HK1237490A - An anti - cheating network research method, device and system based on bioassay - Google Patents

Description

Anti-cheating network investigation method, device and system based on in-vivo detection

Technical Field

The invention relates to the technical field of image recognition, in particular to an anti-cheating network investigation method, device and system based on in-vivo detection.

Background

At present, with the development of the internet, network research has become one of the main ways for the market research to acquire data. How to identify the real effectiveness of the user in the investigation process is a primary problem in judging whether a questionnaire data sample obtained by network investigation is effective. The existing network questionnaire investigation system mainly performs effectiveness identification judgment in a user registration link, and issues a verification code to allow a user to submit verification, and asks questions to the user from multiple angles based on the effectiveness of the answer of the judgment question. Because the technology of simulating human beings to identify the verification codes and submitting the verification codes is mature at present, and the situation that the machines replace the human beings to answer the questionnaires also happens frequently at present, the real effectiveness of the sample data of network questionnaire investigation is greatly reduced.

Disclosure of Invention

In order to overcome the disadvantages of the prior art, it is an object of the present invention to provide an anti-cheating network investigation method based on liveness detection, which can check the authenticity of a user.

It is another object of the present invention to provide an anti-cheating network investigation apparatus based on liveness detection, which can check authenticity of a user.

It is a further object of the present invention to provide an anti-cheating network investigation system based on liveness detection, which can check authenticity of a user.

One of the purposes of the invention is realized by adopting the following technical scheme:

an anti-cheating network investigation method based on in-vivo detection comprises the following steps:

a model establishing step: establishing an action recognition model library;

an information acquisition step: acquiring action identification information of a user, wherein the action identification information comprises a current feature vector of a human face;

a characteristic comparison step: and comparing the action identification information of the user with the verification characteristic vector in the action identification model base, and if the comparison result is consistent, passing the verification.

Further, the model establishing step specifically includes the following substeps:

an action acquisition step: obtaining verification action information, wherein the verification action information comprises a verification feature vector of a human face, and the verification feature vector is the displacement change of a verification feature point;

establishing an action model library: and establishing an action model library according to the verification action information and the corresponding operation instruction.

Further, the model building step further comprises a face recognition step: and constructing a face recognition model library according to the obtained face recognition information of the user.

Further, the feature comparison step specifically includes the following substeps:

and (3) similarity judgment: and judging whether the similarity between the action identification information and the verification action information in the action identification model library is greater than a preset value, and if so, passing the verification.

Further, the feature alignment step further comprises a face alignment step: and comparing the acquired face recognition information with data in a face recognition model library, and if the comparison result is consistent, executing a similarity judgment step.

The second purpose of the invention is realized by adopting the following technical scheme:

an anti-cheating network investigation device based on in-vivo detection comprises the following modules:

a model building module: the method is used for establishing an action recognition model library;

an information acquisition module: the system comprises a data processing unit, a data processing unit and a data processing unit, wherein the data processing unit is used for acquiring action identification information of a user, and the action identification information comprises a current feature vector of a human face;

a feature comparison module: and the verification module is used for comparing the action identification information of the user with the verification characteristic vector in the action identification model base, and if the comparison result is consistent, the verification is passed.

Further, the model building module specifically includes the following sub-modules:

an action acquisition module: the verification action information comprises a verification feature vector of the human face, and the verification feature vector is the displacement change of the verification feature point;

an action model library establishing module: and the action model base is established according to the verification action information and the operation instruction corresponding to the verification action information.

Further, the model building module further comprises a face recognition module: and the face recognition model library is constructed according to the obtained face recognition information of the user.

Further, the feature comparison module specifically includes the following sub-modules:

a similarity judging module: and the method is used for judging whether the similarity between the action identification information and the verification action information in the action identification model library is greater than a preset value or not, and if so, the verification is passed.

The third purpose of the invention is realized by adopting the following technical scheme:

an anti-cheating network investigation system based on in-vivo detection comprises an actuator, wherein the actuator is used for executing the anti-cheating network investigation method based on in-vivo detection.

Compared with the prior art, the invention has the beneficial effects that:

the invention introduces face recognition to carry out living body detection in a network questionnaire investigation system, and completes the user verification link of facial action according to prompts by combining with the living body detection technology, thereby improving the validity and authenticity of questionnaire sample data and avoiding the occurrence of a large number of invalid questionnaires which are answered fraudulently by a machine.

Drawings

FIG. 1 is a flow chart of the anti-cheating network investigation method based on in-vivo detection according to the present invention;

fig. 2 is a block diagram of an anti-cheating network investigation apparatus based on liveness detection according to the present invention.

Detailed Description

The present invention will be further described with reference to the accompanying drawings and the detailed description, and it should be noted that any combination of the embodiments or technical features described below can be used to form a new embodiment without conflict.

The anti-cheating network investigation system based on living body detection mainly comprises: smart machine, camera, server.

The intelligent equipment: a computer connected with the camera or a mobile terminal with the camera, such as a mobile phone. The user accesses the network questionnaire through the intelligent device to perform related operations, such as registration, login, questionnaire setting, answer and the like.

A camera: for capturing video images of the face of a user during use of the questionnaire system.

The server is provided with: the system comprises a user management module, a questionnaire module and a user verification module; the server is connected with the intelligent equipment through a wireless network or an optical cable;

a user management module: acquiring management user data and authority distribution. The method comprises a registration part, a login part and a user right management part 3:

registering: through the registration process, the user is guided to submit basic identity data information, a password is set, the user is prompted to do designated action through a camera so as to obtain the facial video image data of the registered user, the information is sent to a user authority management module, and a facial recognition model and basic data information of each user are correspondingly established and stored.

Logging in: through the login process, the identity information of the user is verified, the basic data of the user are matched, the user is verified if necessary, and the user information is sent to the user authority management module after the user successfully logs in so as to judge the user authority;

and (3) user authority management: storing and managing basic data information of a user, corresponding face recognition model information and questionnaire setting management authority or questionnaire answering authority; configuring the questionnaire setting or answering authority corresponding to the user through the data information submitted during user registration and the selected account type information, and judging and distributing the authority after the user logs in; establishing a face model corresponding to the user through a face video acquired during user registration for verifying the consistency of the user;

a questionnaire module: the method comprises three parts of questionnaire setting, network questionnaire and questionnaire data analysis; questionnaire setting: the questionnaire management user configures the questionnaire content, the investigation question type and the matched user type through the questionnaire setting module, and sets and finishes issuing the questionnaire.

Network questionnaires: a user checks the question content through a network questionnaire, performs corresponding operation to answer questions and submits information; the web questionnaire includes questionnaire questions set by a questionnaire management user and randomly inserted user authentication questions. Random insertion of user authentication questions can effectively improve the authenticity of questionnaire data. The method mainly comprises the steps that in the process of answering questions by a user, a face action instruction configured by an identification and verification module is randomly extracted, a face appointed action video finished by the user according to prompts is obtained through a camera, the consistency of the user is verified through comparing a face model of the user, and the authenticity of the user is verified through comparing the identification model.

Questionnaire data: and after the questionnaire data analysis module acquires the answer information submitted by the user, analyzing and processing the answer information, and displaying the questionnaire data result to a questionnaire management user for viewing.

A user authentication module: including user consistency verification and user authenticity verification.

User consistency verification: extracting key frames of face video images of users during questionnaire investigation or login verification, extracting face features of the users, comparing face feature models of the users established according to the face video images submitted during user registration, verifying the consistency of the users, and considering the users as the same users when the similarity is greater than 80%.

The main principle process is as follows:

analyzing facial video images submitted by users during registration, extracting key frames, constructing 72 key points of each part of the face according to attributes such as shapes, sizes, positions, distances and the like of facial image outlines such as eye irises, nose wings, mouth corners and the like, then calculating geometrical characteristic quantities of the key points, forming characteristic vectors describing the facial images through the characteristic quantities, establishing a facial characteristic vector set of each registered user as a facial characteristic model of the user, correspondingly storing the facial characteristic vector set to a user authority management module, and comparing the facial characteristic vectors with the facial characteristic model when the consistency of the user is verified in the later period;

when the user is authenticated in the process of user login or answering, extracting a user face video key frame in the authentication process, comparing the feature vectors of 72 points of the face with the face feature model of the corresponding user, and judging the consistency of the user.

The user authenticity verification comprises: identifying the model and verifying authenticity. And analyzing the user face action video image obtained by the camera by using the established identification model, extracting the characteristic vector of the user face action change, and verifying the authenticity of the user by comparing the identification model.

Identifying a model: extracting key frames from a face video image, constructing face key points, and extracting features of the key points; a training set template library is established according to the change information of key points corresponding to the action instruction information when the facial action of a user changes by learning facial action videos of a large number of users and serves as an identification model for verifying the authenticity of the user.

As shown in fig. 1, the present invention provides a method for investigating anti-cheating network based on in-vivo detection, comprising the following steps:

s1: establishing an action recognition model library; the model establishing step specifically comprises the following substeps:

s11: constructing a face recognition model library according to the obtained face recognition information of the user; the facial structure and facial morphology combination of the human has significant changing characteristics when different facial movements change. Through learning and continuous correction, 72 stable key points which can reflect the change of human facial actions and can be used for each angle deviation of the human face under the influence of various light projection external environments are found out according to the attributes such as the shape, the size, the position, the combined distance and the like of facial features, and a recognition model base is established based on the 72 key points; the step is mainly to collect the main face information of the person for later face identification and verification;

s12: obtaining verification action information, wherein the verification action information comprises a verification feature vector of a human face, and the verification feature vector is the displacement change of a verification feature point; dividing the user verification action into 5 instructions of nodding, turning the head to the left, turning the head to the right, blinking and opening the mouth, and establishing a recognition model according to coordinate offset vectors of 72 points of the face of each action instruction;

s13: and establishing an action model library according to the verification action information and the corresponding operation instruction. Through machine learning training and analysis of a large number of videos of facial action changes of users, point coordinate information change data of 72 key points in different facial action changes are counted, coordinate offset vectors of the key points under different action instructions are calculated, and face feature vectors of the action instructions are formed; and storing the extracted verification feature vectors of the human face corresponding to each action instruction template library so as to establish an identification model for user verification. The training process needs to correct the vector set of each instruction by continuously comparing the recognition results; this step mainly identifies whether the real person is answering by verifying the action;

s2: acquiring action identification information of a user, wherein the action identification information comprises a current feature vector of a human face; the step is mainly used for collecting data, and can be arranged in the login period of the user in an interspersed mode or in the process of answering questions of the questionnaire of the user in an interspersed mode;

s3: and comparing the action identification information of the user with the verification characteristic vector in the action identification model base, and if the comparison result is consistent, passing the verification. The feature comparison step specifically comprises the following substeps:

s31: and comparing the acquired face recognition information with data in a face recognition model library, and if the comparison result is consistent, executing a similarity judgment step.

S32: judging whether the similarity between the action identification information and the verification action information in the action identification model library is greater than a preset value, if so, passing the verification; to verify the acquired information.

The main application flow of the invention is as follows:

when accessing a network questionnaire investigation system, a user clicks registration, submits identity information and account type (common answering users and questionnaire management users) related information and starts to establish a user account;

the user verification module randomly generates a group of living body detection user verification instructions, and obtains a face video image of a user for completing a specified face action according to a prompt through a camera;

extracting a feature vector of the user face action in the video image, comparing the feature vector with a feature vector set of corresponding action in the recognition model, and passing verification if the similarity reaches more than 80%;

extracting and storing facial feature vectors of the users, establishing a facial feature model of the users, and storing registration related identity information submitted by the corresponding users to the user management module. The user registration is completed through the above steps.

After registration, the user initiates a login process when he/she needs to use the questionnaire survey system. When the normal account logs in, the user can complete the login only by verifying the password of the user account. If the account is abnormal (if the password passes the password authentication after inputting wrong passwords for many times), the user authentication process is started.

Prompting the user to complete related facial actions such as mouth opening and blinking according to instructions through the camera. After the user verification module obtains the user face video, extracting the face feature vectors of the users in the video, comparing the extracted face feature vectors with the face feature models of the corresponding users stored in the system to judge the consistency of the users, judging the consistency of the appointed actions after the judgment is passed, verifying the authenticity of the users, and if all the verification passes, the users log in successfully, and entering a questionnaire investigation system to perform related operations.

When a user enters the questionnaire system to answer questions, the questionnaire can be randomly inserted with specified facial action question types so as to improve the authenticity of the questionnaire sample. Specifically, after the answer of the general question is completed, the face action question is entered. The system acquires a video image of the appointed facial action finished by the user according to the prompt through a camera, extracts the facial feature vectors of the user in the video, compares the facial feature models of the user to verify the consistency of the user, compares the corresponding action feature vector sets in the identification model to verify the authenticity of the user, finishes answering the question after the verification is passed, and enters the next answering link.

And after the questionnaire data analysis module acquires the complete answer information submitted by the user, analyzing and processing the answer information, and displaying the questionnaire data result to a questionnaire management user for viewing.

The method is the most preferable scheme of the embodiment, and the verification can be performed in other manners, for example, when the user consistency is judged, the method is used for verifying whether the user is the user when submitting a questionnaire or other related operations. In the step, the basic information of the user can be verified in other questioning modes without judgment. In the user authenticity verification link, some non-questions can be randomly generated, so that the user answers in a head nodding and head shaking mode, and meanwhile, the correctness of the answer and the consistency of the feature vector of the user facial action video image and the identification model are judged to finish the real user verification.

As shown in fig. 2, the present invention provides an anti-cheating network investigation apparatus based on in-vivo detection, comprising the following modules:

a model building module: the method is used for establishing an action recognition model library; the model building module specifically comprises the following sub-modules:

a face recognition module: the face recognition model library is constructed according to the obtained face recognition information of the user;

an action acquisition module: the verification action information comprises a verification feature vector of the human face, and the verification feature vector is the displacement change of the verification feature point;

an action model library establishing module: and the action model base is established according to the verification action information and the operation instruction corresponding to the verification action information.

An information acquisition module: the system comprises a data processing unit, a data processing unit and a data processing unit, wherein the data processing unit is used for acquiring action identification information of a user, and the action identification information comprises a current feature vector of a human face;

a feature comparison module: and the verification module is used for comparing the action identification information of the user with the verification characteristic vector in the action identification model base, and if the comparison result is consistent, the verification is passed. The characteristic comparison module specifically comprises the following sub-modules;

a face comparison module: the face recognition module is used for comparing the acquired face recognition information with data in the face recognition model library, and if the comparison result is consistent, the similarity judgment module is executed;

a similarity judging module: and the method is used for judging whether the similarity between the action identification information and the verification action information in the action identification model library is greater than a preset value or not, and if so, the verification is passed.

The above embodiments are only preferred embodiments of the present invention, and the protection scope of the present invention is not limited thereby, and any insubstantial changes and substitutions made by those skilled in the art based on the present invention are within the protection scope of the present invention.

Claims (10)

1. An anti-cheating network investigation method based on in-vivo detection is characterized by comprising the following steps:

a model establishing step: establishing an action recognition model library;

an information acquisition step: acquiring action identification information of a user, wherein the action identification information comprises a current feature vector of a human face;

a characteristic comparison step: and comparing the action identification information of the user with the verification characteristic vector in the action identification model base, and if the comparison result is consistent, passing the verification.

2. The in-vivo detection-based anti-cheating network investigation method according to claim 1, wherein the model building step specifically comprises the following sub-steps:

an action acquisition step: obtaining verification action information, wherein the verification action information comprises a verification feature vector of a human face, and the verification feature vector is the displacement change of a verification feature point;

establishing an action model library: and establishing an action model library according to the verification action information and the corresponding operation instruction.

3. The in-vivo detection-based anti-cheating network investigation method according to claim 2, wherein said model building step further comprises a face recognition step of: and constructing a face recognition model library according to the obtained face recognition information of the user.

4. The anti-cheating network investigation method based on in-vivo detection as claimed in claim 3, wherein the feature comparison step specifically comprises the following sub-steps:

and (3) similarity judgment: and judging whether the similarity between the action identification information and the verification action information in the action identification model library is greater than a preset value, and if so, passing the verification.

5. The fraud network investigation method based on liveness detection according to claim 4, wherein the feature comparison step further comprises a face comparison step of: and comparing the acquired face recognition information with data in a face recognition model library, and if the comparison result is consistent, executing a similarity judgment step.

6. An anti-cheating network investigation device based on in-vivo detection is characterized by comprising the following modules:

a model building module: the method is used for establishing an action recognition model library;

an information acquisition module: the system comprises a data processing unit, a data processing unit and a data processing unit, wherein the data processing unit is used for acquiring action identification information of a user, and the action identification information comprises a current feature vector of a human face;

a feature comparison module: and the verification module is used for comparing the action identification information of the user with the verification characteristic vector in the action identification model base, and if the comparison result is consistent, the verification is passed.

7. The in-vivo detection-based anti-cheating network investigation apparatus according to claim 6, wherein the model building module specifically comprises the following sub-modules:

an action acquisition module: the verification action information comprises a verification feature vector of the human face, and the verification feature vector is the displacement change of the verification feature point;

an action model library establishing module: and the action model base is established according to the verification action information and the operation instruction corresponding to the verification action information.

8. The liveness detection based anti-cheating network investigation apparatus of claim 7, wherein said model building module further comprises a facial recognition module: and the face recognition model library is constructed according to the obtained face recognition information of the user.

9. The in-vivo detection-based anti-cheating network investigation apparatus according to claim 8, wherein the feature comparison module specifically comprises the following sub-modules:

a similarity judging module: and the method is used for judging whether the similarity between the action identification information and the verification action information in the action identification model library is greater than a preset value or not, and if so, the verification is passed.

10. An anti-cheating network investigation system based on in-vivo detection, comprising an actuator for executing the anti-cheating network investigation method based on in-vivo detection according to any one of claims 1-5.