WO2014182014A1

WO2014182014A1 - Method and system for recognizing face

Info

Publication number: WO2014182014A1
Application number: PCT/KR2014/003931
Authority: WO
Inventors: 정규택
Original assignee: 주식회사 파이브지티
Priority date: 2013-05-07
Filing date: 2014-05-02
Publication date: 2014-11-13
Also published as: KR101297736B1

Abstract

The present invention provides a method and a system for recognizing a face. The present invention stores the number of matchings directly used for face authentication, which is established when a face image in association with face images previously registered for face recognition matches with a face image for authentication, and first compares a face image having the large number of matchings with the face image for the authentication in a face recognition process. Accordingly, the present invention can greatly improve the face recognition speed even when the number of face images increases according to an increase in the number of users. Also, the present invention replaces an image having the small number of matchings by a new face image in a predetermined cycle. Therefore, the present invention can adaptively and naturally reflect a face of a user, which changes according to the growth or aging of the user, in the face recognition.

Description

Facial recognition method and system

The present invention relates to a face recognition method and system, and more particularly to a face recognition method and system capable of updating a face image to the latest image.

The face recognition algorithm of the existing face recognition system has an inconvenience of re-registering the face image for face authentication according to the growth of the person, resulting in problems of mis-authentication and authentication mismatch.

For example, while a five-year-old child registers a face with a face recognition terminal and uses it, after about a year, the face shape changes partly as the child grows. In this case, the face recognition terminal recognizes the changed face of the child. The problem does not occur.

Similarly, in the case of the elderly, the initial facial image registration is normally well authenticated, but as time goes by, the eyelids are stretched or wrinkles are made, so that the shape of the face changes little by little, there is a problem that authentication errors occur in the same way as children.

Accordingly, in order to solve this problem, conventional face recognition apparatuses require periodically registering the latest face image. However, such a process of periodically registering the face image causes inconvenience to users and becomes a major factor that inhibits the commercialization of the face recognition system.

In particular, in the conventional face recognition method, a plurality of face images are registered in each memory for each face and stored in a memory, and when a face image for recognition is input, each face image is compared with each other. Therefore, assuming that there are 10 registered images per person, as the number of registered arguments increases, not only the memory burden is increased, but also the processing speed is increased in a method of authenticating a face by comparing input face images with stored face images. There is a problem that is significantly reduced.

The problem to be solved by the present invention is that the face image is automatically updated to the latest face image even if the user does not need to periodically register the latest face image, and the face is quickly recognized even if the registered face image increases as the user increases. To provide a face recognition method and system that can be.

According to an aspect of the present invention, there is provided a face recognition method comprising: (a) constructing a database storing a plurality of face images for each user; (b) performing an authentication by comparing the authentication image with a plurality of face images stored in the database when an image for face recognition (authentication image) is input; (c) updating the number of matching for the face image used for face recognition by matching with the image for authentication among the plurality of stored face images; And (d) replacing the face image having the smallest number of matches among the plurality of face images at a predetermined time period with an image for authentication.

In addition, the step (b) in the face recognition method according to the preferred embodiment of the present invention may be compared with the image for authentication from the registered face image with a large number of matching times.

In addition, in the face recognition method according to the preferred embodiment of the present invention, the face images of each user stored in the database are classified into a plurality of groups according to the order of the number of matching, and the step (b) is the authentication. The dragon images may be compared in order of face images of a plurality of users belonging to a category having the highest matching number, in order of face images of a plurality of users belonging to a category having the smallest matching number.

Also, in the face recognition method according to the preferred embodiment of the present invention, the step (b) may be performed from the registered face images having the most matching number to the registered face images having the smallest matching number among all the registered face images stored in the database. It can be compared with the image for authentication.

On the other hand, the face recognition method according to another embodiment of the present invention for solving the above problems, (a) building a database storing a plurality of face images for each user; (b) performing an authentication by comparing the authentication image with a plurality of face images stored in the database when an image for face recognition (authentication image) is input; (c) updating the number of matching for the face image used for face recognition by matching with the image for authentication among the plurality of stored face images; And (d) when the face image for authentication is input, performing authentication by comparing the face image with a large number of matching times with the image for authentication, and updating the matching number with respect to the matched face image.

In addition, in the face recognition method according to another preferred embodiment of the present invention, the face images of each user stored in the database are classified into a plurality of groups according to the order of the number of matching, the step (d), The image for authentication may be compared in the order of the face images of the plurality of users belonging to the category with the highest matching number, from the face images of the plurality of users belonging to the category with the smallest matching number.

In the face recognition method according to another preferred embodiment of the present invention, the step (d) may be performed from the registered face images having the highest matching number to the registered face images having the smallest matching number among all the registered face images stored in the database. It can be compared with the image for authentication.

In addition, the recording medium according to a preferred embodiment of the present invention for solving the above-described problems, the face recognition method of the above-described preferred embodiments of the present invention can be read by a computer, and recorded as executable program code.

On the other hand, the face recognition system according to a preferred embodiment of the present invention for solving the above problems, the database for storing a plurality of face images for each user; Shooting unit for taking a face image; An authentication unit configured to perform authentication by comparing the authentication image with a plurality of face images stored in the database when an image for authentication (face for authentication) is input from the photographing unit; A matching number manager configured to update a matching number of face images used for face recognition by matching with the authentication image among the plurality of stored face images; And a database manager that replaces the face image having the least matching number among the plurality of face images with an image for authentication at a predetermined time period.

In addition, the authentication unit of the face recognition system according to an embodiment of the present invention described above may be compared with the image for authentication from the face image with a large number of matching.

In addition, the face images of the users stored in the database of the face recognition system according to an embodiment of the present invention described above are classified into a plurality of groups according to the order of the number of matching, the authentication unit, the image for authentication Are compared in the order of the face images of the plurality of users belonging to the category with the highest number of matching, from the face images of the plurality of users belonging to the category with the smallest number of matching.

In addition, the authentication unit of the face recognition system according to a preferred embodiment of the present invention described above, the image for authentication in the order of registration face images with the smallest number of matching face images from the number of matching faces among all the registered face images stored in the database Can be compared with

On the other hand, the face recognition system according to another preferred embodiment of the present invention for solving the above-described problems, the number of matching matched with a plurality of face images and authentication images for each user indicating the number of times used for face recognition mutually Stored database; Shooting unit for taking a face image; An authentication unit configured to perform authentication by comparing a face image having a large number of matches among the plurality of face images stored in the database with an authentication image when an image for authentication (face for authentication) is input from the photographing unit; And a matching number manager configured to update the number of matchings with respect to the face image used for face recognition by matching the image for authentication among the stored plurality of face images.

In addition, according to the above-described face recognition system according to another embodiment of the present invention, the face images of the users stored in the database are classified into a plurality of groups in order of increasing number of matching, and the authentication unit is configured for the authentication. The images may be compared in the order of the face images of the plurality of users belonging to the category having the highest matching number, from the face images of the plurality of users belonging to the category having the smallest matching number.

In addition, the authentication unit of the face recognition system according to another preferred embodiment of the present invention described above, authentication for authentication from the registered face image with the highest number of matching faces among the registered face images stored in the database in order Can be compared with the image.

The present invention stores the number of matching times used for face recognition by matching the face image with the face image for authentication in association with a face image registered in advance for face recognition. By first comparing with the image for authentication, the face recognition speed can be greatly improved.

In addition, according to the present invention, a face having a small number of matching is replaced with a new face image at a predetermined time period, thereby adaptively and naturally reflecting a face changing due to growth or aging of a user.

1 is a view for explaining the configuration of a face recognition system according to a preferred embodiment of the present invention.

2 is a view for explaining a face recognition method according to a preferred embodiment of the present invention.

3 is a view for explaining a face authentication method according to the prior art.

4A to 4C illustrate a face authentication method and a face image update process according to a preferred embodiment of the present invention.

Hereinafter, preferred embodiments of the present invention will be described with reference to the accompanying drawings.

1 is a view for explaining the configuration of a face recognition system according to an embodiment of the present invention.

First, referring to FIG. 1, a face recognition system according to an exemplary embodiment of the present invention includes a photographing unit 110, an authenticating unit 130, a matching number managing unit 140, a database managing unit 150, and a database. And a preprocessing unit 120 configured to perform preprocessing such as size control, pixel conversion, and face region extraction on the face image photographed by the photographing unit 110.

First, the photographing unit 110 photographs a face image and outputs it to the preprocessor 120. The photographing unit 110 may be configured as a single camera, but may be implemented as a multi-camera including an infrared camera and a color camera. The photographing unit 110 may detect a user around the lens and illuminate the light when the user is detected. It may further include.

The preprocessor 120 outputs a face image by extracting a face region of the user from the input face image and performing a preprocessing process such as converting the face image to a predefined size or converting pixel values. As a pretreatment process performed by the preprocessor 120, a pretreatment method known in the art may be applied.

In particular, the preprocessing unit 120 outputs the preprocessed image to the database manager 150 in the user image registration mode, and to the authentication unit 130 in the authentication mode. The preprocessor 120 of the present invention may exist independently in a separate configuration from the photographing unit 110, or may be integrated into the photographing unit 110 and implemented.

The database 160 stores face images registered by users under the control of the database manager 150. In this case, the database 160 may store the face image itself, may separately extract and store only characteristic information used for face recognition, and may store both the face image itself and feature information extracted therefrom. In addition, the database 160 of the present invention stores the matching number of each registered image together. It should be noted that the "face image" used below is a concept that includes all of these.

Meanwhile, the database manager 150 manages face images of users stored in the database 160. In the user image registration mode, the database manager 150 receives a face image photographed at various angles or a face image photographed at a predetermined time interval. Stored in the base 160 for each user. In addition, the database manager 150 replaces the face image having the lowest matching number among the registered face images of the users with a new face image at a predetermined time period and stores the same in the database 160.

In addition, the database manager 150 may classify the images stored in the database 160 into a plurality of groups according to the matching number representing the number of times the face image of the user is used for authentication for each user.

4A to 4C illustrating a face authentication method and a face image updating process according to an exemplary embodiment of the present invention, among the plurality of face images stored for each user, the number of times each face image is used for user authentication is shown. The matching number of times indicated is stored in association with each other, the face image with the most matching number is classified into the first group, and the face image with the most matching number is classified into the second group. When all the face images are classified in this manner, as many groups as the number of registered face images for each user are generated, and the face images belonging to the last group are replaced with the latest images at regular intervals.

As described above, in general, the user's face changes over time, and the image with the least number of matchings substantially contributes to the user's authentication, and in particular, does not reflect the latest appearance of the user as an old face image. Most likely not. Therefore, the present invention can naturally update the image reflecting the latest face of the user by replacing the face image with the fewest matching times with a new image at a predetermined time period (for example, 1 month, 6 months, 1 year, etc.). In this case, although the image for authentication is preferably used for the new face image to be replaced, the newly registered face image may be used in the face image registration mode.

When the authentication unit 130 receives an image (authentication image) for face recognition from the preprocessor 120 or the imaging unit 110, the authentication unit 130 compares the authentication image with a plurality of face images stored in the database 160. The authentication is performed, and the matching number managing unit 140 updates the matching number of the pre-registered face image used for user authentication by the authenticating unit 130 by +1 and stores the matching number in the database 160.

A preferred embodiment of the present invention improves recognition speed by first comparing face images that are frequently used for performing authentication. Referring to the face authentication method according to the related art shown in FIG. 3, the conventional face authentication method authenticates a user by comparing the input authentication image with all registered face images for each user stored in the database 160. do.

In the example shown in FIG. 3, assuming that 10 users (n = 10) are registered in the database 160, and 6 pictures are registered for each user, the conventional method is taken authentication for authentication. The dragon face image is compared from the first face image of the user 1 to the sixth face image of the user, and then again from the first face image of the user 2 to the sixth face image. However, if the user who is waiting for authentication is the user 10 and the authentication image and the registration face image are matched and approved in the sixth face image among the registered face images of the user 10, 60 face images must be compared during the authentication process. Too much time is required for facial authentication.

Therefore, according to an exemplary embodiment of the present invention, as shown in FIGS. 4A to 4C, the number of times of matching is performed by matching the registered face image and the corresponding face image with the face image for authentication (that is, the face authentication). The number of times of direct use) is stored in association with each other and compared with the face image with a large number of matching times from the image for authentication.

When a face image is registered according to a preferred embodiment of the present invention, as shown in FIG. 4A, all face images are stored in order according to the registration order, and the matching number of each face image is initially set to zero.

Then, each time the authentication process is performed in the authentication unit 130, the matching frequency management unit 140 updates the matching number of the registered face image used for authentication by +1 and registers it in the database 160. The group of face images is adjusted such that a face image with a large number of matching belongs to an upper group and a face image with a small number of matching belongs to a lower group. If this process is repeated, in the example shown in FIG. 4B, for each user, the face image with the most matching number belongs to the first group that is the upper group, and the sixth group with the lowest number of face images is the lowest group. Belong to As a result, more authentication is performed for the face image belonging to the upper group, and the first group includes n number of face images with the best authentication of n users.

In addition, when the authentication image is input and the face recognition is performed, the authentication unit 130 performs authentication by comparing the image for authentication and the registered face image in the order of the upper group to the lower group. That is, as shown in FIG. 4B, when an image for authentication is input, the authentication unit 130 first compares the face images of the most authenticated n users included in the first group, and then performs authentication in the first group. If not, the face authentication is performed by comparing with the registered face images included in the second group.

In this way, by first comparing the face images well authenticated with the image for authentication, the face recognition method of the present invention can significantly improve the face recognition speed compared to the prior art.

In addition, when face recognition is continuously performed, the matching frequency of each registered face image is changed, and movement between groups of registered face images occurs. When the matching frequency is high, when the matching frequency is low, the matching number is moved to a lower group. do.

Meanwhile, the database manager 150 updates the new face image at a predetermined time period. As described above, when the user's face is changed due to growth or aging, since the registered face image that does not properly reflect the current face of the user will be naturally classified into subgroups, the database manager 150 may perform a predetermined time period. The face image of the lowest subgroup (sixth group) is replaced with the new face image. As described above, in the preferred embodiment of the present invention, the registered face image is replaced with the image for authentication for natural replacement. However, the registered face image of the lowest group may be replaced with the image photographed in the separate face registration mode.

The face recognition system according to the preferred embodiment of the present invention has been described so far.

Hereinafter, a face recognition method according to a preferred embodiment of the present invention will be described with reference to FIG. 2.

2 is a view for explaining a face recognition method according to a preferred embodiment of the present invention. However, all functions performed in the face recognition method illustrated in FIG. 2 are performed in the face recognition system described with reference to FIG. 1, and thus, all functions described with reference to FIG. 1 are preferred embodiments of the present invention without an explicit description. It should be noted that all functions performed in the face recognition method according to FIG. 2 and described with reference to FIG. 2 are performed in the face recognition system according to the preferred embodiment of the present invention.

Referring to FIG. 2, in order to use the face recognition method of the present invention, all the users register their face images in the database 160 to construct the database 160 (S210). At this time, the face image may go through the above-described preprocessing process, and the feature information extracted from the face image may be registered together with the face image.

After the database 160 is constructed, the user photographs the face image for authentication by using a camera for face recognition (S220).

After the authentication face image photographed in step S220 is subjected to a preprocessing process (S230), the face image authentication process is performed (S240). The face image authentication process is performed by comparing the registered face image of the group having a large number of matching with the image for authentication, and detailed description thereof is omitted with reference to FIGS. 4A to 4C.

When authentication is performed in step S240, the matching number of registered face images used for authentication is updated, and if necessary, the group of registered images is changed according to the updated matching number (S250).

On the other hand, after the matching number is updated, it is checked whether an image replacement cycle has arrived, and the flow proceeds to step S220 described above where the image update cycle has not arrived (S260).

In operation S260, when the image update period arrives, the current image of the user is reflected in the database 160 by replacing and storing the image having the least matching number as the image for authentication (S270).

In the preferred embodiment of the present invention described above, the registered face images of each user are classified into a plurality of groups according to the number of matching, and the face images belonging to the upper group are compared with the image for authentication.

According to a modified embodiment of the present invention, authentication may be performed by comparing an image having a high number of matches among all registered images with an image for authentication. For example, in the example shown in FIG. 4B, the image for authentication in the order of the face image 3 of the user 1, the face image 5 of the user 1, the face image 1 of the user 2, the face image 1 of the user 1, and the face image 2 of the user 2. Can be compared with In this case, the higher the number of matching, the more the face recognition of the present invention has been performed, and the recognition performance is improved compared to the prior art in that there is a high possibility of using face recognition again.

Further, in another modified embodiment of the present invention, after comparing all of the authentication image in the order of the number of matching the registered face images of one user, compare all of the registered face images of another user in the order of the number of matching Authentication can also be performed.

The invention can also be embodied as computer readable code on a computer readable recording medium. The computer-readable recording medium includes all kinds of recording devices in which data that can be read by a computer system is stored. Examples of computer-readable recording media include ROM, RAM, CD-ROM, magnetic tape, floppy disk, optical data storage, and the like, which are also implemented in the form of a carrier wave (for example, transmission over the Internet). It also includes. The computer readable recording medium can also be distributed over network coupled computer systems so that the computer readable code is stored and executed in a distributed fashion.

So far I looked at the center of the preferred embodiment for the present invention. Those skilled in the art will appreciate that the present invention can be implemented in a modified form without departing from the essential features of the present invention. Therefore, the disclosed embodiments should be considered in descriptive sense only and not for purposes of limitation. The scope of the present invention is shown in the claims rather than the foregoing description, and all differences within the scope will be construed as being included in the present invention.

Claims

(a) constructing a database storing a plurality of face images for each user;

(b) performing an authentication by comparing the authentication image with a plurality of face images stored in the database when an image for face recognition (authentication image) is input;

(c) updating the number of matching for the face image used for face recognition by matching with the image for authentication among the plurality of stored face images; And

and (d) replacing the face image having the lowest matching number among the plurality of face images with an image for authentication at a predetermined time period.
The method of claim 1,

Step (b) is a face recognition method characterized in that the comparison with the image for authentication from the registered face image with a large number of matching.
The method of claim 2,

Face images of the users stored in the database are classified into a plurality of groups in order of increasing number of matching,

In the step (b), the face recognition of the authentication images is compared in order of face images of a plurality of users belonging to a category having the smallest matching number from the face images of the plurality of matching categories. Way.
The method of claim 2, wherein step (b)

The face recognition method of claim 1, wherein the registered face images of the registered face images having the highest number of matching starts from the registered face images having the smallest number of matching among the registered face images stored in the database.
(a) constructing a database storing a plurality of face images for each user;

(b) performing an authentication by comparing the authentication image with a plurality of face images stored in the database when an image for face recognition (authentication image) is input;

(c) updating the number of matching for the face image used for face recognition by matching with the image for authentication among the plurality of stored face images; And

(d) when the face image for authentication is input, performing a verification by comparing the face image having a large number of matching with the image for authentication, and updating the matching number with respect to the matched face image. .
The method of claim 5,

Face images of the users stored in the database are classified into a plurality of groups in order of increasing number of matching,

In the step (d), the face recognition of the authentication images is compared in the order of face images of a plurality of users belonging to a category having the smallest matching number from the face images of the plurality of matching categories. Way.
The method of claim 5, wherein step (d)

The face recognition method of claim 1, wherein the registered face images of the registered face images having the highest number of matching starts from the registered face images having the smallest number of matching among the registered face images stored in the database.
A recording medium in which the face recognition method according to any one of claims 1 to 7 can be read by a computer and recorded in executable program code.
A database storing a plurality of face images for each user;

Shooting unit for taking a face image;

An authentication unit configured to perform authentication by comparing the authentication image with a plurality of face images stored in the database when an image for authentication (face for authentication) is input from the photographing unit;

A matching number manager configured to update a matching number of face images used for face recognition by matching with the authentication image among the plurality of stored face images; And

And a database manager configured to replace the face image having the smallest number of matches among the plurality of face images at a predetermined time period with an image for authentication.
The method of claim 9,

The authentication unit is a face recognition system, characterized in that for comparing the number of face images with a large number of matching images for authentication.
The method of claim 10,

Face images of the users stored in the database are classified into a plurality of groups in order of increasing number of matching,

And the authentication unit compares the image for authentication in order of face images of a plurality of users belonging to a category having the smallest matching number, from face images of a plurality of users belonging to a category having the highest matching number.
The method of claim 10, wherein the authentication unit

The face recognition system of claim 1, wherein the registered face images of the registered face images having the highest number of matching, from the registered face images having the smallest number of matching, are compared with the image for authentication in all registered face images stored in the database.
A database storing matching numbers representing the number of times used for face recognition by matching the plurality of face images and authentication images for each user;

Shooting unit for taking a face image;

An authentication unit configured to perform authentication by comparing a face image having a large number of matches among the plurality of face images stored in the database with an authentication image when an image for authentication (face for authentication) is input from the photographing unit; And

And a matching number manager configured to update a matching number with respect to a face image used for face recognition by matching the image for authentication among the stored plurality of face images.
The method of claim 13,

Face images of the users stored in the database are classified into a plurality of groups in order of increasing number of matching,

And the authentication unit compares the image for authentication in order of face images of a plurality of users belonging to a category having the smallest matching number, from face images of a plurality of users belonging to a category having the highest matching number.
The method of claim 13, wherein the authentication unit

The face recognition system of claim 1, wherein the image is compared with the image for authentication in the order of the registered face image having the most matching number from the registered face image having the smallest number of matching among all the registered face images stored in the database.