JP5508088B2 - Face authentication apparatus, method, and program - Google Patents

Description

  The present invention relates to a face authentication apparatus, method, and program for performing face authentication on a person image taken by an imaging means such as a camera.

  In general, there is known a system that performs face authentication by matching a human image taken by a camera or the like with a database prepared in advance and obtaining a correlation value thereof.

  As such a face authentication system, there is one that attempts to reduce false authentication by performing a weighting calculation in consideration of an outlier (see, for example, Patent Document 1).

  Further, as such a face authentication system, when a correlation value by matching is less than a threshold value for authentication, a new target image is obtained from the imaging unit and matching is repeated again up to a predetermined number of times, and erroneous authentication is performed. Some attempt to reduce this (see, for example, Patent Document 2).

JP 2006-178651 A JP 2009-54052 A

  However, in the case of counting the number of people passing through a certain point, for example, face authentication is used to prevent duplicate counting of the same person. The database for can not be prepared in advance. Thus, under the situation where the database cannot be prepared in advance, there has been a problem that the conventional erroneous authentication reduction as described in Patent Documents 1 and 2 cannot be performed.

  The present invention has been made in view of such a situation, and even if the database for face authentication cannot be prepared in advance, the same person as the person registered in that way while registering in the database It is an object of the present invention to provide a face authentication apparatus, method, and program that can reduce false authentication for face authentication for authenticating whether or not a face is authenticated.

  In order to achieve such an object, the face authentication apparatus according to the present invention includes a feature of the first face image stored in association with the person identification information and a second face image detected from the photographed person image. If the second face image is not determined to be the same person as the person in the person identification information and the predetermined condition is satisfied, the feature information of the second face image is stored, If the feature of the face image is compared with the feature of the third face image detected from an image different from the image, and the third face image is determined to be the same person as the person in the person identification information, The feature of the face image and the feature of the third face image are compared to determine whether or not the second face image is the same person as the person in the person identification information.

  Further, the face authentication apparatus according to the present invention includes a detection unit that detects a face image from a person image captured by the imaging unit, a feature point extraction unit that extracts a feature point from the face image detected by the detection unit, A storage unit; a registration unit that registers the extracted data from which feature points have been extracted by the feature point extracting unit; and a correlation value between the extracted data stored in the storage unit and the extracted data from the feature point extracting unit. A face authentication unit that performs face authentication to determine whether or not the person is the same person, and a re-determination unit that determines that the face authentication unit is not the same person and re-determines the extracted data that satisfies a predetermined condition. It is characterized by that.

  Further, the face authentication method according to the present invention includes a detection step of detecting a face image from a human image photographed by the imaging means, a feature point extraction step of extracting a feature point from the face image detected by the detection step, The registration process for registering the extracted data from which the feature points have been extracted in the feature point extraction process in the storage means, and the correlation between the extracted data stored in the storage means and the extracted data in the feature point extraction process is the same person. A face authentication step for performing face authentication on whether or not, and a re-determination step for determining again that the extracted data satisfying a predetermined condition are determined not to be the same person in the face authentication step. To do.

  Further, the face authentication program according to the present invention includes a detection procedure for detecting a face image from a human image photographed by the imaging means, a feature point extraction procedure for extracting a feature point from the face image detected by the detection procedure, Registration procedure for registering the extracted data from which feature points have been extracted by the feature point extraction procedure in the storage means, and the same person by calculating the correlation value between the extracted data stored in the storage means and the extracted data in the feature point extraction procedure A face authentication procedure for performing face authentication on whether or not, and a re-determination procedure for performing re-determination on extracted data that satisfies a predetermined condition and is determined not to be the same person in the face authentication procedure. Features.

  As described above, according to the present invention, even if the database for face authentication cannot be prepared in advance, it is authenticated whether or not it is the same person as that registered person while performing registration in the database. For face recognition to be performed, false authentication can be reduced.

It is a block diagram which shows the structural example of the people count system as embodiment of this invention. It is a figure which shows the example of a data structure of the registrant database 25. FIG. It is a flowchart which shows the outline of the authentication and registration operation | movement by this embodiment. It is a figure which shows the specific example of the new registration to the registrant database 25, and an additional registration. It is a flowchart which shows the operation example in case registration data is stored to the upper limit in a database. It is a figure which shows the specific example of update operation | movement of the registrant database. It is a flowchart which shows the outline of authentication, a count, and registration operation | movement by this embodiment. It is a flowchart which shows the operation example in the state in which the data temporarily stored by the gray zone exist. It is a figure which shows one specific example when the temporarily stored data is also the same person. It is a figure which shows one specific example when the temporarily preserve | saved data is another person.

  Next, an embodiment in which the face authentication apparatus, method, and program according to the present invention are applied to a people counting system will be described in detail with reference to the drawings.

First, an outline of the present embodiment will be described.
In the present embodiment, for example, when the number of people passing through a certain point is counted, the database for face authentication is not prepared in advance, and the registered person is registered with the database while performing registration in the database. Face-authenticates whether they are the same person.

In addition, when determining whether or not the continuously detected person is the same person, the matching result is not at a level where the person can be determined to be the same person. Temporary storage (determination pending state) is provided, and a function for additionally performing redetermination as a gray zone is provided.
This avoids the problem of mis-authentication that would be counted as multiple people when, for example, one person walks around while walking or looks around, so that the correct number of people can be counted. It is.

  Next, the configuration of the people counting system as the present embodiment will be described with reference to FIG.

  As shown in FIG. 1, the number counting system according to the present embodiment is configured by connecting a camera 1 to an information processing device (face authentication device) 2. The information processing apparatus 2 may be a PC on which the program according to the present embodiment is installed.

  The information processing apparatus 2 includes a face detection unit 21 that detects a human face image from a person-captured image taken by the camera 1, a feature point extraction unit 22 that extracts a feature point from the detected face image, and a reliability that calculates reliability. A degree calculating unit 23, a registration / updating unit 24 for registering or updating a database, a registrant database 25, a face authenticating unit 26, and a counting unit 27 for counting the number of people.

The reliability calculation unit 23 determines how highly the extracted data from which the feature points are extracted by the feature point extraction unit 22 is a human face image based on, for example, the pupil interval and the degree of parallelism of the pupils. Is calculated as follows.
Various known techniques may be used for the face image detection processing by the face detection unit 21, the feature point extraction processing by the feature point extraction unit 22, and the reliability calculation processing by the reliability calculation unit 23, and detailed description thereof is omitted.

  The face authentication unit 26 calculates the similarity between the extracted data registered in the registrant database 25, the extracted data temporarily stored as a gray zone, and the extracted data extracted by the feature point extracting unit 22 as a correlation value. . The correlation value calculation method may use various known techniques, and a detailed description thereof will be omitted.

  When the face authentication unit 26 determines that the person is different, the registration / update unit 24 newly registers the extracted data determined to be another person in the registrant database 25, and the count unit 27 adds 1 to the number of people.

Next, the data structure of the registrant database 25 will be described with reference to the example of FIG.
The registrant database 25 includes, for each registrant determined to be a different person, two kinds of verification data and registration data in person identification information assigned to each registrant (Ms. A and B in the example of FIG. 2). Store a combination of extracted data in association.

  The collation data is data in which the reliability calculated by the reliability calculation unit is associated with the extracted data from which the feature point extraction unit 22 simply extracted the feature points. It is mainly used for verification data with registered data.

  The registered data is data obtained by performing image processing again at the feature point extraction unit 22 for registration and extracting feature points in greater detail (a larger number) than the collation data. It is mainly used for matching with newly extracted data newly extracted by the feature point extracting unit 22.

Next, the operation according to the present embodiment will be described.
First, an outline of the operation from image acquisition by the camera 1 will be described with reference to the flowchart of FIG.

  The camera 1 captures an image, and the face detection unit 21 acquires a moving image, and detects a human face image from the acquired image data. The feature point extraction unit 22 simply extracts the feature points from the face image thus detected, and the reliability calculation unit 23 calculates the reliability (step S1).

  When the reliability calculated in this way is equal to or higher than a predetermined reliability threshold (step S2; Yes), the face authentication unit 26 performs matching with data registered in the registrant database 25 (step S3), and registration is performed. / The update unit 24 performs registration and update in the registrant database 25 (step S4).

  In the face authentication in step S3, when it is determined that the new extracted data by the feature point extracting unit 22 is the same person as the data of any person identification information registered in the registrant database 25, the counting unit 27 counts the number of people. Is not added. Also, the registration / update unit 24 registers the collation data and registration data for the new extracted data that is the determination target in the registrant database 25 in association with the person identification information determined to be the same person.

  In the face authentication of step S3, when it is determined that the new extracted data by the feature point extracting unit 22 is a different person from the data of any person identification information registered in the registrant database 25, the counting unit 27 counts the number of people. Add to +1. Further, the registration / update unit 24 registers the collation data and registration data for the new extracted data that is the determination target in the registrant database 25 in association with the new person identification information.

  If the calculated reliability is lower than a preset reliability threshold (step S2; No), the face authentication unit 26 does not perform face authentication or registration in the registrant database 25, and the extracted data is discarded (step S5). ).

Next, operations of new registration and additional registration in the registrant database 25 will be described with reference to a specific example of FIG.
As a specific example, a case will be described in which the confidence threshold is set in advance as 70%, the authentication threshold for determining the same person as 70%, and the gray threshold temporarily stored as a gray zone as 65%.

  FIG. 4A shows an example in which new registration is performed from a state in which no data is registered in the registrant database 25.

First, for newly extracted data that is detected as a face image by the face detection unit 21 and from which a feature point is newly extracted by the feature point extraction unit 22, data having a reliability of less than 70% that is a reliability threshold is discarded. If the reliability is 70% or more, the feature point extraction unit 22 again extracts feature points in more detail from the face image from which the newly extracted data is extracted, and generates registration data.
The registration / update unit 24 associates the collation data in which the reliability is associated with the newly extracted data and the registration data as one set of data, and associates the collation data and the registration data with the person identification information (A in the example of FIG. 4). ) And register.

  FIG. 4B shows an example in which additional registration is performed on data already registered in the registrant database 25.

  If the reliability of the newly extracted data is 70% or more of the reliability threshold, the face authentication unit 26 calculates a correlation value with the registered data of each person identification information registered in the registrant database 25, and performs face authentication. I do. If the authentication result (correlation value) is 70% or more, which is the authentication threshold, the face authentication unit 26 determines that they are the same person (step S3 in FIG. 3).

  When it is determined that they are the same person, the registration / update unit 24 associates the collation data in which the reliability is associated with the newly extracted data and the registration data as one set of data, and sets the same combination of the collation data and the registration data. Additional registration is performed in association with the person identification information (Mr. A in the example of FIG. 4) that is a person (step S4 in FIG. 3).

When there are a plurality of pairs of verification data and registration data associated with one person identification information in the registrant database 25, one person identification is obtained in the comparison between the authentication result by the face authentication unit 26 and the authentication threshold. You may determine with the same person, when the authentication result of either of the registration data linked | related with information and newly extracted data is more than an authentication threshold value.
In addition, when the verification result of the verification data and the newly extracted data, which is the combination of the verification data with the highest reliability among the verification data and registration data associated with one person identification information, is equal to or greater than the authentication threshold May be determined as the same person.
Further, various determination methods may be used, such as weighting according to preset settings.

  Next, the update operation of the registration information in the registrant database 25 will be described with reference to FIGS. FIG. 6 shows a specific example in the update operation.

  When the upper limit value is not set for the number of registered data in the registrant database 25, the set of collation data and registration data to be registered is registered as it is, but the upper limit value is already set and the upper limit value has already been set. If the registration data has been registered until then, either the overwriting registration or the non-registration operation is performed.

  FIG. 5 is a flowchart showing an operation example when registration data is stored in the database up to the upper limit value.

  First, when the reliability of the newly extracted data is equal to or higher than the reliability threshold and the face authentication unit 26 determines that the registered person is the same as any registered person registered in the registrant database 25, the registration / update unit 24 The collation data with the lowest reliability among all the collation data registered for the person identification information of the same person is compared with the reliability (step S11).

  When the newly extracted data has a lower reliability than the collation data having the lowest reliability in the registered data, as illustrated in FIG. 6B, the registration / update unit 24 causes the registrant database 25. Are not updated and the newly extracted data is discarded (step S13).

  When the newly extracted data has higher reliability than the collation data having the lowest reliability in the registered data, the registration / update unit 24 has the reliability as illustrated in FIG. The lowest verification data is deleted, and the combination of verification data and newly registered data based on the newly extracted data is overwritten and registered in association with the person identification information determined to be the same person (step S14).

  When the reliability of the newly extracted data is the same as the reliability of the collation data having the lowest reliability in the registered data, the registration / update unit 24 registers all the collations registered for the person identification information of the same person Refer to the authentication result of the matching data with the highest reliability in the data and the registered data as a pair, and compare the authentication result with the matching data with the lowest reliability and the authentication result for the newly extracted data ( Step S12).

  When the newly extracted data has a lower authentication result than the registered data that has the lowest authentication result among the registered data, the registration / update unit 24 uses the registrant database 25 as illustrated in FIG. 6B. Are not updated and the newly extracted data is discarded (step S13).

  When the newly extracted data has a higher authentication result than the registered data having the lowest authentication result among the registered data, as illustrated in FIG. 6A, the registration / update unit 24 indicates that the authentication result is The lowest verification data is deleted, and the combination of verification data and newly registered data based on the newly extracted data is overwritten and registered in association with the person identification information determined to be the same person (step S14).

  By performing the database update operation in this manner, the registrant database 25 can be configured with data with higher reliability and matching level.

  Note that the collation data having the lowest reliability among all the collation data registered for the person identification information determined to be the same person as the newly extracted data in the comparison in step S11 described above, and the reliability of the newly extracted data When the difference in the comparison result is less than a predetermined threshold, the determination as to whether or not to update the registrant database 25 is weighted using the authentication result and the reliability according to the preset setting contents. Various determination methods such as performing the above may be used.

  Next, an example of an operation for performing face authentication according to the present embodiment and performing counting of the number of people, registration in the registrant database 25, and the like will be described with reference to the flowchart of FIG.

  First, if the degree of reliability of newly extracted data detected as a face image by the face detection unit 21 and newly extracted feature points by the feature point extraction unit 22 is equal to or higher than a predetermined reliability threshold, the face authentication unit 26 A correlation value between the newly extracted data and the registered data for each person identification information registered in the registrant database 25 is calculated. If the authentication result (correlation value) is equal to or greater than a predetermined authentication threshold (step S21), the face authentication unit 26 determines that the person is the same as the person identification information associated with the registration data that is the authentication result (step S22). . Therefore, the registration / update unit 24 additionally registers a set of collation data and registration data based on the newly extracted data in association with the person identification information determined to be the same person (step S23).

  When the authentication result in step S21 is lower than the predetermined authentication threshold and equal to or higher than the predetermined gray threshold (step S24; Yes), the registration / update unit 24 sets the combination of the collation data and the registration data based on the newly extracted data to gray. Temporarily save as a zone (step S25).

  When the authentication result in step S21 is lower than the predetermined gray threshold value (step S24; No), the face authentication unit 26 determines that each person identification information registered in the registrant database 25 is a different person (step S26). For this reason, the count unit 27 adds 1 to the number of people count. Also, the registration / update unit 24 newly registers a set of collation data and registration data based on the newly extracted data in association with the new person identification information (step S27).

Next, an example of operation in a state where data temporarily stored in the gray zone in step S25 described above will be described with reference to the flowchart of FIG. 8 and the specific examples of FIGS.
FIG. 9 shows a specific example when the newly extracted data temporarily stored as a gray zone is also determined to be the same person.
FIG. 10 shows a specific example when the newly extracted data temporarily stored as a gray zone is determined to be another person after being temporarily stored.

  When newly extracted data whose reliability is equal to or higher than the reliability threshold is acquired within a predetermined time (3 seconds in the examples of FIGS. 9 and 10) after the data is temporarily stored in the gray zone (step S31; Yes), The face authentication unit 26 uses the registrant database 25 to store the person identification information (Mr. A in the examples of FIGS. 9 and 10) of the registration data whose authentication result with the data temporarily stored in the gray zone is equal to or greater than the gray threshold. A correlation value between the associated registration data and the next newly extracted data of the gray zone temporarily stored is calculated (step S32).

  If the authentication result (correlation value) is equal to or greater than a predetermined authentication threshold (step S32; Yes), the face authentication unit 26 newly extracts data temporarily stored in the gray zone and the next new extraction of the gray zone temporarily stored. Both of the data and the data are determined to be the same person as the person identification information (Mr. A in the example of FIG. 9) of the registered data whose authentication result with the data temporarily stored in the gray zone is equal to or greater than the gray threshold (step S33). ).

  Thereafter, the face authentication unit 26 performs matching with the registration data for each of the newly extracted data temporarily stored in the gray zone and the next newly extracted data of the gray zone temporarily stored, and the authentication result is a predetermined authentication threshold value. If it is above (step S34; Yes), both the group of newly extracted data and registration data temporarily stored in the gray zone and the next group of newly extracted data and registration data temporarily stored in the gray zone, A registration / update unit as a set of collation data and registration data associated with person identification information (Mr. A in the example of FIG. 9) of registration data whose authentication result with data temporarily stored in the gray zone is equal to or greater than the gray threshold value 24 additionally registers in the registrant database 25 (step S35).

  If the authentication result in step S34 is lower than the predetermined authentication threshold, the authentication result of only the next newly extracted data and registration data for temporary storage in the gray zone and the data temporarily stored in the gray zone is equal to or greater than the gray threshold. The registration / update unit 24 additionally registers in the registrant database 25 as a set of collation data and registration data associated with the personal identification information of the registered data (step S36). That is, a set of newly extracted data and registration data temporarily stored in the gray zone is discarded without being registered.

If the authentication result in step S32 is lower than the predetermined authentication threshold value and greater than or equal to the predetermined gray threshold value (step S37), temporary storage in the gray zone from step S25 is repeated again.
In this case, data of a plurality of gray zones temporarily stored in this way may be handled together, and weighted according to the authentication result by matching with the registered data of the next newly extracted data of gray zone temporary storage. (For example, only those whose authentication result is equal to or greater than a predetermined threshold may be registered in the database).

  In addition, when temporary storage in the gray zone is repeated in Step S37; Yes, an upper limit value may be set in advance for the number of repeated temporary storages. In this case, when the number of temporary storages in the gray zone exceeds the upper limit value, the oldest temporary storage data is deleted in order.

  When the authentication result in step S32 is lower than the predetermined gray threshold value (step S37; No), the face authentication unit 26 includes the new extracted data temporarily stored in the gray zone and the next new extracted data of the gray zone temporarily stored. Both are determined to be a new person (Mr. B in the example of FIG. 10) that is different from the person identification information of the registered data whose authentication result with the data temporarily stored in the gray zone is equal to or greater than the gray threshold (step S38). The count unit 27 adds 1 to the number of people count.

  Thereafter, the face authentication unit 26 performs matching with the registration data for each of the newly extracted data temporarily stored in the gray zone and the next newly extracted data of the gray zone temporarily stored, and the authentication result is a predetermined authentication threshold value. If it is above (step S39; Yes), both the group of newly extracted data and registration data temporarily stored in the gray zone, and the next group of newly extracted data and registration data temporarily stored in the gray zone, The registration / update unit 24 newly registers in the registrant database 25 as a set of collation data and registration data associated with the person identification information (Mr. B in the example of FIG. 10) determined in step S38 (step S40).

  If the authentication result in step S39 is lower than the predetermined authentication threshold, the collation data and registration for associating only the next newly extracted data and registration data for temporary storage in the gray zone with the person identification information of the new person determined in step S38 The registration / update unit 24 newly registers in the registrant database 25 as a data set (step S41). That is, a set of newly extracted data and registration data temporarily stored in the gray zone is discarded without being registered.

In step S31, after the data is temporarily stored in the gray zone, if new extracted data whose reliability is equal to or higher than the reliability threshold is not acquired within a predetermined time (3 seconds in the examples of FIGS. 9 and 10), Returning to S26, the face authentication unit 26 determines that the data temporarily stored in the gray zone is a different person.
For this reason, the count unit 27 adds 1 to the number of people count. In step S27, the registration / update unit 24 newly registers a set of collation data and registration data temporarily stored in the gray zone in association with new person identification information.

  As described above, according to the above-described embodiment, a database for face authentication does not exist in advance, and it is authenticated whether or not it is the same person as that registered person while registering in the database. Regarding face authentication, it is possible to reduce erroneous authentication such as counting by a plurality of people even though they are the same person.

  Further, for each face, two types of data, that is, detection data for determining reliability and registration data extracted in detail for matching are registered in the registrant database 25 as a set, thereby achieving higher accuracy. You can build a database.

  Further, a database with higher accuracy can be constructed by holding a plurality of data based on the matching result for each person's face and constantly matching and updating the detected data.

  Also, for data extracted continuously for a short time, even if the authentication result is slightly lower than the authentication threshold, the data is temporarily stored as a gray zone and re-determined to widen the range for determining the same person within an appropriate range. be able to. For this reason, it is possible to reduce erroneous authentication such that the face of the same person is continuously detected but newly registered with the face of another person.

  Each of the above-described embodiments is a preferred embodiment of the present invention, and the present invention is not limited to this, and various modifications can be made based on the technical idea of the present invention.

  For example, the trust threshold value, the authentication threshold value, and the gray threshold value are not limited to the values in the above-described example, and may be arbitrary numerical values that are appropriately set by a system administrator or the like.

In addition, regarding the configuration for temporarily storing as a gray zone, the re-determination method is not limited to the above-described one as long as it is a configuration for performing re-determination.
For example, in step S39, matching is performed with registration data for each of the newly extracted data temporarily stored in the gray zone and the next newly extracted data of the gray zone temporarily stored. If the authentication result is lower than a predetermined authentication threshold, The data of the gray zone temporary storage and the next newly extracted data of the gray zone temporary storage may be different persons (B and C).

Further, by recording a processing procedure for realizing the face authentication apparatus as the above-described embodiment on a recording medium as a program, the above-described functions according to each embodiment of the present invention are supplied from the recording medium. Thus, the processing can be realized by causing the CPU of the computer constituting the system to perform processing.
In this case, the present invention can be applied even when an information group including a program is supplied to the output device from the above recording medium or from an external recording medium via a network.
That is, the program code itself read from the recording medium realizes the novel function of the present invention, and the recording medium storing the program code and the signal read from the recording medium constitute the present invention. It will be.
As this recording medium, for example, a hard disk, an optical disk, a magneto-optical disk, a floppy (registered trademark) disk, a magnetic tape, a nonvolatile memory card, a ROM, or the like may be used.

  According to the program according to the present invention, each function in the above-described embodiment can be realized by a computer controlled by the program.

  The present invention relates to a system that does not have a database for face authentication in advance and performs face authentication while constructing a new face database, for example, face authentication used for a store visitor count system. The present invention can be applied to various face authentication devices such as a system and a digital signage viewer count system. Therefore, the present invention can be similarly applied to a system that targets the faces of an unspecified number of people.

1 Camera 2 Information processing device (an example of a face authentication device)
DESCRIPTION OF SYMBOLS 21 Face detection part 22 Feature point extraction part 23 Reliability calculation part 24 Registration / update part 25 Registrant database 26 Face authentication part 27 Count part

Claims (9)

And features of the first face image pre-stored as personal identification information that is hit split in person, by comparing the features of the second face image detected from the image of the subject acquired, the second If the face image is not determined to be the same person as the person in the person identification information and if a predetermined condition is satisfied, the feature information of the second face image is temporarily stored;
Comparing the feature of the first face image with the feature of the third face image detected from an image different from the image;
By determining whether or not the third face image is the same person as the person in the person identification information, the characteristics of the second face image and the characteristics of the third face image are Whether or not the person is the same person is authenticated, and based on the authentication result, both the feature of the second face image and the feature of the third face image or only the feature of the third face image are newly added. A face recognition apparatus that judges whether or not to newly register as person identification information assigned to a registered person . Detecting means for detecting a face image from an image of a person photographed by the imaging means;
Feature point extraction means for extracting feature points of the face image from the face image detected by the detection means;
Storage means for storing registration data as person identification information assigned to a person in advance ;
Registration means for registering in the storage means extracted data obtained by extracting feature points of the face image by the feature point extracting means;
Face authentication means for performing face authentication as to whether or not they are the same person by calculating a correlation value between registration data stored in the storage means and extracted data from the feature point extraction means;
In the face authentication device, comprising: a re-determination unit that determines that the face authentication unit is not the same person and performs re-determination on the extracted data that satisfies a predetermined condition .
The re-determination means compares the registration data by the face authentication means with another extraction data extracted by the feature point extraction means from an image different from the extraction data,
It is authenticated whether the extracted data and the separate extracted data are the same person as the person of the person identification information by determining whether the separate extracted data is the same person as the person of the person identifying information. And
Whether to newly register both the extracted data and the separate extracted data or only the separate extracted data as person identification information assigned to a newly registered person based on the authentication result is determined. Authentication device. The face authentication unit determines that the person is the same person when the correlation value is equal to or greater than a predetermined authentication threshold by matching the registration data stored in the storage unit and the extracted data from the feature point extraction unit,
3. The face according to claim 2, wherein the re-determination means performs the re-determination when the correlation value is lower than the authentication threshold and equal to or higher than a preset gray threshold lower than the authentication threshold. Authentication device.
The registration means registers the extracted data in the storage means in association with person identification information, and the registered data registered in the storage means and the extracted data from the feature point extraction means are the same person by the face authentication means. 4. The face authentication apparatus according to claim 3, wherein when it is determined, the extracted data determined as the same person is registered in the storage unit in association with the same person identification information. The re-determination means, by matching the extracted data extracted from the stored registration data and said feature point extraction means in the storage means as a single personal identification information, the correlation value calculated by the face authentication means the When it is determined that it is lower than the authentication threshold and equal to or higher than the gray threshold, the extracted data to be determined is placed in a determination pending state,
The face is determined by matching newly extracted data newly extracted from the feature point extracting means and registration data stored in the storage means as the one person identification information within a predetermined time after the determination is suspended. When it is determined that the correlation value calculated by the authentication unit is higher than the authentication threshold , a person related to both the new extracted data extracted from the feature point extracting unit and the extracted data in the determination pending state Determining the same person as the registration data stored in the storage means as the one person identification information,
If the registration means determines that the same person, the new extraction data extracted from the feature point extraction means and the extraction data put in the determination pending state are additionally registered in the storage means as the same person identification information The face authentication apparatus according to claim 4, wherein: Comprising reliability calculation means for calculating the reliability of the extracted data from the feature point extraction means;
6. The face authentication apparatus according to claim 4, wherein the registration means performs registration and the face authentication means performs face authentication when the reliability by the reliability calculation means is equal to or greater than a predetermined confidence threshold. .   The registration means is configured such that an upper limit value of the number of extracted data that can be registered in association with one person identification information is preset, and the number of extracted data registered in association with the one person identification information has reached the upper limit value. When the extracted data registered in association with the one person identification information by the face authentication means and the extracted data from the feature point extraction means are determined to be the same person, the extracted data is associated with the one person identification information. If the reliability of the extracted data having the lowest reliability among the extracted data registered is higher than the reliability of the extracted data from the feature point extracting means, the extracted data from the feature point extracting means is not registered. When the reliability of the extracted data from the feature point extracting means is lower, the extracted data with the lowest reliability is deleted, and the extracted data from the feature point extracting means is overwritten and registered. Face authentication apparatus of Motomeko 6 described. A detection step of detecting a face image from an image of a person photographed by the imaging means; a feature point extraction step of extracting a feature point of the face image from the face image detected by the detection step;
A storage step of storing registration data as person identification information assigned to a registered person in advance ;
A registration step of registering in the storage means extracted data obtained by extracting feature points of the face image in the feature point extraction step;
A face authentication step of performing face authentication as to whether or not they are the same person by calculating a correlation value between the registration data stored in the storage means and the extracted data from the feature point extraction means;
In the face authentication method, comprising: a re-determination step in which it is determined that the person is not the same person in the face authentication step and re-determination is performed on the extracted data satisfying a predetermined condition .
In the face authentication step, the registration data is compared with other extracted data extracted by the feature point extracting means from an image different from the extracted data,
It is authenticated whether the extracted data and the separate extracted data are the same person as the person of the person identification information by determining whether the separate extracted data is the same person as the person of the person identifying information. And
Based on the authentication result, in the re-determination step, it is determined whether or not to newly register both the extracted data and the separate extracted data or only the separate extracted data as person identification information assigned to a newly registered person. face authentication method, characterized in that. A detection process for detecting a face image from a human image captured by the imaging means; a feature point extraction process for extracting a feature point of the face image from the face image detected by the detection process ;
A storage process for storing registration data as person identification information assigned to a registered person in advance ;
A registration process for registering extracted data obtained by extracting feature points of a face image in the storage means by the feature point extraction process ;
Face authentication processing for performing face authentication as to whether or not they are the same person by calculating a correlation value between registration data stored in the storage means and extracted data from the feature point extraction means;
A re-determination process in which it is determined that the person is not the same person in the face authentication process and re-determination is performed on the extracted data that satisfies a predetermined condition;
In the face authentication process, the registration data is compared with other extracted data extracted by the feature point extracting means from an image different from the extracted data,
It is authenticated whether the extracted data and the separate extracted data are the same person as the person of the person identification information by determining whether the separate extracted data is the same person as the person of the person identifying information. And
Based on the authentication result, in the re-determination step, it is determined whether or not to newly register both the extracted data and the separate extracted data or only the separate extracted data as person identification information assigned to a newly registered person. A face authentication program for causing a computer to execute processing .

Images