WO2018192406A1 - Identity authentication method and apparatus, and storage medium - Google Patents

Abstract

Disclosed are an identity authentication method and apparatus, and a storage medium. The identity authentication method comprises: providing action prompt information for an object to be verified; acquiring video stream data of the object to be verified, wherein the video stream data is continuous frames of a human face image collected when the object to be verified executes a corresponding action according to the action prompt information; determining a target human face image according to the video stream data; determining the reliability that the object to be verified is a living body according to the target human face image; and carrying out, according to the reliability and the target human face image, identity authentication on the object to be verified.

Description

Authentication method and device, storage medium

The present application claims priority to Chinese Patent Application No. JP-A No. No. No. No. No. No. No. No. No. No. No. .

Technical field

The embodiments of the present invention relate to the field of computer technologies, and in particular, to an identity verification method and apparatus, and a storage medium.

background

With the continuous development and popularization of terminal technologies, applications in terminals are also increasing. Some applications involve a large amount of privacy of users. To ensure information security, a login operation is required during use to confirm the identity of the user.

Technical content

The invention provides an identity verification method and device, and a storage medium.

The embodiments of the present invention provide the following technical solutions:

An authentication method, the method is performed by the network device, including: providing action prompt information to the object to be verified; acquiring video stream data of the object to be verified, the video stream data being the object to be verified according to the action a continuous frame face image acquired when the prompting action is performed; determining a target face image according to the video stream data; determining, according to the target face image, the credibility of the object to be verified as a living body; The credibility and the target face image authenticate the object to be verified.

The embodiment of the invention further provides the following technical solutions:

An authentication device comprising: at least one memory; at least one processor; wherein the at least one memory stores at least one instruction, the at least one instruction being executed by the at least one processor and implementing a method of: The verification object provides the action prompt information; the video stream data of the object to be verified is obtained, and the video stream data is a continuous frame face image collected by the object to be verified according to the action prompt information; Determining a target face image according to the video stream data; determining, according to the target face image, the credibility of the object to be verified as a living body; performing the object to be verified according to the credibility and the target face image Authentication.

The embodiment of the invention further provides the following technical solutions:

A computer storage medium having stored therein computer readable instructions or programs, the computer readable instructions or programs being executed by a processor.

BRIEF DESCRIPTION OF THE DRAWINGS

The technical solutions and other advantageous effects of the present invention will be apparent from the following detailed description of embodiments of the invention.

1 is a schematic flowchart of an identity verification method according to an embodiment of the present invention;

2a is a schematic flowchart of an identity verification method according to an embodiment of the present invention;

FIG. 2b is a schematic flowchart of user identity verification in a conference sign-in system according to an embodiment of the present invention;

3a is a schematic structural diagram of an identity verification apparatus according to an embodiment of the present invention;

FIG. 3b is a schematic structural diagram of another identity verification apparatus according to an embodiment of the present disclosure;

3c is a schematic structural diagram of a verification submodule according to an embodiment of the present invention;

FIG. 4 is a schematic structural diagram of a network device according to an embodiment of the present invention;

Implementation

The technical solutions in the embodiments of the present invention are clearly and completely described in the following with reference to the accompanying drawings in the embodiments of the present invention. It is obvious that the described embodiments are only a part of the embodiments of the present invention, but not all embodiments. All other embodiments obtained by a person skilled in the art based on the embodiments of the present invention without creative efforts are within the scope of the present invention.

The embodiment of the invention provides an identity verification method, device and system. The details are described below separately. It should be noted that the numbers of the following examples are not intended to limit the preferred order of the embodiments.

First embodiment

This embodiment will be described from the perspective of an identity verification device, which may be implemented as an independent entity or integrated into a network device, such as a terminal or a server.

An authentication method includes: providing action prompt information to an object to be verified, and acquiring video stream data of the object to be verified, where the video stream data is continuously collected when the object to be verified performs corresponding action according to the action prompt information. Frame a face image, and then determining a target face image according to the video stream data, and determining, according to the target face image, the credibility of the object to be verified as a living body, and then, according to the credibility and the target face image pair The object to be verified is authenticated.

As shown in Figure 1, the specific process of the identity verification method can be as follows:

S101. Provide action prompt information to the object to be verified.

In this embodiment, the action prompt information is mainly used to prompt the user to perform some specified actions, such as shaking the head or blinking, etc., and the display may be displayed through a prompt box or a prompt interface. When the user clicks a button on the interactive interface, such as "brush face login", the action of providing the action prompt information may be triggered.

S102. Obtain video stream data of the object to be verified, where the video stream data is a continuous frame face image collected by the object to be verified according to the action prompt information.

In this embodiment, the video stream data may be a video captured in a predetermined time (for example, one minute), and is mainly used for image data of a user's face, and may be collected by a video capture device such as a camera.

S103. Determine a target face image according to the video stream data.

For example, the foregoing step S103 may specifically include:

1-1. Obtain a key point set of each frame face image in the video stream data, and location information of each key point in the key point set.

In this embodiment, the key points in the key point focus mainly on feature points in the face image, that is, points where the image gray value changes drastically, or points with large curvature on the edge of the image (ie, two edges) Intersection), such as the eyes, eyebrows, nose, mouth and facial contours. Specifically, some deep learning models, such as ASM (Active Shape Model) or AAM (Active Appearance Model), can be used to extract key points. The location information is mainly a two-dimensional coordinate for a certain reference coordinate system, such as a face acquisition interface displayed by the terminal.

1-2. Determine a motion track of the object to be verified according to a key point set and position information of each frame of the face image.

In this embodiment, the motion trajectory mainly refers to a route formed by the entire face or a partial region from the start motion to the end motion when the object to be verified performs corresponding action according to the motion prompt information, such as a blink track, a shake track, and the like. Specifically, the position change information of some important key points (such as eyes, mouth corners, cheek edges, and nose) in each frame of the face image, and the angles and relative distances between the important key points are determined. The three-dimensional face model of the object to be verified is obtained with three-dimensional coordinates of each key point, and then the motion trajectory is determined according to the three-dimensional coordinates of any key point.

1-3. Determine a target face image from the video stream data according to the motion trajectory.

For example, the foregoing steps 1-3 may specifically include:

Determining whether the motion trajectory satisfies a preset condition;

If yes, selecting a face image corresponding to the preset track point from the video stream data as the target face image;

If not, a verification result indicating that the object to be verified is an illegal user is generated.

It can be understood that the preset track point is a preset position point in the motion track.

In this embodiment, the preset condition is mainly determined according to the characteristics of the human body motion. Considering that the human motion is coherent, the preset condition may be set as: the motion track includes a plurality of specified track points, such as a 5° deflection angle. Point, 15° deflection angle point and 30° deflection angle point, etc., or the preset condition can be set as follows: the number of track points in the motion track reaches a certain value, for example, 10. The preset track point may be determined according to actual needs. For example, considering that the more key points on the face image, the more accurate the conclusion is, the 0° deflecting corner point may be selected as the preset track point, that is, the front side is selected. The face image serves as the target face image. Of course, considering that the user may not start collecting from a 0° deflection corner, the preset trajectory point may be a point of a smaller interval range including a 0° deflection angle point, rather than a single point.

There are two main types of illegal users: unknown live users and virtual users. The unknown live users mainly refer to live users who are not registered or authenticated on the system platform. The virtual users mainly refer to a group of photos taken by lawless elements using legitimate users. A pseudo-live user who is forged by a video or a human head model (that is, a screen remake). Specifically, when the motion trajectory meets the specified condition, it indicates that the target face image is not a single photo or multiple photos remake. In this case, it is necessary to further confirm whether the target face image is remake by video according to the image texture feature. Or the forged model of the human head. When the motion trajectory does not meet the specified conditions, such as only a small number of two or three track points, it indicates that the object to be verified is most likely a fake living user falsified by staking a single photo or multiple photos of the user. When it is determined, it can be directly determined as an illegal user, and the user is prompted to perform the detection again.

S104. Determine, according to the target face image, the credibility of the object to be verified as a living body.

In this embodiment, the credibility mainly refers to the degree of credibility of the object to be verified as a living body, which may be expressed in the form of a probability value or a fractional value. The texture of the image that is reticled by the screen is different from the texture of the normal image. Therefore, the reliability of the user to be authenticated may be determined by performing feature analysis on the target facial image, that is, the step S104 may specifically include:

2-1. Determine at least one target key point from the key points of the target face image.

In this embodiment, the target key points mainly include some feature points that are relatively stable in relative position and have distinct distinguishing features, such as two left and right pupils, two left and right mouth corners, and a nose tip, etc., which may be determined according to actual needs.

2-2. Determine a normalized image according to the location information of the target key point and the target face image.

For example, the foregoing step 2-2 may specifically include:

Get the preset position of each target key point;

Calculating the Euclidean distance between each preset position and corresponding position information;

The target face image is similarly transformed according to the Euclidean distance to obtain a normalized image.

In this embodiment, the preset position may be obtained according to a standard face model, where the Euclidean distance refers to a distance between a preset position and position information corresponding to each target key point. The similarity transformation may include operations such as rotation, translation, and scaling. Generally, the image before the similar transformation and the similarly transformed image have the same graphics, that is, the shape of the included graphics does not change. Specifically, by continuously adjusting the size, the rotation angle, and the coordinate position of the target face image, the distance between the preset position of the target key point and the corresponding position information can be minimized, that is, the target face image is unified. Normalize the image to a standard face model.

2-3. Calculating the normalized image by using a preset classification model, and obtaining the reliability of the object to be verified as a living body.

In this embodiment, the preset classification model mainly refers to a trained deep neural network, which can be trained by some deep training models, such as CNN (Convolutional Neural Networks), wherein CNN is a multi-layer The neural network is composed of an input layer, a convolution layer, a pooling layer, a fully connected layer and an output layer, which supports inputting an image of a multi-dimensional input vector directly into the network, thereby avoiding data reconstruction in feature extraction and classification, and greatly reducing The complexity of image processing. When the normalized image is input into the CNN network, the information is transformed from the input layer to the output layer, and the calculation process performed by the CNN network is actually the input (normalized image) and the weight of each layer. The value matrix is multiplied by the point to obtain the final output (ie, the confidence of the object to be verified).

It is easy to understand that the preset classification model needs to be trained according to the sample and the classification information in advance, that is, before the normalized image is calculated by using the preset classification model, the identity verification method may further include:

Obtaining a preset face image set, and category information of each preset face image in the preset face map set;

The convolutional neural network is trained according to the preset image set and category information to obtain a preset classification model.

In this embodiment, since the preset classification model is mainly used to identify whether the user to be verified is a virtual user forged by a screen remake, the preset face image set may include a screen remake photo sample (negative sample) and Normal photo samples (positive samples), the specific sample size can be determined according to actual needs. This category information is usually manually labeled, which can include both remake photos and normal photos.

The training process mainly includes two stages: a forward propagation phase and a backward propagation phase. In the forward propagation phase, each sample X _i (that is, a preset face image) can be input into the n-layer convolutional neural network. to give actual output O _{i, where, O i = F n (...} (F 2 (F 1 (X i W (1)) W (2)) ...) W (n)), i is a positive integer, W ⁽ⁿ⁾ is the weight of the nth layer, and F is an activation function (such as a sigmoid function or a hyperbolic tangent function). By inputting the preset face image set to the convolutional neural network, a weight matrix can be obtained, and then In the backward propagation phase, the difference between each actual output O _i and the ideal output Y _i can be calculated, and the adjustment weight matrix is back-propagated according to the method of minimizing the error, wherein Y _i is obtained according to the category information of the sample X _i For example, if the sample X _i is a normal photo, Y _i can be set to 1. If the sample X _i is a remake photo, Y _i can be set to 0, and finally, the trained volume is determined according to the adjusted weight matrix. The neural network, which is the preset classification model.

S105. Perform identity verification on the object to be verified according to the credibility and the target face image.

For example, the foregoing step S105 may specifically include:

Determining whether the credibility is greater than a first preset threshold;

If yes, the object to be verified is authenticated according to the target face image;

If not, a verification result indicating that the object to be verified is an illegal user is generated.

In this embodiment, the first preset threshold may be determined according to an actual application domain. For example, when the identity verification method is mainly applied to a financial domain with high security requirements, the first preset threshold may be set to be compared. Large, such as 0.9, when the authentication method is mainly applied to a field such as a conference sign-in system that has relatively low security requirements, the first preset threshold may be set to be relatively small, such as 0.5.

Specifically, when the calculated reliability is less than or equal to the first preset threshold, it indicates that the object to be verified is likely to be a virtual user whose screen is remake. At this time, in order to reduce the false positive rate, the user may be prompted. Perform face image acquisition again. When the calculated reliability is greater than the first preset threshold, it indicates that the object to be verified is very likely to be a living user. In this case, it is necessary to further analyze whether the living user is an unknown living user or a registered or authenticated living user. That is, the above step "identifying the object to be verified according to the target face image" may specifically include:

3-1. Divide the target face image into a plurality of face regions according to the key point set of the target face image.

In this embodiment, the face area mainly refers to the facial features, such as eyes, mouth, nose, eyebrows, and cheeks, etc., which are mainly based on the relative positional relationship between the key points to segment the target face image.

3-2. Determine target feature information according to the plurality of face regions.

For example, the foregoing step 3-2 may specifically include:

Performing a feature extraction operation on the face region to obtain a plurality of feature information, each face region corresponding to one feature information;

The plurality of pieces of feature information are reorganized to obtain target feature information.

In this embodiment, the feature extraction may be performed on the face region through the deep learning network, and the extracted features are recombined to obtain the feature string (that is, the target feature information). Since the geometric models corresponding to different face regions are different, In order to improve extraction efficiency and accuracy, different depth learning networks can be used to extract different face regions.

3-3. Perform identity verification on the object to be verified according to the target feature information.

For example, the foregoing step 3-3 may specifically include:

3-3-1. Acquire a stored feature information set, and a user identifier corresponding to each stored feature information in the stored feature information set.

In this embodiment, the user identifier is a unique identifier of the user, which may include a registered account. The stored feature information set includes at least one stored feature information, and the different stored feature information is obtained according to face images of different registered users. In the actual application process, the user identifier of each registered user and the stored feature information are associated in advance, that is, before the foregoing step 3-3-1, the identity verification method may further include:

Obtaining a user registration request, where the user registration request carries a user identifier to be registered and a face image to be registered;

Determining feature information to be registered according to the image of the face to be registered;

The to-be-registered feature information is associated with the to-be-registered user identifier, and the to-be-registered feature information is added to the stored feature information set.

In this embodiment, the to-be-registered face image may be processed by using the methods involved in steps 3-1 and 3-2 to obtain the to-be-registered feature information. The user registration request may be automatically triggered. For example, after the user's face image is collected, the user registration request may be automatically generated, or may be generated by the user, for example, when the user clicks the “finish” button, the user may generate the request. The user registration request may be determined according to actual needs. The face image to be registered may be collected on the spot, or may be uploaded after the user has taken the image in advance.

3-3-2. Calculate the similarity between each stored feature information and the target feature information by using a preset algorithm.

In this embodiment, the preset algorithm may include a joint Bayesian algorithm, which is a classification method of statistics. The main idea is to treat a face as a two-part composition, and a part is a difference between people. The other part is the individual's own differences (such as changes in expression), and the overall similarity is calculated based on the difference between the two parts.

3-3-3. Perform identity verification on the object to be verified according to the similarity and the corresponding user identifier.

For example, the foregoing step 3-3-3 may specifically include:

Determining whether there is a similarity that is not less than a second preset threshold among all calculated similarities;

If yes, the user identifier corresponding to the similarity of the second preset threshold is used as the target user identifier, and the verification result indicating that the object to be verified is the target user identifier is generated;

If not, generate a verification result indicating that the object to be verified is an illegal user.

In this embodiment, the second preset threshold may be determined according to actual needs. For example, a face image of a large number of users may be collected in advance, and each user collects two face images, and then, according to two people collected by each user. The face image calculates the corresponding similarity, and the average value is counted, and the average value is used as the second preset threshold. Generally, the similarity of the two face images taken by the same user at different times due to the individual's own differences The degree is usually slightly less than 1, so that the second predetermined threshold can also be set to be slightly less than 1, such as 0.8.

It should be noted that when the verification result indicating that the object to be verified is the target user identifier is generated, the living user is a registered or authenticated living user. In this case, the user can be directly logged in according to the target user identifier, without the user manually. Enter your password and account number, the method is simple, convenient and fast.

It can be seen from the above that the identity verification method provided by the embodiment provides the action prompt information to the object to be verified, and acquires the video stream data of the object to be verified, and the video stream data is made by the object to be verified according to the action prompt information. a continuous frame face image acquired during the corresponding action, and then determining a target face image according to the video stream data, and determining, according to the target face image, the credibility of the object to be verified as a living body, and then, according to the credibility And the target face image authenticates the object to be verified, and can effectively block various types of attacks such as photos, videos, and human head models in the face recognition process, and the method is simple and the security is high.

Second embodiment

According to the method described in Embodiment 1, the following will be exemplified in further detail.

In this embodiment, the integration of the identity verification device in the network device will be described in detail as an example.

As shown in Figure 2a, an authentication method can be as follows:

S201. The network device acquires a user registration request, where the user registration request carries the to-be-registered user identifier and the to-be-registered face image.

For example, when a user first registers an application system (such as a conference check-in system), the user may be required to provide an account to be registered and a face image to be registered, and the image to be registered may be collected on site, or may be taken in advance by the user. After uploading, after the user clicks the "Finish" button, the user registration request can be generated.

S202. The network device determines the to-be-registered feature information according to the to-be-registered face image, and then associates the to-be-registered feature information with the to-be-registered user identifier, and adds the to-be-registered feature information to the stored feature information set.

For example, a key point extraction may be performed on the face image to be registered, and the face image to be registered is segmented into multiple regions according to the extracted key points, and then the feature extraction is performed on the segmented region by using multiple depth learning networks. And reorganizing these features to obtain the feature information to be registered. The user identifier and the feature information of each registered user are stored in association, so that in the subsequent login process, the network device can verify the identity of the user according to the stored information.

S203. The network device acquires a login request, and provides action prompt information to the object to be verified according to the login request.

For example, referring to FIG. 2b, when the object to be verified clicks the “face login” button on the interactive interface, the login request may be generated. At this time, an action prompt box may be displayed on the interaction interface to prompt the object to be verified. Make specific actions, such as shaking your head.

S204: The network device acquires video stream data of the object to be verified, where the video stream data is a continuous frame face image collected by the object to be verified according to the action prompt information.

For example, the video stream data may be face data collected within a specified time (for example, 1 minute). During the actual collection process, a detection frame can be displayed on the interactive interface, and the user is prompted to put the face into the detection frame to guide the user to stand in a suitable position for collecting video stream data.

S205. The network device acquires a key point set of each frame face image in the video stream data, and location information of each key point in the key point set.

For example, the ASM algorithm can be used to extract a set of key points for each frame of the face image, which can include 88 key points such as eyes, eyebrows, nose, mouth, and facial contours. The location information may be display coordinates of each key point in the detection frame. When the user's face is located in the detection frame, the display coordinates of each key point may be automatically located.

S206. The network device determines, according to a key point set and location information of each frame of the face image, a motion trajectory of the object to be verified.

For example, the positional change information of important key points such as the eyes, the corners of the mouth and the nose, and the angles and relative distances between the important key points can be used to determine the three-dimensional face model of the object to be verified, and The three-dimensional coordinates of each key point, and then the motion trajectory is determined according to the three-dimensional coordinates of any key point.

S207. The network device determines whether the motion track meets the preset condition. If yes, the following step S208 is performed. If not, the verification result indicating that the object to be verified is an illegal user may be generated, and the process returns to step S203.

For example, if the preset condition is: including a 5° deflection angle point, a 10° deflection angle point, a 15° deflection angle point, and a 30° deflection angle point, when the motion trajectory is the user's head from the front side (ie, 0°) When the rotation is 40°, that is, when the motion trajectory includes a deflection angle of 0 to 40°, it can be determined that the preset condition is satisfied. When the motion trajectory is formed by the user's head rotated by 15° from the front side (ie, 0°), that is, when the motion trajectory only includes a deflection angle point of 0 to 15°, it can be determined that the preset condition is not satisfied. You can further prompt the user to verify the failure and inform the reason for the failure, such as the current photo does not meet the requirements, etc., so that the user can shoot again.

S208. The network device selects, from the video stream data, a face image corresponding to the preset track point as the target face image.

For example, the preset track point may be a 0° deflecting corner point. At this time, the target face image is also a face image corresponding to a 0° deflecting corner point in the video stream data.

S209. The network device determines at least one target key point from the key points of the target face image, and determines a normalized image according to the location information of the target key point and the target face image.

For example, the above step S209 may specifically include:

Get the preset position of each target key point;

Calculating the Euclidean distance between each preset position and corresponding position information;

The target face image is similarly transformed according to the Euclidean distance to obtain a normalized image.

For example, the target key point may be five points of left and right pupils, two left and right mouth corners, and a nose point, and the preset position may be two-dimensional coordinates of the five points in the standard face model for the same reference coordinate system. By placing the target face image and the standard face model in the same reference coordinate system, and adjusting the target face image through similar transformation methods such as rotation, translation and zoom, the five in the target face image are made. The points are as close as possible to the corresponding points in the standard face model, and the normalized image of the target face image can be realized to obtain a normalized image.

S210: The network device calculates the normalized image by using a preset classification model, and obtains the credibility of the object to be verified as a living body, and determines whether the credibility is greater than a first preset threshold, and if yes, performs the following In step S211, if not, a verification result indicating that the object to be verified is an illegal user may be generated, and the process returns to step S203.

For example, the preset classification model may be obtained by training a CNN with a large number of screen remake photo samples (negative samples) and normal photo samples (positive samples) in advance. When the normalized image is input into the trained CNN, the image information is transformed from the input layer to the output layer, and finally output through the output layer is a probability value, that is, the reliability. The first preset threshold may be 0.5. In this case, if the reliability is 0.7, the determination may be YES, and if the reliability is 0.3, the determination may be NO.

S211. The network device divides the target face image into a plurality of face regions according to the key point set of the target face image, and determines target feature information according to the plurality of face regions.

For example, the target face image can be segmented based on the relative positional relationship between the key points to obtain a plurality of face regions including eyes, mouth, nose, eyebrows, and cheeks, and then, through different depth learning. The network extracts features from different face regions and reorganizes the extracted features to obtain the target feature information.

S212. The network device calculates, by using a preset algorithm, a similarity between each stored feature information and the target feature information in the stored feature information set, and determines whether the calculated similarity is not less than a second preset threshold. The similarity, if yes, performs the following step S213, and if not, it can generate a verification result indicating that the object to be verified is an illegal user, and returns to the above step S203.

For example, the similarity between each stored feature information and the target feature information can be calculated by the joint Bayesian algorithm to obtain a plurality of similarities {A1, A2...An}, at this time, if {A1, A2. If there is Ai greater than or equal to the second preset threshold in An.., it can be determined as YES. If it does not exist, it can be determined as NO, where i∈(1, 2...n), when it is determined as Otherwise, you can further prompt the user to verify the failure and inform the reason for the failure, such as the inability to find the user.

S213. The network device uses the user identifier corresponding to the similarity of the second preset threshold as the target user identifier, and generates a verification result indicating that the object to be verified is the target user identifier.

For example, the user identifier corresponding to the similarity Ai (that is, the target user identifier) may be used as the authentication result of the object to be verified, and the result may be displayed to the user in the form of prompt information to inform the user that the login is successful.

According to the above, the identity verification method provided by the embodiment, wherein the network device obtains a user registration request, the user registration request carries the to-be-registered user identifier and the to-be-registered face image, and determines the to-be-registered feature according to the to-be-registered face image. Information, after which the feature information to be registered is associated with the user identity to be registered, and the feature information to be registered is added to the stored feature information set, and then a login request is obtained, and an action is provided to the object to be verified according to the login request. After the information is obtained, the video stream data of the object to be verified is obtained, and the video stream data is a continuous frame face image collected by the object to be verified according to the motion prompt information, and each video stream data is acquired. The key point set of the face image and the position information of each key point in the key point set, and then determining the motion track of the object to be verified according to the key point set and position information of each frame face image, and determining Whether the motion track satisfies a preset condition, and if not, it may generate an indication that the object to be verified is not The verification result of the method user, if yes, selecting a face image corresponding to the preset track point from the video stream data as the target face image, and then determining at least one target key point from the key point of the target face image And determining a normalized image according to the location information of the target key point and the target face image, and then calculating the normalized image by using a preset classification model to obtain the reliability of the object to be verified as a living body, and Determining whether the credibility is greater than a first preset threshold, and if yes, dividing the target face image into a plurality of face regions according to the key point set of the target face image, and determining target feature information according to the plurality of face regions, Then, using a preset algorithm, calculating a similarity between each stored feature information and the target feature information in the stored feature information set, and determining whether there is a similarity between the calculated similarities that is not less than a second preset threshold. If yes, the user identifier corresponding to the similarity of the second preset threshold is used as the target user identifier, and the object to be verified is generated. The verification result of the target user identifier can effectively block various types of attacks such as photos, videos and human head models in the face recognition process, and the method is simple, high in security, and the identity can be realized without the user manually inputting the password and the account number. Verification, convenient and fast.

Third embodiment

According to the method described in Embodiment 1 and Embodiment 2, the present embodiment will be further described from the perspective of an identity verification device, which can be integrated in a network device.

Referring to FIG. 3a, FIG. 3a specifically describes an identity verification apparatus according to a third embodiment of the present invention, which may include:

At least one memory;

At least one processor;

Wherein the at least one memory stores at least one instruction module configured to be executed by the at least one processor; wherein the at least one instruction module comprises:

The module 10, the obtaining module 20, the first determining module 30, the second determining module 40, and the verifying module 50 are provided, wherein:

(1) Providing module 10

The module 10 is configured to provide action prompt information to the object to be verified.

In this embodiment, the action prompt information is mainly used to prompt the user to perform some specified actions, such as shaking the head or blinking, etc., and the display may be displayed through a prompt box or a prompt interface. When the user clicks a button on the interactive interface, such as "brush face login", the providing module 10 can be triggered to provide the action prompt information.

(2) acquisition module 20

The obtaining module 20 is configured to acquire video stream data of the object to be verified, where the video stream data is a continuous frame face image collected by the object to be verified according to the motion prompt information.

In this embodiment, the video stream data may be a video captured in a predetermined time (for example, one minute), and is mainly used for image data of the user's face, and the acquiring module 20 may perform the collection by using a video capturing device such as a camera.

(3) First determining module 30

The first determining module 30 is configured to determine a target face image according to the video stream data.

For example, the first determining module 30 can be specifically configured to:

1-1. Obtain a key point set of each frame face image in the video stream data, and location information of each key point in the key point set.

In this embodiment, the key points in the key point focus mainly on feature points in the face image, that is, points where the image gray value changes drastically, or points with large curvature on the edge of the image (ie, two edges) Intersection), such as the eyes, eyebrows, nose, mouth and facial contours. The first determining module 30 can perform key point extraction operations through some deep learning models, such as an ASM (Active Shape Model) or an AAM (Active Appearance Model). The location information is mainly a two-dimensional coordinate for a certain reference coordinate system, such as a face acquisition interface displayed by the terminal.

1-2. Determine a motion track of the object to be verified according to a key point set and position information of each frame of the face image.

In this embodiment, the motion trajectory mainly refers to a route formed by the entire face or a partial region from the start motion to the end motion when the object to be verified performs corresponding action according to the motion prompt information, such as a blink track, a shake track, and the like. Specifically, the first determining module 30 may firstly change the position change information of some important key points (such as eyes, mouth corners, cheek edges, and noses) in each frame of the face image, and the angles between the important key points. And the relative distance to determine the three-dimensional face model of the object to be verified, and obtain the three-dimensional coordinates of each key point, and then determine the motion trajectory according to the three-dimensional coordinates of any key point.

1-3. Determine a target face image from the video stream data according to the motion trajectory.

For example, the foregoing steps 1-3 may specifically include:

Determining whether the motion trajectory satisfies a preset condition;

If yes, selecting a face image corresponding to the preset track point from the video stream data as the target face image;

If not, a verification result indicating that the object to be verified is an illegal user is generated.

In this embodiment, the preset condition is mainly determined according to the characteristics of the human body motion. Considering that the human motion is coherent, the preset condition may be set as: the motion track includes a plurality of specified track points, such as a 5° deflection angle. Point, 15° deflection angle point and 30° deflection angle point, etc., or the preset condition can be set as follows: the number of track points in the motion track reaches a certain value, for example, 10. The preset track point may be determined according to actual needs. For example, considering that the more key points on the face image, the more accurate the conclusion is, the 0° deflecting corner point may be selected as the preset track point, that is, the front side is selected. The face image serves as the target face image. Of course, considering that the user may not start collecting from a 0° deflection corner, the preset trajectory point may be a point of a smaller interval range including a 0° deflection angle point, rather than a single point.

There are two main types of illegal users: unknown live users and virtual users. The unknown live users mainly refer to live users who are not registered or authenticated on the system platform. The virtual users mainly refer to a group of photos taken by lawless elements using legitimate users. A pseudo-live user who is forged by a video or a human head model (that is, a screen remake). Specifically, when the motion trajectory meets the specified condition, it indicates that the target face image is not a single photo or multiple photos remake. In this case, it is necessary to further confirm whether the target face image is remake by video according to the image texture feature. Or the forged model of the human head. When the motion trajectory does not meet the specified conditions, such as only a small number of two or three track points, it indicates that the object to be verified is most likely a fake living user falsified by staking a single photo or multiple photos of the user. When it is determined, it can be directly determined as an illegal user, and the user is prompted to perform the detection again.

(4) Second determining module 40

The second determining module 40 is configured to determine, according to the target face image, the credibility of the object to be verified as a living body.

In this embodiment, the credibility mainly refers to the degree of credibility of the object to be verified as a living body, which may be expressed in the form of a probability value or a fractional value. The second determining module 40 can determine the credibility of the user to be verified by performing feature analysis on the target face image, that is, see the texture of the picture that is retries by the screen. 3b, the second determining module 40 may specifically include a first determining submodule 41, a second determining submodule 42 and a calculating submodule 43, wherein:

The first determining sub-module 41 is configured to determine at least one target key point from the key point of the target face image.

In this embodiment, the target key points mainly include some feature points that are relatively stable in relative position and have distinct distinguishing features, such as two left and right pupils, two left and right mouth corners, and a nose tip, etc., which may be determined according to actual needs.

The second determining sub-module 42 is configured to determine a normalized image according to the location information of the target key point and the target facial image.

For example, the second determining sub-module 42 can be specifically used to:

Get the preset position of each target key point;

Calculating the Euclidean distance between each preset position and corresponding position information;

The target face image is similarly transformed according to the Euclidean distance to obtain a normalized image.

In this embodiment, the preset position may be obtained according to a standard face model, where the Euclidean distance refers to a distance between a preset position and position information corresponding to each target key point. The similarity transformation may include operations such as rotation, translation, and scaling. Generally, the image before the similar transformation and the similarly transformed image have the same graphics, that is, the shape of the included graphics does not change. Specifically, the second determining sub-module 42 can minimize the distance between the preset position of the target key point and the corresponding position information by continuously adjusting the size, the rotation angle, and the coordinate position of the target face image, that is, the distance The target face image is normalized to the standard face model to obtain a normalized image.

The calculation sub-module 43 is configured to calculate the normalized image by using a preset classification model to obtain the reliability of the object to be verified as a living body.

In this embodiment, the preset classification model mainly refers to a trained deep neural network, which can be trained by some deep training models, such as CNN (Convolutional Neural Networks), wherein CNN is a multi-layer The neural network is composed of an input layer, a convolution layer, a pooling layer, a fully connected layer and an output layer, which supports inputting an image of a multi-dimensional input vector directly into the network, thereby avoiding data reconstruction in feature extraction and classification, and greatly reducing The complexity of image processing. When the normalized image is input into the CNN network, the information is transformed from the input layer to the output layer, and the calculation process performed by the CNN network is actually the input (normalized image) and the weight of each layer. The value matrix is multiplied by the point to obtain the final output (ie, the confidence of the object to be verified).

It is easy to understand that the preset classification model needs to be trained according to the sample and the classification information in advance, that is, the identity verification apparatus may further include a training module 60 for:

Before the calculating sub-module 43 calculates the normalized image by using the preset classification model, acquiring a preset face image set and category information of each preset face image in the preset face map set;

The convolutional neural network is trained according to the preset image set and category information to obtain a preset classification model.

In this embodiment, since the preset classification model is mainly used to identify whether the user to be verified is a virtual user forged by a screen remake, the preset face image set may include a screen remake photo sample (negative sample) and Normal photo samples (positive samples), the specific sample size can be determined according to actual needs. This category information is usually manually labeled, which can include both remake photos and normal photos.

The training process mainly includes two phases: a forward propagation phase and a backward propagation phase. In the forward propagation phase, the training module 60 can input each sample X _i (that is, a preset face image) into the n-layer convolution. neural networks, and the actual output O _{i, where, O i = F n (...} (F 2 (F 1 (X i W (1)) W (2)) ...) W (n)), i is A positive integer, W ⁽ⁿ⁾ is the weight of the nth layer, and F is an activation function (such as a sigmoid function or a hyperbolic tangent function). By inputting the preset face image set to the convolutional neural network, a weight matrix can be obtained. Then, in the backward propagation phase, the training module 60 can calculate the difference between each actual output O _i and the ideal output Y _i , and backproper the adjustment weight matrix according to the method of minimizing the error, wherein Y _i is based on the sample X _i type information obtained, for example, if the sample is a normal picture X _i, Y _i is set to be 1, if X _i is a sample picture photographing, the Y _i may be set to 0, and finally, the weights adjusted according to The matrix determines the trained convolutional neural network, which is the pre-defined classification model.

(5) Verification module 50

The verification module 50 is configured to perform identity verification on the object to be verified according to the credibility and the target face image.

For example, the verification module 50 may specifically include a determination sub-module 51, a verification sub-module 52, and a generation sub-module 53, wherein:

The determining sub-module 51 is configured to determine whether the credibility is greater than a first preset threshold.

In this embodiment, the first preset threshold may be determined according to an actual application domain. For example, when the identity verification method is mainly applied to a financial domain with high security requirements, the first preset threshold may be set to be compared. Large, such as 0.9, when the authentication method is mainly applied to a field such as a conference sign-in system that has relatively low security requirements, the first preset threshold may be set to be relatively small, such as 0.5.

The verification sub-module 52 is configured to perform identity verification on the object to be verified according to the target face image if the reliability is greater than the first preset threshold.

In this embodiment, when the calculated reliability is greater than the first preset threshold, it indicates that the object to be verified is very likely to be a living user. At this time, the verification sub-module 52 needs to further analyze that the living user is an unknown living user. The active user who is already registered or authenticated, that is, referring to FIG. 3c, the verification sub-module 52 may specifically include a dividing unit 521, a determining unit 522, and a verifying unit 523, where:

The dividing unit 521 is configured to divide the target face image into a plurality of face regions according to the key point set of the target face image.

In this embodiment, the face area mainly refers to the facial features, such as eyes, mouth, nose, eyebrows, and cheeks, etc., which are mainly based on the relative positional relationship between the key points to segment the target face image.

The determining unit 522 is configured to determine target feature information according to the plurality of face regions.

For example, the determining unit 522 can be specifically configured to:

Performing a feature extraction operation on the face region to obtain a plurality of feature information, each face region corresponding to one feature information;

The plurality of pieces of feature information are reorganized to obtain target feature information.

In this embodiment, the determining unit 522 may perform feature extraction on the face region through the deep learning network, and recombine the extracted features to obtain a feature string (that is, the target feature information), because the geometry corresponding to different face regions Different models, in order to improve extraction efficiency and accuracy, different depth learning networks can be used to extract different face regions.

The verification unit 523 is configured to perform identity verification on the object to be verified according to the target feature information.

For example, the verification unit 523 can be specifically configured to:

3-3-1. Acquire a stored feature information set, and a user identifier corresponding to each stored feature information in the stored feature information set.

In this embodiment, the user identifier is a unique identifier of the user, which may include a registered account. The stored feature information set includes at least one stored feature information, and the different stored feature information is obtained according to face images of different registered users.

In the actual application process, the user identifier of each registered user and the stored feature information are associated in advance, that is, the identity verification device may further include an association module, configured to:

Before the verification unit 523 acquires the stored feature information set and the user identifier corresponding to each stored feature information in the stored feature information set, acquiring a user registration request, the user registration request carrying the to-be-registered user identifier and the to-be-registered face image;

Determining feature information to be registered according to the image of the face to be registered;

The to-be-registered feature information is associated with the to-be-registered user identifier, and the to-be-registered feature information is added to the stored feature information set.

In this embodiment, the association module may process the to-be-registered face image by referring to the method used by the dividing unit 521 and the determining unit 522 to obtain the to-be-registered feature information. The user registration request may be automatically triggered. For example, after the user's face image is collected, the user registration request may be automatically generated, or may be generated by the user, for example, when the user clicks the “finish” button, the user may generate the request. The user registration request may be determined according to actual needs. The face image to be registered may be collected on the spot, or may be uploaded after the user has taken the image in advance.

3-3-2. Calculate the similarity between each stored feature information and the target feature information by using a preset algorithm.

In this embodiment, the preset algorithm may include a joint Bayesian algorithm, which is a classification method of statistics. The main idea is to regard a face as a two-part composition, and a part is a difference between people. The other part is the individual's own differences (such as changes in expression), and the overall similarity is calculated based on the difference between the two parts.

3-3-3. Perform identity verification on the object to be verified according to the similarity and the corresponding user identifier.

Further, the verification unit 523 can be used to:

Determining whether there is a similarity that is not less than a second preset threshold among all calculated similarities;

If yes, the user identifier corresponding to the similarity of the second preset threshold is used as the target user identifier, and the verification result indicating that the object to be verified is the target user identifier is generated;

If not, generate a verification result indicating that the object to be verified is an illegal user.

In this embodiment, the second preset threshold may be determined according to actual needs. For example, a face image of a large number of users may be collected in advance, and each user collects two face images, and then, according to two people collected by each user. The face image calculates the corresponding similarity, and the average value is counted, and the average value is used as the second preset threshold. Generally, the similarity of the two face images taken by the same user at different times due to the individual's own differences The degree is usually slightly less than 1, so that the second predetermined threshold can also be set to be slightly less than 1, such as 0.8.

It should be noted that, when the verification unit 523 generates a verification result indicating that the object to be verified is the target user identifier, it indicates that the living user is a registered or authenticated living user, and at this time, the login can be directly performed according to the target user identifier. No need for users to manually enter passwords and accounts, the method is simple, convenient and fast.

The generating sub-module 53 is configured to generate a verification result indicating that the object to be verified is an illegal user, if the credibility is not greater than the first preset threshold.

In this embodiment, when the calculated reliability is less than or equal to the first preset threshold, it indicates that the object to be verified is likely to be a virtual user whose screen is retrased. At this time, in order to reduce the false positive rate, Prompt the user to re-collect the face image.

In the specific implementation, the foregoing units may be implemented as a separate entity, or may be implemented in any combination, and may be implemented as the same or a plurality of entities. For the specific implementation of the foregoing, refer to the foregoing method embodiments, and details are not described herein.

It can be seen that the identity verification method provided in the above embodiment can be implemented when the processor executes the instructions stored in the memory.

It can be seen from the above that the identity verification device provided by the embodiment provides the action prompt information to the object to be verified by the providing module 10, and the obtaining module 20 acquires the video stream data of the object to be verified, and the video stream data is the object to be verified according to the The action prompt information is used to make a continuous frame face image collected when the action is performed. Then, the first determining module 30 determines the target face image according to the video stream data, and the second determining module 40 determines the object to be verified according to the target face image. For the credibility of the living body, the verification module 50 then authenticates the object to be verified according to the credibility and the target face image, and can effectively block various types of photos, videos, and head models in the face recognition process. The attack is simple and safe.

Fourth embodiment

Correspondingly, the embodiment of the present invention further provides an identity verification system, which includes any of the identity verification devices provided by the embodiments of the present invention. For details, refer to the third embodiment.

The network device may provide action prompt information to the object to be verified; and obtain video stream data of the object to be verified, where the video stream data is a continuous frame face image collected by the object to be verified according to the motion prompt information. Determining a target face image according to the video stream data; determining, according to the target face image, the credibility of the object to be verified as a living body; and authenticating the object to be verified according to the credibility and the target face image.

For the specific implementation of the foregoing devices, refer to the foregoing embodiments, and details are not described herein again.

The system for generating the traffic information may include any of the identity verification devices provided by the embodiments of the present invention. Therefore, the beneficial effects of any of the identity verification devices provided by the embodiments of the present invention can be implemented. The embodiment is not described here.

Fifth embodiment

Correspondingly, the embodiment of the present invention further provides a network device, as shown in FIG. 4, which shows a schematic structural diagram of a network device according to an embodiment of the present invention, specifically:

The network device can include a processor 701 of one or more processing cores, a memory 702 of one or more computer readable storage media, a radio frequency (RF) circuit 703, a power source 704, an input unit 705, and a display unit 707 And other components. It will be understood by those skilled in the art that the network device structure illustrated in FIG. 4 does not constitute a limitation to the network device, and may include more or less components than those illustrated, or some components may be combined, or different component arrangements. among them:

The processor 701 is the control center of the network device, interconnecting various portions of the entire network device using various interfaces and lines, by running or executing software programs and/or modules stored in the memory 702, and recalling stored in the memory 702. Data, performing various functions of the network device and processing data, thereby performing overall monitoring of the network device. Optionally, the processor 701 may include one or more processing cores; preferably, the processor 701 may integrate an application processor and a modem processor, where the application processor mainly processes an operating system, a user interface, an application, and the like. The modem processor primarily handles wireless communications. It can be understood that the above modem processor may not be integrated into the processor 701.

The memory 702 can be used to store software programs and modules, and the processor 701 executes various functional applications and data processing by running software programs and modules stored in the memory 702. The memory 702 may mainly include a storage program area and a storage data area, wherein the storage program area may store an operating system, an application required for at least one function (such as a sound playing function, an image playing function, etc.), and the like; the storage data area may be stored according to Data created by the use of network devices, etc. Moreover, memory 702 can include high speed random access memory, and can also include non-volatile memory, such as at least one magnetic disk storage device, flash memory device, or other volatile solid state storage device. Accordingly, memory 702 can also include a memory controller to provide processor 701 access to memory 702.

The RF circuit 703 can be used for receiving and transmitting signals during the process of transmitting and receiving information. Specifically, after receiving the downlink information of the base station, it is processed by one or more processors 701; in addition, the data related to the uplink is sent to the base station. Generally, the RF circuit 703 includes, but is not limited to, an antenna, at least one amplifier, a tuner, one or more oscillators, a Subscriber Identity Module (SIM) card, a transceiver, a coupler, and a Low Noise Amplifier (LNA). , duplexer, etc. In addition, the RF circuit 703 can also communicate with the network and other devices through wireless communication. The wireless communication can use any communication standard or protocol, including but not limited to Global System of Mobile communication (GSM), General Packet Radio Service (GPRS), and Code Division Multiple Access (CDMA). Code Division Multiple Access), Wideband Code Division Multiple Access (WCDMA), Long Term Evolution (LTE), e-mail, Short Messaging Service (SMS), and the like.

The network device also includes a power source 704 (such as a battery) that supplies power to the various components. Preferably, the power source 704 can be logically coupled to the processor 701 through the power management system to manage functions such as charging, discharging, and power management through the power management system. . The power supply 704 can also include any one or more of a DC or AC power source, a recharging system, a power failure detection circuit, a power converter or inverter, a power status indicator, and the like.

The network device can also include an input unit 705 that can be used to receive input numeric or character information and to generate keyboard, mouse, joystick, optical or trackball signal inputs related to user settings and function controls. In particular, in one particular embodiment, input unit 705 can include a touch-sensitive surface as well as other input devices. Touch-sensitive surfaces, also known as touch screens or trackpads, collect touch operations on or near the user (such as the user using a finger, stylus, etc., any suitable object or accessory on a touch-sensitive surface or touch-sensitive Operation near the surface), and drive the corresponding connecting device according to a preset program. Alternatively, the touch sensitive surface may include two parts of a touch detection device and a touch controller. Wherein, the touch detection device detects the touch orientation of the user, and detects a signal brought by the touch operation, and transmits the signal to the touch controller; the touch controller receives the touch information from the touch detection device, converts the touch information into contact coordinates, and sends the touch information. The processor 701 is provided and can receive commands from the processor 701 and execute them. In addition, touch-sensitive surfaces can be implemented in a variety of types including resistive, capacitive, infrared, and surface acoustic waves. In addition to the touch-sensitive surface, the input unit 705 can also include other input devices. Specifically, other input devices may include, but are not limited to, one or more of a physical keyboard, function keys (such as volume control buttons, switch buttons, etc.), trackballs, mice, joysticks, and the like.

The network device can also include a display unit 706 that can be used to display information entered by the user or information provided to the user and various graphical user interfaces of the network device, the graphical user interface can be represented by graphics, text, icons, Video and any combination of them. The display unit 706 can include a display panel. Optionally, the display panel can be configured in the form of a Liquid Crystal Display (LCD), an Organic Light-Emitting Diode (OLED), or the like. Further, the touch-sensitive surface may cover the display panel, and when the touch-sensitive surface detects a touch operation thereon or nearby, it is transmitted to the processor 701 to determine the type of the touch event, and then the processor 701 displays the type according to the touch event. A corresponding visual output is provided on the panel. Although in FIG. 4, the touch-sensitive surface and display panel are implemented as two separate components to perform input and input functions, in some embodiments, the touch-sensitive surface can be integrated with the display panel to implement input and output functions.

Although not shown, the network device may further include a camera, a Bluetooth module, and the like, and details are not described herein again. Specifically, in this embodiment, the processor 701 in the network device loads the executable file corresponding to the process of one or more applications into the memory 702 according to the following instructions, and is stored and stored by the processor 701. The application in memory 702, thereby implementing various functions, as follows:

Providing action prompt information to the object to be verified;

Obtaining video stream data of the object to be verified, where the video stream data is a continuous frame face image collected when the object to be verified performs corresponding action according to the action prompt information;

Determining a target face image according to the video stream data;

Determining, according to the target face image, the credibility of the object to be verified as a living body;

The object to be verified is authenticated according to the credibility and the target face image.

For the implementation of the foregoing operations, refer to the foregoing embodiments, and details are not described herein again.

The network device can implement the effective effects of any of the identity verification devices provided by the embodiments of the present invention. For details, refer to the foregoing embodiments, and details are not described herein again.

A person skilled in the art may understand that all or part of the various steps of the foregoing embodiments may be performed by a program to instruct related hardware. The program may be stored in a computer readable storage medium, and the storage medium may include: Read Only Memory (ROM), Random Access Memory (RAM), disk or optical disk.

Embodiments of the present invention also provide a computer storage medium having stored therein computer readable instructions or programs, the computer readable instructions or programs being executed by a processor.

The foregoing describes an identity verification method, apparatus, and system provided by the embodiments of the present invention. The principles and implementation manners of the present invention are described in the following. The description of the foregoing embodiments is only for helping understanding. The method of the present invention and its core idea; at the same time, those skilled in the art, according to the idea of the present invention, there are some changes in the specific implementation manner and application scope. In summary, the content of the present specification should not be understood as Limitations of the invention.

Claims (25)

1. An authentication method is performed by a network device, and the method includes:

   Providing action prompt information to the object to be verified;

   Acquiring the video stream data of the object to be verified, where the video stream data is a continuous frame face image collected when the object to be verified performs corresponding action according to the action prompt information;

   Determining a target face image according to the video stream data;

   Determining, according to the target face image, the credibility of the object to be verified as a living body;

   And authenticating the object to be verified according to the credibility and the target face image.
2. The identity verification method according to claim 1, wherein the determining the target face image according to the video stream data comprises:

   Obtaining a key point set of each frame face image in the video stream data, and location information of each key point in the key point set;

   Determining a motion trajectory of the object to be verified according to a key point set and position information of each frame of the face image;

   Determining the target face image from the video stream data according to the motion trajectory.
3. The identity verification method according to claim 2, wherein the determining the target face image from the video stream data according to the motion trajectory comprises:

   Determining whether the motion trajectory meets a preset condition;

   If yes, selecting a face image corresponding to the preset track point from the video stream data, and using the selected face image as the target face image; the preset track point is preset in the motion track Location point

   If not, generating a verification result indicating that the object to be verified is an illegal user.
4. The identity verification method according to claim 2, wherein the determining, according to the target face image, the credibility of the object to be verified as a living body comprises:

   Determining at least one target key point from a key point of the target face image;

   Determining a normalized image according to the location information of the target key point and the target face image;

   The normalized image is calculated by using a preset classification model to obtain the reliability of the object to be verified as a living body.
5. The identity verification method according to claim 4, wherein the determining the normalized image according to the location information of the target keypoint and the target facial image comprises:

   Get the preset position of each target key point;

   Calculating an Euclidean distance between a preset position of each target key point and corresponding position information in the target face image;

   Performing a similar transformation on the target face image according to the Euclidean distance to obtain the normalized image.
6. The identity verification method according to claim 4, further comprising: before calculating the normalized image by using a preset classification model, further comprising:

   Obtaining a preset face image set, and category information of each preset face image in the preset face map set;

   The convolutional neural network is trained according to the preset image set and the category information to obtain the preset classification model.
7. The identity verification method according to any one of claims 2-6, wherein the authenticating the object to be verified according to the credibility and the target face image comprises:

   Determining whether the credibility is greater than a first preset threshold;

   If yes, performing identity verification on the object to be verified according to the target face image;

   If not, generating a verification result indicating that the object to be verified is an illegal user.
8. The identity verification method according to claim 7, wherein the authenticating the object to be verified according to the target face image comprises:

   Dividing the target face image into a plurality of face regions according to a key point set of the target face image;

   Determining target feature information according to the plurality of face regions;

   And authenticating the object to be verified according to the target feature information.
9. The identity verification method according to claim 8, wherein the determining the target feature information according to the plurality of face regions comprises:

   Performing a feature extraction operation on the plurality of face regions to obtain a plurality of pieces of feature information; each face region corresponding to one piece of feature information;

   Recombining the plurality of pieces of feature information to obtain the target feature information.
10. The identity verification method according to claim 8, wherein the authenticating the object to be verified according to the target feature information comprises:

    Obtaining a stored feature information set, and a user identifier corresponding to each stored feature information in the stored feature information set;

    Calculating a similarity between each stored feature information and the target feature information by using a preset algorithm;

    And authenticating the object to be verified according to the similarity and the corresponding user identifier.
11. The identity verification method according to claim 10, wherein the authenticating the object to be verified according to the similarity and the corresponding user identifier comprises:

    Determining whether there is a similarity that is not less than a second preset threshold among all calculated similarities;

    If yes, the user identifier corresponding to the similarity of the second preset threshold is used as the target user identifier, and the verification result indicating that the object to be verified is the target user identifier is generated;

    If not, generate a verification result indicating that the object to be verified is an illegal user.
12. The identity verification method according to claim 10, further comprising: before acquiring the stored feature information set and the user identifier corresponding to each stored feature information in the stored feature information set, further comprising:

    Obtaining a user registration request, where the user registration request carries a user identifier to be registered and a face image to be registered;

    Determining feature information to be registered according to the face image to be registered;

    Associating the to-be-registered feature information with the to-be-registered user identifier, and adding the to-be-registered feature information to the stored feature information set.
13. An authentication device includes:

    At least one memory;

    At least one processor;

    Wherein the at least one memory stores at least one instruction, the at least one instruction being executed by the at least one processor and implementing the following method:

    Providing action prompt information to the object to be verified;

    Acquiring the video stream data of the object to be verified, where the video stream data is a continuous frame face image collected when the object to be verified performs corresponding action according to the action prompt information;

    Determining a target face image according to the video stream data;

    Determining, according to the target face image, the credibility of the object to be verified as a living body;

    And authenticating the object to be verified according to the credibility and the target face image.
14. The identity verification device according to claim 13, wherein the determining the target face image according to the video stream data comprises: acquiring a key point set of each frame face image in the video stream data, and the Key points are collected for position information of each key point; determining a motion trajectory of the object to be verified according to a key point set and position information of each frame of the face image; determining a target person from the video stream data according to the motion trajectory Face image.
15. The identity verification device according to claim 14, wherein the determining, according to the target face image, the credibility of the object to be verified as a living body comprises: determining at least a key point set of the target face image a target key point; determining a normalized image according to the position information of the target key point and the target face image; calculating the normalized image by using a preset classification model, and obtaining the object to be verified as a living body Reliability.
16. The identity verification device according to claim 14, wherein the determining the target face image from the video stream data according to the motion trajectory comprises: determining whether the motion trajectory satisfies a preset condition; if so, Selecting a face image corresponding to the preset track point in the video stream data, and using the selected face image as the target face image; the preset track point is a preset position point in the motion track; No, a verification result indicating that the object to be verified is an illegal user is generated.
17. The identity verification device according to claim 15, wherein the determining the normalized image according to the location information of the target key point and the target facial image comprises: acquiring a preset position of each target key point; Calculating an Euclidean distance between a preset position of each target key point and corresponding position information in the target face image; performing similar transformation on the target face image according to the Euclidean distance, to obtain the Normalize the image.
18. The identity verification device according to claim 15, wherein before calculating the normalized image by using a preset classification model, the method further comprises: acquiring a preset face image set, and the preset face map set Type information of each preset face image; training the convolutional neural network according to the preset image set and the category information to obtain the preset classification model.
19. The identity verification device according to any one of claims 14 to 18, wherein

    The authenticating the object to be verified according to the credibility and the target face image includes: determining whether the credibility is greater than a first preset threshold; if yes, according to the target face The image is authenticated to the object to be verified; if not, a verification result indicating that the object to be verified is an illegal user is generated.
20. The identity verification device according to claim 19, wherein the authenticating the object to be verified according to the target face image comprises:

    Dividing the target face image into a plurality of face regions according to the key point set of the target face image; determining target feature information according to the plurality of face regions; and performing the to-be-verified object according to the target feature information Authentication.
21. The identity verification device according to claim 20, wherein the determining the target feature information according to the plurality of face regions comprises: performing a feature extraction operation on the plurality of face regions to obtain a plurality of pieces of feature information; each face The area corresponds to a piece of feature information; the plurality of pieces of feature information are reorganized to obtain the target feature information.
22. The identity verification device according to claim 20, wherein the authenticating the object to be verified according to the target feature information comprises: acquiring a stored feature information set, and each of the stored feature information sets The user identifier corresponding to the feature information is stored; the similarity between each stored feature information and the target feature information is calculated by using a preset algorithm; and the identity of the object to be verified is determined according to the similarity and the corresponding user identifier. verification.
23. The identity verification device according to claim 22, wherein the authenticating the object to be verified according to the similarity and the corresponding user identifier comprises: determining whether the calculated similarity is not less than And a similarity of the preset threshold; if yes, the user identifier corresponding to the similarity of the second preset threshold is used as the target user identifier, and the verification that the object to be verified is the target user identifier is generated result;

    If not, generate a verification result indicating that the object to be verified is an illegal user.
24. The identity verification device according to claim 22, further comprising: obtaining a user registration request, before acquiring the stored feature information set and the user identifier corresponding to each stored feature information in the stored feature information set The user registration request carries the to-be-registered user identifier and the to-be-registered face image; the to-be-registered feature information is determined according to the to-be-registered face image; the to-be-registered feature information is associated with the to-be-registered user identifier, and the The registration feature information is added to the stored feature information set.
25. A computer storage medium having stored therein computer readable instructions or programs, the computer readable instructions or programs being executed by a processor as claimed in any one of claims 1-12.

Images