Identity verification method and device, electronic equipment and computer readable storage medium

Abstract

The application relates to the technical field of data processing, and is applied to an insurance business scene, and provides an identity verification method and device, an electronic device and a computer readable storage medium, the method comprising the following steps: acquiring a static figure image; converting the static figure image into a dynamic figure video; performing basic action comparison processing on a collected real-time dynamic video to be verified and the dynamic figure video to obtain a basic action comparison result; performing micro-expression comparison processing on the real-time dynamic video to be verified and the dynamic figure video to obtain a micro-expression comparison result; performing three-dimensional posture comparison processing on the real-time dynamic video to be verified and the dynamic figure video to obtain a three-dimensional posture comparison result; and determining an identity verification result according to the basic action comparison result, the micro-expression comparison result and the three-dimensional posture comparison result. Through the technical scheme, the problem that a static verification is easy to break through in the industry can be solved, and the healthy development of the insurance business is thus facilitated.

Description

Technical Field

[0001] The embodiments of this application relate to, but are not limited to, the field of data processing technology, and in particular to an identity verification method, apparatus, electronic device, and computer-readable storage medium. Background Technology

[0002] With the continuous development of society and economy and the advancement of technology, people's living standards have been greatly improved. To ensure their livelihoods, the insurance business has also been widely promoted and developed. However, in core business processes such as insurance underwriting, claims settlement, and policy maintenance, identity verification and anti-fraud are crucial for ensuring business compliance and asset security. Currently, the industry faces the following pain points: traditional methods relying on static photos of ID cards and real-time facial recognition cannot effectively resist insurance fraud methods such as "re-photographing ID cards" and "AI-synthesized static faces." Summary of the Invention

[0003] The following is an overview of the subject matter described in detail herein. This overview is not intended to limit the scope of the claims.

[0004] To address the problems mentioned in the background section, this application provides an identity verification method, apparatus, electronic device, and computer-readable storage medium, which can solve the problem of easy breach of existing static verification in the industry, thereby contributing to the healthy development of insurance business.

[0005] In a first aspect, embodiments of this application provide an identity verification method, the identity verification method comprising: Obtain static images of people; Convert the static human image into a dynamic human video; The collected real-time dynamic video to be verified is compared with the dynamic character video by basic motion comparison to obtain the basic motion comparison result. The real-time dynamic video to be verified is compared with the dynamic character video using micro-expression analysis to obtain micro-expression comparison results. The real-time dynamic video to be verified is compared with the dynamic character video in three dimensions to obtain the three-dimensional posture comparison result. Based on the comparison results of the basic actions, the comparison results of the micro-expressions, and the comparison results of the three-dimensional postures, the identity verification result is determined.

[0006] Secondly, embodiments of this application also provide an identity verification device, the identity verification device comprising: The acquisition unit is used to acquire static images of people. A conversion unit is used to convert the static human image into a dynamic human video; The first comparison unit is used to perform basic motion comparison processing on the collected real-time dynamic video to be verified and the dynamic character video to obtain the basic motion comparison result. The second comparison unit is used to perform micro-expression comparison processing on the real-time dynamic video to be verified and the dynamic character video to obtain micro-expression comparison results. The third comparison unit is used to perform three-dimensional posture comparison processing on the real-time dynamic video to be verified and the dynamic character video to obtain a three-dimensional posture comparison result. The judgment unit is used to determine the identity verification result based on the basic action comparison result, the micro-expression comparison result, and the three-dimensional posture comparison result.

[0007] Thirdly, embodiments of this application also provide an electronic device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the identity verification method described in the first aspect above.

[0008] Fourthly, embodiments of this application also provide a computer-readable storage medium storing computer-executable instructions for performing the identity verification method described in the first aspect above.

[0009] The identity verification method according to the embodiments provided in this application has at least the following beneficial effects: In the identity verification process, a static image of a person is first acquired; then, the static image is converted into a dynamic video; next, the acquired real-time dynamic video to be verified is compared with the dynamic video to obtain a basic action comparison result; furthermore, the real-time dynamic video to be verified and the dynamic video to be verified can be compared with micro-expressions to obtain a micro-expression comparison result; and the real-time dynamic video to be verified and the dynamic video to be verified can be compared with three-dimensional postures to obtain a three-dimensional posture comparison result. Finally, the corresponding identity verification result can be determined based on the basic action comparison result, the micro-expression comparison result, and the three-dimensional posture comparison result. Through the above technical solution, a static image of a person is converted into a dynamic video, and then the acquired real-time dynamic video to be verified is compared with the converted dynamic video to obtain an identity verification result through multi-dimensional comparison. This solves the problem of easy breach of static verification in the industry, thus contributing to the healthy development of the insurance business. Attached Figure Description

[0010] The accompanying drawings are used to provide a further understanding of the technical solutions of this application and constitute a part of the specification. They are used together with the embodiments of this application to explain the technical solutions of this application and do not constitute a limitation on the technical solutions of this application.

[0011] Figure 1 This is a flowchart illustrating an identity verification method provided in one embodiment of this application; Figure 2 yes Figure 1 A schematic diagram of a specific implementation method of step S200; Figure 3 yes Figure 2 A flowchart illustrating a specific implementation of step S240; Figure 4 yes Figure 1 A schematic diagram of a specific implementation of step S300; Figure 5 yes Figure 1 A schematic diagram of a specific implementation of step S400; Figure 6 yes Figure 1 A schematic diagram of a specific implementation of step S500; Figure 7 This is a flowchart illustrating an identity verification method provided in another embodiment of this application; Figure 8 This is a schematic diagram of an identity verification device provided in one embodiment of this application; Figure 9 This is a schematic diagram of an electronic device provided in one embodiment of this application. Detailed Implementation

[0012] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0013] It should be noted that although functional modules are divided in the device schematic diagram and a logical order is shown in the flowchart, in some cases, the steps shown or described may be performed in a different order than the module division in the device or the order in the flowchart. The terms "first," "second," etc., in the specification, claims, and the aforementioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence.

[0014] It should be noted that, unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only and is not intended to limit this application.

[0015] The embodiments of this application can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence (AI) refers to the theories, methods, technologies, and application systems that use digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results.

[0016] AI is a new technical science that studies and develops theories, methods, technologies, and application systems for simulating, extending, and expanding human intelligence. Artificial intelligence is a branch of computer science that attempts to understand the essence of intelligence and produce new intelligent machines that can react in a way similar to human intelligence. Research in this field includes robotics, speech recognition, image recognition, natural language processing, and expert systems. Artificial intelligence can simulate the information processes of human consciousness and thought. Furthermore, artificial intelligence utilizes digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceiving the environment, acquiring knowledge, and using that knowledge to achieve optimal results—the theories, methods, technologies, and application systems available for use.

[0017] Foundational technologies for artificial intelligence generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing, operating / interactive systems, and mechatronics. AI software technologies mainly encompass computer vision, robotics, biometrics, speech processing, natural language processing, and machine learning / deep learning.

[0018] Artificial intelligence, or AI, is the theory, method, technology, and application system that uses digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results.

[0019] The servers involved in artificial intelligence technology can be standalone servers or cloud servers that provide basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDN), and big data and artificial intelligence platforms.

[0020] This application provides an identity verification method, apparatus, electronic device, and computer-readable storage medium. In the identity verification process, a static image of a person is first acquired; then, the static image is converted into a dynamic video; next, the acquired real-time dynamic video to be verified is compared with the dynamic video to obtain a basic action comparison result; furthermore, micro-expression comparison can be performed between the real-time dynamic video to be verified and the dynamic video to obtain a micro-expression comparison result; and finally, three-dimensional posture comparison can be performed between the real-time dynamic video to be verified and the dynamic video to obtain a three-dimensional posture comparison result. Finally, the corresponding identity verification result can be determined based on the basic action comparison result, micro-expression comparison result, and three-dimensional posture comparison result. Through the above technical solution, a static image of a person is converted into a dynamic video, and then the acquired real-time dynamic video to be verified is compared with the converted dynamic video to obtain an identity verification result. This solves the problem of easy breach of static verification in the industry, thus contributing to the healthy development of the insurance business.

[0021] The identity verification method provided in this application relates to the field of data processing technology. This identity verification method can be applied to a terminal, a server, or software running on either a terminal or a server. In some embodiments, the terminal can be a smartphone, tablet, laptop, desktop computer, etc.; the server can be configured as an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms.

[0022] This application can be used in a wide variety of general-purpose or special-purpose computer system environments or configurations. Examples include: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, and distributed computing environments including any of the above systems or devices. This application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc., that perform specific tasks or implement specific abstract data types. This application can also be practiced in distributed computing environments where tasks are performed by remote processing devices connected via a communication network. In distributed computing environments, program modules can reside in local and remote computer storage media, including storage devices.

[0023] It should be noted that in all specific embodiments of this application, when processing data related to user identity or characteristics, such as user information, user behavior data, user historical data, and user location information, user permission or consent is obtained first. Furthermore, the collection, use, and processing of this data comply with relevant laws, regulations, and standards. In addition, when embodiments of this application require access to sensitive personal information of users, separate permission or consent from the user is obtained through pop-ups or redirection to confirmation pages. Only after obtaining the user's separate permission or consent is the necessary user-related data required for the proper functioning of these embodiments acquired.

[0024] The embodiments of this application will be further described below with reference to the accompanying drawings.

[0025] like Figure 1 As shown, Figure 1 This is a flowchart illustrating an identity verification method provided in one embodiment of this application. The identity verification method includes the following steps: Step S100: Obtain a static image of a person.

[0026] The identity verification method provided in this application first acquires a static image of a person during the identity verification process to provide data for subsequent identity verification processing. This static image can be a portrait on an ID card or a facial image uploaded by the user during the insurance application process. For example, during the insurance application process, during identity verification, a user can upload an ID card image or a real-time portrait image taken with a mobile phone to the insurance system.

[0027] It is worth noting that in the process of acquiring static images of people, when it involves processing data related to user identity or characteristics, such as user information, user behavior data, user historical data, and user location information, the user's permission or consent is always obtained first. Furthermore, the collection, use, and processing of this data comply with relevant laws, regulations, and standards. In addition, when this application embodiment needs to obtain sensitive personal information of users, separate permission or consent from the user is obtained through pop-ups or redirects to confirmation pages. Only after explicitly obtaining the user's separate permission or consent is the necessary user-related data for the normal operation of this application embodiment acquired.

[0028] Step S200: Convert the static human image into a dynamic human video.

[0029] The identity verification method provided in this application embodiment can convert a static image of a person into a dynamic video of a person after obtaining the static image. Converting a static image of a person into a dynamic video of a person can solve the problem of easy breach of static verification in the industry during subsequent identity verification and improve the security of identity verification.

[0030] It is worth noting that in the process of converting static human images into dynamic human videos, the process first involves extracting the face region from the static human image based on a pre-defined lightweight face detection network model to obtain a face region image. Next, feature extraction is performed on the face region image to obtain a facial feature vector. Then, 3D pose estimation is performed on the face region image to obtain pose estimation parameters. Following this, based on a pre-defined dynamic action instruction library, the facial feature vector and pose estimation parameters are dynamically transformed to obtain multiple consecutive dynamic frames. Finally, optimization processing of these multiple consecutive dynamic frames yields the dynamic human video. This approach addresses the existing problem of easily bypassing static verification in the industry, improving the security and accuracy of identity verification.

[0031] like Figure 2 As shown, converting a static image of a person into a dynamic video of a person can include the following steps: Step S210: Based on the preset lightweight face detection network model, the face region of the static human image is extracted to obtain the face region image. Step S220: Perform feature extraction processing on the face region image to obtain a face feature vector; and perform three-dimensional pose estimation processing on the face region image to obtain pose estimation parameters. Step S230: Based on the preset dynamic action instruction library, the facial feature vector and pose estimation parameters are dynamically transformed to obtain multiple continuous dynamic frames. Step S240: Optimize multiple consecutive dynamic frames to obtain a dynamic character video.

[0032] For steps S210 to S240, in the process of converting a static person image into a dynamic person video, firstly, a face region extraction process is performed on the static person image based on a preset lightweight face detection network model to obtain a face region image; then, feature extraction processing is performed on the face region image to obtain a facial feature vector; and then, three-dimensional pose estimation processing is performed on the face region image to obtain pose estimation parameters; next, based on a preset dynamic action instruction library, dynamic conversion processing is performed on the facial feature vector and pose estimation parameters to obtain multiple continuous dynamic frames; finally, optimization processing is performed on multiple continuous dynamic frames to obtain a dynamic person video, thereby solving the problem of easy breach of static verification in the industry and improving the security and accuracy of identity verification.

[0033] For example, converting static human images into dynamic human videos aims to transform static ID photos (such as ID card portraits or facial photos uploaded during insurance applications) submitted by customers into dynamic human video videos that comply with insurance verification standards, generating "controllable and differentiated" dynamic templates to provide a standard reference for subsequent liveness detection.

[0034] In the process of converting static human images into dynamic human videos, a dynamic motion instruction library is first built, and a dedicated dynamic motion library for insurance verification is designed, covering basic movements (slowly blinking twice, opening and closing the mouth at a uniform speed, turning the head left and right by 15°) and micro-expression movements (raising one corner of the mouth, rotating the iris of the eye). The movements are designed according to ergonomics, suitable for all age groups, and "motion parameter thresholds" are set; for example, the blinking closure degree must reach more than 80% of the eye pixels, and the head rotation angular velocity must not exceed 5° / second.

[0035] Logic for dynamic generation of static portraits: Input layer: Receives static ID photos of customers, extracts face regions through a lightweight face detection network, and automatically crops and aligns them to a standard size of 256×256 pixels; Feature extraction layer: An improved convolutional network is used, with a new "facial detail feature branch". When extracting facial texture and organ morphology features, a 3D pose prior is introduced. The head's 3D angles are estimated by simplifying the optimization framework of fitting a 3D human model to 2D observations. The feature extraction process can be represented as:

[0036] in, For the input image, This is the feature extraction function for convolutional networks. This is a 3D pose estimation function that ultimately outputs a joint representation of "facial feature vector + pose parameters". .

[0037] Dynamic Generation Layer: Based on parameters from the motion instruction library, dynamic keypoint offsets are generated through a 4-layer lightweight multilayer perceptron. Combined with the optical flow warping module, only the optical flow of key regions such as the eyes and lips is calculated, transforming static features into continuous dynamic frames. The optical flow calculation uses an optimized version of the sparse optical flow estimation algorithm, reducing computational load by 60%. The dynamic frame generation formula is as follows:

[0038] in, for Static frame at any moment for Set the frame rate to 25 fps and the duration to 3-5 seconds for key moments, ensuring that facial details in the dynamic video are highly consistent with those in the still photo.

[0039] Dynamic template quality control: A "dynamic consistency checker" is introduced, which calculates the facial feature similarity between adjacent frames in a dynamic video (using the LPIPS loss function).

[0040] in, The feature distance metric function is... For the first Layer network feature extraction function, The weighting coefficient is used to verify whether the action parameters meet the threshold (e.g., blink closure rate ≥ 90%), and to filter out dynamic templates that do not meet the quality standards, so as to ensure the accuracy of subsequent verification.

[0041] like Figure 3 As shown, optimizing multiple consecutive dynamic frames to obtain a dynamic character video can include the following steps: Step S241: For each consecutive dynamic frame, perform similarity calculation on the facial features of the consecutive dynamic frame and the facial features of adjacent consecutive dynamic frames to obtain facial feature similarity. Step S242: If the facial feature similarity does not meet the preset similarity requirement, the corresponding continuous dynamic frames are removed to obtain a dynamic character video.

[0042] For steps S241 to S242, in the process of optimizing multiple consecutive dynamic frames to obtain a dynamic character video, for each consecutive dynamic frame, the facial features of the consecutive dynamic frame are similar to the facial features of adjacent consecutive dynamic frames to obtain the facial feature similarity. Then, if the facial feature similarity does not meet the preset similarity requirement, the corresponding consecutive dynamic frame is removed to obtain the dynamic character video. Through the above technical solution, the generated dynamic character video can be more accurate and reasonable.

[0043] It is worth noting that multiple consecutive dynamic frames generated are optimized based on the similarity between two adjacent consecutive dynamic frames, so that the generated dynamic character videos can be more reasonable, thereby greatly improving the accuracy of subsequent identity verification.

[0044] Step S300: Perform basic motion comparison processing on the collected real-time dynamic video to be verified and the dynamic character video to obtain the basic motion comparison results.

[0045] The identity verification method provided in this application converts a static image of a person into a dynamic video of a person. Then, it compares the acquired real-time dynamic video to be verified with the dynamic video of the person to be verified using basic actions, thereby obtaining a comparison result. Furthermore, if the basic action comparison result does not meet the requirements, it can be directly determined that there is a problem with the real-time dynamic video to be verified, and the identity verification can be directly deemed unsuccessful. This technical solution effectively avoids the problem of easy bypassing of traditional static verification, thus improving the accuracy of identity verification.

[0046] It is worth noting that in the process of comparing the collected real-time dynamic video to be verified with the dynamic person video to obtain the basic action comparison result, the following steps are taken: First, the real-time dynamic video to be verified is processed to extract temporal features, resulting in the first action temporal features; then, temporal features are extracted from the dynamic person video to obtain the second action temporal features; next, the first and second action temporal features are compared to obtain the temporal similarity; finally, the basic action comparison result can be determined based on the temporal similarity. This technical solution makes the determination of the basic action comparison result simpler, faster, and more accurate.

[0047] like Figure 4 As shown, the basic motion comparison process involves comparing the acquired real-time dynamic video to be verified with the dynamic character video to obtain the basic motion comparison results. This process may include the following steps: Step S310: Perform temporal feature extraction processing on the real-time dynamic video to be verified to obtain the first action temporal feature; Step S320: Perform temporal feature extraction processing on the dynamic character video to obtain the second action temporal feature; Step S330: Perform similarity calculation on the temporal features of the first action and the temporal features of the second action to obtain temporal similarity; Step S340: Determine the basic action comparison results based on temporal similarity.

[0048] For steps S310 to S340, in the process of comparing the collected real-time dynamic video to be verified with the dynamic person video to obtain the basic action comparison result, firstly, the real-time dynamic video to be verified is subjected to temporal feature extraction processing to obtain the first action temporal feature; then, the dynamic person video is subjected to temporal feature extraction processing to obtain the second action temporal feature; then, the first action temporal feature and the second action temporal feature are subjected to similarity calculation processing to obtain the temporal similarity; finally, the basic action comparison result can be determined based on the temporal similarity. Through the above technical solution, the determination of the basic action comparison result can be made simpler and more accurate.

[0049] It is worth noting that the real-time dynamic video to be verified can be a real-time facial video taken by the user using their mobile phone; for example, during the insurance application process, the user may take videos of blinking, turning their head, and opening their mouth according to the insurance requirements. By comparing the real-time dynamic video to be verified with the dynamic video of the person previously converted from a static image, accurate identity verification can be performed, overcoming the problem of low accuracy in identity verification based on static images.

[0050] It is worth noting that action timing features may include, but are not limited to, the frame interval of blinking and the curve of head rotation angle change; by comparing the basic actions of the real-time dynamic video to be verified with the previously generated dynamic character video, it is possible to identify whether the similarity of the action timing features of the real-time dynamic video to be verified meets the requirements, and then identity verification can be performed.

[0051] Step S400: Perform micro-expression comparison processing on the real-time dynamic video to be verified and the dynamic human video to obtain the micro-expression comparison results.

[0052] The identity verification method provided in this application converts a static image of a person into a dynamic video of a person. It then performs micro-expression comparison processing on the acquired real-time dynamic video to be verified and the dynamic video of the person, thereby obtaining a micro-expression comparison result. Furthermore, if the micro-expression comparison result does not meet the requirements, it can be directly determined that there is a problem with the real-time dynamic video to be verified, and the identity verification can be directly determined to have failed. Through the above technical solution, the problem of easy bypassing of traditional static verification can be effectively avoided, further improving the accuracy of identity verification.

[0053] It is worth noting that in the process of comparing the micro-expression features of the real-time dynamic video to be verified with the video of a moving person, the process involves first extracting micro-expression features from the real-time dynamic video to obtain the first micro-expression feature; then calculating the target region feature distance from the first micro-expression feature to obtain the first region feature distance; next, extracting micro-expression features from the video of the moving person to obtain the second micro-expression feature; then calculating the target region feature distance from the second micro-expression feature to obtain the second region feature distance; finally, comparing the first and second region feature distances yields the micro-expression comparison result. This technical solution makes the process of comparing the micro-expression features of the real-time dynamic video to be verified with the video of a moving person simpler, more reasonable, and more accurate.

[0054] like Figure 5 As shown, comparing the micro-expression of the real-time dynamic video to be verified with the dynamic video of the person in question to obtain the micro-expression comparison results may include the following steps: Step S410: Extract micro-expression features from the real-time dynamic video to be verified to obtain the first micro-expression feature; and calculate the target region feature distance from the first micro-expression feature to obtain the first region feature distance. Step S420: Extract micro-expression features from the dynamic human video to obtain the second micro-expression feature; and calculate the target region feature distance from the second micro-expression feature to obtain the second region feature distance. Step S430: Compare the feature distance of the first region and the feature distance of the second region to obtain the micro-expression comparison result.

[0055] For steps S410 to S430, the process of comparing the micro-expression features of the real-time dynamic video to be verified with the dynamic person video to obtain the micro-expression comparison result involves firstly extracting micro-expression features from the real-time dynamic video to obtain the first micro-expression feature; then calculating the target region feature distance of the first micro-expression feature to obtain the first region feature distance; next, extracting micro-expression features from the dynamic person video to obtain the second micro-expression feature; then calculating the target region feature distance of the second micro-expression feature to obtain the second region feature distance; finally, comparing the first region feature distance and the second region feature distance yields the micro-expression comparison result. This technical solution makes the process of comparing the micro-expression features of the real-time dynamic video to be verified with the dynamic person video simpler, more reasonable, and more accurate.

[0056] It is worth noting that during the process of calculating the feature distance of the target region for micro-expression features, the eye region can be selected, and the distance of the eye region features can then be calculated and processed, making the comparison of micro-expressions simpler and more reliable.

[0057] Step S500: Perform 3D pose comparison processing on the real-time dynamic video to be verified and the dynamic character video to obtain the 3D pose comparison result.

[0058] The identity verification method provided in this application converts a static image of a person into a dynamic video of a person. Then, it performs a 3D pose comparison between the acquired real-time dynamic video to be verified and the dynamic video of the person, thereby obtaining a 3D pose comparison result. Furthermore, if the 3D pose comparison result does not meet the requirements, it can be directly determined that there is a problem with the real-time dynamic video to be verified, and the identity verification can be directly determined to have failed. Through the above technical solution, the problem of easy bypassing of traditional static verification can be effectively avoided, and the accuracy of identity verification is further improved.

[0059] It is worth noting that in the process of comparing the real-time dynamic video to be verified with the dynamic person video to obtain the 3D pose comparison result, the following steps are taken: First, the real-time dynamic video to be verified is processed to extract 3D pose parameters, thus obtaining the first 3D pose parameters; then, the dynamic person video is processed to extract 3D pose parameters, thus obtaining the second 3D pose parameters; next, the distance between the first and second 3D pose parameters is calculated to obtain the pose parameter distance; finally, the 3D pose comparison result can be determined based on the pose parameter distance. This technical solution makes the process of comparing the real-time dynamic video to be verified with the dynamic person video in 3D pose more convenient, reasonable, and accurate.

[0060] like Figure 6 As shown, the process of comparing the real-time dynamic video to be verified with the dynamic character video in three dimensions to obtain the three-dimensional pose comparison result can include the following steps: Step S510: Extract three-dimensional pose parameters from the real-time dynamic video to be verified to obtain the first three-dimensional pose parameters. Step S520: Extract three-dimensional pose parameters from the dynamic character video to obtain the second three-dimensional pose parameters; Step S530: Perform parameter distance calculation on the first three-dimensional attitude parameters and the second three-dimensional attitude parameters to obtain the attitude parameter distance; Step S540: Determine the three-dimensional attitude comparison result based on the attitude parameter distance.

[0061] For steps S510 to S540, in the process of comparing the three-dimensional pose of the real-time dynamic video to be verified with the dynamic person video to obtain the three-dimensional pose comparison result, firstly, the real-time dynamic video to be verified is processed to extract three-dimensional pose parameters, thereby obtaining the first three-dimensional pose parameters; then, the dynamic person video is processed to extract three-dimensional pose parameters, thereby obtaining the second three-dimensional pose parameters; next, the first three-dimensional pose parameters and the second three-dimensional pose parameters are processed to calculate the pose parameter distance; finally, the three-dimensional pose comparison result can be determined based on the pose parameter distance. Through the above technical solution, the process of comparing the three-dimensional pose of the real-time dynamic video to be verified with the dynamic person video becomes simpler, more reasonable, and more accurate.

[0062] It is worth noting that by extracting three-dimensional pose parameters from the real-time dynamic video to be verified, the three-dimensional pose parameters of the head in the real-time dynamic video to be verified can be obtained; by extracting three-dimensional pose parameters from the dynamic person video, the three-dimensional pose parameters of the head in the dynamic person video can be obtained; and finally, by comparing the extracted three-dimensional pose parameters, the three-dimensional pose comparison results can be obtained conveniently and quickly.

[0063] Step S600: Based on the comparison results of basic movements, micro-expressions, and three-dimensional postures, determine the identity verification result.

[0064] The identity verification method provided in this application involves comparing the acquired real-time dynamic video to be verified with a dynamic person video using basic motion comparison to obtain a basic motion comparison result, comparing the real-time dynamic video to be verified with the dynamic person video using micro-expression comparison to obtain a micro-expression comparison result, and comparing the real-time dynamic video to be verified with the dynamic person video using three-dimensional pose comparison to obtain a three-dimensional pose comparison result. The identity verification result is considered successful only if all three comparison results (basic motion comparison, micro-expression comparison, and three-dimensional pose comparison) meet their respective requirements; otherwise, the identity verification result is considered unsuccessful. This technical solution makes the identity verification process more accurate, reasonable, and secure.

[0065] For example, in the process of multi-dimensionally comparing the collected real-time dynamic video to be verified with the dynamic person video, the purpose is to compare the dynamic action video captured by the customer in real time with the standard dynamic template (i.e., dynamic person video) generated by the "static portrait animation module" in multiple dimensions, accurately identify fraudulent behaviors such as "dynamic video forgery" and "inadequate action imitation", and at the same time lower the action execution threshold for special groups.

[0066] Real-time dynamic data acquisition optimization: Client compatibility: Supports mobile and PC data acquisition, automatically adapts to the camera resolution of different devices, and provides "real-time guidance prompts" during the acquisition process to assist customers in completing actions and reduce operational errors.

[0067] Anti-interference processing: For scenes with low lighting and complex backgrounds, adaptive lighting compensation (adjusting brightness through histogram equalization) and background blurring (using a semantic segmentation network to isolate face regions) are added. The semantic segmentation network is based on a U-shaped network architecture and minimizes the cross-entropy loss function. Training was conducted, among which For real labels, To predict probabilities and avoid the influence of environmental factors on the comparison results.

[0068] Multi-dimensional feature comparison logic: Basic motion comparison: Extract the temporal features of motion from real-time video and standard templates (such as the frame interval of blinking and the angle change curve of head rotation), and use a dynamic time warping algorithm to calculate the temporal similarity between the two. :

[0069] in, and These are feature sequences from real-time video and standard templates, respectively. For path planning, For local distance measurement, a similarity of ≥0.85 is considered passing (the threshold can be adjusted according to the business risk level, such as raising the threshold for high-insurance-amount insurance to 0.9).

[0070] Micro-expression detail comparison: Micro-expression features are extracted from real-time video and standard templates using 106 high-resolution facial feature points, and feature distances in local regions are calculated (e.g., LPIPS is used to measure the distances in the eye region). ( ), filtering out fake videos that "look similar in action but do not match the details".

[0071] 3D Pose Consistency Comparison: Utilizing a simplified optimized framework that fits a 3D human body model to 2D observations, head 3D pose parameters (neck rotation angle) are extracted from real-time video and a standard template. Shoulder tilt angle (etc.), calculate the L1 distance of the attitude parameters:

[0072] Avoid scams involving "flat photo reproductions (without 3D pose changes)".

[0073] Specialized group adaptation strategy: For groups with weaker motor skills, such as elderly customers and children, a "difficulty reduction option" is provided, while the comparison threshold is dynamically adjusted based on a Bayesian risk model. The fraud probability is determined by the feature conditions. The prior probability is updated using historical data, and the threshold is dynamically adjusted to balance security and verification pass rate.

[0074] like Figure 7 As shown, the identity verification method may also include the following steps: Step S610: Store the dynamic template features of static human images and dynamic human videos into a preset historical feature library. The dynamic template features include action sequence features, micro-expression features and three-dimensional pose features. Step S620: Perform multi-dimensional matching processing on the newly collected real-time dynamic features and historical features in the historical feature database to obtain multi-dimensional matching results; Step S630: When the appearance is represented by the multi-dimensional matching result, the changed area of ​​the real-time dynamic feature is masked based on the preset semantic segmentation network in order to determine the unchanged area of ​​the real-time dynamic feature. Step S640: Perform regional feature similarity calculation on the unchanged regions of real-time dynamic features and the corresponding historical features to obtain the regional feature similarity. Step S650: Determine the cross-scene comparison results based on the similarity of regional features.

[0075] For steps S610 to S650, the dynamic template features of static human images and dynamic human videos are stored in a pre-defined historical feature library. These dynamic template features include action sequence features, micro-expression features, and 3D pose features. Next, the newly acquired real-time dynamic features are matched with historical features in the historical feature library in a multi-dimensional manner to obtain a multi-dimensional matching result. If the multi-dimensional matching result indicates changes in appearance, a pre-defined semantic segmentation network is used to mask the changed areas of the real-time dynamic features to determine the unchanged areas. Then, the unchanged areas of the real-time dynamic features are compared with the corresponding historical features to calculate the regional feature similarity. Finally, the cross-scene comparison result can be determined based on the regional feature similarity. This technical solution makes cross-scene identity verification more accurate and effectively prevents identity verification failures due to changes in user appearance over time.

[0076] It is understandable that the dynamic template features of static images and dynamic videos of people are stored in a pre-set historical feature library, which can then serve as a data verification library for subsequent identity verification. After a long period, when a user needs to verify their identity again, the newly collected real-time dynamic features are matched with historical features in the historical feature library in a multi-dimensional process to obtain a multi-dimensional matching result. Furthermore, if the multi-dimensional matching result indicates changes in appearance, a semantic segmentation network is used to mask the changed areas of the real-time dynamic features to identify the unchanged areas. After masking the changed areas, the real-time dynamic features and corresponding historical features in the unchanged areas are processed to calculate regional feature similarity. Finally, the identity verification result can be determined based on the regional feature similarity, thus effectively avoiding identity verification failures due to a lack of verification over a long period.

[0077] For example, in the process of cross-scenario identity verification, the identity data link of the entire insurance business process is connected to achieve cross-scenario consistency verification of "historical static ID photo - current real-time face - dynamic features of business scenario", and solve the problem of "misjudgment and rejection due to appearance change".

[0078] Construction of an identity feature knowledge base: Data storage: When a customer applies for insurance for the first time, the "static ID photo + dynamic template features (action sequence, micro-expression, 3D posture)" are encrypted and stored in the insurance private cloud to build a customer-exclusive "identity feature knowledge base".

[0079] Feature Update: In scenarios such as subsequent policy maintenance and claims, the dynamic features in the knowledge base are automatically updated. An "incremental learning" strategy is adopted, using gradient descent to fine-tune the upper-layer parameters of the feature extraction network. The loss function is:

[0080] in, For new data loss, Due to the loss of historical data, The weighting factor is ≤10 seconds, and it adapts to changes in the customer's appearance.

[0081] Cross-scene comparison logic: Feature matching: In new business scenarios (such as face-to-face claims verification), after extracting the customer's real-time dynamic features, a "multi-dimensional matching" is performed with historical features in the knowledge base. If the appearance changes, the changed areas are masked using a semantic segmentation network, and cosine similarity is used to calculate the feature similarity of the unchanged areas.

[0082] in, and These are real-time and historical feature vectors, respectively.

[0083] Time decay compensation: For historical ID photos, a "time decay coefficient" is introduced. (e.g., annual decay of 0.05, ), dynamically adjust the comparison threshold :

[0084] in, This serves as the initial threshold to avoid false positives or rejections due to photo aging.

[0085] Multi-source data fusion: If a customer submits multiple static photos in different scenarios, the dynamic template features of these photos will be fused into a "comprehensive feature template" through a weighted average.

[0086] in, For the first The characteristics of the photograph Weights are used to improve the robustness of cross-scenario comparisons.

[0087] In addition, a collaborative mechanism of "technical initial screening + manual review" is provided for high-risk scenarios such as large claims and high-insurance coverage. Visual tools are used to assist manual judgment of identity verification results, shortening the review cycle from 1-3 working days to within 30 minutes.

[0088] Verification feature visualization tool: Feature Comparison Interface: Develop a web-based review interface to intuitively display the multi-dimensional feature comparison results of "real-time video - standard template", including action timeline curve comparison, micro-expression detail annotation, 3D posture heatmap, etc.

[0089] Suspicious region localization: Based on anomaly detection algorithms, suspicious regions in the comparison results are automatically identified by calculating the distance between the feature vector and the normal sample distribution. Locate the anomalies and mark them with flashing prompts to guide reviewers to focus on them.

[0090] Review and decision support functions: Automated risk scoring: Based on multi-dimensional comparison results, a logistic regression model is used to generate a "risk score". :

[0091] in, For eigenvalues, As weight, For bias, This is a sigmoid function. When the score is ≥70, a manual review is triggered, and the scoring criteria are provided.

[0092] Historical case reference: If the feature similarity between the current verification scenario and a historical fraud case is ≥0.7 (calculated using cosine similarity), the historical case will be automatically pushed to the review interface to assist reviewers in judging the risk type.

[0093] Review result traceability: Record the operation log of manual review, bind and store it with the verification feature data, support subsequent audit and risk analysis, and ensure that the review process is traceable and supervised.

[0094] Encrypted data transmission: Sensitive data such as customer still photos and videos are encrypted during transmission using encryption protocols, and session keys are generated through key exchange algorithms. To ensure secure data transmission.

[0095] Storage encryption: Identity feature data is stored in the private cloud using the "AES-256" encryption algorithm, and a "data anonymization" strategy is adopted to comply with the requirements of the "Personal Information Protection Law" and the "Measures for the Administration of Insurance Data Security".

[0096] Data access control: Adopt the "principle of least privilege" to set data access permissions, allowing only the verification system and review personnel to access data in business scenarios, and the access behavior must be subject to "two-factor authentication" (such as password + dynamic token).

[0097] Dynamic template watermark embedding: An "invisible digital watermark" is embedded in the generated standard dynamic template, and the watermark information is transmitted based on a spread spectrum watermarking algorithm. Embedded image high-frequency texture regions:

[0098] in, For the original image, For embedding strength, As a high-frequency filter, if fraudsters forge dynamic templates, the source of the template can be traced through watermark extraction tools.

[0099] Fraud Feature Database Update: Regularly collect new fraud methods in the industry, use unsupervised learning algorithms to cluster and analyze new fraud features, update the system's "fraud feature database", and optimize the feature extraction and comparison model through incremental training to improve the ability to identify new frauds.

[0100] This solution is based on the core logic of "dynamic facial image generation + multi-dimensional feature comparison," relying on cutting-edge AI visual recognition and biometric analysis technologies. Through a lightweight model architecture deeply adapted to insurance business scenarios, it effectively balances verification efficiency and computational resource consumption. On this basis, it constructs an identity verification and anti-fraud system covering the entire lifecycle of insurance business: "prevention before the event, verification during the event, and traceability after the event." Before the event, high-risk customers are identified through a risk assessment model; during the event, dynamic liveness detection and multi-modal feature fusion comparison technologies are used to verify customer identity in real time; and after the event, blockchain evidence storage and big data analysis enable full-process traceability.

[0101] In addition, such as Figure 8 As shown, one embodiment of this application also provides an identity verification device 10, which includes: Acquisition unit 100 is used to acquire static human images; The conversion unit 200 is used to convert static human images into dynamic human videos; The first comparison unit 300 is used to perform basic motion comparison processing on the collected real-time dynamic video to be verified and the dynamic human video to obtain the basic motion comparison result. The second comparison unit 400 is used to perform micro-expression comparison processing on the real-time dynamic video to be verified and the dynamic human video to obtain the micro-expression comparison result. The third comparison unit 500 is used to perform three-dimensional pose comparison processing on the real-time dynamic video to be verified and the dynamic human video to obtain the three-dimensional pose comparison result. The judgment unit 600 is used to determine the identity verification result based on the comparison results of basic actions, micro-expressions, and three-dimensional posture.

[0102] The specific implementation of the identity verification device 10 is basically the same as the specific embodiment of the identity verification method described above, and will not be repeated here.

[0103] In addition, such as Figure 9 As shown, one embodiment of this application also provides an electronic device 700, which includes: a memory 720, a processor 710, and a computer program stored on the memory 720 and executable on the processor 710.

[0104] The processor 710 and memory 720 can be connected via a bus or other means.

[0105] The non-transient software program and instructions required to implement the identity verification method of the above embodiments are stored in the memory 720. When executed by the processor 710, the identity verification method of each of the above embodiments is executed.

[0106] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0107] Furthermore, one embodiment of this application also provides a computer-readable storage medium storing computer-executable instructions that are executed by a processor 710 or a controller, for example, by a processor 710 in the above-described device embodiment, which enables the processor 710 to perform the identity verification method in the above-described embodiment.

[0108] The above embodiments can be used in combination, and modules with the same name in different embodiments may be the same or different.

[0109] The foregoing has described specific embodiments of this application; other embodiments are within the scope of the appended claims. In some cases, the actions or steps described in the claims may be performed in a different order than those shown in the embodiments and may still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily have to follow the specific or sequential order shown to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0110] The various embodiments in this application are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the embodiments of apparatus, devices, and computer-readable storage media are basically similar to the method embodiments, so the descriptions are relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0111] The apparatus, device, computer-readable storage medium and method provided in the embodiments of this application are corresponding. Therefore, the apparatus, device and non-volatile computer storage medium also have similar beneficial technical effects as the corresponding method. Since the beneficial technical effects of the method have been described in detail above, the beneficial technical effects of the corresponding apparatus, device and computer storage medium will not be described again here.

[0112] In the 1990s, improvements to a technology could be clearly distinguished as either hardware improvements (e.g., improvements to the circuit structure of diodes, transistors, switches, etc.) or software improvements (improvements to the methodology). However, with technological advancements, many methodological improvements today can be considered direct improvements to the hardware circuit structure. Designers almost always obtain the corresponding hardware circuit structure by programming the improved methodology into the hardware circuit. Therefore, it cannot be said that a methodological improvement cannot be implemented using hardware physical modules. For example, a Programmable Logic Device (PLD) (such as a Field Programmable Gate Array (FPGA)) is such an integrated circuit whose logic function is determined by the user programming the device. Designers can program a digital system themselves to "integrate" it onto a PLD, without needing chip manufacturers to design and manufacture dedicated integrated circuit chips. Moreover, nowadays, instead of manually manufacturing integrated circuit chips, this programming is mostly implemented using "logic compiler" software, which is similar to the software compiler used when writing program development code. The original code before compilation must also be written in a specific programming language, which is called a Hardware Description Language (HDL). There is not just one HDL, but many kinds, such as ABEL (Advanced Boolean Expression Language), AHDL (Altera Hardware Description Language), Confluence, CUPL (Cornell University Programming Language), HDCal, JHDL (Java Hardware Description Language), Lava, Lola, MyHDL, PALASM, RHDL (Ruby Hardware Description Language), etc. Currently, the most commonly used are VHDL (Very-High-Speed ​​Integrated Circuit Hardware Description Language) and Verilog. Those skilled in the art should also understand that by simply performing some logic programming on the method flow using the aforementioned hardware description languages ​​and programming it into an integrated circuit, the hardware circuit that implements the logic method flow can be easily obtained.

[0113] The controller can be implemented in any suitable manner. For example, it can take the form of a microprocessor or processor and a computer-readable medium storing computer-readable program code (e.g., software or firmware) executable by the (micro)processor, logic gates, switches, application-specific integrated circuits (ASICs), programmable logic controllers, and embedded microcontrollers. Examples of controllers include, but are not limited to, the following microcontrollers: ARC625D, Atmel AT91SAM, Microchip PIC18F26K20, and Silicon Labs C8051F320. A memory controller can also be implemented as part of the control logic of the memory. Those skilled in the art will also recognize that, in addition to implementing the controller in purely computer-readable program code form, the same functionality can be achieved by logically programming the method steps to make the controller take the form of logic gates, switches, ASICs, programmable logic controllers, and embedded microcontrollers. Therefore, such a controller can be considered a hardware component, and the means included therein for implementing various functions can also be considered as structures within the hardware component. Alternatively, the means for implementing various functions can be considered as both software modules implementing the method and structures within the hardware component.

[0114] The systems, devices, modules, or units described in the above embodiments can be implemented by computer chips or entities, or by products with certain functions. A typical implementation device is a computer. Specifically, a computer can be, for example, a personal computer, laptop computer, cellular phone, camera phone, smartphone, personal digital assistant, media player, navigation device, email device, game console, tablet computer, wearable device, or any combination of these devices.

[0115] For ease of description, the above apparatus is described by dividing it into various functional units. Of course, in implementing the embodiments of this application, the functions of each unit can be implemented in one or more software and / or hardware.

[0116] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, embodiments of this application can take the form of entirely hardware embodiments, entirely software embodiments, or embodiments combining software and hardware aspects. Furthermore, embodiments of this application can take the form of computer program products implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0117] This specification is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0118] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0119] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0120] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0121] Memory may include non-persistent storage in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash memory (FlashRAM). Memory is an example of computer-readable media.

[0122] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information by any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0123] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0124] In this application embodiment, "at least one" refers to one or more, and "more than one" refers to two or more. "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent the existence of A alone, A and B simultaneously, or B alone. A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. "At least one of the following" and similar expressions refer to any combination of these items, including any combination of singular or plural items. For example, at least one of a, b, and c can represent: a, b, c, a and b, a and c, b and c, or a and b and c, where a, b, and c can be single or multiple.

[0125] The embodiments of this application can be described in the general context of computer-executable instructions, such as program modules, that are executed by a computer. Generally, program modules include routines, programs, objects, components, data structures, etc., that perform a specific task or implement a specific abstract data type. The embodiments of this application can also be practiced in distributed computing environments where tasks are performed by remote processing devices connected via a communication network. In a distributed computing environment, program modules can reside in local and remote computer storage media, including storage devices.

[0126] The various embodiments in this application are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the system embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0127] The above description is merely an embodiment of this application and is not intended to limit this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principle of this application should be included within the scope of the claims of this application.

Claims

1. An identity verification method, characterized in that, The identity verification method includes: Obtain static images of people; Convert the static human image into a dynamic human video; The collected real-time dynamic video to be verified is compared with the dynamic character video by basic motion comparison to obtain the basic motion comparison result. The real-time dynamic video to be verified is compared with the dynamic character video using micro-expression analysis to obtain micro-expression comparison results. The real-time dynamic video to be verified is compared with the dynamic character video in three dimensions to obtain the three-dimensional posture comparison result. Based on the comparison results of the basic actions, the comparison results of the micro-expressions, and the comparison results of the three-dimensional postures, the identity verification result is determined.

2. The identity verification method according to claim 1, characterized in that, The process of converting the static image of a person into a dynamic video of a person includes: The static human image is processed by extracting the face region based on a preset lightweight face detection network model to obtain a face region image. The face region image is subjected to feature extraction processing to obtain a face feature vector; and the face region image is subjected to three-dimensional pose estimation processing to obtain pose estimation parameters. Based on a preset dynamic action instruction library, the facial feature vector and the pose estimation parameters are dynamically transformed to obtain multiple continuous dynamic frames. The multiple consecutive dynamic frames are optimized to obtain the dynamic character video.

3. The identity verification method according to claim 2, characterized in that, The optimization processing of multiple consecutive dynamic frames to obtain the dynamic character video includes: For each of the continuous dynamic frames, the facial features of the continuous dynamic frame are compared with the facial features of adjacent continuous dynamic frames to obtain the facial feature similarity. If the facial feature similarity does not meet the preset similarity requirement, the corresponding continuous dynamic frames are removed to obtain the dynamic character video.

4. The identity verification method according to claim 1, characterized in that, The step of comparing the acquired real-time dynamic video to be verified with the dynamic character video using basic motion comparison processing to obtain the basic motion comparison result includes: The real-time dynamic video to be verified is subjected to temporal feature extraction processing to obtain the first action temporal feature; The dynamic character video is subjected to temporal feature extraction processing to obtain the second action temporal feature; The temporal features of the first action and the temporal features of the second action are processed to calculate the temporal similarity. The comparison result of the basic actions is determined based on the temporal similarity.

5. The identity verification method according to claim 1, characterized in that, The step of performing micro-expression comparison processing on the real-time dynamic video to be verified and the dynamic character video to obtain micro-expression comparison results includes: The real-time dynamic video to be verified is subjected to micro-expression feature extraction processing to obtain the first micro-expression feature; and the first micro-expression feature is subjected to target region feature distance calculation processing to obtain the first region feature distance. The dynamic human video is subjected to micro-expression feature extraction processing to obtain a second micro-expression feature; and the second micro-expression feature is subjected to target region feature distance calculation processing to obtain a second region feature distance. The feature distances of the first region and the feature distances of the second region are compared to obtain the micro-expression comparison results.

6. The identity verification method according to claim 1, characterized in that, The step of performing a three-dimensional pose comparison between the real-time dynamic video to be verified and the dynamic character video to obtain a three-dimensional pose comparison result includes: The real-time dynamic video to be verified is subjected to three-dimensional pose parameter extraction processing to obtain the first three-dimensional pose parameters; The dynamic character video is processed to extract three-dimensional pose parameters to obtain the second three-dimensional pose parameters; The first three-dimensional attitude parameters and the second three-dimensional attitude parameters are processed to calculate the parameter distance, and the attitude parameter distance is obtained. The three-dimensional attitude comparison result is determined based on the attitude parameter distance.

7. The identity verification method according to claim 1, characterized in that, The identity verification method also includes: The static human image and the dynamic template features of the dynamic human video are stored in a preset historical feature library, wherein the dynamic template features include action sequence features, micro-expression features and three-dimensional posture features. The newly collected real-time dynamic features are matched with the historical features in the historical feature library in a multi-dimensional way to obtain the multi-dimensional matching results. When the multi-dimensional matching results indicate changes in appearance, the changed areas of the real-time dynamic features are masked based on a preset semantic segmentation network in order to determine the unchanged areas of the real-time dynamic features. The region feature similarity is calculated by performing region feature similarity calculation on the unchanged region of the real-time dynamic feature and the corresponding historical feature; The cross-scene comparison results are determined based on the similarity of the regional features.

8. An identity verification device, characterized in that, The identity verification device includes: The acquisition unit is used to acquire static images of people. A conversion unit is used to convert the static human image into a dynamic human video; The first comparison unit is used to perform basic motion comparison processing on the collected real-time dynamic video to be verified and the dynamic character video to obtain the basic motion comparison result. The second comparison unit is used to perform micro-expression comparison processing on the real-time dynamic video to be verified and the dynamic character video to obtain micro-expression comparison results. The third comparison unit is used to perform three-dimensional posture comparison processing on the real-time dynamic video to be verified and the dynamic character video to obtain a three-dimensional posture comparison result. The judgment unit is used to determine the identity verification result based on the basic action comparison result, the micro-expression comparison result, and the three-dimensional posture comparison result.

9. An electronic device, comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that the processor, when executing the computer program, implements the identity verification method as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing computer-executable instructions, characterized in that, The computer-executable instructions are used to execute the identity verification method according to any one of claims 1 to 7.

Images