JP6427248B2 - Authentication database management method, authentication database management device and authentication database management program - Google Patents

Description

  The present invention relates to partial image data of a user (for example, partial image data including a face portion; hereinafter referred to as “face image data”) extracted from image data captured by an imaging device such as a camera. The present invention relates to an authentication database management method, an authentication database management apparatus, and an authentication database management program, which are managed in association with user identification information (hereinafter referred to as "user ID") of the above.

  Conventionally, an ID card (Identity Card) such as a membership card or employee ID card has been issued to the user who is the authentication target person, and the user ID recorded in the ID card at the time of entering and leaving each user There is known a personal identification system for reading and personal identification. In addition, there is also known a personal identification technology that carries out personal identification using a fingerprint or the like of a user in order to cope with frauds such as forgery of the ID card.

  However, when using the above-described personal identification technology, it is necessary to stop the user and collect fingerprints and the like each time the user is authenticated, which is not preferable from the viewpoint of convenience of the user. For this reason, in recent years, face recognition technology that can perform personal identification without stopping the walking user has attracted attention. For example, in Patent Document 1, the feature amount of the user's face image data is registered in advance as dictionary information, and when the user while walking approaches the camera, the camera captures an image of the user's face part There is disclosed a technique of acquiring face image data, collating the acquired feature amount of the face image data of the user with the contents of the dictionary information, and performing face authentication of the user.

  However, according to the conventional face recognition system represented by this patent document 1, the face image data of the user must be registered in the face authentication database in advance, which places a burden on the user himself and the worker of the face authentication database. Result. This is because it is necessary for the user and the operator to register the face image data obtained by imaging the face portion in the face authentication database, and it is necessary for the operator to manually edit the face image data. In particular, the above-mentioned burden occurs periodically because the human face has a time-dependent nature.

  Therefore, in Patent Document 2, a card reader and a camera are installed near the entrance, and when an ID card to which a user ID is assigned is read by the card reader, the face of a person appearing on the camera. There is disclosed an entry / exit management device configured to automatically register image data in a face authentication database in association with a user ID. Using this patent document 2, if the face image data of the user is registered in association with the user ID only for the first time, the user can pass through the entrance without performing the reading operation of the ID card for the second and subsequent times. Further, since the face image data of the user is automatically registered in the face authentication database, the burden on the user and the worker can be reduced.

JP, 2009-104599, A JP 2005-146709 A

  However, when the prior art represented by the said patent document 2 is used, the problem of the misregistration which another person's face is misregistered arises. For example, when a plurality of persons appear in the camera when reading an ID card, face image data of a person different from the original user is associated with the user ID, which may cause misregistration.

  When the face image data thus erroneously registered is used at the time of authentication, it becomes a cause of the erroneous authentication and causes a decrease in the authentication accuracy. In order to avoid erroneous registration, it may be considered to make the conditions for registration strict, but when the conditions for registration are strict, the face image data to be registered decreases, so the construction of the database becomes slow. That is, the strictness of the registration condition is difficult to be uniquely determined because the construction speed of the database and the risk of misregistration are in a trade-off relationship.

  Furthermore, even under strict conditions, it is difficult to completely eliminate the misregistration of another person's face. Once an incorrect registration occurs, in the prior art, the face image data that has been incorrectly registered has to be manually deleted by the administrator, resulting in a heavy workload.

  From these things, when constructing an authentication database, it has been an important issue how to solve the problem of misregistration.

  The present invention has been made to solve the above-described problems of the prior art, and an authentication database management method, an authentication database management apparatus, and an authentication database management program that can solve the problem of misregistration in an authentication database. Intended to be provided.

  In order to solve the problems described above and achieve the object, the invention according to claim 1 is an authentication database that stores user identification information uniquely identifying a user and registered image data of the user in association with each other. An authentication database management method for managing, comprising: collating a user identification information acquiring step of acquiring the user identification information; input image data captured by an imaging device and registered image data stored in the authentication database And an image authentication step of outputting, as image authentication identification information, user identification information associated with registered image data similar to the input image data, and user identification information acquired by the user identification information acquisition step. Registration using the comparison result of certain acquired identification information and the image authentication identification information output in the image authentication step in the matching process And a database updating step of updating the authentication database by deleting the registered image data associated with the image data in association with image data and accumulating and when the accumulated comparison result satisfies a predetermined condition. I assume.

  Also, in the invention according to claim 2, in the invention according to claim 1, the database updating step compares the non-coincidence count indicating the number of times the acquired identification information and the image authentication identification information become non-coincident. As a result, the registered image data is stored in association with the registered image data, and the registered image data is deleted when the non-matching count satisfies the predetermined condition.

  Also, in the invention according to claim 3, in the invention according to claim 2, the database updating step includes a coincidence count indicating the number of times the acquired identification information and the image authentication identification information coincide. A condition which is further stored in association with each other and which defines a relationship between the match count and the mismatch count associated with the same registered image data is used as the predetermined condition.

  Also, in the invention according to claim 4, in the invention according to claim 3, the database updating step is performed when the nonconformity count associated with the same registered image data exceeds the coincidence count. It is characterized by deleting image data.

  Also, in the invention according to claim 5, in the invention according to claim 4, the database updating step is performed when the deleted registered image data is similar to registered image data associated with other user identification information. Is characterized by re-registering the deleted registered image data in association with the other user identification information.

  In the invention according to claim 6, in the invention according to any one of claims 1 to 5, the authentication database stores an image of a face portion of the user as registered image data, and the image The authentication step is characterized in that a partial image in which a face portion is detected from an image captured by the imaging device is used as the input image data.

  Also, in the invention according to claim 7, in the invention according to claim 6, the image authentication step is carried out in each of the input image data from a plurality of images continuously taken by the image pickup device of a moving person. It is characterized by detecting.

  The invention according to claim 8 is the invention according to any one of claims 1 to 7, wherein the user identification information acquiring step acquires and acquires the user identification information from a predetermined recording medium. It is characterized by

  The invention according to claim 9 is an authentication database management apparatus for managing an authentication database which stores user identification information uniquely identifying a user and registered image data of the user in association with each other. User identification information acquisition means for acquiring user identification information, input image data captured by an imaging device and registered image data stored in the authentication database are collated, and a registered image similar to the input image data Image authentication means for outputting user identification information associated with data as image authentication identification information, acquisition identification information which is user identification information acquired by the user identification information acquisition means, and the image authentication means The comparison result with the output image authentication identification information is stored in association with the registered image data used in the matching process, and the stored comparison When the result satisfies a predetermined condition, characterized by comprising a database update means for updating the authentication database by deleting the registration image data associated with.

  The invention according to claim 10 is an authentication database management program for managing an authentication database that stores user identification information uniquely identifying a user and registered image data of the user in association with each other. A user identification information acquisition procedure for acquiring user identification information, input image data captured by an imaging device, and registered image data stored in the authentication database are collated, and a registered image similar to the input image data An image authentication procedure that outputs user identification information associated with data as image authentication identification information, acquisition identification information that is user identification information acquired by the user identification information acquisition procedure, and the image authentication procedure The comparison result with the output image authentication identification information is stored in association with the registered image data used in the matching process, It was compared to the case where the result satisfies a predetermined condition, characterized in that to execute and the database update procedure to a computer to update the authentication database by deleting the registration image data associated with.

  According to the present invention, the user identification information is acquired, and the input image data captured by the imaging device and the registered image data stored in the authentication database are collated to generate registered image data similar to the input image data. The associated user identification information is output as image authentication identification information. Then, the comparison result of the acquired identification information, which is the acquired user identification information, and the image authentication identification information is stored in association with the registered image data used for the matching process, and the stored comparison result satisfies the predetermined condition. In this case, the authentication database is updated by deleting the associated registered image data. Such a configuration can eliminate the problem of misregistration.

FIG. 1 is an explanatory diagram for describing additional registration of face image data according to the first embodiment. FIG. 2 is a block diagram showing an internal configuration of the entry and exit management device according to the first embodiment. FIG. 3 is an explanatory diagram for explaining data stored in the storage unit of the room entry / exit management device shown in FIG. FIG. 4 is a flow chart showing the processing procedure of the additional registration by the entry and exit management device shown in FIG. FIG. 5 is a flow chart showing a processing procedure of gate control by the entry and exit management device shown in FIG. FIG. 6 is an explanatory diagram for describing deletion of misregistered registered face image data according to the second embodiment. FIG. 7 is a block diagram showing an internal configuration of the entry and exit management device according to the second embodiment. FIG. 8 is an explanatory diagram for explaining data stored in the storage unit of the entry and exit management device shown in FIG. 7. FIG. 9 is an explanatory diagram for describing a case where a deletion chain does not occur. FIG. 10 is an explanatory diagram for describing a case where a deletion sequence occurs. FIG. 11 is a flowchart showing a processing procedure of erroneous registration data deletion by the entry and exit management device shown in FIG.

  Hereinafter, preferred embodiments of an authentication database management method, an authentication database management apparatus, and an authentication database management program according to the present invention will be described in detail with reference to the attached drawings. In the first and second embodiments described below, the authentication database management method, authentication database management apparatus, and authentication database management program according to the present invention will be described as applied to a room entry and exit management apparatus.

  In the first embodiment, additional registration of face image data using reading of an ID from a card and image authentication will be described. FIG. 1 is an explanatory diagram for describing additional registration of face image data according to the first embodiment.

  First, the card reader 10 is installed in front of the gate device 30 to be subjected to the entry / exit management. Further, the camera 20 is installed at a position where it is possible to image a person approaching the gate device 30 and the card reader 10.

  As shown in FIG. 1, when the user registered in advance approaches the card reader 10 and holds the ID card over the card reader 10, the card reader 10 reads the user ID from the ID card. FIG. 1 shows the case where the read user ID is "0001".

  In addition, the camera 20 continuously captures the user approaching the card reader 10. The entry and exit management device 40 detects face images from a plurality of images with different imaging timings, using the output of the camera 20.

  The entry and exit management device 40 uses the detected face image as input face image data, and performs an authentication process using the face authentication database 42a. The face authentication database 42 a is a database in which the degree of maturity, registered face image data, and the like are associated with the user ID.

  The maturity level is an evaluation value obtained by evaluating the registration state of the registered face image data for the user ID. The maturity level is calculated using the registration record (the number of registrations, the number of registrations, etc.) of the registered face image data, the use record in authentication, and the like. For example, if the number of registrations of registered face image data is less than 10% of the upper limit number defined for the user ID, the degree of maturity "C", if 10% or more and less than 30%, the degree of maturity If it is "B" and 30% or more, the maturity level is "A". In the following description, the degree of maturity associated with the user ID is referred to as "the degree of maturity of the user ID".

  In the authentication process, each of the registered face image data registered in the face authentication database 42a is collated with the input face image data, and the user ID of the registered face image data having the highest similarity is used as the authentication ID.

  The room entry / exit management device 40 determines whether the user ID read by the card reader 10 and the authentication ID match, and whether to additionally register the input face image data as registration face image data based on the determination result Decide whether or not.

  The criteria for additional registration are defined by the registration processing table 42b. Specifically, if the maturity of the read user ID is "A", the read user ID and the authentication ID match, and the similarity at the time of authentication is a predetermined threshold (first registration Only in the case where it is the threshold value or more, the input face image data is additionally registered as the registered face image data of the user ID that has read the input face image data.

  If the read user ID has a maturity of "B", and if the read user ID and the authentication ID match, the user ID that read the input face image data regardless of the similarity at the time of authentication It is additionally registered as registration face image data of. On the other hand, when the read user ID and the authentication ID do not match, the input face image data is read only when the similarity at the time of authentication is less than a predetermined threshold (second registration threshold). It is additionally registered as registration face image data of the person ID. Here, in the case where the degree of similarity at the time of authentication is low, the registered face image data having the most similar input face image data corresponds to another user ID, but the degree of similarity is low. . When the maturity of the read user ID is "B", such input face image data is additionally registered. The second registration threshold may be the same as the first registration threshold. In the first embodiment, it is assumed that the second registration threshold and the first registration threshold are the same, and the case where the similarity exceeds the registration threshold is the similarity “high” and the similarity is the registration threshold or less. Let the similarity be "low".

  If the maturity level of the read user ID is "C", the input face image data regardless of whether the read user ID matches the authentication ID or regardless of the similarity at the time of authentication Is additionally registered as registered face image data of the user ID that has read.

  That is, according to the registration processing table 42b, when the degree of maturity is low, priority is given to loosening the conditions for additional registration and increasing registered face image data, and as the degree of maturity increases, the conditions for additional registration are tightened and authenticated. We try to carefully select data that contributes to the improvement of accuracy.

  In the example of FIG. 1, the user ID read by the card reader 10 is "0001", and the maturity degree of the user ID "0001" is "B". Therefore, if the authentication ID of the input face image data is "0001", the input face image data is additionally registered as registered face image data of the user ID "0001" regardless of the degree of similarity. Further, if the authentication ID of the input face image data is “0002” and the similarity is low, the input face image data is additionally registered as registered face image data of the user ID “0001”. If the authentication ID of the input face image data is “0002” and the degree of similarity is high, the additional registration of the registered face image data of the input face image data is not performed.

  As described above, the room entrance / exit management device 40 according to the first embodiment reads the user ID by the card reader 10 and obtains the authentication ID by image authentication, and compares the comparison result of the ID and the read user ID Since it is determined whether the input face image data is additionally registered depending on the degree, priority is given to increasing registered face image data when the degree of maturity is low, and misregistration is strictly prevented as the degree of maturity increases. It is possible to construct a high accuracy face recognition database early.

  Next, the internal configuration of the entry and exit management device 40 will be described. FIG. 2 is a block diagram showing an internal configuration of the entry and exit management device 40. As shown in FIG. As shown in FIG. 2, the entry and exit management device 40 is connected to the card reader 10, the camera 20 and the gate device 30. The room entry and exit management device 40 further includes an input / output unit 41, a storage unit 42, and a control unit 43 therein.

  The card reader 10 is a device for reading out a user ID from an ID card. As the ID card, any recording medium such as a magnetic card or a near field communication type card can be used regardless of the recording and reading method. Moreover, the recording medium of a shape other than a card may be used.

The camera 20 is an imaging device including a CCD (Charge Coupled Device) element or the like that images a user and outputs the image to the room access control device 40. The camera 20 is set in advance to an angle of view capable of capturing an image of the user's face on the way to the card reader 10 and the gate device 30. The gate device 30 opens and closes under the control of the entry and exit management device 40.

  The input / output unit 41 is an input / output device group such as a liquid crystal panel, a keyboard, and a mouse. The storage unit 42 is a storage device such as a hard disk drive or a non-volatile memory, and stores the face authentication database 42 a and the registration processing table 42 b.

  The control unit 43 is a control unit that entirely controls the entry and exit management device 40, and includes a user ID acquisition unit 43a, a face image detection unit 43b, an authentication processing unit 43c, a gate control unit 43d, a registration processing unit 43e, and a sorting processing unit It has 43f. In practice, programs corresponding to these functional units are stored in a ROM (Read Only Memory) or a non-volatile memory (not shown), and these programs are loaded into a central processing unit (CPU) and executed. The user ID acquisition unit 43a, the face image detection unit 43b, the authentication processing unit 43c, the gate control unit 43d, the registration processing unit 43e, and the selection processing unit 43f execute the corresponding processes.

  The user ID acquisition unit 43a is a processing unit that acquires a user ID read from the ID card by the card reader 10. The face image detection unit 43 b detects the face image from each of a plurality of images with different imaging timings, using the output of the camera 20. Specifically, the face image detection unit 43b detects a face image present in the image by shifting and matching the template using a template (mask pattern) indicating the outline of a general face of a person. In addition to such template matching, various known face detection algorithms (image processing techniques) can be used. The face image detection unit 43b outputs the detected face image as input face image data.

  The authentication processing unit 43 c collates the input face image data output from the face image detection unit 43 b with each of the registered face image data registered in the face authentication database 42 a to obtain registered face image data having the highest similarity. Output user ID as authentication ID.

  The gate control unit 43d instructs the gate device 30 to open the gate when either of the authentication by the ID card and the authentication by the face image is established. Specifically, when the user ID read by the user ID acquisition unit 43a is the user ID registered in the face authentication database 42a, the gate control unit 43d opens the gate to the gate device 30. To indicate. Further, the gate control unit 43d instructs the gate device 30 to open the gate also when the authentication ID is output from the authentication processing unit 43c and the similarity is equal to or higher than the authentication threshold.

  The registration processing unit 43e is a processing unit that performs additional registration processing according to the registration processing table 42b when the user ID is acquired by the user ID acquisition unit 43a and the input face image data is output by the face image detection unit 43b. is there.

  The selection processing unit 43f is a processing unit that selects registered face image data. More specifically, the sorting processing unit 43f deletes registered face image data for which a predetermined time or more has passed since the user ID whose maturity level is A is registered and which contributes little to authentication. Select registered face image data.

  Next, a specific example of data stored in the storage unit 42 of the entry and exit management device 40 will be described. FIG. 3 is an explanatory diagram for explaining data stored in the storage unit 42 of the entry and exit management device 40. As shown in FIG.

  As shown in FIG. 3, the face authentication database 42 a is a database in which the user ID is associated with maturity, registered face image data, registration date and time, number of times of authentication use, and cumulative authentication score.

  The registration date, the number of times of authentication use, and the cumulative authentication score are associated with each registered face image data. The registration date indicates the date when the registered face image data was registered. The number of times of authentication use indicates the number of times the degree of similarity with the registered face image data is maximized at the time of authentication. The cumulative authentication score is a value obtained by adding the degree of similarity each time the degree of similarity with the registered face image data is maximized at the time of authentication.

  The accumulated authentication score is a larger value as the number of times of authentication use is higher and the degree of similarity is higher, and therefore, it is an index of the degree of contribution to authentication. Furthermore, by dividing the accumulated authentication score by the number of times of authentication use, it is possible to obtain an average authentication score which is the average similarity when the similarity with the registered face image data is maximized. Based on the average authentication score, it is possible to evaluate the accuracy of authentication when the registered face image data is used.

  For example, when the sorting processing unit 43f sorts registered face image data, it is sufficient to delete registered face image data for which a predetermined time or more has passed from the registration date and which has a cumulative authentication score equal to or less than the sorting threshold. Instead of the accumulated authentication score, whether to delete may be determined by the average authentication score.

  Further, the cumulative authentication score or the average authentication score may be used to determine the degree of maturity. By considering the cumulative authentication score and the average authentication score, it is possible to determine the degree of maturity taking into consideration not only the number of registered face image data but also the quality thereof.

  The face authentication database 42a shown in FIG. 3 has a state in which the user ID "0001" is associated with maturity "B", registered face image data "G001001.jpg" and registered face image data "G001002.jpg". It shows. Further, the registration date and time of the registered face image data “G001001.jpg” is “2011/05/02 11:42”, the number of times of authentication use is “9”, and the cumulative authentication score is “540”. The registration date and time of the registered face image data "G001002.jpg" is "2011/05/02 11:42", the number of times of authentication use is "8", and the cumulative authentication score is "640".

  In the face authentication database 42a shown in FIG. 3, the user ID "0002" is associated with maturity "A", registered face image data "G002001.jpg" and registered face image data "G002002.jpg". It shows the state. Further, the registration date and time of the registered face image data “G002001.jpg” is “2011/10/10 15:10”, the number of times of authentication use is “4”, and the cumulative authentication score is “320”. The registration date and time of the registered face image data “G002002.jpg” is “2011/10/11 09:30”, the number of times of authentication use is “7”, and the cumulative authentication score is “630”.

  As shown in FIG. 3, the registration processing table 42b is a table for determining whether or not additional registration is possible according to the comparison result of the ID, the similarity of the authentication ID, and the maturity of the user ID. Here, as the maturity of the user ID, the maturity associated with the user ID read from the ID card by the card reader 10 is used. Specifically, if the maturity level of the user ID is "A", the input is performed only when the read user ID and the authentication ID match and the similarity at the time of authentication is "high". The face image data is additionally registered as registered face image data of the user ID read.

  If the maturity level of the user ID is "B", and if the read user ID and the authentication ID match, registration of the user ID that read the input face image data regardless of the similarity at the time of authentication Register additionally as face image data. On the other hand, when the read user ID and the authentication ID do not match, only when the similarity at the time of authentication is "low", it is registered face image data of the user ID that has read the input face image data. Register additionally.

  If the maturity level of the user ID is "C", the input face image data is read regardless of whether the read user ID matches the authentication ID or regardless of the similarity at the time of authentication The user ID is additionally registered as registered face image data of the user ID.

  Next, the processing procedure of the additional registration by the room entry / exit management device 40 will be described. FIG. 4 is a flowchart showing the processing procedure of the additional registration by the entry and exit management device 40. First, the user ID acquisition unit 43a acquires the user ID read from the ID card by the card reader 10 (step S101). Further, the user ID acquisition unit 43a acquires the degree of maturity of the acquired user ID from the face authentication database 42a (step S102).

  The face image detection unit 43b detects the face image from each of a plurality of images different in imaging timing using the output of the camera 20, and outputs the detected face image as input face image data (step S103). The registration processing unit 43e selects one unprocessed input face image data (step S104). The registration processing unit 43e determines whether the maturity degree of the user ID is "A" (step S105), and if the maturity degree is "A" (step S105; Yes), the process proceeds to step S106. .

  In step S106, the authentication processing unit 43c collates the input face image data with each of the registered face image data registered in the face authentication database 42a, and uses the user ID of the registered face image data having the highest similarity. Output as authentication ID.

  After step S106, the registration processing unit 43e determines whether the user ID acquired in step S101 matches the authentication ID (step S107). If the IDs match (step S107; Yes), it is determined whether the degree of similarity of the authentication ID obtained in step S106 is "high" (step S108). If the similarity is "high" (step S108; Yes), the input face image data is additionally registered as registered face image data of the user ID acquired in step S101 (step S109).

  If the maturity level is not "A" (step S105; No), the registration processing unit 43e determines whether the maturity level of the user ID is "B" (step S111), and the maturity level is "B". If it is "" (step S111; Yes), it will transfer to step S112.

  In step S112, the authentication processing unit 43c collates the input face image data with each of the registered face image data registered in the face authentication database 42a, and uses the user ID of the registered face image data having the highest similarity. Output as authentication ID.

  After step S112, the registration processing unit 43e determines whether the user ID acquired in step S101 matches the authentication ID (step S113). If the IDs match (step S113; Yes), the input face image data is additionally registered as registered face image data of the user ID acquired in step S101 (step S114). If the IDs do not match (step S113; No), the registration processing unit 43e determines whether the degree of similarity of the authentication ID obtained in step S112 is "high" (step S115). Then, if the similarity is not "high" (step S115; No), the input face image data is additionally registered as registered face image data of the user ID acquired in step S101 (step S114).

  If the maturity level is not "B" (step S111; No), that is, if the maturity level is "C", the registration processing unit 43e registers the input face image data in the registered face image data of the user ID acquired in step S101. Are additionally registered (step S116).

  After the input face image data is registered in step S109, step S114 or step S116, the process proceeds to step S110. If the IDs do not match in step S107 (step S107; No), if the similarity is "low" in step S108 (step S108; No), if the similarity is "high" in step S115 (step S108) Also in step S115, the process proceeds to step S110.

  In step S110, the registration processing unit 43e determines whether all input face image data has been processed. Then, when unprocessed input face image data remains (step S110; No), the process proceeds to step S104, and when all input face image data is processed (step S110; Yes), additional registration processing Finish.

  Next, a processing procedure of gate control by the entry and exit management device 40 will be described. FIG. 5 is a flow chart showing the processing procedure of gate control by the entry and exit management device 40. The gate control unit 43d determines whether the user ID acquisition unit 43a has acquired the user ID from the ID card (step S201).

  If the user ID acquisition unit 43a acquires the user ID (step S201; Yes), the gate control unit 43d determines that the acquired user ID is a user ID registered in the face authentication database 42a. It is verified whether or not it is (step S202). If the user ID is not registered (step S202; No), the gate control unit 43d outputs an authentication error to notify the user (step S203), and ends the gate control process. On the other hand, if it is a registered user ID (step S202; Yes), the gate control unit 43d instructs the gate apparatus 30 to open the gate (step S204), and ends the gate control process.

  If the user ID acquisition unit 43a does not acquire the user ID (step S201; No), the gate control unit 43d determines whether the face image detection unit 43b detects a face image from the output of the camera 20. (Step S205). If the face image detection unit 43b does not detect a face image (Step S205; No), the gate control unit 43d ends the gate control process.

  If the face image detection unit 43b detects a face image (step S205; Yes), the authentication processing unit 43c receives input face image data including the detected face image and a registered face image registered in the face authentication database 42a. The data is collated with each other, and the user ID of the registered face image data having the highest similarity is output as an authentication ID (step S206). Further, the authentication processing unit 43c determines whether the similarity of the registered face image data having the highest similarity is equal to or higher than the authentication threshold. If the similarity is equal to or higher than the authentication threshold, the authentication processing unit 43c determines that the authentication is successful. If the authentication is successful (step S207; Yes), the gate control unit 43d instructs the gate device 30 to open the gate (step S204), and ends the gate control process. If the authentication is not successful (Step S207; No), the gate control unit 43d ends the gate control process.

  As described above, the room entry / exit management device 40 according to the first embodiment reads the user ID by the card reader 10 and obtains the authentication ID by image authentication, and compares the ID comparison result with the read user ID. Since it is decided whether or not to additionally register the input face image data according to the degree of maturity of, the priority is given to increasing the registered face image data when the degree of maturity is low, and the misregistration is increased as the degree of maturity is higher. By strictly preventing this, a high accuracy face recognition database can be constructed early.

  In the second embodiment, deletion of misregistered registered face image data will be described. FIG. 6 is an explanatory diagram for describing deletion of misregistered registered face image data according to the second embodiment.

First, the card reader 10 is installed in front of the gate device 30 to be subjected to the entry / exit management. Further, the camera 20 is installed at a position where it is possible to image a person approaching the gate device 30 and the card reader 10.

  As shown in FIG. 6, when the user registered in advance approaches the card reader 10 and holds the ID card over the card reader 10, the card reader 10 reads the user ID from the ID card. FIG. 6 shows the case where the read user ID is "0001".

  In addition, the camera 20 continuously captures the user approaching the card reader 10. The entry and exit management device 140 detects the face image from each of a plurality of images with different imaging timings using the output of the camera 20.

  The entry and exit management device 140 uses the detected face image as input face image data, and performs an authentication process using the face authentication database 142a. The face authentication database 142a is a database in which registered face image data and the like are associated with the user ID. Further, a match count NT and a non-match count NF are associated with each registered face image data.

  In the authentication process, each of the registered face image data registered in the face authentication database 142a is collated with the input face image data, and the user ID of the registered face image data having the highest similarity is used as the authentication ID.

  The entry and exit management device 140 determines whether the user ID read by the card reader 10 matches the authentication ID, and if the IDs match, the similarity to the input face image data is the highest. The match count NT of the registered face image data is incremented by one. If the IDs do not match, one is added to the mismatch count NF of the registered face image data having the highest degree of similarity with the input face image data. Then, the registered face image data whose non-match count NF has become larger than the match count NT is deleted.

  It is shown that registered face image data having a large non-coincidence count is highly similar to registered face image data registered in another user ID, not the user ID of the registration destination. That is, it is highly likely that an inappropriate face image has been incorrectly registered. The registered face image data incorrectly registered in this manner causes a decrease in authentication accuracy. Therefore, registered face image data having a large non-coincidence count is deleted.

  In the example of FIG. 6, the user ID read by the card reader 10 is "0001". Therefore, if the authentication ID of the input face image data is "0001", the coincidence count NT of the registered face image data having the highest degree of similarity is incremented by one. If the authentication ID of the input face image data is "0002", then the non-coincidence count NF of the registered face image data having the highest degree of similarity is incremented by one.

  As described above, the room entry / exit management device 140 according to the second embodiment reads the user ID by the card reader 10 and obtains the authentication ID by image authentication, and the match count of registered face image data according to the comparison result of the ID. And the unmatched count. Then, since the registered face image data in which the non-match count NF is larger than the match count NT is deleted, the misregistered registered face image data can be automatically deleted, which contributes to the improvement of the authentication accuracy.

  Next, the internal configuration of the entry and exit management device 140 will be described. FIG. 7 is a block diagram showing an internal configuration of the entry and exit management device 140. As shown in FIG. As shown in FIG. 7, the entry and exit management device 140 is connected to the card reader 10, the camera 20 and the gate device 30. The room entry and exit management device 40 further includes an input / output unit 41, a storage unit 42, and a control unit 143 therein.

  The storage unit 42 is different from the first embodiment in that the face authentication database 142 a is stored instead of the face authentication database 42 a and the registration processing table 42 b. The control unit 143 is different from the first embodiment in that the misregistration deletion processing unit 143g is included instead of the registration processing unit 43e and the sorting processing unit 43f. The other configuration and operation are the same as those of the room entrance / exit management device 40 shown in the first embodiment, so the same components are attached with the same components and the description will be omitted.

  The face authentication database 142a is, as shown in FIG. 8, a database in which registered face image data, a match count NT and a mismatch count NF are associated with the user ID. Specifically, a state in which the registered face image data "G001001.jpg" and the registered face image data "G001002.jpg" are associated with the user ID "0001" is shown. The match count NT of the registered face image data “G001001.jpg” is “9”, and the mismatch count NF is “1”. The match count NT of the registered face image data “G001002.jpg” is “8”, and the mismatch count NF is “2”.

  The face authentication database 142a shown in FIG. 8 shows a state in which the registered face image data "G002001.jpg" and the registered face image data "G002002.jpg" are associated with the user ID "0002". The match count NT of the registered face image data “G002001.jpg” is “4”, and the mismatch count NF is “6”. The match count NT of the registered face image data “G002002.jpg” is “7”, and the mismatch count NF is “3”.

  The misregistration deletion processing unit 143g is a processing unit that deletes the registered face image data that has been misregistered. Specifically, when the user ID is read by the card reader 10 and the authentication ID is obtained by image authentication, the IDs are compared, and the match count NT and the mismatch count NF of the registered face image data are compared according to the comparison result. to add. Then, the registered face image data whose non-match count NF has become larger than the match count NT is deleted.

  Here, deletion of registered face image data will be further described. The misregistration deletion processing unit 143 g sets registered face image data in which the mismatch count NF is larger than the match count NT as deletion target data, and calculates the degree of similarity between the deletion target data and each user ID.

  Then, if the deletion target data is similar to a plurality of user IDs, the deletion target data is deleted from the face authentication database 142a. On the other hand, when the deletion target data indicates a high degree of similarity with a specific user, the deletion target data is moved to the user ID.

  For example, when any of the registered face image data of the user ID "0001" is the deletion target data, the degree of similarity with the other registered face image data in the user ID "0001" of the deletion target data is 40, Similarity to registered face image data of person ID “0002” is 20, similarity to registered face image data of user ID “0003” is 20, similar to registered face image data of user ID “0004” If is 20, the deletion target data is deleted from the face authentication database 142a because the degree of similarity with various users is high.

  On the other hand, if the similarity of the deletion target data with the other registered face image data in the user ID “0001” is 40 and the similarity with the registered face image data of the user ID “0002” is 60, then The degree of similarity with the user ID "0002" is higher than the user ID "0001", and is not similar to other users. Therefore, the deletion target data is deleted from the user ID "0001" and moved to the user ID "0002".

  9 and 10 are explanatory diagrams for explaining the deletion chain. Whether the erroneous registration deletion processing unit 143g deletes the deletion target data in series depending on how similar the plurality of registered face image data related to the user ID of the deletion source are to each other when deleting the deletion target data Decide whether or not.

  Specifically, the misregistration deletion processing unit 143g calculates the degree of similarity among the plurality of registered face image data associated with the same user ID, and the degree of similarity becomes equal to or higher than a predetermined value (for example, an authentication threshold). Link between registered face image data. Then, when there is isolated registered face image data due to the deletion of the deletion target data, the isolated registered face image data is also deleted. This is because isolated registered face image data is highly likely to be erroneously registered because the degree of similarity with the deletion target data is high.

  FIG. 9 shows the case where the deletion chain does not occur. Even in the case of registered face image data linked with deletion target data, if the registered face image data has a link with other registered face image data, only the deletion target data may be deleted. . It is because isolation of registration face image data does not occur by this deletion.

  Specifically, the registered face image data "G001003.jpg" is linked to the registered face image data "G001001.jpg" and the registered face image data "G001002.jpg". In addition, the registered face image data “G001001.jpg” and the registered face image data “G001002.jpg” are linked. Therefore, even if the registered face image data "G001003.jpg" is to be deleted, the registered face image data "G001001.jpg" and the registered face image data "G001002.jpg" are not isolated. Therefore, only the registered face image data "G001003.jpg" is deleted.

  Further, when there is only one link between the deletion target data and the other registered face image data, that is, the deletion target data is located at the end of the link relationship, only the deletion target data is deleted.

  Specifically, registered face image data "G001004.jpg" is linked to only registered face image data "G001001.jpg". Therefore, when the registered face image data "G001004.jpg" is used as the deletion target data, only the registered face image data "G001004.jpg" is deleted.

  FIG. 10 shows the case where a deletion chain occurs. As shown in FIG. 10, when there is registered face image data isolated in the same user ID due to deletion of deletion target data, deletion is also made in a chain with deletion of deletion target data. Do.

  Specifically, the registered face image data "G001001.jpg" is linked to the registered face image data "G001002.jpg", "G001003.jpg" and the registered face image data "G001004.jpg". The registered face image data “G001002.jpg” and the registered face image data “G001003.jpg” are linked. Assuming that the registered face image data "G001001.jpg" is to be deleted, the registered face image data "G001002.jpg" and the registered face image data "G001003.jpg" are not isolated, but the registered face image data "G001004.jpg" Be isolated. Therefore, the registered face image data "G001004.jpg" is also deleted in tandem with the deletion of the registered face image data "G001001.jpg".

  Next, a processing procedure of erroneous registration data deletion by the room entry and exit management device 140 will be described. FIG. 11 is a flowchart showing a processing procedure of erroneous registration data deletion by the entry / exit management device 140. First, the user ID acquisition unit 43a acquires the user ID read from the ID card by the card reader 10 (step S301).

  The face image detection unit 43b detects the face image from each of a plurality of images with different imaging timings using the output of the camera 20, and outputs it as input face image data (step S302). The misregistration deletion processing unit 143g selects one unprocessed input face image data (step S303). The authentication processing unit 43c collates the input face image data selected in step S303 with each of the registered face image data registered in the face authentication database 42a, and uses the registered face image data having the highest similarity. An authentication process is performed to output the ID as an authentication ID (step S304).

  The misregistration deletion processing unit 143g determines whether the user ID acquired in step S301 and the authentication ID match (step S305). If the IDs match (step S305; Yes), the misregistration deletion processing unit 143g adds one to the match count NT of the registered face image data having the highest similarity to the input face image data (step S306). .

  If the IDs do not match (step S305; No), the misregistration deletion processing unit 143g adds one to the mismatch count NF of the registered face image data having the highest similarity to the input face image data (step S308). . Then, it is determined whether the non-match count NF of the registered face image data has become larger than the match count NT (step S309).

  If the non-coincidence count NF becomes larger than the non-coincidence count NT (step S309; Yes), the misregistration deletion processing unit 143g deletes the registered face image data as deletion target data (step S310). Further, in step S310, it is determined whether the deletion target data is to be moved to another user ID or the necessity for deletion chaining, and movement or chaining is performed as necessary.

  After step S306, after step S310, or when the mismatch count NF is less than or equal to the match count NT (step S309; No), the false registration deletion processing unit 143g determines whether all input face image data has been processed. It is determined (step S307). Then, when unprocessed input face image data remains (step S307; No), the process proceeds to step S303, and when all input face image data is processed (step S307; Yes), erroneous registration data End the deletion process. The processing procedure of the gate control is the same as that of the first embodiment, so the description will be omitted.

  As described above, the room entrance and exit management device 140 according to the second embodiment reads the user ID by the card reader 10 and obtains the authentication ID by image authentication, and the registered face image data is obtained according to the comparison result of the IDs. The match count NT and the non-match count NF are added. Then, since the registered face image data in which the non-match count NF is larger than the match count NT is deleted, the misregistered registered face image data can be automatically deleted, which contributes to the improvement of the authentication accuracy.

  The second embodiment exemplifies the case where deletion of registered face image data is performed on the condition that the noncoincidence count NF is larger than the coincidence count NT, but the relationship between the noncoincidence count NF and the coincidence count NT is appropriately defined and deleted. It can be a condition of For example, the ratio of non-coincidence count to coincidence count may be defined as the condition for deletion. Further, a predetermined number of registered face image data whose match count is high in the same user ID may not be deleted.

  Although the first embodiment for describing the additional registration and the second embodiment for describing the deletion of the misregistration data are described separately for convenience of explanation, both of the additional registration and the deletion of the misregistration data are described. It goes without saying that the present invention may be implemented as a room entry and exit management device. In this case, in the entry and exit management device 40 disclosed in the first embodiment, the face authentication database 42a may further associate the match count NT and the non-match count NF, and the control unit 43 may further include the erroneous registration deletion processing unit 143g. .

  In the first and second embodiments, the case where the present invention is applied to the entry and exit management apparatus 40 and the entry and exit management apparatus 140 has been described. However, a program that causes a computer to execute each processing unit of the control unit 43 and the control unit 143 The invention may be practiced as

  In the first and second embodiments, the present invention is applied to the room entry and exit management system, but the present invention is not limited to this and can be applied to various systems such as attendance management system. . In this way, face authentication can be introduced into entry and exit management and attendance management. Also, it may be implemented as an apparatus or program that only manages the database without performing room entry / exit management or attendance management.

  In the first and second embodiments, the case where the image of the face portion is used for authentication has been described as an example, but the present invention is not limited to this, and partial images other than the face of a person, whole body The present invention is also applicable to the case where an image, an upper body image or the like is used for authentication.

  As described above, the authentication database management method, authentication database management apparatus, and authentication database management program according to the present invention are suitable for solving the problem of misregistration in the authentication database.

10 card reader 20 camera 30 gate device 40, 140 entry and exit management device 41 input / output unit 42 storage unit 42a, 142a face authentication database 42b registration processing table 43, 143 control unit 43a user ID acquisition unit 43b face image detection unit 43c authentication Processing unit 43d Gate control unit 43e Registration processing unit 43f Sorting processing unit 143g False registration deletion processing unit

Claims (10)

An authentication database management method for managing an authentication database that stores user identification information uniquely identifying a user and registered image data of the user in association with each other.
A user identification information acquisition step of acquiring the user identification information;
The input image data captured by the imaging device and the registered image data stored in the authentication database are collated, and user identification information associated with registered image data similar to the input image data is used as image authentication identification information. An image authentication step to be output as
The comparison result between the acquired identification information which is the user identification information acquired in the user identification information acquisition step and the image authentication identification information output in the image authentication step is associated with the registered image data used in the collation process Database updating step of updating the authentication database by deleting the associated registered image data when the stored comparison result satisfies a predetermined condition. Management method.   The database updating step stores a mismatch count indicating the number of times the acquired identification information and the image authentication identification information do not match in association with the registered image data as the comparison result, and the mismatch count satisfies the predetermined condition. The authentication database management method according to claim 1, wherein the registered image data is deleted when the following condition is satisfied.   The database updating step further associates and stores a match count indicating the number of times the acquired identification information and the image authentication identification information match with the registered image data, and stores the match count associated with the same registered image data. 3. The authentication database management method according to claim 2, wherein a condition that defines a relationship with the non-coincidence count is used as the predetermined condition.   4. The authentication database management method according to claim 3, wherein the database updating step deletes the registered image data when the mismatch count associated with the same registered image data exceeds the match count. .   In the database updating step, when the deleted registered image data is similar to the registered image data associated with other user identification information, the deleted registered image data is re-associated in association with the other user identification information. The authentication database management method according to claim 4, characterized in that registration is performed. The authentication database stores an image of a face portion of the user as registered image data,
The authentication database management according to any one of claims 1 to 5, wherein the image authentication step uses, as the input image data, a partial image obtained by detecting a face part from an image captured by the imaging device. Method.   7. The authentication database management method according to claim 6, wherein the image authentication step detects the input image data from each of a plurality of images continuously captured by the imaging device with a moving person.   The authentication database management method according to any one of claims 1 to 7, wherein the user identification information acquisition step reads and acquires the user identification information from a predetermined recording medium. An authentication database management apparatus that manages an authentication database that stores user identification information that uniquely identifies a user and registered image data of the user in association with each other.
User identification information acquisition means for acquiring the user identification information;
The input image data captured by the imaging device and the registered image data stored in the authentication database are collated, and user identification information associated with registered image data similar to the input image data is used as image authentication identification information. Image authentication means for outputting as
The comparison result between the acquired identification information which is the user identification information acquired by the user identification information acquisition means and the image authentication identification information outputted by the image authentication means is associated with the registered image data used in the collation process Database updating means for updating the authentication database by deleting the associated registered image data when the stored comparison results satisfy a predetermined condition. Management device. An authentication database management program that manages an authentication database that stores user identification information that uniquely identifies a user and registered image data of the user in association with each other.
A user identification information acquisition procedure for acquiring the user identification information;
The input image data captured by the imaging device and the registered image data stored in the authentication database are collated, and user identification information associated with registered image data similar to the input image data is used as image authentication identification information. Image authentication procedure to be output as
The comparison result between the acquired identification information which is the user identification information acquired by the user identification information acquisition procedure and the image authentication identification information output by the image authentication procedure is associated with the registered image data used in the collation process And causing the computer to execute a database update procedure for updating the authentication database by deleting the associated registered image data when the stored comparison result satisfies a predetermined condition. Authentication database management program.

Images