JP5747713B2 - Authentication device, authentication method, authentication program - Google Patents

Description

  The present invention relates to an authentication apparatus, an authentication method, and an authentication program for authenticating an object such as a person using feature information of the object such as face feature information registered in advance.

  As an example of an authentication apparatus, a face authentication apparatus that personally authenticates a person using a face image photographed by a camera or the like has been conventionally installed and used in various products. In the face authentication device, face feature information generated from a face image of a predetermined person is registered in a database in advance, and face feature information generated from a face image of a person to be authenticated and face feature information registered in the database In general, it is determined whether or not the authentication target is the same as the registered person by obtaining the similarity.

Japanese Patent Laid-Open No. 9-35068 JP 2006-127236 A

  When registering facial feature information of a face image of a predetermined person in the database in the face authentication device, try to register it based on the face orientation, facial expression, lighting status, out-of-focus, exposure shift, camera shake, etc. May be registered in a state that is not suitable for authentication. In this case, even if the face image to be authenticated is that of the person, it may not be determined that the person is the person. Therefore, there is a need for a face authentication device, a face authentication method, and a face authentication program that can register face feature information suitable for authentication and can authenticate a face image with high accuracy.

  Here, the face authentication device is taken as an example, but the same applies to a case where a person is authenticated using images of hands (palmprints), fingers (fingerprints), eyes (retinas), and the like. Furthermore, the same applies to the case of authenticating not only a person but also an animal or an object other than a person. That is, there is a demand for an authentication device that can register feature information suitable for authentication and can authenticate an object with high accuracy. In order to meet such a demand, the present invention provides an authentication apparatus, an authentication method, and an authentication program that can register feature information suitable for authentication and can authenticate an object with high accuracy. And

  In order to solve the above-described problems of the related art, the present invention provides a feature information extraction unit (2) that extracts feature information from an image obtained by photographing an object to be registered, and a plurality of second images obtained by photographing the object. The holding unit (3) that holds the plurality of first feature information extracted by the feature information extraction unit from each image of one frame, the plurality of first feature information, and the first frame are different Feature information collation that is a frame and calculates first similarity by collating with second feature information extracted by the feature information extraction unit from one or a plurality of second frame images obtained by photographing the object And a control unit (1) that causes the holding unit (4) to hold one feature information having the highest first similarity among the plurality of first feature information as registration information. An authentication device is provided.

  In the above authentication apparatus, the second frame is plural, and the feature information matching unit calculates a similarity between each of the plurality of first feature information and each of the plurality of second feature information, Each of the plurality of first feature information further includes a detection unit (6) that detects an average similarity obtained by averaging similarities with the plurality of second feature information, and the control unit includes the plurality of first feature information Of the first feature information, the feature information having the highest average similarity detected by the detection unit is preferably held in the holding unit as registration information.

  In the authentication apparatus, the feature information matching unit calculates a second similarity between the plurality of first feature information and third feature information of an object different from the object to be registered. The control unit causes the holding unit to hold the feature information having the largest difference between the first similarity and the second similarity among the plurality of first feature information as registration information. May be.

  In addition, in order to solve the above-described problems of the related art, the present invention extracts feature information from each of a plurality of first frame images obtained by photographing an object to be registered, and the plurality of first feature information. And holding step (S101 to S103, S201 to S203) and the second feature information from the image of one or a plurality of second frames that are different from the first frame and photograph the object. Extracting step (S104, S204), and a first calculating step (S105, S205) for calculating the first similarity by collating the plurality of first feature information and the second feature information. And a registration step (S110, S216) for registering, as registration information, one feature information having the highest first similarity among the plurality of first feature information. To provide an authentication method to.

  In the above authentication method, the second frame is plural, and the first calculation step calculates a similarity between each of the plurality of first feature information and each of the plurality of second feature information. For each of the plurality of first feature information, the method further includes a detection step (S108, S208) for detecting an average similarity obtained by averaging the similarity with the plurality of second feature information, and the registration step includes: Of the plurality of first feature information, the feature information having the highest average similarity detected in the detection step is preferably registered as registration information.

  In the authentication method, a second calculation step of calculating a second similarity between the plurality of first feature information and third feature information of an object different from the object to be registered ( S212), and the registration step registers, as registration information, feature information having a largest difference between the first similarity and the second similarity among the plurality of first feature information. It may be.

Furthermore, in order to solve the above-described problems of the conventional technology, the present invention provides a computer,
A holding step (S101 to S103, S201 to S203) for extracting feature information from each of a plurality of first frame images obtained by photographing an object to be registered and holding the plurality of first feature information, and the first An extraction step (S104, S204) for extracting second feature information from one or a plurality of second frame images obtained by photographing the object, and a plurality of the first frames. A first calculation step (S105, S205) for calculating a first similarity by collating feature information with the second feature information, and a first similarity among the plurality of first feature information An authentication program characterized by causing a registration step (S110, S216) for registering one piece of feature information having the largest value as registration information is provided.

In the above authentication program, the second frame is plural, and the computer
As the first calculation step, a process of calculating a similarity between each of the plurality of first feature information and each of the plurality of second feature information is executed, and for each of the plurality of first feature information , Further executing a detection step (S108, S208) for detecting an average similarity obtained by averaging the similarity with the plurality of second feature information, and as the registration step, among the plurality of first feature information, It is preferable to execute a process of registering the feature information having the highest average similarity detected in the detection step as registration information.

  In the authentication program, the computer calculates a second similarity between the plurality of first feature information and third feature information of an object different from the object to be registered. A calculation step (S212) is further executed, and as the registration step, feature information having the largest difference between the first similarity and the second similarity is registered information as the registration information. It is also possible to execute the process of registering as

  According to the face authentication apparatus, authentication method, and authentication program of the present invention, feature information suitable for authentication can be registered, and an object can be authenticated with high accuracy.

It is a block diagram which shows 1st Embodiment of the authentication apparatus of this invention. It is a figure for demonstrating operation | movement of each embodiment. It is a flowchart for demonstrating 1st Embodiment. It is a block diagram which shows 2nd Embodiment of the authentication apparatus of this invention. It is a flowchart for demonstrating 2nd Embodiment. FIG. 10 is a diagram for explaining an operation of the first modification. FIG. 10 is a diagram for explaining an operation of a second modification.

  Hereinafter, embodiments of an authentication device, an authentication method, and an authentication program of the present invention will be described with reference to the accompanying drawings. Each embodiment, as a suitable example of an authentication device, an authentication method, and an authentication program, registers a face image obtained by photographing a person's face and personally authenticates the person based on facial feature information, an authentication method, and an authentication program Will be described. However, the image to be registered is not limited to a face image, and the object to be authenticated is not limited to a person. The present invention may be an arbitrary authentication device, an authentication method, and an authentication program for authenticating an object using feature information of the object registered in advance.

<< First Embodiment >>
In the first embodiment, when registering a facial image of a predetermined person, facial feature information extracted from the facial image is provisionally registered, and after the temporary registration, facial feature information is further extracted from the facial image to obtain facial features. The information is configured to be fully registered. When the main registration of the face feature information is completed, the face feature information of the indefinite person is checked against the registered face feature information to perform personal authentication. Hereinafter, a temporary registration operation, a main registration operation, and an authentication operation will be described in this order using FIG.

<Tentative registration operation>
In FIG. 1, face image data obtained by capturing a predetermined person to register a face image with a camera (not shown) is input to the face feature information extraction unit 2. The person here is assumed to be a person Pa. The face feature information extraction unit 2 extracts face feature information from each face image data for a predetermined N frames obtained by photographing a face of a predetermined person based on control by the control unit 1. N is a natural number, preferably 2 or more. In the first embodiment, N is 6. Since the face feature information extraction unit 2 extracts face feature information from each frame image data for six frames, it generates six pieces of face feature information and outputs them to the temporary registration information holding unit 3. The temporary registration information holding unit 3 holds the inputted six face feature information as temporary registration information based on the control by the control unit 1. Although the number of pieces of face feature information is expressed here as individual pieces, each piece of face feature information includes one or more pieces of information indicating features for specifying a face.

<Main Registration Operation>
As described above, the main registration operation is executed as a series of operations following the provisional registration operation. In the main registration operation, the face feature information extraction unit 2 is based on control by the control unit 1 for a predetermined M frames obtained by photographing a face of a predetermined person (person Pa) who is the same person as the person who performed the temporary registration operation. Facial feature information is extracted from each face image data. M is a natural number, preferably 2 or more. More preferably, M is greater than N. In the first embodiment, M is 12. Since the face feature information extraction unit 2 extracts face feature information from each of the 12 frames of face image data, the face feature information extraction unit 2 generates 12 pieces of face feature information and outputs them to the face feature information matching unit 5. The temporary registration information holding unit 3 outputs the held six face feature information to the face feature information matching unit 5 based on the control by the control unit 1.

  As shown in FIG. 2A, six pieces of face feature information supplied from the temporary registration information holding unit 3 to the face feature information matching unit 5 are conceptually represented by FI1 to FI6, and the face feature information extracting unit 2 The face feature information of 12 pieces of face feature information supplied to the face feature information matching unit 5 is conceptually represented by FI11 to FI22. The face feature information FI1 to FI6 and the face feature information FI11 to FI22 are all face feature information of the person Pa. Face feature information FI1 to FI6 and face feature information FI11 to FI22 are input to the face feature information matching unit 5 through a series of operations from the temporary registration operation to the main registration operation.

  Based on the control by the control unit 1, the face feature information matching unit 5 performs matching with each of the face feature information FI 11 to FI 22 on the basis of each of the face feature information FI 1 to FI 6 as shown in FIG. Calculate similarity. FIG. 2A shows only the collation between the face feature information FI1 and the face feature information FI11 to FI22, but the face feature information FI2 to FI6 is collated in the same manner.

  FIG. 2B conceptually shows the similarity obtained by the face feature information matching unit 5. As shown in FIG. 2B, the face feature information matching unit 5 has similarity information S111 to S122 indicating the similarity between the face feature information FI1 and the face feature information FI11 to FI22 with respect to the face feature information FI1. Is generated. The face feature information matching unit 5 generates, for the face feature information FI2, similarity information S211 to S222 indicating the similarity between the face feature information FI2 and the face feature information FI11 to FI22. The face feature information matching unit 5 generates similarity information S311 to S322 indicating the similarity between the face feature information FI3 and the face feature information FI11 to FI22 with respect to the face feature information FI3.

  The face feature information matching unit 5 generates similarity information S411 to S422 indicating the similarity between the face feature information FI4 and the face feature information FI11 to FI22, with respect to the face feature information FI4. The face feature information matching unit 5 generates similarity information S511 to S522 indicating the similarity between the face feature information FI5 and the face feature information FI11 to FI22 with respect to the face feature information FI5. The face feature information matching unit 5 generates similarity information S611 to S622 indicating the similarity between the face feature information FI5 and the face feature information FI11 to FI22, with respect to the face feature information FI6.

  Similarity information S111 to S122, S211 to S222, S311 to S322, S411 to S422, S511 to S522, S611 to S622 for the face feature information FI1 to FI6 shown in FIG. Is output to the principal similarity detection unit 6. Based on the control by the control unit 1, the identity similarity detection unit 6 performs similarity information S111 to S122, S211 to S222, S311 to S322, S411 to S422, S511 to S522, as shown in FIG. By accumulating each of S611 to S622 and taking the average, average similarity information Savg1 to Savg6 indicating the average similarity is calculated for each of the face feature information FI1 to FI6. The average similarity information Savg1 to Savg6 indicates the similarity when the same person is photographed in the temporary registration operation and the main registration operation, and is the person similarity. Average similarity information Savg1 to Savg6 indicating the identity similarity is output to the control unit 1.

  The control unit 1 is similar to the person more than a predetermined registration threshold set to a value larger than the authentication threshold among the input person similarity (average similarity information Savg1 to Savg6) from the face feature information FI1 to FI6. One face feature information indicating the average similarity information having the highest degree is specified. Then, the control unit 1 outputs the facial feature information indicating the specified average similarity information to the registered information holding unit 4 among the six pieces of facial feature information FI1 to FI6 held by the temporary registration information holding unit 3. Let The registration information holding unit 4 holds the input facial feature information as main registration information based on the control by the control unit 1.

  FIG. 2D shows an example of face feature information output from the temporary registration information holding unit 3 to the registration information holding unit 4. For example, it is assumed that the average similarity information Savg3 of the face feature information FI3 is the maximum among the average similarity information Savg1 to Savg6. In the registration information holding unit 4, the face feature information FI3 is held and registered through the above-described series of temporary registration operations for the person Pa to the main registration operation. When the same operation is performed on the other persons Pb, Pc..., For example, the face feature information FI5 is registered for the person Pb in the registered information holding unit 4, and for example, the face feature information FI1 is stored for the person Pc. Is registered.

  The control unit 1 controls the temporary registration information holding unit 3 and the registration information holding unit 4 so that none of the facial feature information FI1 to FI6 is fully registered if the average similarity information Savg1 to Savg6 is less than the registration threshold. . That is, any one of the facial feature information FI1 to FI6 is not output from the temporary registration information holding unit 3 and is not held in the registration information holding unit 4.

  When the main registration is completed, the control unit 1 displays on the display unit 9 that the registration is successful. When the main registration is not completed, the control unit 1 displays on the display unit 9 that the registration is not successful. Display.

<Authentication operation>
The face feature information extraction unit 2 extracts face feature information from face image data obtained by photographing an indefinite person's face based on control by the control unit 1. In the authentication operation, face feature information may be extracted from one or a plurality of frames. It is preferable to extract face feature information from a plurality of frames because the accuracy of authentication is higher when face feature information is extracted from a plurality of frames. The indefinite person here is one of the above-described persons Pa, Pb, Pc, or any other arbitrary person. The face feature information extraction unit 2 outputs the face feature information of the indefinite person to the face feature information matching unit 5.

  The registered information holding unit 4 outputs the held one or more face feature information to the face feature information matching unit 5 based on the control by the control unit 1. In the example of FIG. 2D, face feature information FI3 for the person Pa, face feature information FI5 for the person Pb, face feature information FI1 for the person Pc, etc. are output to the face feature information matching unit 5. The face feature information collation unit 5 collates the face feature information output from the face feature information extraction unit 2 with the face feature information output from the registered information holding unit 4 based on the control by the control unit 1. Then, the similarity is calculated, and the calculated similarity is output to the authentication result determination unit 7.

  The authentication result determination unit 7 determines from the control unit 1 the degree of similarity between the face feature information of the indefinite person output from the face feature information matching unit 5 and the face feature information that is registration information held in the registration information holding unit 4. It is determined whether there is face feature information that is equal to or greater than a specified authentication threshold value. If the authentication result determination unit 7 determines that there is face feature information greater than or equal to the authentication threshold, the authentication result determination unit 7 determines that the indefinite person corresponds to a person registered in the registration information holding unit 4, and facial feature information greater than or equal to the authentication threshold If it is determined that there is no person, it is determined that the person does not correspond to the person to which the indefinite person is registered. The authentication result determination unit 7 outputs the determination result to the outside as a face authentication result. The outside is an external device that uses the face authentication result of the face authentication device, such as a release device that unlocks the door.

  A series of operations from the temporary registration operation to the main registration operation according to the first embodiment will be further described with reference to FIG. Each operation of the flowchart shown in FIG. 3 is executed based on control by the control unit 1. In FIG. 3, the face feature information extraction unit 2 extracts face feature information from face image data of a predetermined person in step S101. In step S102, the temporary registration information holding unit 3 temporarily registers the facial feature information extracted in step S101 as temporary registration information. In step S103, the control unit 1 determines whether or not provisional registration of N-frame face feature information has been completed. If it is not determined that the process has been completed (NO), the process returns to step S101, and steps S101 to S103 are repeated. If it is determined in step S103 that the temporary registration of facial feature information for N frames has been completed (YES), the process proceeds to step S104. Steps S101 to S103 are temporary registration operations.

  Next, in step S104, the face feature information extraction unit 2 newly extracts face feature information from the newly inputted face image data of the same predetermined person. In step S105, the face feature information collation unit 5 collates the face feature information extracted in step S104 and the plurality of temporary registration information provisionally registered in step S102, and calculates the similarity. In step S106, the identity similarity detection unit 6 accumulates the similarities calculated in step S105. In step S107, the control unit 1 determines whether or not collation of face feature information for M frames has been completed. If it is not determined that the process has been completed, the process returns to step S104, and steps S104 to S107 are repeated. If it is determined in step S107 that collation of face feature information for M frames has been completed, the process proceeds to step S108.

  In step S108, the person similarity detection unit 6 calculates average similarities Savg1 to Savg6 obtained by dividing the accumulated similarity for each temporary registration information by the number of frames M, and calculates the calculated average similarities Savg1 to Savg6 for each temporary. The identity similarity of registered information. In step S <b> 109, the control unit 1 determines whether there is provisional registration information whose identity similarity is equal to or greater than a predetermined registration threshold. If it is determined that there is provisional registration information whose identity similarity is equal to or greater than the predetermined registration threshold (YES), the control unit 1 selects and selects provisional registration information having the highest identity similarity in step S110. Temporary registration information (face feature information) is permanently registered in the registration information holding unit 4. The facial feature information selected in step S110 is facial feature information that most appropriately represents facial features of a predetermined person, and is optimal face feature information for authenticating the predetermined person as the person. In step S111, the control unit 1 notifies the operator of the success of registration through screen display by the display unit 9, and ends.

  On the other hand, if it is not determined in step S109 that there is provisional registration information whose identity similarity is equal to or higher than a predetermined registration threshold (NO), the control unit 1 is operated by screen display by the display unit 9 in step S112. The user is notified of the registration failure and ends. The fact that there is no provisional registration information that is equal to or greater than the registration threshold in step S109 means that face feature information that appropriately represents the facial features of the predetermined person cannot be obtained from the input face image. That is, face feature information suitable for authentication could not be extracted. In this case, the processing from step S101 may be performed again. Steps S104 to S112 are main registration operations.

<< Second Embodiment >>
The configuration and operation of the second embodiment will be described with reference to FIG. Also in the second embodiment, the temporary registration operation, the main registration operation, and the authentication operation are performed as in the first embodiment. The temporary registration operation and the authentication operation in the second embodiment are the same as those in the first embodiment, and only the main registration operation is different. Accordingly, the description of the temporary registration operation and the authentication operation is omitted, and the main registration operation will be mainly described.

<Main Registration Operation>
In the second embodiment, information for specifying a person, such as the name of a predetermined person to register a face image, is input at the time of registration. In FIG. 4, the face feature information extraction unit 2 extracts face feature information from each frame image data for 12 frames and outputs it to the face feature information matching unit 5. The temporary registration information holding unit 3 outputs the facial feature information that is temporary registration information to the facial feature information matching unit 5. When the registration information holding unit 4 holds registration information of a person different from the predetermined person to be registered based on the control by the control unit 1, the registration information holding unit 4 displays the registration information of the different person as the face feature information matching unit 5. To output.

  Based on the control by the control unit 1, the face feature information matching unit 5 includes a plurality of temporary features of the predetermined person output from the facial feature information output from the facial feature information extraction unit 2 and the temporary registration information holding unit 3. The registration information is collated with each other to calculate the similarity, and the calculated similarity is output to the identity similarity detection unit 6. Further, the facial feature information matching unit 5 is different from the temporary registration information of the predetermined person output from the temporary registration information holding unit 3 and the predetermined person output from the registration information holding unit 4 based on the control by the control unit 1. The similarity is calculated by collating each of the registered information of the person, and the calculated similarity is output to the stranger similarity detection unit 8.

  Based on the control by the control unit 1, the identity similarity detection unit 6 accumulates the similarity between the facial feature information of the predetermined person output from the facial feature information matching unit 5 and the provisional registration information for each provisional registration information. The average similarity Savg1 to Savg6 is calculated, and the average similarity Savg1 to Savg6 is output to the control unit 1 as the individual similarity of each temporary registration information. The stranger similarity detection unit 8 has a maximum similarity for each temporary registration information among the similarities between the temporary registration information of the predetermined person output from the face feature information matching unit 5 and the registration information of a person different from the predetermined person. The degree of similarity is detected, and the maximum degree of similarity is output to the control unit 1 as the degree of others similarity of each temporary registration information.

  The control unit 1 determines the principal similarity according to the principal similarity of each temporary registration information output from the principal similarity detection unit 6 and the other person similarity of each temporary registration information output from the stranger similarity detection unit 8. Temporary registration information having the largest difference between the person's similarity and the others' similarity among the temporary registration information having a predetermined registration threshold or more. When the registration information holding unit 4 does not hold registration information of a person different from the predetermined person to be registered, the stranger similarity detection unit 8 cannot detect the others similarity, so the control unit 1 As in the first embodiment, temporary registration information having the highest principal similarity among temporary registration information having a principal similarity equal to or higher than a predetermined registration threshold is permanently registered. The control unit 1 controls the temporary registration information holding unit 3 and the registration information holding unit 4 so that no temporary registration information is permanently registered if the identity similarity is less than the registration threshold.

  The main registration operation according to the second embodiment will be further described with reference to FIG. The processing in steps S201 to S209 in FIG. 5 is the same as the processing in steps S101 to S109 in FIG. 3, and description thereof is omitted here. In FIG. 5, if the control unit 1 determines in step S209 that there is provisional registration information whose identity similarity is equal to or higher than a predetermined registration threshold (YES), in step S210, the predetermined person to be registered is determined. It is determined whether there is registration information of a different person, that is, another person. If it is determined that there is registration information of another person (YES), the control unit 1 selects provisional registration information whose identity similarity is equal to or greater than a predetermined registration threshold in step S211.

  In step S212, the face feature information collation unit 5 collates the selected temporary registration information with other person's registration information, and calculates the similarity. In step S213, the stranger similarity detection unit 8 detects the maximum value of the similarity calculated in step S212, and sets the detected maximum value as the other person similarity of each temporary registration information. In step S214, the control unit 1 calculates the difference between the principal similarity and the others similarity of each temporary registration information. In step S215, the control unit 1 selects temporary registration information having the largest difference between the principal similarity and the others similarity and performs permanent registration. The provisional registration information selected in step S215 is suitable for authenticating a predetermined person as the person, that is, the face characteristic information that appropriately represents the face characteristic of the predetermined person and is most easily distinguished from other person's facial features. In addition, it is face feature information that is unlikely to be mistaken for others.

  On the other hand, if it is not determined in step S210 that there is other person's registration information (NO), in step S216, the control unit 1 selects temporary registration information having the highest individual similarity and performs main registration. Although it is not known whether the temporary registration information selected in step S216 is easily distinguishable from other people's facial features, the facial feature information that best represents the facial features of the predetermined person, that is, the predetermined person It is face feature information suitable for authenticating as.

  After steps S215 and S216, the control unit 1 notifies the operator of the success of registration through screen display by the display unit 9 and ends in step S218. On the other hand, if it is not determined in step S209 that there is provisional registration information whose identity similarity is equal to or higher than the predetermined registration threshold (NO), the control unit 1 is operated by screen display by the display unit 9 in step S217. The user is notified of the registration failure and ends.

  In the first and second embodiments described above, six face feature information is extracted from each face image data for six frames in the temporary registration operation and temporarily registered, and each face for 12 frames in the main registration operation. The example in which 12 pieces of face feature information are extracted from the image data and the similarity between the 6 pieces of face feature information and the 12 pieces of face feature information is calculated has been described. For simplification, six face feature information is extracted from each face image data for six frames in the temporary registration operation and temporarily registered, and one face feature is obtained from each face image data for one frame in the main registration operation. It is also possible to extract information and calculate the similarity between six pieces of face feature information and one piece of face feature information.

<< Modification 1 >>
A configuration for calculating the similarity between six pieces of face feature information and one piece of face feature information will be described as a first modification with reference to FIG. The face feature information matching unit 5 receives six pieces of face feature information FI1 to FI6 by the temporary registration operation from the temporary registration information holding unit 3, and one piece of face feature information by the main registration operation from the face feature information extraction unit 2. Face feature information FI10 is input. As shown in FIG. 6A, the face feature information matching unit 5 compares the face feature information FI1 to FI6 and the face feature information FI10 to calculate the similarity. In this case, the face feature information matching unit 5 generates similarity information S10 to S60 for each of the face feature information FI1 to FI6 as shown in FIG.

  In the first modification, since one piece of similarity information S10 to S60 corresponds to each of the face feature information FI1 to FI6, the process of obtaining the average similarity in the person similarity detection unit 6 is unnecessary. Accordingly, in the first modification, the person similarity detection unit 6 can be deleted. As shown in FIG. 6C, the control unit 1 extracts the facial feature information having the highest similarity among the similarity information S10 to S60 from the facial feature information FI1 to FI6, and extracts the extracted face. The feature information is read from the temporary registration information holding unit 3 and held in the registration information holding unit 4. Here, a case is shown in which the similarity information S30 of the face feature information FI3 is the maximum for the person Pa.

  In the first and second embodiments and the first modification, the facial feature information FI11 to FI22 or the facial feature information FI10 extracted by the main registration operation is based on the facial feature information FI1 to FI6 extracted by the temporary registration operation. The similarity is calculated, and one of the facial feature information FI1 to FI6 is stored in the registration information holding unit 4 as main registration information based on the calculated similarity. The face feature information extracted in the main registration operation may be selected from the face feature information extracted in the main registration operation using the face feature information extracted in the main registration operation as a reference.

<< Modification 2 >>
Using FIG. 7, the facial feature information extracted by the main registration operation is selected from the facial feature information extracted by the main registration operation, using the facial feature information extracted by the main registration operation as a reference. The configuration will be described as a second modification. The face feature information matching unit 5 receives one piece of face feature information FI0 from the provisional registration information holding unit 3 through the provisional registration operation, and the face feature information extraction unit 2 from the six face feature information through the main registration operation. The feature information FI11 to FI16 is input. As shown in FIG. 7, the face feature information collating unit 5 collates the face feature information FI0 and each of the face feature information FI11 to FI16 to calculate the similarity. In the second modification, the facial feature information FI11 to FI16 are also held in the temporary registration information holding unit 3.

  As in the first modification, the control unit 1 extracts face feature information having a high similarity to the face feature information FI0 from the face feature information FI11 to FI16, and the extracted face feature information is temporarily registered information holding unit. 3 and stored in the registration information storage unit 4. In FIG. 7, for simplification, the facial feature information FI0 is extracted with only one frame in the temporary registration operation, but a plurality of facial feature information may be extracted from a plurality of frames as in FIG.

  As can be seen from the above description, it is preferable that there are a plurality of N frames in the temporary registration operation and M frames in the main registration operation. For simplification, one of the frame from which the facial feature information is extracted by the temporary registration operation and the frame from which the facial feature information is extracted by the main registration operation may be one frame. In this case, if a plurality of face feature information for selecting face feature information to be finally registered in the registration information holding unit 4 is selected as a plurality of frames from the temporary registration operation and the main registration operation. Good.

  The present invention is not limited to the first embodiment, the second embodiment, the first modification, and the second modification described above, and various modifications can be made without departing from the scope of the present invention. 1 and 4, the temporary registration information holding unit 3 and the registration information holding unit 4 are described separately for convenience, but may be a single holding unit (memory). It has been described that one of the face feature information FI1 to FI6 held in the temporary registration information holding unit 3 is selected and held in the registration information holding unit 4, but the face feature information to be registered is left in the holding unit, Other facial feature information may be deleted. 1 and 4 are conceptual block diagrams of the present invention, and the present invention is not limited to the configurations of FIGS.

  It is also possible to install an authentication program in a computer installed in the authentication apparatus and execute the processing described with reference to FIG. The authentication program can be provided by being recorded on a recording medium such as an optical disk, or can be provided via a communication line such as the Internet. It is possible to appropriately mix software processing by the authentication program and hardware processing.

DESCRIPTION OF SYMBOLS 1 Control part 2 Face feature information extraction part 3 Temporary registration information holding part 4 Registration information holding part 5 Face feature information collation part 6 Personal person similarity detection part 7 Authentication result determination part 8 Others person similarity detection part 9 Display part

Claims (6)

A feature information extraction unit that extracts feature information from an image of an object to be registered;
A holding unit for holding a plurality of first feature information extracted by the feature information extraction unit from each of a plurality of first frame images obtained by photographing the object;
The plurality of first feature information and the second frame are different from the first frame, and the feature information extraction unit extracts the second information extracted from the image of one or more second frames obtained by photographing the object. A feature information collating unit that collates feature information and calculates a first similarity;
A control unit that causes the holding unit to hold one feature information having the largest first similarity among the plurality of first feature information as registration information;
Equipped with a,
The feature information matching unit calculates a second similarity between the plurality of first feature information and third feature information of an object different from the object to be registered,
The control unit causes the holding unit to hold feature information having the largest difference between the first similarity and the second similarity among the plurality of first feature information as registration information.
Authentication device comprising a call. A plurality of the second frames;
The feature information matching unit calculates a similarity between each of the plurality of first feature information and each of the plurality of second feature information;
For each of the plurality of first feature information, further comprising a detection unit that detects an average similarity obtained by averaging the similarity with the plurality of second feature information,
The authentication unit according to claim 1, wherein the control unit causes the holding unit to hold, as registration information, feature information having the highest average similarity detected by the detection unit among the plurality of first feature information. apparatus. A holding step of extracting feature information from each of a plurality of first frame images obtained by photographing the object to be registered, and holding the extracted feature information as a plurality of first feature information;
An extraction step of extracting second feature information from an image of one or a plurality of second frames that are different from the first frame and photographed the object;
A first calculation step of calculating a first similarity by collating the plurality of first feature information and the second feature information;
A registration step of registering, as registration information, one feature information having the highest first similarity among the plurality of first feature information;
A second calculation step of calculating a second similarity between the plurality of first feature information and third feature information of an object different from the object to be registered;
It includes,
The registration step registers, as registration information, feature information having the largest difference between the first similarity and the second similarity among the plurality of first feature information.
Authentication wherein a call. A plurality of the second frames;
The first calculating step calculates a similarity between each of the plurality of first feature information and each of the plurality of second feature information;
For each of the plurality of first feature information, the method further includes a detection step of detecting an average similarity obtained by averaging similarities with the plurality of second feature information,
The authentication method according to claim 3 , wherein the registration step registers, as registration information, feature information having the highest average similarity detected in the detection step among the plurality of first feature information. On the computer,
A holding step of extracting feature information from each of a plurality of first frame images obtained by photographing the object to be registered, and holding the extracted feature information as a plurality of first feature information;
An extraction step of extracting second feature information from an image of one or a plurality of second frames that are different from the first frame and photographed the object;
A first calculation step of calculating a first similarity by collating the plurality of first feature information and the second feature information;
A registration step of registering, as registration information, one feature information having the highest first similarity among the plurality of first feature information;
A second calculation step of calculating a second similarity between the plurality of first feature information and third feature information of an object different from the object to be registered;
Was executed,
As the registration step, a process of registering, as registration information, feature information having the largest difference between the first similarity and the second similarity among the plurality of first feature information is executed.
Authentication program which is characterized a call. A plurality of the second frames;
On the computer,
As the first calculation step, a process of calculating a similarity between each of the plurality of first feature information and each of the plurality of second feature information is executed.
For each of the plurality of first feature information, a detection step of detecting an average similarity obtained by averaging similarities with the plurality of second feature information is further executed.
Examples registration step, the plurality of first characteristic information, according to claim 5, characterized in that to execute a process of registering the largest characteristic information average degree of similarity is detected by said detecting step as registration information Certification program.

Images