines that the registered face information in the wearing state and the face information of the person to be collated are similar in a region where the wearing object is not present. The face authentication apparatus according to claim 1, wherein the supplementary determination is performed for a condition.

Images