JP2012133411A - Face collation system, face collation device and face collation method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To reduce a processing load on a face collation device for collating a face template with input face image data obtained by many monitoring cameras, and to efficiently suppress the leakage of the face template in the collation processing.SOLUTION: A primary collation device 30 placed in each area collates input face image data imaged by cameras 10 and detected by face detection devices 20 with an average face template 32a. When the collation result is equal to or higher than a predetermined threshold, the primary collation device transmits the input face image data to a secondary collation device 60. The secondary collation device collates the input face image data with a face template.

Description

  In the present invention, when input face image data is acquired by a large number of surveillance cameras and this input face image data is collated with a face template, the face template is leaked while reducing the processing load of the apparatus that performs the collation process. The present invention relates to a face matching system, a face matching device, and a face matching method that are efficiently suppressed.

  Conventionally, reference data (hereinafter referred to as “face template”) of the face image of the person to be monitored is registered in advance, and the face image data of the person (hereinafter referred to as “input face image data”) from the video captured by the surveillance camera And the input face image data thus cut out is collated with a face template.

  Here, as the number of monitoring cameras connected to the face image matching device that performs face image matching processing increases, the processing load on the face image matching device increases, so the processing load on the face image matching device is reduced. Prior art has been proposed. For example, in Patent Document 1, a plurality of surveillance cameras are connected to a computer, a face template is registered in the surveillance camera, and input face image data is transmitted to the computer when the matching process is successful. A technique for reducing the computer labor required for monitoring is disclosed.

  Patent Document 2 discloses a technique for performing high-speed and high-precision collation processing by putting an index into a video frame that needs to be processed on the monitoring camera side and processing the indexed image on the center side. ing. Furthermore, in Patent Document 3, when a person authenticated from an image acquired by a surveillance camera is associated with a supervisor, privacy is protected by publishing the person's image to the supervisor. Techniques that can be used are disclosed.

JP 2007-158421 A JP 2007-233495 A JP 2007-318333 A

  However, the thing of the said patent document 1 may leak the face template registered in the surveillance camera, when a surveillance camera is installed in places other than a security area. Here, when the face template is feature amount data, a third party cannot see the face itself at first glance, but such feature amount data is still data that can identify an individual. Leakage to a third party is a problem. In particular, the risk of such leakage increases as the number of surveillance cameras installed increases. When it is necessary to strictly manage the face template of the person to be monitored, the above-mentioned Patent Document 1 is insufficient.

  Moreover, although the thing of the said patent document 2 does not have the risk of the leak of the said face template, since the face collation process will be finally performed by the center side, if the number of the monitoring cameras installed increases, the processing load of a center Will increase. In particular, when using surveillance cameras installed in stations, airports, and streets, thousands to tens of thousands of surveillance cameras are targeted. The processing load is enormous.

  Furthermore, according to the above-mentioned Patent Document 3, privacy can be protected by restricting the public images provided to the observer, but when the number of surveillance cameras connected to the network increases, images acquired by the surveillance cameras There is a problem that the processing load at the center that performs the verification processing increases.

  For these reasons, how to reduce the processing load of the face matching device that performs the matching process when the input face image data is acquired by a large number of surveillance cameras and this input face image data is matched with the face template. How to prevent template leakage is an important issue.

  The present invention has been made in order to solve the above-described problems. When input face image data is acquired by a number of surveillance cameras and the input face image data is checked against a face template, matching processing is performed. An object of the present invention is to provide a face matching system, a face matching device, and a face matching method that can efficiently suppress leakage of a face template while reducing the processing load of the face matching device to be performed.

  In order to solve the above-described problems and achieve the object, the present invention is a face matching system that performs face matching processing on face templates of a plurality of monitoring subjects registered in advance and input face image data captured by an imaging device. A template generation device that generates a plurality of face templates and a combined face template based on face image data of a plurality of monitoring subjects, a plurality of imaging devices respectively disposed at a plurality of bases, and an image captured by the imaging device A primary collation device that performs a collation process on the input face image data thus generated with the composite face template generated by the template generation device, and transmits input face image data having a collation rate equal to or higher than a predetermined value; A secondary collation device for collating input face image data received from the collation device with the face template generated by the template generation device Characterized by comprising and.

  According to the present invention, a plurality of face templates and a combined face template are generated based on face image data of a plurality of monitoring subjects, and input face image data captured by one or a plurality of imaging devices is combined with a combined face template and 1 Since the collation process is performed by the secondary collation device and the input face image data having a collation rate equal to or higher than a predetermined value is collated by the face template and the secondary collation device, the processing load of the secondary collation device is reduced. It is possible to efficiently suppress the leakage of the face template in the next verification device.

FIG. 1 is a block diagram illustrating a system configuration of the face matching system according to the present embodiment. FIG. 2 is a functional block diagram showing an internal configuration of the operation management apparatus shown in FIG. FIG. 3 is a diagram illustrating an example of the face template and the average face template illustrated in FIG. FIG. 4 is a functional block diagram showing an internal configuration of the primary collation apparatus shown in FIG. FIG. 5 is an explanatory diagram for explaining automatic threshold setting. FIG. 6 is a functional block diagram showing an internal configuration of the secondary verification apparatus shown in FIG. FIG. 7 is a functional block diagram showing an internal configuration of the face detection apparatus shown in FIG. FIG. 8 is a flowchart showing a registration process procedure of the face template and the average face template by the operation management apparatus shown in FIG. FIG. 9 is a flowchart showing a processing procedure until the input face image data is displayed through the collation processing by the primary collation device and the secondary collation device shown in FIG. FIG. 10 is a block diagram illustrating another example of the operation management apparatus.

  Exemplary embodiments of a face matching system, a face matching device, and a face matching method according to the present invention will be described below in detail with reference to the accompanying drawings. In this embodiment, a case where an average face template is used as an example of a synthetic face template will be described. In the present embodiment, for convenience of explanation, a case where a simple configuration using four cameras, two primary collation devices, and one secondary collation device will be described. The present invention is not limited to this, and can also be applied to a case where a large number of cameras are accommodated.

  First, the system configuration of the face matching system according to the present embodiment will be described. FIG. 1 is a block diagram illustrating a system configuration of the face matching system according to the present embodiment. As shown in the figure, in this face collation system, the operation management device 50 and the secondary collation device 60 disposed in the center, and the primary collation device 30a and the base B disposed in the base A are disposed. The primary verification device 30 b is connected via the network 40. The primary verification device 30a is connected to the cameras 10a and 10b via the face detection devices 20a and 20b, and the primary verification device 30b is connected to the cameras 10c and 10d via the face detection devices 20c and 20d. Yes.

  Here, such a center is an area where security measures such as entrance / exit management are strictly taken, and unauthorized removal of face templates by unauthorized persons is suppressed. On the other hand, the bases A and B are areas where security measures are not taken as much as the center. The cameras 10a to 10d are collectively referred to as the camera 10, the face detection devices 20a to 20d are collectively referred to as the face detection device 20, and the primary verification devices 30a to 30b are collectively referred to as the primary verification device 30. The cameras 10a to 10d, the face detection devices 20a to 20d, and the primary verification devices 30a to 30b are devices having the same functions.

  The camera 10 is a camera composed of a CCD element (Charge Coupled Device Image Sensor) that can capture a moving image. The camera 10 is provided with a control mechanism for automatically controlling the mounting angle, and the angle of view and the like are controlled by a control signal from the face detection device 20.

  The face detection device 20 is a device that detects a human face image from the moving image data received from the camera 10 and transmits the detected face image data to the primary verification device 30. When face image data is detected from moving image data, still image data is sequentially extracted from the moving image data, and whether or not a human face portion exists in the extracted still image is determined by a template matching technique or the like. If it is determined that a human face portion exists, partial image data of the face portion (hereinafter referred to as “input face image data”) is cut out from the still image data and transmitted to the primary verification device 30. .

  Further, the face detection apparatus 20 registers in advance the reference coordinate position of the reference object that should exist at a predetermined position, and the reference object coordinate position and the reference coordinate position that exist in the still image data extracted from the moving image data. If there is a positional deviation greater than or equal to a predetermined value, a control signal for calculating the positional deviation amount and correcting the positional deviation amount is transmitted to the control mechanism of the camera 10. The control mechanism that has received this control signal automatically changes the angle of view of the camera 10. In this way, by automatically changing the angle of view of the camera 10 or the like, when the camera 10 is operated to look at another direction, or when the direction of the camera 10 is shifted due to an object hitting, It is possible to return to the direction to be monitored.

  The primary verification device 30 stores an average face template 32a, which is transmitted from the operation management device 50 and averages face image data of a plurality of monitoring subjects, in the storage unit 32 in advance, and is detected by the face detection device 20. A collation value between the face image data and the average face template 32a stored in the storage unit 32 is calculated, and when the collation value is equal to or greater than a predetermined threshold value, the input face image data is sent to the secondary collation device 60. A processing unit for transmission. The collation value becomes higher as the input face image data is closer to the average face template 32a, and can be calculated using a conventional technique such as a cross-correlation value.

  As described above, the primary collation device 30 stores not the face image data of a plurality of monitoring subjects themselves but the average face template 32a obtained by averaging the face image data of the monitoring subjects in the storage unit 32. Therefore, even when the primary verification device 30 is placed in an area where security measures are insufficient, it is possible to avoid a situation in which the face image data of the monitoring subject itself leaks.

  In addition, the primary verification device 30 images a non-monitoring target person that does not include a monitoring target person who passes in front of the camera 10 as a learning period when the camera 10 is installed or when the operation of the camera 10 is started. A collation value between the input face image data received from each camera 10 via the face detection device 20 and the average face template 32a is calculated, and a threshold value is automatically set based on the distribution of the collation values. Thereby, after learning is completed, the optimal threshold value for every camera 10 can be automatically set.

  The secondary collation device 60 stores a plurality of face templates 62 a composed of face image data of a plurality of monitoring subjects in advance in the storage unit 62 and received from the primary collation device 30. When a collation value between the image data and the face template 62a stored in the storage unit 62 is calculated and the collation value is equal to or greater than a predetermined threshold value, the monitoring target person corresponding to the input face image data and the face template 62a Is a processing unit that transmits the identification information to the operation management apparatus 50.

  As described above, since the secondary collation device 60 collates the face template 62 of a plurality of monitoring subjects, that is, the face image data itself and the input face image data, it can perform accurate collation. At this time, since only the input face image data filtered by the primary collation device 30 is a collation target, the processing load can be reduced. This is particularly effective when the number of cameras 10 is enormous.

  The operation management device 50 is a device that performs operation management of the entire face matching system. Specifically, the image data of the person to be monitored is registered together with identification information and registration information such as name, address, and reason for verification. The image data of the monitoring subject is acquired from another server device or storage device.

  In addition, the operation management device 50 generates a face template and an average face template from the image data of the monitoring target person and transmits them to the secondary verification device 60, and averages the face templates of the plurality of monitoring target persons to average face values. A template is generated and transmitted to the primary verification device 30. The face template and the average face template are also stored in the operation management apparatus 50.

  Further, when the operation management apparatus 50 receives the input face image data and the identification information from the secondary verification apparatus 60, the operation management apparatus 50 searches the face template and the attribute information using the identification information, and inputs the face image data and the face template. And the attribute information is displayed on the display unit 52. Note that the position information of the camera 10 that captured the input face image data can also be displayed. In such a case, the position information of the camera 10 needs to be acquired from the primary verification device 30.

  Next, the configuration of the operation management apparatus 50 shown in FIG. 1 will be specifically described. FIG. 2 is a functional block diagram showing an internal configuration of the operation management apparatus 50 shown in FIG. As shown in the figure, the operation management apparatus 50 includes an input unit 51, a display unit 52, an interface unit 53, a storage unit 54, and a control unit 55. The input unit 51 is an input device including a keyboard and a mouse, and the display unit 52 is a display device including a liquid crystal panel and a display device. The interface unit 53 is an interface unit for performing data communication with other devices via the LAN and the network 40.

  The storage unit 54 is a storage device including a flash memory, a hard disk device, and the like, and stores registration data 54a, a face template 54b, and an average face template 54c. The registration data 54a is data in which each monitoring target person's face image data and attribute information such as name, address, and registration reason are associated with the monitoring target person's identification information. The face template 54b is face image data obtained by viewing the face of the person to be monitored from a plurality of directions as shown in FIG. The average face template 54c is face image data obtained by averaging a plurality of identification target person's face images as shown in FIG.

  The control unit 55 is a control device that controls the entire operation management device 50, and includes a face template generation unit 55a, an average face template generation unit 55b, and an operation management unit 55c. As shown in FIG. 3A, the face template generation unit 55a cuts out partial image data of the face portion from the image data of the face image of the person to be monitored, and uses a known image processing technique to detect the face in the partial image data. A face template composed of 10 face image data is generated while changing the direction of.

  As shown in FIG. 3B, the average face template generation unit 55b cuts out partial facial image data from a plurality of monitoring subject's facial photographs, and facial parts (eyes, nose, mouth) of the extracted partial image data. Etc. are averaged. Various conventional techniques can be used when generating the average face.

  The operation management unit 55c is a processing unit that performs operation management of the entire face matching system. The operation management unit 55c newly registers the camera 10, the face detection device 20, the primary matching device 30, or the secondary matching device. Process to erase the device. Further, when the operation management unit 55c receives the input face image data and the identification information from the secondary verification device 60, the operation management unit 55c searches for the face image data, the face template, and the registration information corresponding to the identification information, and inputs the input face image. A process of displaying on the display unit 52 together with the data is performed.

  Next, the internal configuration of the primary verification device 30 shown in FIG. 1 will be described. FIG. 4 is a functional block diagram showing the internal configuration of the primary verification device 30 shown in FIG. As shown in the figure, the primary verification device 30 includes an interface unit 31, a storage unit 32, and a control unit 33. The interface unit 31 is an interface unit for performing data communication with other devices via the network 40.

  The storage unit 32 is a storage device including a flash memory, a hard disk device, and the like, and stores an average face template 32a. The average face template 32a is transmitted from the operation management apparatus 50 shown in FIG. 2, and is the same as the average face template 54c stored in the storage unit 54 of the operation management apparatus 50.

  The control unit 33 is a control unit that controls the primary collation device 30 as a whole, and includes an average face collation processing unit 33a and an automatic threshold setting unit 33b. The average face matching processing unit 33a performs a matching process between the input face image data received from the face detection device 20 and the average face template 32a stored in the storage unit 32, and the matching value is equal to or greater than a predetermined threshold value. And a processing unit that transmits the input face image data to the secondary collation device 60. Note that various conventional techniques such as cross-correlation processing can be used for such verification processing.

  The automatic threshold value setting unit 33b captures a non-monitoring target person that does not include a monitoring target person who passes in front of the camera 10 as a learning period when the camera 10 is installed or when the operation of the camera 10 is started. This is a processing unit that calculates a collation value between the input face image data received from the camera 10 via the face detection device 20 and the average face template 32a, and automatically sets a threshold value based on the distribution of the collation value.

  Specifically, when the histogram of the matching value between the input face image data and the average face template 32a becomes a distribution function as shown in FIG. 5, the average, variance, and standard deviation σ of this distribution function are obtained. For example, 3σ may be set as the threshold value.

  Thereby, after learning is completed, the optimal threshold value for every camera 10 can be automatically set. This is because the performance, shooting angle, lighting environment, and the like may be different for each camera 10. When there are a plurality of average face templates 32a for each face orientation, a threshold value can be automatically set for each average face template 32a.

  Next, the internal configuration of the secondary verification device 60 shown in FIG. 1 will be described. FIG. 6 is a functional block diagram showing the internal configuration of the secondary verification device 60 shown in FIG. As shown in the figure, the secondary verification device 60 includes an interface unit 61, a storage unit 62, and a control unit 63. The interface unit 61 is an interface unit for performing data communication with other devices via the LAN and the network 40.

  The storage unit 62 is a storage device including a flash memory, a hard disk device, and the like, and stores a plurality of face templates 62a. The face template 62a is transmitted from the operation management apparatus 50 shown in FIG. 2, and is the same as the face template 54b stored in the storage unit 54 of the operation management apparatus 50.

  The control unit 63 is a control unit that totally controls the secondary collation device 60, and includes a face collation processing unit 63a. The face matching processing unit 63a performs matching processing between the input face image data received from the primary matching device 30 and the plurality of face templates 62a stored in the storage unit 62, and the matching value is equal to or greater than a predetermined threshold value. In this case, the processing unit transmits the input face image data and the identification information to the operation management apparatus 50. Note that various conventional techniques such as cross-correlation processing can also be used for such verification processing.

  Next, the internal configuration of the face detection apparatus 20 shown in FIG. 1 will be described. FIG. 7 is a functional block diagram showing an internal configuration of the face detection apparatus 20 shown in FIG. As shown in the figure, the face detection apparatus 20 includes an interface unit 21, a storage unit 22, and a control unit 23. The interface unit 21 is an interface unit for performing data communication with the camera 10 and the primary verification device 30.

  The storage unit 22 is a storage device including a flash memory, a hard disk device, and the like, and temporarily stores moving image data received from the camera 10. The temporarily stored moving image data is transmitted to the operation management apparatus 50 when requested by the operation management apparatus 50.

  The control unit 23 is a control unit that controls the entire face detection device 20, and includes a face detection processing unit 23a and a camera position control unit 23b. The face detection processing unit 23 a stores the moving image data received from the camera 10 in the storage unit 22, detects a human face image from the moving image data, and transmits the detected face image data to the primary verification device 30. Is a processing unit. Specifically, still image data is sequentially extracted from the moving image data, and whether or not a human face portion exists in the extracted still image is determined by a template matching technique or the like.

  The camera position control unit 23b registers in advance a reference coordinate position of a reference object that should exist at a predetermined position, and between the reference object coordinate position and the reference coordinate position that exist in the still image extracted from the moving image data. If there is a positional deviation greater than or equal to a predetermined value, a control signal for calculating the positional deviation amount and correcting the positional deviation amount is transmitted to the control mechanism of the camera 10.

  Next, the registration process procedure of the face template and the average face template will be described. FIG. 8 is a flowchart showing the registration processing procedure of the face template and the average face template by the operation management apparatus 50 shown in FIG.

  As shown in FIG. 8, when the operation management apparatus 50 captures face photo data from another server apparatus or storage device and completes registration of attribute information while providing identification information (step S101), the operation management apparatus 50 selects a face template. Generate (step S102). Specifically, as shown in FIG. 3A, partial image data of a face is cut out from face photo data by using template matching or the like, and the orientation of the face in the partial image data is changed by an existing image processing technique. Ten face image data are generated while this is used as a face template.

  Thereafter, an average face template is generated (step S103). Specifically, as shown in FIG. 3B, partial image data of a face is cut out from a plurality of face photographs, and an average face is generated while averaging face parts using an existing technique.

  Then, after the generated face template 54b and average face template 54c are stored in the storage unit 54 (step S104), if other face photo data exists (step S105; Yes), the process proceeds to step S101 and the same Repeat the process.

  On the other hand, when there is no other face photo data (step S105; No), the face template is transmitted to the secondary collation device 60 (step S106), and the average face template is transmitted to the primary collation device 30. (Step S107).

  If the secondary collation device 60 receives such a face template (step S108; Yes), it registers the received face template in the storage unit 62 (step S109). Similarly, when the primary collation device 30 receives the average face template (step S110; Yes), it registers the received average face template in the storage unit 32 (step S111).

  By performing the above-described series of processing, the average face template 32a is registered in the primary collation device 30, and the face template 62a is registered in the secondary collation device 60, and preparation for performing the face collation processing is completed. .

  Next, a processing procedure until the input face image data is displayed through the collation processing by the primary collation device 30 and the secondary collation device 60 shown in FIG. 1 will be described. FIG. 9 is a flowchart showing a processing procedure until the input face image data is displayed after the collation processing by the primary collation device 30 and the secondary collation device 60 shown in FIG.

  First, the moving image data captured by the camera 10 is received by the face detection device 20, and the face detection device 20 extracts still image data from the moving image data to extract human face image data, which is input face image. The data is transmitted to the primary verification device 30 as data.

  When the primary collation device 30 receives the input face image data (step S201; Yes), the primary collation device 30 performs an average face collation process for collating the input face image data with the average face template 32a (step S202), and a collation value. Is equal to or greater than the predetermined threshold value (step S203; Yes), the input face image data is transmitted to the secondary verification device 60 (step S204). If the collation value is less than the predetermined threshold value (step S203; No), the input face image data is not transmitted to the secondary collation device 60.

  If the input face image data is received (step S205; Yes), the secondary matching device 60 performs face matching processing for matching the input face image data with the face template 62a (step S206), and the matching value is predetermined. If it is equal to or greater than the threshold value (step S207; Yes), the input face image data and the identification information are transmitted to the operation management apparatus 50 (step S208). If the collation value is less than the predetermined threshold (step S207; No), the input face image data and the identification information are not transmitted to the operation management apparatus 50.

  If the operation management device 50 receives the input face image data and the identification information (step S209; Yes), the operation management device 50 searches the face template and the attribute information using the identification information, and inputs the face image data, the face template, and the attribute information. Is displayed (step S210). The attribute information includes face photograph data in addition to the name and address of the person to be monitored, the reason for the subject of verification, and the like.

  As shown in FIG. 10, a tracking processing unit 72 and a body template generation unit 73 can be provided in the control unit 55 of the operation management apparatus 70. The tracking processing unit 72 is a processing unit that performs tracking processing of the location of the monitoring target person. If there is a monitoring subject in the moving image data captured by the camera 10 through the series of processes described above, the operation management device 70 is notified of the input face image data of the monitoring subject.

  Here, if the operation management apparatus 70 can receive the identification information and the photographing time of the camera 10 at the time of receiving the final input face image data, “when and where the corresponding monitoring subject was located”. Information is available. For this reason, the tracking processing unit 72 collects the behavior trajectory of the monitoring target person using the information notified from the secondary verification device 60 and can use it for the tracking process for grasping the movement route of the monitoring target person. it can. In order to perform such tracking processing, the face detection device 20 notifies the primary collation device 30 of the identification information and the photographing time of the camera 10 together with the input face image data. It is necessary to notify the next verification device of the identification information of the camera 10 and the photographing time together with the input face image data. As for details of the tracking process, an existing technique can be used, but it is desirable to perform the tracking process while performing route prediction.

  The body template generation unit 73 is a processing unit that generates the body template 71 of the person to be monitored. For example, the body template generation unit 73 stores the features of appearance features such as the whole figure, height, and color, and the features of movement features such as walking. A template 71 can be used.

  Specifically, a face template similar to that shown in FIG. 3A is generated from a photograph of the entire body, and the generated body template 71 is transmitted to the primary verification device 30 and stored in the storage unit 54. To remember. The primary verification device 30 that has received the body template 71 notifies the operation management device 70 of the result when a result equal to or higher than a predetermined verification rate is obtained by the verification process using the body template 71. become. By performing the process using the body template 71, the tracking process can be performed even when the face of the person to be monitored is not captured in the moving image data. In addition, while generating an average body template and transmitting it to the primary collation apparatus 30, the body template 71 can also be transmitted to the secondary collation apparatus 60, and the process similar to the collation process using the said face image can also be performed. Then, the movement path can be estimated by collating and detecting the images of the camera 10 installed within a certain range from the camera 10 from which the face is detected, for a predetermined time.

  As described above, in the present embodiment, the primary collation device 30 disposed at each site collates the input face image data captured by the camera 10 and detected by the face detection device 20 with the average face template 32a. The input face image data is transmitted to the secondary collation device 60 when the input face image data is equal to or greater than the predetermined threshold value, and the secondary collation device is configured to perform collation processing on the input face image data with the face template. Since it is no longer necessary to place a face template at a location where no is taken, information leakage can be prevented. In addition, even when the number of cameras 10 installed is increased, the processing load of the secondary verification device 60 can be reduced.

  In the above embodiment, the case where four cameras 10, four face detection devices 20, two primary collation devices 30 and one secondary collation device 60 are used is shown. The present invention is not limited to this, and can be applied when more cameras 10 or the like are provided. In particular, it is suitable when thousands to tens of thousands of cameras 10 are provided. Even in this case, it is possible to reduce the processing load of the secondary verification device 60 and prevent information leakage of the monitoring target person.

  In the above embodiment, when generating an average face template, a case is shown in which the parts of the faces of a plurality of monitoring subjects are averaged. In order to prevent loss of personal characteristics of the monitoring subjects due to averaging. The unique face parts among the face parts of the monitoring subject can be weighted. In addition, if only one average face template is generated using the face image data of a large number of monitoring subjects, the possibility of losing personal characteristics increases, so an average face template is generated for each predetermined number of people (for example, 10 people). It is possible. In this case, it is desirable to group the extracted features that are similar. For example, it is desired to generate an average face with faces within a predetermined distance in a feature amount space.

  Furthermore, in the above embodiment, reference to specific cases was omitted. It can be used to detect suspects.

  At this time, the processing of the server can be reduced by narrowing down the persons to be processed in advance. For example, when a person who passes an automatic ticket gate at a station is photographed by the camera 10, it is possible to narrow down by excluding the person with the commuter pass from the processing target. In this case, the commuter pass purchaser is pre-registered as a non-detection target in advance when purchasing the commuter pass, and the commuter pass purchaser is excluded from the verification process as a non-detection target person among the people read by the ticket gate Will do. Further, for impersonation by commuter pass exchange, since the ID is known from the commuter pass, one-to-one verification can be performed, so the load on the server is relatively low. In addition, for business travelers and travelers from the local area, if it is determined that the person is a non-detection target after being processed by a server at a local station with a small number of traffic, the person who owns the ticket is verified. Can be excluded from the target.

  Moreover, although the case where the average face template is used as an example of the synthetic face template is shown in the above embodiment, the present invention is not limited to this, and montage or morphing generated from a plurality of face image data is used. You can also.

  As described above, the face matching system, the face matching device, and the face matching method according to the present invention acquire input face image data with a large number of monitoring cameras, and when this input face image data is checked against a face template, It is suitable for efficiently suppressing the leakage of the face template while reducing the processing load of the face matching device that performs the matching process.

DESCRIPTION OF SYMBOLS 10 Camera 20 Face detection apparatus 21 Interface part 22 Storage part 23 Control part 23a Face detection process part 23b Camera position control part 30 Primary collation apparatus 31 Interface part 32 Storage part 32a Average face template 33 Control part 33a Average face collation process part 33b Automatic threshold setting unit 33b
DESCRIPTION OF SYMBOLS 40 Network 50 Operation management apparatus 51 Input part 52 Display part 53 Interface part 54 Storage part 54a Registration data 54b Face template 54c Average face template 55 Control part 60 Secondary collation apparatus 61 Interface part 62 Storage part 62a Face template 63 Control part 63a Face Verification processing unit 70 Operation management device 71 Body template 72 Tracking processing unit 73 Body template generation unit

Claims (7)

A face matching system that performs face matching processing on a plurality of pre-registered face templates of monitoring subjects and input face image data captured by an imaging device,
A template generation device that generates a plurality of face templates and a combined face template based on face image data of a plurality of monitoring subjects;
A plurality of imaging devices respectively disposed at a plurality of bases;
A primary collation device that performs collation processing on input face image data captured by the image capturing device with the composite face template generated by the template generation device, and transmits input face image data having a collation rate equal to or higher than a predetermined value. When,
A face collation system, comprising: a secondary collation device that collates input face image data received from the primary collation device with a face template generated by the template generation device.   The primary verification device is disposed at each base together with one or a plurality of imaging devices, and the secondary verification device is disposed at a center where security measures such as entry / exit management are taken. The face matching system described.   The primary collation device automatically sets a predetermined value as a transmission processing target based on a distribution of a collation rate obtained by collating input face image data captured by the imaging device for a predetermined period and the synthetic face template. The face collation system according to claim 1 or 2, wherein   The primary verification device notifies the imaging device of a positional deviation amount between a reference object located in the input face image data imaged by the imaging device and a normal location of the reference object, and the imaging device The face collation system according to claim 1, wherein the angle of view is controlled based on a positional deviation amount received from the primary collation device.   The secondary collation device generates tracking information of a corresponding monitoring target person based on a collation processing result obtained by collating the input face image data received from the primary collation device and the face template. The face collation system according to any one of claims 1 to 4. A face matching method for performing face matching processing on a plurality of pre-registered face templates of monitoring subjects and input face image data captured by an imaging device,
A template generating step for generating a plurality of face templates and a combined face template based on the face image data of a plurality of monitoring subjects;
A primary matching step of matching input face image data captured by one or more imaging devices with the synthetic face template generated in the template generation step;
A face matching method, comprising: a secondary matching step of matching input face image data having a matching rate equal to or higher than a predetermined value in the primary matching step with the face template generated in the template generation step. A face matching device that performs face matching processing on a plurality of pre-registered face templates of monitoring subjects and input face image data captured by an imaging device,
A storage unit for storing a composite face template generated based on face image data of a plurality of monitoring subjects;
An extraction processing unit for extracting input face image data having a matching rate equal to or greater than a predetermined value by comparing input face image data captured by a predetermined imaging device with the composite face template stored in the storage unit; A face collation apparatus comprising:

Images