CN105005779A - Face verification anti-counterfeit recognition method and system thereof based on interactive action - Google Patents

Abstract

A face verification anti-counterfeiting recognition method and system based on interactive actions, including initial entry of information on registered static face images and information on multiple registered face action images; When shooting the static face image of the user to be verified, compare and match the features with the stored features. If there is a stored feature with a matching degree that reaches the preset threshold, it meets the verification requirements; it is randomly selected according to the entered face action And prompt the user to be verified to complete the corresponding face action, extract the feature of the action image of the user to be verified, compare and match with the historical verification feature information of the corresponding face action, if the matching rate reaches the threshold, complete the face identity verification, and Add the matching image to the historical verification feature information, otherwise the matching fails or the expected effect is not achieved, return to select the next face action to continue matching, if the preset number of times is exceeded, it will be considered a matching failure and the verification will end. <b />

Description

Face verification anti-counterfeiting recognition method and system based on interactive actions

technical field

The present invention relates to the field of face verification and identification anti-counterfeiting technologies, in particular to a face verification and anti-counterfeiting recognition technology based on the interaction between facial expression actions and users.

Background technique

Face recognition usually uses image acquisition equipment to obtain face image information, extracts facial features, and compares them with historical facial feature information to achieve the purpose of face recognition and complete person identity confirmation or person search. .

The existing face recognition technology is mainly face recognition based on visible light images. This technology has a history of more than 30 years, but this method still has an insurmountable defect, that is, the illumination has a huge impact on the recognition rate.

There are currently two types of face recognition systems: one is face recognition based on static images, and the other is face recognition based on static images based on facial activity.

The first and most common face recognition method. This method is entirely based on static images, first extracting facial features, and then comparing with the feature library. It is characterized by fast recognition speed and high recognition accuracy: at present, the recognition rate for static images has reached more than 97.5%. However, this method cannot confirm the authenticity of the image, so it is easy to be fooled by camouflaged static pictures such as photos and fake faces, and cannot truly prevent counterfeiting and achieve high-level face verification.

The second method is to add target activity detection on the basis of the above-mentioned static face recognition: to judge human activity through thermal induction imaging and modeling a large number of fake face image data. However, the accuracy of human activity detection in this way is not high, and it is easy to be deceived by the activity of static images simulated in various ways, so its anti-counterfeiting effect is still very poor.

Contents of the invention

The technical problem solved by the present invention is to improve the accuracy and robustness of face recognition anti-counterfeiting technology, so that the face itself can be truly recognized.

The problem further solved by the present invention is to greatly improve the accuracy of face anti-counterfeiting, thereby popularizing face passwords to wider application scenarios.

The further problem to be solved by the present invention is to reduce hardware requirements and algorithm complexity.

The technical solution of the present invention provides a face verification anti-counterfeiting recognition method based on interactive actions, including the following steps,

Step 1, for the registered user, the information of the registered static face image and the information of multiple registered face action images are initialized and entered, and the face registration is completed; the information entered includes the image itself and features extracted from the image;

Step 2, wait for the static face image to be tested to be read in;

Step 3: When it is detected that the static face image to be tested is captured by the user to be verified, the features of the static face image to be tested are extracted, and the currently extracted features are compared and matched with the entered features. If the matching degree reaches the preset If the storage characteristics of the threshold meet the verification requirements, go to step 4;

Step 4. According to the characteristics of the static face image to be tested, find the matching registered user and the information of the registered face action image entered, randomly select and prompt the user to be verified to complete the corresponding face action according to the entered face action, and start Detect the action image of the user to be verified;

Step 5: Extract the feature of the action image of the user to be verified, and compare and match it with the historical verification feature information of the corresponding face action. If the matching rate reaches the threshold, the face identity verification will be completed, and the matched image will be added to the history Verify the feature information, otherwise the matching fails or does not achieve the expected effect, return to step 4 to select the next face action to continue matching, if the preset number of times is exceeded, the matching will be considered as a failure and the verification will end.

Moreover, the input of multiple registered facial action images in step 1 includes the input of corresponding human facial action image information of system-specified actions and the corresponding input of human facial action image information of user-defined actions.

Moreover, the feature information of the historical verification in step 5 includes the features of the registered face action image of the corresponding face action and the features of the action image to be tested when the previous verification is passed.

The present invention also provides a face verification anti-counterfeiting recognition system based on interactive actions, which includes the following modules,

The registration input module is used for registering the information of the static face image and the initial input of the information of a plurality of registered face action images for the registered user, and completes the face registration; the information entered includes the image itself and features extracted from the image;

A real-time anti-counterfeiting start-up module, which is used to wait for the static face image to be read in;

The static matching module is used to extract the features of the static face image to be tested when it is detected that the static face image to be tested is captured by the user to be verified, and compare and match the currently extracted features with the entered features. The storage characteristics of the preset threshold meet the verification requirements, and the command action prompts the module to work;

The action prompt module is used to find the matching registered user and the information of the registered face action image entered according to the characteristics of the static face image to be tested, randomly select and prompt the user to be verified to complete the corresponding face action according to the entered face action , and start to detect the action image of the user to be verified;

The verification and judgment module is used to extract features of the action image to be tested by the user to be verified, and compare and match with the historical verification feature information of the corresponding face action. If the matching rate reaches the threshold, the face identity verification is completed, and the matched Add historical verification feature information to the image. Otherwise, the matching fails or the expected effect is not achieved. The command action prompt module selects the next face action to continue matching. If the preset number of times is exceeded, the matching will be deemed as a failure and the verification will end.

Moreover, the entry of multiple registered face action images in the registration entry module includes the entry of corresponding face action image information of the system-specified action and the entry of corresponding face action image information of user-defined actions.

Moreover, the historical verification feature information of the verification judgment module includes the features of the registered face action image of the corresponding face action and the feature of the action image to be tested when the previous verification is passed.

The invention combines self-defined personalized action and interactive recognition and applies it to face recognition anti-counterfeiting, thereby improving the accuracy and robustness of face recognition. The advantages of the present invention are:

(1) The present invention ensures the vitality of the human body through interactive verification, thereby ensuring the robustness of recognition, and largely eliminating false verification of fake faces and static images;

(2) The present invention uses personalized custom action registration to ensure the distinction of individuals, thus enriching the characteristics of individuals, not just identifying and anti-counterfeiting through facial features, thereby improving the accuracy and reliability of verification ;

(3) The present invention adopts an incremental learning and matching method, and continuously improves user action recognition based on historical face recognition verification, thereby better improving recognition accuracy and reliability;

(4) The present invention does not require additional equipment other than the camera, and is simple and convenient to operate, with low system cost and high robustness.

Description of drawings

FIG. 1 is a schematic diagram of a typical implementation environment of an embodiment of the present invention;

Fig. 2 is the specific flow diagram of the embodiment of the present invention;

Fig. 3 is a schematic diagram of the principle of an embodiment of the present invention.

Detailed ways

In order to make the object, technical solution and advantages of the present invention more clear, the present invention will be further described in detail below in combination with specific implementation cases and with reference to the accompanying drawings.

The typical implementation environment of the anti-counterfeiting recognition based on interactive face action verification according to the present invention is shown in Figure 1. The camera is aimed at the face, and the user is prompted to complete the corresponding action on the device screen, and the user cooperates to complete the corresponding action. match recognition. The present invention can be applied to mobile devices, computer devices with cameras, and other devices with image acquisition and certain computing capabilities.

In order to solve the above problems, the present invention discloses a technical method based on interactive face action anti-counterfeiting recognition. The solution adopted is to use the camera to obtain the initial static feature information of the face, and perform a preliminary comparison. If the comparison passes, Prompt the face to do the corresponding action, select the appropriate image from the action image for processing to obtain the corresponding features, compare it with the historical action image features, and judge whether it is the same person, so as to achieve face anti-counterfeiting recognition. During specific implementation, those skilled in the art can use software technology to realize the technical solution of the present invention, for example, setting an information entry program and a face recognition anti-counterfeiting program in the device, so as to facilitate information initialization input and real-time face recognition anti-counterfeiting in advance, and to set the storage information face registration database, etc.

Referring to Fig. 3, the basic steps that the method provided by the embodiment comprises are as follows:

Step 1, for the registered user, pre-register static facial feature image information and multiple registered facial action image information for initial entry, which can be added to the face registration database to complete the face registration; during specific implementation, the information entry program can be started, Perform static face feature image information entry, user system face action image information entry, and user-defined face action image information entry. The user system face action image information input is for the preset system required action input, and for flexible implementation, it also supports the user-defined action input. The entered information includes the image itself and the features extracted from the image. The images and image features generally need to be encrypted and stored to prevent user information from being leaked. During specific implementation, those skilled in the art can set specific feature extraction types and extraction methods by themselves, such as extracting the distance between the eyes under different facial expressions, the shape of the triangle formed by the eyes and the center of the lips under different facial expressions, and so on.

Step 2, wait for the static face image information to be tested to be read in. During specific implementation, the face verification anti-counterfeiting recognition program (ie face recognition anti-counterfeiting program) can be started to recognize the input static face image information in real time, see Figure 3.

Step 3, when it is detected that the static face image of the user to be verified is captured, the static face feature information is extracted, that is, the feature of the static face image to be tested, and the currently extracted static face feature information is combined with the database feature information ( The features stored in the face registration database) are compared and matched. If there is a stored feature with a matching degree reaching the preset threshold, it meets the verification requirements: in specific implementation, the camera can be used to obtain static face image information, extract its features and compare with The registered face is matched to determine whether the face has been registered. If the feature value matching degree reaches a certain preset threshold, it meets the verification requirements; if the face does not meet the requirements, it ends, and if it meets the requirements, go to step 4.

Step 4. According to the static face feature information (features of the static face image to be tested), find the matching registered user and the face action information entered, and prompt the user to be verified to complete the corresponding face action according to the face action entered by the registered user. Action, and start to detect the motion image information of the user to be verified: According to the preset system requirements, the action prompts the verification personnel to perform corresponding actions on the face, such as blinking, sticking out the tongue, turning the head (shaking the head), etc., and according to the pre-defined registration personnel Any action prompts the verifier to perform corresponding actions on the face, such as moving ears, etc., and then proceed to step 5.

Step 5: Process and feature extract the acquired user action image information, that is, extract features from the action image of the user to be verified, and compare and match it with the historical verification feature information of the corresponding face action. If the matching rate reaches the threshold, Then the face identity verification is completed, and the matched image is added to the historical verification feature information, otherwise the matching fails or the expected effect is not achieved, return to step 4 to select the next face action to continue matching, if it exceeds the preset number of times, it will be considered as Match failed, end verification:

The historical verification feature information includes the feature of the registered face action image of the corresponding face action and the feature of the action image to be tested when the previous verification is passed. Of course, when the first verification is successful, the historical verification feature information only has the features of the registered face action image, and the feature matching the action image to be tested when the previous verification is passed has not yet appeared. Select one frame or fuse multiple frames from the face action image information as the image to be matched. The principle of selection is that the image is clear, the illumination is sufficient and uniform, and the contrast with other user images is obvious, and its feature information is extracted. Action image features are jointly compared. Specifically, the image to be compared is compared with the registered image corresponding to the action prompted by the system and all historical images. If the feature comparison reaches a certain threshold, the match is considered passed, and the The matching images and their features are added to the historical image database to complete the face recognition, adding historical images and features to give the system higher robustness, for example, it can reflect the natural and continuous changes of the user's face in different seasons; otherwise, the matching If it does not pass or does not achieve the expected effect, return to step 4 to strengthen the matching result. If it exceeds a certain number of times, such as three times, it is considered a matching failure and the verification ends.

For the convenience of implementation, the present invention further improves the specific implementation process of the anti-counterfeiting technology based on face interactive action identity verification, as shown in Figure 2. After the information entry program is enabled for initial entry, the steps of real-time verification application are as follows:

(1) First turn on the device, and start the anti-counterfeiting program of face recognition, and aim the camera and screen at the face;

(2) The system starts face detection and waits for the static face image information to be read. If no face is detected within a certain period of time, such as 30 seconds, the system completes the anti-counterfeiting process of face verification. The identity verification fails and there is no verification output , and close the face verification program; if a face is detected, go to step (3);

(3) Extract the face feature information, compare the face feature information with the face registration database, if it is detected to match a certain face in the face registration database, go to step (4), otherwise output the user does not exist, If the identity verification fails, the matching is completed and the verification procedure is closed;

(4) Extract the matching user's action set (including system or custom actions) from the face registration database, randomly select the user's action, prompt the user to complete the corresponding action, and wait for the user's action image information to be obtained, and the matching times will increase automatically. Go to step (5);

(5) Detecting user actions: first select a representative user action image fused with one or more frames, then extract the feature information of the user action image, and compare it with the historical image database (historical image database, face registration database). The feature information of the action image and the feature information of the registered action image are jointly matched to determine whether the matching degree reaches the expected threshold. If yes, the output verification is successful, the matching is completed, and the verification program is closed; After the number of invalid matches exceeds a preset certain number of times, such as three times, the output verification fails, the matching is completed, and the verification program is closed. During specific implementation, it can also be designed to determine whether the current invalid matching times have reached the upper limit after the number of matching times has been incremented, such as four times.

During specific implementation, those skilled in the art can use software methods to realize the automatic operation of the above process, and can also use modularization to realize. The embodiment of the present invention also provides a face verification anti-counterfeiting recognition system based on interactive actions, including the following modules,

The registration input module is used for registering the information of the static face image and the initial input of the information of a plurality of registered face action images for the registered user, and completes the face registration; the information entered includes the image itself and features extracted from the image;

A real-time anti-counterfeiting start-up module, which is used to wait for the static face image to be read in;

The static matching module is used to extract the features of the static face image to be tested when it is detected that the static face image to be tested is captured by the user to be verified, and compare and match the currently extracted features with the entered features. The storage characteristics of the preset threshold meet the verification requirements, and the command action prompts the module to work;

The action prompt module is used to find the matching registered user and the information of the registered face action image entered according to the characteristics of the static face image to be tested, randomly select and prompt the user to be verified to complete the corresponding face action according to the entered face action , and start to detect the action image of the user to be verified;

The verification and judgment module is used to extract features of the action image to be tested by the user to be verified, and compare and match with the historical verification feature information of the corresponding face action. If the matching rate reaches the threshold, the face identity verification is completed, and the matched Add historical verification feature information to the image. Otherwise, the matching fails or the expected effect is not achieved. The command action prompt module selects the next face action to continue matching. If the preset number of times is exceeded, the matching will be deemed as a failure and the verification will end.

For the specific implementation of each module, reference may be made to the corresponding steps, which will not be described in detail in the present invention.

The specific embodiments described above have further described the purpose, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above descriptions are only specific embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention shall be included in the protection scope of the present invention.

Claims (6)

1., based on a face verification Antiforge recognizing method for interactive action, it is characterized in that: comprise the following steps,

Step 1, for registered user, carries out registering the information of Static Human Face image and the information initializing typing of multiple registration face motion images, completes face registration; The feature that the information of typing comprises image itself and extracts from image;

Step 2, waits for and reads in Static Human Face image to be measured;

Step 3, when shooting user's gained to be verified Static Human Face image to be measured being detected, extracts the feature of Static Human Face image to be measured, the feature of current extraction is compared with the feature of typing and mates, if there is matching degree to reach the storage feature of default threshold value, then conforms with checking requirement, enter step 4;

Step 4, the information of the registered user of coupling and the registration face motion images of typing thereof is found according to the feature of Static Human Face image to be measured, point out user to be verified to complete corresponding human face action according to the human face action random selecting of typing, and start the motion images to be measured detecting user to be verified;

Step 5, the motion images to be measured treating authentication of users extracts feature, verify that characteristic information comparison is mated with the history of corresponding human face action, if matching rate reaches threshold value, then complete face authentication, and this image mated is added history checking characteristic information, otherwise coupling is not passed through or fallen flat, return step 4 and choose next human face action continuation coupling, exceed preset times, then be considered as that it fails to match, terminate checking.

2. according to claim 1 based on the face verification Antiforge recognizing method of interactive action, it is characterized in that: the typing of multiple registration face motion images in step 1, comprise the corresponding face motion images Data Enter of system required movement and the corresponding face motion images Data Enter of User Defined action.

3. according to claim 1 or 2 based on the face verification Antiforge recognizing method of interactive action, it is characterized in that: history checking characteristic information described in step 5, when comprising the feature of the registration face motion images of corresponding human face action and be verified before, mate the feature of motion images to be measured used.

4. based on a face verification anti-counterfeit recognition system for interactive action, it is characterized in that: comprise with lower module,

Registration typing module, for for registered user, carries out registering the information of Static Human Face image and the information initializing typing of multiple registration face motion images, completes face registration; The feature that the information of typing comprises image itself and extracts from image;

Real-time false proof startup module, reads in Static Human Face image to be measured for waiting for;

Static matching module, during for shooting user's gained to be verified Static Human Face image to be measured being detected, extract the feature of Static Human Face image to be measured, the feature of current extraction is compared with the feature of typing and mates, if there is matching degree to reach the storage feature of default threshold value, then conform with checking requirement, command action reminding module works;

Action prompt module, for finding the information of the registered user of coupling and the registration face motion images of typing thereof according to the feature of Static Human Face image to be measured, point out user to be verified to complete corresponding human face action according to the human face action random selecting of typing, and start the motion images to be measured detecting user to be verified;

Checking judge module, motion images to be measured for treating authentication of users extracts feature, verify that characteristic information comparison is mated with the history of corresponding human face action, if matching rate reaches threshold value, then complete face authentication, and this image mated is added history checking characteristic information, otherwise coupling is not passed through or fallen flat, command action reminding module is chosen next human face action and is continued coupling, exceedes preset times, then be considered as that it fails to match, terminate checking.

5. according to claim 4 based on the face verification anti-counterfeit recognition system of interactive action, it is characterized in that: the typing of multiple registration face motion images in registration typing module, comprises the corresponding face motion images Data Enter of system required movement and the corresponding face motion images Data Enter of User Defined action.

6. according to claim 4 or 5 based on the face verification anti-counterfeit recognition system of interactive action, it is characterized in that: history checking characteristic information described in checking judge module, when comprising the feature of the registration face motion images of corresponding human face action and be verified before, mate the feature of motion images to be measured used.