Remote verification method and system for access authority of terminal equipment

Abstract

The invention relates to the field of data processing, and particularly provides a remote verification method and system for access permission of terminal equipment, and the method comprises the steps: carrying out the biological feature extraction operation of a sign collection image of the terminal equipment, and obtaining an initial permission verification feature; performing feature decomposition on the initial permission verification feature to obtain a plurality of initial verification sub-features; performing feature extraction error adjustment on each initial verification sub-feature according to a terminal equipment physical sign acquisition image to obtain a verification adjustment sub-feature; determining a verification adjustment feature according to the plurality of verification adjustment sub-features; performing feature improvement on the verification adjustment feature according to the terminal equipment physical sign collection image to obtain an authority verification improvement feature; and determining a target permission verification feature according to the permission verification perfection feature, and performing permission verification based on the target permission verification feature. According to the method, the permission verification characteristics with higher characterization and accuracy can be obtained, and the result obtained by performing permission verification based on the permission verification characteristics has higher precision and reliability.

Description

Technical Field

[0001] The present invention relates to the field of data processing, and in particular to a remote verification method and system for access rights of a terminal device. Background Art

[0002] In today's digital age, remote operation and maintenance terminals for government and enterprise customers play a vital role in various businesses. With the rapid development of information technology, the data processed and tasks undertaken by remote operation and maintenance terminals are becoming more and more critical, and the security verification of their access rights has become the core link to ensure the normal operation of government and enterprise businesses and data security. Traditional terminal device access rights verification methods have exposed many limitations in the face of increasingly complex network environments and diverse security threats. Early verification methods were often based on simple passwords or token mechanisms, which were easily cracked or misused and could not provide sufficient security. For example, passwords may be leaked because the user's settings are too simple or in an unsafe environment, and tokens may be lost or copied, allowing illegal users to easily obtain access rights, bringing serious data leakage and business interruption risks to government and enterprise customers. With the advancement of technology, biometric recognition technology has gradually been applied to the field of access rights verification. However, existing biometric-based verification methods still have significant shortcomings. On the one hand, the accuracy of biometric extraction needs to be improved. Due to the accuracy differences of acquisition equipment, interference from environmental factors, and the diversity of biometrics themselves, errors are prone to occur when extracting biometrics. Summary of the invention

[0003] In view of this, the embodiment of the present invention at least provides a remote verification method and system for access rights of a terminal device. The technical solution of the present invention is implemented as follows:

[0004] On the one hand, the present invention provides a remote verification method for access rights of a terminal device, the method comprising: performing a biometric feature extraction operation on a vital sign collection image of the terminal device to obtain an initial access verification feature; performing feature decomposition on the initial access verification feature to obtain a plurality of initial verification sub-features; performing feature extraction error adjustment on each of the initial verification sub-features based on the vital sign collection image of the terminal device to obtain a corresponding verification adjustment sub-feature; determining a verification adjustment feature based on the plurality of verification adjustment sub-features; performing feature improvement on the verification adjustment feature based on the vital sign collection image of the terminal device to obtain an access verification improvement feature; determining a target access verification feature based on the access verification improvement feature; performing identification based on the target access verification feature to obtain a verification result of the access rights of the terminal device.

[0005] On the other hand, the present invention provides a remote verification system for access rights of a terminal device, comprising a memory and a processor, wherein the memory stores a computer program executable on the processor, and the processor implements the steps in the above method when executing the program.

[0006] The remote verification method for access rights of a terminal device provided by the present invention performs a biometric feature extraction operation on a vital sign collection image of the terminal device to obtain an initial permission verification feature. The initial permission verification feature is feature-decomposed according to the vital sign part to obtain a plurality of initial verification sub-features. For each of the plurality of initial verification sub-features, the extraction error of the initial verification sub-feature is determined according to the vital sign collection image of the terminal device, and the extraction error is adjusted to obtain a corresponding verification adjustment sub-feature, and a plurality of verification adjustment sub-features are obtained. Then, a verification adjustment feature is obtained according to the plurality of verification adjustment sub-features. The verification adjustment feature is feature-perfected according to the vital sign collection image of the terminal device to obtain a permission verification improvement feature. A target permission verification feature is generated according to the permission verification improvement feature.

[0007] After obtaining the initial permission verification feature, the embodiment of the present invention performs feature extraction error adjustment and feature improvement on the initial permission verification feature to obtain the permission verification improvement feature, and then obtains the target permission verification feature based on the permission verification improvement feature. The feature extraction error adjustment process is used to adjust the extraction error of the initial permission verification feature for the vital sign collection image of the terminal device, and the feature improvement process verifies the adjustment feature for the missing features of the terminal device vital sign collection image. Based on this, a permission verification feature with higher representation and accuracy can be obtained, and the result obtained by performing permission verification based on the permission verification feature has higher accuracy and reliability. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] The accompanying drawings herein are incorporated into and constitute a part of the specification. These drawings illustrate embodiments consistent with the present invention and, together with the specification, are used to explain the technical solutions of the present invention.

[0009] Figure 1 A schematic diagram of the implementation flow of a method for remotely verifying access rights of a terminal device provided in an embodiment of the present invention.

[0010] Figure 2 A schematic diagram of hardware entities of a remote verification system for access rights of a terminal device provided in an embodiment of the present invention.

[0011] Remote verification system for access rights of terminal device-1000; processor-1001; memory-1002. DETAILED DESCRIPTION

[0012] In order to make the purpose, technical solutions and advantages of the present invention clearer, the technical solutions of the present invention are further elaborated in detail below in conjunction with the drawings and embodiments. The described embodiments should not be regarded as limiting the present invention. All other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of the present invention.

[0013] The embodiment of the present invention provides a method for remote verification of access rights of a terminal device, which can be executed by a processor of a remote verification system for access rights of a terminal device. The remote verification system for access rights of a terminal device can refer to a device with data processing capabilities, such as a server, a laptop, a tablet computer, or a desktop computer.

[0014] Figure 1 A schematic diagram of a remote verification method for access rights of a terminal device provided by an embodiment of the present invention is shown in FIG. Figure 1 As shown, the method comprises the following steps:

[0015] Step S100: Perform a biometric feature extraction operation on the terminal device's vital sign collection image to obtain an initial permission verification feature.

[0016] In step S100, the terminal device vital sign collection image is obtained by collecting vital sign information related to the remote operation and maintenance terminal of government and enterprise customers through specific equipment and technology. For example, a high-definition camera may be used to shoot the face, fingerprints and other vital signs of the operation and maintenance personnel, so as to obtain images containing key information. The quality and accuracy of these images directly affect the effect of subsequent biometric extraction. After obtaining the terminal device vital sign collection image, the remote verification system of the terminal device access rights starts the biometric extraction operation. This operation aims to accurately extract the biometric information related to the permission verification from the collected image. To achieve this goal, a feasible implementation method is to use a permission verification neural network that has been pre-adjusted and converged. The neural network has been trained and optimized with a large amount of data, and can effectively identify and extract biometric features in the image.

[0017] For example, by learning from a large number of images containing different facial features, fingerprint features, etc., the permission verification neural network can learn how to accurately distinguish key biometric feature points and feature lines from the image. When the remote verification system for terminal device access rights inputs the terminal device vital signs collection image into this pre-adjusted and converged permission verification neural network, the network will analyze and process the image according to the patterns and rules it has learned.

[0018] Specifically, the remote verification system for access rights of terminal devices first pre-processes the terminal device vital signs collection image. This may include grayscale processing of the image, that is, converting the color image into a grayscale image to simplify subsequent calculations and analysis. At the same time, image noise reduction processing may also be performed to remove noise in the image caused by factors such as shooting equipment and environmental interference, thereby improving the clarity and quality of the image. For example, a Gaussian filter algorithm is used to reduce the noise of the image, and the noise interference is removed by performing a weighted average calculation on the grayscale values ​​of each pixel in the image and its surrounding pixels.

[0019] Next, the remote verification system for access rights of the terminal device inputs the preprocessed image into the permission verification neural network. The neural network will perform a series of feature extraction operations on the image. For example, when processing facial images, the neural network can identify the position and shape information of key facial features such as eyes, nose, and mouth; when processing fingerprint images, it will extract fingerprint lines, breakpoints, bifurcation points and other features. These extracted feature information will constitute an important part of the initial permission verification features.

[0020] Assume that the terminal device vital sign collection image is a facial image of an operation and maintenance personnel. The remote verification system for terminal device access rights first grayscales and de-noises it, and then inputs it into the permission verification neural network. The neural network analyzes the image layer by layer through its internal neurons and connection weights. In the first layer, edge information in the image can be identified, such as the outline edge of the eye; the second layer can further extract more complex features, such as the shape and relative position of the eye; as the network layer deepens, more advanced and abstract facial features will be continuously extracted, and finally the initial permission verification feature containing the overall facial feature information will be formed. For another example, if the fingerprint image is collected, the remote verification system for terminal device access rights will also pre-process it and input it into the permission verification neural network. The network will extract the detailed features of the fingerprint from the image, such as the location of the breakpoint, the direction of the bifurcation point, etc. These features are combined to form the initial permission verification feature related to the fingerprint.

[0021] Step S200: Decomposing the initial authority verification feature to obtain a plurality of initial verification sub-features.

[0022] In step S200, the initial permission verification features processed by the remote verification system for terminal device access rights are a set of features with permission verification related information extracted from the terminal device vital sign collection image. These features may contain a variety of biometric elements. For example, when processing facial signs, they may cover feature information of different parts such as eyes, nose, mouth, etc.; when processing fingerprint signs, they may include fingerprint patterns, breakpoints, bifurcation points and other features. The purpose of feature decomposition by the remote verification system for terminal device access rights is to decompose these comprehensive features into multiple more independent and targeted sub-features according to certain rules.

[0023] Decomposition according to the body part is a feasible and effective way. For example, when the initial permission verification feature comes from the facial body part collection image, the remote verification system of the terminal device access rights is divided according to different areas of the face. The eye area, as an important part of the face, its features are of great significance for identity authentication. The remote verification system of the terminal device access rights separates the features related to the eyes from the initial permission verification features to form an initial verification sub-feature about the eyes. This may include the shape features of the eyes, such as whether the eyes are round or almond-shaped; the position features of the eyes, that is, the relative position relationship of the eyes on the face; and the color features of the eyes. By extracting and integrating these eye-related features separately, an independent initial verification sub-feature of the eyes is formed. Similarly, for the nose part, the remote verification system of the terminal device access rights extracts the feature information of the shape, length, width, etc. of the nose, and combines it into an initial verification sub-feature about the nose. For example, whether the nose is straight or collapsed, the length of the nose bridge and the overall proportion of the face, etc., these features are all included in the initial verification sub-feature of the nose. The mouth is no exception. The remote verification system for terminal device access rights focuses on the shape of the mouth, such as the thickness of the lips, the degree of upturn or droop of the corners of the mouth, etc., and organizes these features into initial verification sub-features about the mouth.

[0024] When processing fingerprint vital sign collection images, the remote verification system for terminal device access rights will also decompose according to different feature areas of the fingerprint. The direction of the fingerprint pattern is an important feature. The remote verification system for terminal device access rights will extract the feature information reflecting the overall direction of the fingerprint pattern to form an initial verification sub-feature about the direction of the fingerprint pattern. For example, whether the fingerprint pattern is clockwise or counterclockwise, or whether it presents a complex spiral shape, etc. will be included. The breakpoint feature of the fingerprint is also one of the key information. The remote verification system for terminal device access rights extracts features related to the breakpoint position, the number of breakpoints, etc., to form an initial verification sub-feature about the breakpoint. For the bifurcation point of the fingerprint, the remote verification system for terminal device access rights analyzes the direction of the bifurcation point, the number of bifurcations, and other features, and integrates them into an initial verification sub-feature about the bifurcation point.

[0025] The implementation method of the remote verification system of the terminal device access rights to implement this feature decomposition operation can utilize specific algorithms and models. For example, the convolutional neural network (CNN) technology based on deep learning can be used. By designing appropriate convolution kernels and network structures, the network can learn how to accurately decompose features according to the feature differences of the body parts. When the convolution kernel scans the initial permission verification features, it will extract and classify features of different positions and types according to its set weight parameters. For example, for the feature extraction of the eye part, the convolution kernel will focus on the pixel information related to the shape and color of the eye, and separate these features from the overall initial permission verification features through calculation and analysis. For another example, the remote verification system of the terminal device access rights quantifies each feature element in the initial permission verification feature and converts it into a digital vector form. Then, through a clustering algorithm, such as the K-means clustering algorithm, similar feature elements are clustered together according to the distance metric between feature vectors. In the case of decomposition by body part, the feature vectors related to the eyes will be clustered into one category to form the initial verification sub-features of the eyes; the feature vectors related to the nose will be clustered into another category to form the initial verification sub-features of the nose, and so on. In this way, the remote verification system for the access rights of the terminal device decomposes the initial permission verification feature into multiple initial verification sub-features with clear body part references, providing clear and independent data units for subsequent more accurate feature extraction error adjustment and permission verification operations based on these sub-features, which helps to improve the accuracy and reliability of the entire verification process.

[0026] Step S300: According to the vital sign acquisition image of the terminal device, feature extraction error adjustment is performed on each of the initial verification sub-features to obtain a corresponding verification adjustment sub-feature.

[0027] The initial verification sub-features obtained by the remote verification system for terminal device access rights are obtained after decomposing the initial permission verification features in step S200. Although these initial verification sub-features each represent a part or a characteristic of the physical sign, due to factors such as environmental interference during image acquisition, equipment accuracy limitations, and limitations of feature extraction algorithms, there will inevitably be certain extraction errors. The task of the remote verification system for terminal device access rights is to minimize these errors as much as possible through feature extraction error adjustment.

[0028] Specifically, the remote verification system for the access rights of the terminal device must first determine the extraction error of the initial verification sub-feature. This requires the use of the terminal device vital sign collection image as a reference. For example, when the initial verification sub-feature is a feature about the eye part of the face, some detailed features of the eye shape may not be accurately extracted due to insufficient image light during the extraction process, resulting in shape deformation errors; or when extracting fingerprint breakpoint features, the position of the breakpoint is deviated due to local blurring of the fingerprint image. In order to adjust these errors, the remote verification system for the access rights of the terminal device can adopt multiple implementation methods. One feasible method is to use a deep learning model. By constructing a neural network model specifically for feature extraction error adjustment, the model takes the terminal device vital sign collection image and the initial verification sub-feature as input. After training and optimization of a large amount of data, the neurons and connection weights inside the model can learn the error patterns and corresponding adjustment methods that may occur in different initial verification sub-features under different circumstances. For example, for the initial verification sub-feature of the face and eyes, the remote verification system for the access rights of the terminal device inputs the terminal device vital sign collection image and the initial verification sub-feature into the error adjustment neural network together. The neural network deeply analyzes the pixel information of the eye in the image and compares it with the existing standard eye feature database to identify the error of the current initial verification sub-feature. If it is found that the eye shape feature has a deviation, the neural network will adjust the initial verification sub-feature according to the adjustment rules it has learned, such as changing the parameter value of the eye shape to make it closer to the real eye shape, thereby obtaining the adjusted verification adjustment sub-feature.

[0029] When processing the initial verification sub-feature of the fingerprint breakpoint, the remote verification system of the terminal device access rights also inputs the fingerprint acquisition image and the sub-feature into the error adjustment model. The model analyzes the overall texture of the fingerprint in the image and the detailed information around the breakpoint to determine the accuracy of the breakpoint position. If a deviation is found in the breakpoint position, the model will correct the breakpoint position based on the preset algorithm and training experience to generate a more accurate verification adjustment sub-feature.

[0030] Another implementation method that can be implemented is a statistical method. The remote verification system for terminal device access rights can collect a large number of similar terminal device vital sign collection images and their corresponding accurate initial verification sub-feature data to establish a statistical model. When performing error adjustment on the current initial verification sub-feature, the remote verification system for terminal device access rights compares and analyzes the sub-feature with the data in the statistical model. By calculating statistical indicators such as the similarity and degree of deviation between features, the error range of the initial verification sub-feature is determined. Then, the initial verification sub-feature is adjusted according to a pre-set error adjustment formula. For example, assuming that for a certain type of fingerprint pattern feature extraction error, the error adjustment formula obtained by the statistical model is: adjusted eigenvalue = current eigenvalue + (standard eigenvalue - current eigenvalue) × error correction coefficient. Among them, the standard eigenvalue comes from the accurate data in the statistical model, and the error correction coefficient is an empirical value obtained based on a large amount of experimental data. The remote verification system for terminal device access rights calculates the initial verification sub-feature through this formula to obtain the error-adjusted verification adjustment sub-feature.

[0031] For another example, for the adjustment of facial features, the remote verification system of the terminal device access rights can use the method of geometric transformation. If there is an error in the positional relationship of the facial features in the initial verification sub-feature, the remote verification system of the terminal device access rights can use geometric transformation algorithms such as affine transformation and perspective transformation to adjust the position of the facial features in the initial verification sub-feature according to the actual position and proportional relationship of the facial features in the terminal device vital sign collection image, so that it conforms to the real facial structure, and then obtain a more accurate verification adjustment sub-feature. Through these implementation methods, the remote verification system of the terminal device access rights can effectively adjust the feature extraction error for each initial verification sub-feature based on the terminal device vital sign collection image, and obtain the corresponding verification adjustment sub-feature, which provides a higher quality and more accurate data basis for the subsequent determination of the verification adjustment feature and further permission verification operations, thereby improving the reliability and accuracy of the entire remote verification of the terminal device access rights.

[0032] Step S400: Determine a verification adjustment feature based on the plurality of verification adjustment sub-features.

[0033] Step S400 requires the remote verification system of the terminal device access rights to determine the verification adjustment feature based on multiple verification adjustment sub-features. This step integrates and refines multiple sub-features after feature extraction error adjustment, aiming to generate a more comprehensive, representative and accurate verification feature, providing key data support for the subsequent permission verification process.

[0034] The multiple verification adjustment sub-features processed by the remote verification system of the terminal device access rights are the results obtained after the feature extraction error adjustment is performed on each initial verification sub-feature in step S300. These verification adjustment sub-features respectively describe the terminal device signs more accurately from different angles, different sign parts or different feature dimensions. For example, when processing facial signs, the eye verification adjustment sub-features, nose verification adjustment sub-features and mouth verification adjustment sub-features after error adjustment each contain more accurate feature information about the corresponding facial parts; when processing fingerprint signs, the verification adjustment sub-features in terms of fingerprint pattern direction, breakpoints, bifurcation points, etc. also more accurately reflect the actual characteristics of the fingerprint. The process of determining the verification adjustment feature by the remote verification system of the terminal device access rights is a process of organically integrating these scattered but interrelated verification adjustment sub-features. A feasible implementation method is to adopt a weighted fusion method. The remote verification system of the terminal device access rights assigns corresponding weights to each verification adjustment sub-feature according to its importance to the overall authority verification. For example, in facial feature verification, eye features are of high importance for identity authentication due to their uniqueness and stability. Therefore, the remote verification system for terminal device access rights can assign a higher weight to the eye verification adjustment sub-feature, such as 0.4; the nose feature is slightly less important than the eye feature, and may be assigned a weight of 0.3; the mouth feature weight may be 0.3. The weight distribution is not fixed, but can be dynamically adjusted and optimized through a large amount of experimental data and actual application scenarios.

[0035] After determining the weight, the remote verification system of the terminal device access rights calculates the verification adjustment feature according to the following formula: Verification adjustment feature = eye verification adjustment sub-feature × 0.4 + nose verification adjustment sub-feature × 0.3 + mouth verification adjustment sub-feature × 0.3. The calculation here is not a simple addition of values, but a corresponding weighted operation on the feature vectors or feature data contained in each verification adjustment sub-feature. For example, the eye verification adjustment sub-feature may be a feature vector containing multi-dimensional information such as eye shape, color, and position. The remote verification system of the terminal device access rights multiplies each dimensional value of the vector by 0.4, and the nose and mouth verification adjustment sub-features are also subjected to similar weighted operations. The weighted results are then merged and integrated to obtain a verification adjustment feature that integrates the feature information of multiple facial parts.

[0036] For fingerprint features, the remote verification system for terminal device access rights will also assign weights to verification adjustment sub-features such as fingerprint pattern direction, breakpoints, and bifurcation points. Assume that the weight of the fingerprint pattern direction verification adjustment sub-feature is 0.4, the weight of the breakpoint verification adjustment sub-feature is 0.3, and the weight of the bifurcation point verification adjustment sub-feature is 0.3. The remote verification system for terminal device access rights integrates and calculates these verification adjustment sub-features according to the weighted fusion formula similar to the above, and obtains a verification adjustment feature that reflects the overall characteristics of the fingerprint.

[0037] In addition to the weighted fusion method, the remote verification system for terminal device access rights can also use feature splicing and convolution fusion technology in deep learning. First, the remote verification system for terminal device access rights splices multiple verification adjustment sub-features in the feature dimension. For example, the verification adjustment sub-features of eyes, nose, and mouth are connected together in a certain order to form a longer feature vector. Then, the spliced ​​feature vector is processed by a convolutional neural network. The convolution layer in the convolutional neural network uses a specific convolution kernel to perform a convolution operation on this long feature vector. The convolution kernel slides on the feature vector and extracts higher-level and more abstract feature information by calculating the inner product of the convolution kernel and the local area of ​​the feature vector. For example, the convolution kernel can identify features such as the spatial relationship between eyes, nose, and mouth. After multiple layers of convolution operations, the remote verification system for terminal device access rights finally obtains a fused and refined verification adjustment feature. Through these implementation methods, the remote verification system for terminal device access rights can effectively integrate and process multiple verification adjustment sub-features to generate a verification adjustment feature. This verification adjustment feature not only integrates the information of multiple sub-features, but also further highlights the key features and weakens the secondary features through reasonable weight distribution or deep learning feature fusion methods, thereby more comprehensively and accurately reflecting the true characteristics of the terminal device's vital signs. It provides a solid and reliable data foundation for subsequent feature improvement based on the verification adjustment feature and the final determination of the target permission verification feature, which helps to improve the accuracy and reliability of remote verification of terminal device access rights.

[0038] Step S500: Based on the vital sign acquisition image of the terminal device, the verification adjustment feature is improved to obtain the permission verification improvement feature.

[0039] The verification adjustment feature is obtained by integrating multiple verification adjustment sub-features in step S400. Although these verification adjustment sub-features have been adjusted for feature extraction errors and are more accurate than the initial verification sub-features, due to various complex factors in the vital sign collection process, the verification adjustment feature may still have some missing or inaccurate parts. The task of the remote verification system for terminal device access rights is to make up for these deficiencies through feature improvement operations.

[0040] Specifically, the remote verification system of the terminal device access rights clearly verifies the missing information that may exist in the adjustment feature. This requires the use of the terminal device's vital signs to collect images again. For example, in the facial vital sign verification scenario, although the verification adjustment feature already contains the feature information of the main facial parts such as eyes, nose, and mouth that have been adjusted for errors, it may be due to the angle problem during image acquisition that some detailed features of the facial contour are not fully reflected in the verification adjustment feature; or in fingerprint vital sign verification, the verification adjustment feature may miss some tiny features of the fingerprint edge.

[0041] In order to achieve feature perfection, the remote verification system of the terminal device access rights can adopt the feature generation network technology based on deep learning. By constructing a special feature generation network, the network takes the terminal device vital sign collection image and the verification adjustment feature as input. After training with a large amount of data, this feature generation network can learn the mapping relationship between different vital sign collection images and corresponding complete features. For example, for facial vital sign verification, the remote verification system of the terminal device access rights inputs the facial image collected by the terminal device and the existing verification adjustment feature into the feature generation network. The neurons inside the network will perform a deep analysis of the facial information in the image, and at the same time, combined with the existing information contained in the verification adjustment feature, identify the facial contour detail features missing in the current verification adjustment feature. Then, the network generates the corresponding missing feature information based on the learned patterns and rules, and merges it with the original verification adjustment feature. Assuming that the facial contour features in the original verification adjustment feature are not complete, the feature generation network can generate contour features that conform to the current facial actual situation by learning a large number of facial images, and add them to the verification adjustment feature, thereby obtaining a more complete permission verification and improvement feature.

[0042] Another feasible implementation method is a method based on feature matching and interpolation. The remote verification system for terminal device access rights matches the verification adjustment features with a pre-established physical sign feature library. This physical sign feature library contains a large number of complete physical sign feature samples of different types, different angles, and different states. The remote verification system for terminal device access rights finds the most matching sample set by calculating the similarity between the verification adjustment features and each sample in the feature library. For example, in facial physical sign verification, find multiple facial feature samples that are most similar to the current verification adjustment features in terms of facial structure, expression, etc.

[0043] Then, based on the differences between these matching samples and the verification adjustment features, an interpolation algorithm is used to fill in the missing parts of the verification adjustment features. Assume that the features of the verification adjustment features in a certain area of ​​the face are not clear or are missing, while the matching samples have complete and clear features in this area. The remote verification system for terminal device access rights can use algorithms such as linear interpolation or spline interpolation to calculate and generate supplementary feature values ​​suitable for the current verification adjustment features based on the feature values ​​of the matching samples in this area. For example, for the skin color features of a certain area of ​​the face, an interpolation algorithm is used to generate a skin color feature value that conforms to the current actual facial situation based on the skin color information of the matching samples, and add it to the verification adjustment features to achieve feature perfection.

[0044] Through these implementation methods, the remote verification system for terminal device access rights can collect images based on the terminal device's vital signs, perform effective feature improvement operations on the verification adjustment features, and obtain the permission verification improvement features. This permission verification improvement feature more comprehensively and accurately reflects the true features of the terminal device's vital signs, and provides a higher quality and more reliable data basis for the subsequent determination of the target permission verification features and the final permission verification, greatly improving the accuracy and effectiveness of the remote verification of terminal device access rights.

[0045] Step S600: Determine a target permission verification feature based on the permission verification improvement feature.

[0046] The authority verification improvement feature is the result obtained after the verification adjustment feature is improved in step S500. This feature has already reflected the terminal device's vital information more comprehensively and accurately, but in order to ensure the high accuracy of the authority verification, it is necessary to further determine the target authority verification feature.

[0047] One feasible implementation method is feature selection and fusion. The remote verification system for terminal device access rights analyzes the permission verification improvement features and determines the importance of each sub-feature for permission verification. For example, in facial sign verification, the permission verification improvement features may include multiple sub-features such as facial contour, facial features proportion, skin color, etc. The remote verification system for terminal device access rights determines the contribution of each sub-feature to identity authentication through a large amount of experimental data and machine learning algorithms. For example, after analysis, it was found that facial contour and facial features proportion have higher recognition in identity authentication, while skin color has relatively little effect on the verification results in some cases.

[0048] Based on this analysis, the remote verification system for terminal device access rights adopts feature selection algorithms, such as the information gain-based method. Information gain is used to measure the increase in the amount of information brought by a feature in a classification task. Its calculation formula is: Information gain = entropy (original data set) - entropy (data set divided according to features). The entropy calculation formula is: ,in It is an event By calculating the information gain of each sub-feature, the remote verification system of the terminal device access rights can select sub-features with higher information gain, that is, sub-features that are more important for permission verification. For example, in facial feature verification, sub-features with high information gain such as facial contour and facial features ratio are selected.

[0049] Then, the remote verification system for the access rights of the terminal device fuses the selected sub-features. The fusion method can be weighted average fusion. Assuming that the facial contour sub-feature F1 and the facial features proportion sub-feature F2 are selected, the remote verification system for the access rights of the terminal device assigns weights to them according to their importance in the access rights verification, such as the facial contour sub-feature weight w1=0.6 and the facial features proportion sub-feature weight w2=0.4. Through the weighted average formula: target access rights verification feature = w1F1+w2F2, these sub-features are fused together to obtain the target access rights verification feature. Another implementation method is feature conversion based on deep learning. Deep neural networks, such as autoencoders, can be used. The autoencoder consists of an encoder and a decoder. The remote verification system for the access rights of the terminal device inputs the access rights verification improvement feature into the encoder part of the autoencoder, and the encoder compresses and converts the feature, extracts the most critical and core information, and converts it into a low-dimensional feature representation. This low-dimensional feature representation is the refined target access rights verification feature. For example, when processing the permission verification and improvement features of facial features, the encoder of the autoencoder will process the features containing many facial detail information and convert them into a more compact low-dimensional vector that better represents the essential features of the face. This vector is the target permission verification feature. In fingerprint feature verification, the autoencoder will also process the permission verification and improvement features, converting the complex features of the fingerprint into a concise and critical low-dimensional feature representation as the target permission verification feature. In this way, the remote verification system for the access rights of the terminal device can determine the target permission verification feature based on the permission verification and improvement features using a suitable implementation method, providing more accurate and reliable core data support for the subsequent terminal device access rights verification based on the feature, thereby effectively improving the accuracy and authority of the verification results.

[0050] Step S700: Identify based on the target permission verification feature to obtain a verification result of the terminal device access permission.

[0051] The target permission verification feature is a highly accurate and representative feature set obtained after a series of operations such as biometric feature extraction, feature decomposition, error adjustment, and feature improvement, starting from the terminal device's vital signs collection image. This feature set contains key information that can accurately reflect the identity or permissions of the terminal device user.

[0052] Specifically, the remote verification system for terminal device access rights can adopt a variety of implementation methods to perform identification based on target permission verification features. One feasible method is a matching algorithm based on distance measurement. The remote verification system for terminal device access rights compares the target permission verification feature with the standard permission verification feature template pre-stored in the database. For example, in a facial recognition scenario, the standard permission verification feature template is a set of facial features of legitimate users that have been pre-collected and strictly verified. The remote verification system for terminal device access rights calculates the distance between the target permission verification feature and each standard template, such as the Euclidean distance formula.

[0053] Assume that the target permission verification feature is the currently collected and processed facial feature vector X, and there are multiple standard facial feature template vectors in the database The remote verification system for terminal device access rights calculates X and each The Euclidean distance between If the calculated distance is less than the preset threshold, it is considered that the target permission verification feature matches the standard template successfully, that is, the verification is passed; conversely, if all calculated distances are greater than the threshold, the verification fails. For example, the threshold is set to T, when When , the remote verification system of the terminal device access right determines that the facial features of the current terminal device user match the kth standard template and grants access right; if for all j, , access is denied.

[0054] In the fingerprint recognition scenario, the distance measurement algorithm is also used. The target permission verification feature is the processed fingerprint feature vector, and the standard template is the pre-stored legal fingerprint feature vector. The remote verification system of the terminal device access rights calculates the distance between the two, such as using the Manhattan distance formula: , to determine whether there is a match. If the distance is within the threshold range, the fingerprint verification is passed and access to the terminal device is allowed; otherwise, the verification fails and access is prohibited.

[0055] Another implementation method is a method based on machine learning classifiers. The remote verification system for terminal device access rights utilizes a pre-trained classification model, such as a support vector machine (SVM) classifier. During the training phase, a large amount of known permission verification feature data (including legal and illegal samples) is used to train the classifier so that the classifier learns the patterns and rules of features under different permission states. During the recognition phase, the remote verification system for terminal device access rights inputs the target permission verification features into the trained SVM classifier. The SVM classifier determines the category to which the target feature belongs by calculating the relationship between the target feature and the classification hyperplane. For the linearly separable case, the equation of the classification hyperplane is wT x+b=0, where w is the normal vector of the hyperplane, x is the feature vector, and b is the bias term. The classifier determines whether the target permission verification feature x is on the side of the hyperplane to determine whether it belongs to the legal permission category. If it is determined to belong to the legal permission category, the verification passes and the terminal device is granted access rights; if it is determined to belong to the illegal category, the verification fails and access is denied.

[0056] As an implementation manner, the step S300, based on the vital sign acquisition image of the terminal device, performs feature extraction error adjustment on each of the initial verification sub-features to obtain a corresponding verification adjustment sub-feature, includes:

[0057] Step S310: embedding the terminal device vital sign acquisition image and the proposed processing initial verification sub-feature into a feature extraction error adjustment indication framework to obtain a proposed adjustment image feature, wherein the proposed adjustment image feature is used to instruct the image processing network to perform feature extraction error adjustment on the proposed processing initial verification sub-feature, and the proposed processing initial verification sub-feature is any one of the multiple initial verification sub-features;

[0058] Step S320: Loading the image feature to be adjusted into the image processing network, obtaining the image feature analysis result output by the image processing network for the image feature to be adjusted, wherein the image feature analysis result to be adjusted includes the verification adjustment sub-feature corresponding to the initial verification sub-feature to be processed and the extraction error category, wherein the extraction error category identifies the type corresponding to the extraction feature error.

[0059] Step S310 requires the remote verification system of the terminal device access rights to embed the terminal device vital sign acquisition image and the proposed initial verification sub-features into the feature extraction error adjustment indication framework to obtain the proposed adjustment image features. The terminal device vital sign acquisition image contains a wealth of original information about the user's vital signs, and the proposed initial verification sub-features are information units preliminarily extracted from these images for a specific vital sign part or feature dimension. The role of the feature extraction error adjustment indication framework is to integrate these two types of information and generate a new feature representation that can guide the subsequent image processing network to perform error adjustment.

[0060] For example, in the facial feature verification scenario, the terminal device's feature collection image is a clear face image that contains complete information about various facial organs such as eyes, nose, and mouth. The initial verification sub-feature to be processed is assumed to be a feature about the eye part, which may include preliminary extracted information such as the approximate shape and position of the eye. The remote verification system for the terminal device's access rights inputs this face image and the initial verification sub-feature of the eye into the feature extraction error adjustment indication framework.

[0061] For example, the feature extraction error adjustment indication framework can be constructed based on the attention mechanism in deep learning. The attention mechanism enables the remote verification system of the terminal device access rights to focus on the key parts when processing complex information, so as to better integrate and analyze the information. In this framework, the remote verification system of the terminal device access rights will first perform feature encoding on the terminal device vital sign acquisition image and the initial verification sub-features to be processed. For images, a convolutional neural network (CNN) can be used for feature extraction to convert the image into a high-dimensional feature vector representation. For example, through a series of convolutional layer and pooling layer operations, the original image is gradually converted into a feature map containing rich image semantic information. For the initial verification sub-features to be processed, corresponding encoding operations will also be performed to convert them into dimensions and formats that match the image feature vector. Assume that the image is processed by CNN to obtain a feature vector I, and the initial verification sub-features to be processed are encoded to obtain a feature vector F. In the framework based on the attention mechanism, an attention weight matrix A will be calculated, which represents the degree of correlation between the image features and the initial verification sub-features. A feasible method for calculating the attention weight matrix is ​​through dot product operations and softmax functions. For example, first calculate ,in is the i-th element of the image feature vector I, is the jth element of the initial verification sub-feature vector F to be processed, and n is the dimension of the vector F. This formula calculates the similarity between the image features and each element of the initial verification sub-feature and normalizes them to obtain the attention weight matrix A.

[0062] Then, according to the attention weight matrix A, the remote verification system of the terminal device access rights performs weighted fusion on the image feature vector I and the initial verification sub-feature vector F to be processed. For example, the image feature G to be adjusted can be obtained by the formula Calculated, where m is the dimension of the image feature vector I, Represents a fusion operation, such as element-level multiplication or concatenation operation. In this way, the image feature G to be adjusted not only contains the overall information of the terminal device vital sign acquisition image, but also highlights the key information related to the initial verification sub-feature to be processed, which can effectively instruct the subsequent image processing network to extract feature errors for the initial verification sub-feature. In the fingerprint vital sign verification scenario, the terminal device vital sign acquisition image is a clear fingerprint image, and the initial verification sub-feature to be processed is the preliminary extracted feature about the fingerprint breakpoint. The remote verification system of the terminal device access rights will also input the fingerprint image and the breakpoint initial verification sub-feature into the feature extraction error adjustment indication framework. Through a similar attention mechanism-based method, the fingerprint image is feature encoded, the breakpoint initial verification sub-feature is encoded, the attention weight matrix is ​​calculated, and weighted fusion is performed to obtain the image feature to be adjusted that can indicate the fingerprint breakpoint feature extraction error adjustment.

[0063] Step S320 requires the remote verification system of the terminal device access rights to load the image features to be adjusted into the image processing network to obtain the analysis results of the image features to be adjusted output by the network, which include the verification adjustment sub-features corresponding to the initial verification sub-features to be processed and the extraction error categories. The image processing network is an intelligent model trained with a large amount of data, which can perform in-depth analysis on the input image features to be adjusted, identify the feature extraction errors therein, and generate the corresponding adjusted verification adjustment sub-features.

[0064] Continuing with the facial features verification as an example, the remote verification system for the access rights of the terminal device inputs the image features to be adjusted obtained in step S310 into the image processing network. This image processing network can be a deep convolutional neural network, which uses a large amount of image data containing different facial features and corresponding accurate feature annotations during the training process. The network contains multiple convolutional layers, pooling layers, and fully connected layers.

[0065] When the features of the image to be adjusted are input into the network, the convolution layer will extract local features of the features and capture various details in the image through different convolution kernels. For example, some convolution kernels may be used specifically to extract the contour features of the eye, while others may be used to extract the texture features of the eye. The pooling layer will downsample the features extracted by the convolution layer to reduce the amount of data while retaining the key features. The fully connected layer will integrate and classify the features extracted by the previous layer. When processing the features of the image to be adjusted about the eyes, the image processing network will analyze the differences between the features of the image to be adjusted and the pre-learned standard eye features. If it is found that there is an error in the extraction of the eye shape, for example, the actual eye shape should be almond-shaped, but the shape extracted by the initial verification sub-feature is closer to a circle, the network will adjust the eye shape feature according to the adjustment rules it has learned. At the same time, the network will also recognize that this error belongs to the category of shape extraction error. Finally, the network outputs the adjusted eye verification adjustment sub-features and the extraction error category.

[0066] For example, the training process of the image processing network can use the back propagation algorithm. In the training phase, a large number of facial images with accurate labels and the corresponding initial verification sub-features are input into the network, and the network outputs the predicted verification adjustment sub-features and error categories. By calculating the loss function between the predicted results and the true labels, such as the mean square error loss function. Then, the back propagation algorithm is used to update the weight parameters of the network according to the gradient of the loss function, so that the network can gradually learn accurate feature extraction and error adjustment capabilities.

[0067] The image processing network can also use recurrent neural networks (RNN) or their variants, such as long short-term memory networks (LSTMs), to process the image features to be adjusted. For some vital signs with sequence characteristics, such as the direction of fingerprint lines, RNN or LSTM can better capture the temporal or sequential relationship between features. For example, when processing the image features to be adjusted of the fingerprint line direction, LSTM can remember the starting direction and turning information of the lines through its internal memory units and gating mechanisms, so as to more accurately analyze and adjust the errors in the extraction of line direction features, and output more accurate verification adjustment sub-features and accurate extraction error categories.

[0068] Through steps S310-S320, the remote verification system for terminal device access rights can utilize the feature extraction error adjustment indication framework and image processing network, for each initial verification sub-feature, combined with the terminal device vital sign acquisition image, to effectively identify and adjust the errors generated in the feature extraction process, and obtain the corresponding verification adjustment sub-features and extraction error categories. This not only improves the accuracy of the verification sub-features, but also provides more detailed error information for subsequent verification processes, which helps to further improve the reliability and accuracy of the entire terminal device access rights remote verification method. Whether in different scenarios such as facial vital sign verification or fingerprint vital sign verification, this implementation method can provide high-quality data support for permission verification through precise error adjustment, and ensure the security of remote operation and maintenance terminals for government and enterprise customers and the effectiveness of access control.

[0069] As an implementation manner, the step S500, based on the vital sign acquisition image of the terminal device, performs feature improvement on the verification adjustment feature to obtain the permission verification improvement feature, includes:

[0070] Step S510: embedding the terminal device vital sign acquisition image and the verification adjustment feature into a feature improvement indication framework to obtain a to-be-perfected image feature, wherein the to-be-perfected image feature is used to instruct an image processing network to perform feature improvement on the verification adjustment feature;

[0071] Step S520: loading the image features to be improved into the image processing network, obtaining the image features analysis results output by the image processing network for the image features to be improved, wherein the image features analysis results to be improved include the permission verification improvement features.

[0072] Step S510 requires the remote verification system of the terminal device access rights to embed the terminal device vital sign acquisition image and the verification adjustment features into the feature improvement indication framework, so as to obtain the proposed perfected image features. The terminal device vital sign acquisition image is an intuitive record of the terminal device user's vital sign information, which contains rich details and overall features; the verification adjustment features are relatively more accurate feature representations obtained after the error adjustment of the initial verification sub-features in the previous steps, but there may still be some missing information or inaccuracies. The purpose of the feature improvement indication framework is to generate a new feature representation that can guide the subsequent image processing network to perform feature improvement.

[0073] Taking facial sign verification as an example, the terminal device sign collection image is a clear face image, covering complete information of all parts of the face, such as skin texture, subtle changes in facial contours, etc. After the previous error adjustment, the verification adjustment feature already contains relatively accurate facial key features, such as the shape and position of the eyes, nose, and mouth, but may be insufficient in capturing the details of facial expressions. The remote verification system for terminal device access rights inputs this face image and the existing facial verification adjustment features into the feature improvement indication framework.

[0074] For example, the feature improvement indication framework can be constructed based on the fusion mechanism of deep learning. First, the remote verification system of the terminal device access rights performs feature encoding on the terminal device vital sign collection image and the verification adjustment feature. For the terminal device vital sign collection image, a convolutional neural network (CNN) is used for feature extraction to convert the image into a high-dimensional feature vector representation. For example, through a series of convolutional layer and pooling layer operations, the original face image is gradually converted into a feature map containing rich facial semantic information. It is assumed that the image feature vector I is obtained after processing. For the verification adjustment feature, corresponding encoding is also performed to match the image feature vector in dimension and format to obtain the feature vector V.

[0075] Next, in the feature improvement indication framework, the image feature vector I and the verification adjustment feature vector V are combined through a fusion operation. One feasible fusion method is weighted fusion based on the attention mechanism. The remote verification system for terminal device access rights calculates an attention weight matrix A to determine the relative importance of image features and verification adjustment features in the fusion process. One method of calculating the attention weight matrix is ​​to use a dot product operation and a softmax function.

[0076] Step S520 requires the remote verification system of the terminal device access rights to load the image features to be improved into the image processing network, so as to obtain the analysis results of the image features to be improved output by the network, which include the permission verification improvement features. The image processing network is an intelligent model trained with a large amount of data, with powerful data analysis and feature generation capabilities, and can conduct in-depth analysis of the input image features to be improved, mine the potential information therein, and generate more complete and accurate permission verification improvement features.

[0077] Still taking facial features verification as an example, the remote verification system for terminal device access rights inputs the proposed perfected image features obtained in step S510 into the image processing network. The image processing network can be a deep convolutional neural network, which uses a large amount of image data containing different facial features and corresponding accurate feature annotations during the training process. The network contains multiple convolutional layers, pooling layers, and fully connected layers.

[0078] When the proposed image features are input into the network, the convolution layer will extract local features of the features. Different convolution kernels will capture various details in the proposed image features. For example, some convolution kernels may focus on extracting micro-expression features related to facial expressions, such as fine wrinkles at the corners of the eyes, slight upturn or downturn at the corners of the mouth, etc.; other convolution kernels may be used to extract facial skin texture features, which may not be fully reflected in the previous verification adjustment features. The pooling layer will downsample the features extracted by the convolution layer, retaining key features while reducing the amount of data. The fully connected layer will integrate and comprehensively analyze the features extracted by the previous layers. When processing facial proposed image features, the image processing network will perform in-depth analysis of the input features based on the facial feature patterns and rules learned during the training process. If the verification adjustment feature is found to be missing in facial expression details, the network will generate corresponding facial expression detail features based on the image information contained in the proposed image features, and merge them with the original verification adjustment features. For example, the facial expression features in the original verification adjustment features are relatively vague, while the proposed perfected image features contain richer expression information by integrating with the terminal device's vital signs collection images. The image processing network will analyze this information, generate more accurate facial expression features, and integrate them with other existing facial features, and finally output permission verification perfected features containing more complete facial features.

[0079] For example, the training process of the image processing network can use the back propagation algorithm. In the training phase, a large number of facial images with accurate labels and corresponding verification adjustment features are input into the network, and the network outputs the predicted authority verification improvement features. By calculating the loss function between the predicted result and the true label, such as the mean square error loss function. Then, the back propagation algorithm is used to update the weight parameters of the network according to the gradient of the loss function, so that the network can gradually learn accurate feature improvement capabilities. The image processing network can also use the architecture of the generative adversarial network (GAN) to achieve feature improvement. In this architecture, a generator and a discriminator are included. The role of the generator is to generate authority verification improvement features based on the input quasi-perfected image features, while the discriminator is responsible for judging whether the generated features are real and accurate. During the training process, the generator and the discriminator confront each other and continuously optimize. The generator strives to generate results that are closer to the real authority verification improvement features to deceive the discriminator; the discriminator continuously improves its judgment ability to accurately distinguish between real features and generated features. Through this adversarial training, the generator can learn how to generate high-quality authority verification improvement features based on the quasi-perfected image features.

[0080] Through steps S510-S520, the remote verification system for access rights of the terminal device utilizes the feature improvement indication framework and the image processing network to effectively combine the information of the terminal device's vital sign collection image with the verification adjustment features, perform comprehensive feature improvement on the verification adjustment features, and obtain the permission verification improvement features.

[0081] As an implementation manner, the step S600, determining the target permission verification feature according to the permission verification improvement feature, includes:

[0082] Step S610: Decomposing the permission verification improvement feature to obtain a plurality of permission verification improvement sub-features;

[0083] Step S620: According to the terminal device vital sign collection image, perform feature extraction error adjustment on each of the permission verification improvement sub-features to obtain a corresponding permission verification improvement adjustment sub-feature;

[0084] Step S630: Determine the target permission verification feature based on the plurality of permission verification improvement adjustment sub-features.

[0085] Step S610 requires the remote verification system of the terminal device access rights to decompose the rights verification improvement feature to obtain multiple rights verification improvement sub-features. The rights verification improvement feature is a relatively complete and accurate feature set obtained after the previous multiple steps of processing, but in order to analyze and utilize these features more carefully, the remote verification system of the terminal device access rights needs to decompose them into multiple more targeted sub-features.

[0086] Taking facial sign verification as an example, the permission verification feature may be a vector containing overall facial feature information, covering comprehensive information such as facial contour, facial features, skin color, etc. The remote verification system for terminal device access rights performs feature decomposition based on different facial regions and feature dimensions. For facial contour, features such as face shape (round, square, oval, etc.) and curvature of facial lines are extracted to form permission verification and improvement sub-features about facial contour; for facial features, detailed feature extraction is performed on eyes, nose, mouth, etc. Taking eyes as an example, the shape of the eyes (single eyelid, double eyelid, phoenix eyes, etc.), the size of the eyes, the distance between the eyes, etc. are obtained as permission verification and improvement sub-features of the eyes; for the nose, sub-features such as nose bridge height, nose tip shape, nose wing width, etc. are extracted; for the mouth, relevant sub-features such as lip thickness and mouth corner shape are separated. In addition, skin color will also extract features such as skin color hue and brightness as part of the permission verification and improvement sub-features. For example, the remote verification system for terminal device access rights can use a feature extraction network based on deep learning to perform feature decomposition. For example, a convolutional neural network (CNN) model can be constructed, and the structure of the model is designed to extract features for different feature areas and dimensions. For the permission verification and improvement features corresponding to facial images, different convolution kernels are designed so that some convolution kernels are specifically used to extract facial contour features, while other convolution kernels focus on extracting facial features. When these convolution kernels perform convolution operations on the permission verification and improvement features, they will extract the corresponding feature information according to their weight parameters and separate it into different permission verification and improvement sub-features. For the permission verification and improvement features of fingerprint images, different types of features such as fingerprint patterns, breakpoints, and bifurcation points can also be extracted through carefully designed convolution kernels to achieve feature decomposition.

[0087] Step S620 requires the remote verification system of the terminal device access rights to perform feature extraction error adjustment on each permission verification improvement sub-feature based on the terminal device vital sign acquisition image, and obtain the corresponding permission verification improvement adjustment sub-feature. Although the permission verification improvement sub-feature has gone through the previous error adjustment and feature improvement steps, there may still be a certain extraction error due to the influence of various factors such as the acquisition environment and equipment accuracy. Therefore, the remote verification system of the terminal device access rights needs to fine-tune these sub-features again based on the terminal device vital sign acquisition image.

[0088] Continuing with the facial sign verification as an example, for the permission verification and improvement sub-feature of eye shape, it is assumed that the eye shape extracted through the previous steps is "circular", but in the actual terminal device sign collection image, it is found that the actual shape of the eye is closer to "almond shape", and there is a certain shape extraction error. The remote verification system for terminal device access rights inputs the terminal device sign collection image and the permission verification and improvement sub-feature of the eye shape into a special error adjustment model. This error adjustment model can be a neural network based on deep learning, which learns the accurate features of eye shapes and feasible error patterns in a large number of different facial images during the training process. The model analyzes the actual shape information of the eyes in the image, compares it with the current permission verification and improvement sub-feature, and adjusts the permission verification and improvement sub-feature of the eye shape according to the learned adjustment rules, and corrects it to an "almond shape" that is closer to the actual situation, thereby obtaining the adjusted eye shape permission verification and improvement adjustment sub-feature.

[0089] For example, the error adjustment model can adopt a residual network (ResNet) structure. The residual network can effectively solve the problem of gradient vanishing in deep learning by introducing jump connections, so that the model can learn more complex error adjustment modes. During the training process, a large number of terminal device vital sign acquisition images with accurate labels and corresponding permission verification and improvement sub-features are input into the residual network, and the network outputs the adjusted permission verification and improvement adjustment sub-features. By calculating the loss function between the predicted result and the true label, such as the mean square error loss function. Then, the back propagation algorithm is used to update the network's weight parameters according to the gradient of the loss function, so that the network can learn accurate error adjustment capabilities.

[0090] Step S630 requires the remote verification system of the terminal device access rights to determine the target permission verification feature based on multiple permission verification and adjustment sub-features. This step is to integrate multiple sub-features after error adjustment to form a comprehensive target permission verification feature that best represents the actual situation of the terminal device's vital signs.

[0091] In the facial sign verification, the remote verification system of the terminal device access rights will fuse the adjusted facial contour, eyes, nose, mouth, skin color and other permission verification and adjustment sub-features. A feasible fusion method is weighted fusion. The remote verification system of the terminal device access rights assigns corresponding weights to each sub-feature according to its importance to facial identity authentication. For example, the facial contour feature has a high degree of recognition for the overall facial recognition, and may be assigned a weight of 0.3; the eye feature may have a weight of 0.3 due to its uniqueness and stability; the nose and mouth feature weights are 0.15 respectively; the skin color feature weight is 0.1. Through the weighted fusion formula: target permission verification feature = facial contour permission verification and adjustment sub-feature × 0.3 + eye permission verification and adjustment sub-feature × 0.3 + nose permission verification and adjustment sub-feature × 0.15 + mouth permission verification and adjustment sub-feature × 0.15 + skin color permission verification and adjustment sub-feature × 0.1, each sub-feature is integrated to obtain the final facial target permission verification feature.

[0092] In addition to weighted fusion, the remote verification system for terminal device access rights can also use feature splicing and convolution fusion technology in deep learning. First, multiple permission verification and improvement adjustment sub-features are spliced ​​in the feature dimension. For example, the permission verification and improvement adjustment sub-features of various parts of the face are connected together in a certain order to form a longer feature vector. Then, the spliced ​​feature vector is processed by a convolutional neural network. The convolution layer in the convolutional neural network uses a specific convolution kernel to perform a convolution operation on this long feature vector to extract more advanced and abstract feature information. After multiple layers of convolution operations, the remote verification system for terminal device access rights finally obtains a fused and refined target permission verification feature.

[0093] As an implementation mode, the step S100, performing a biometric feature extraction operation on the vital sign collection image of the terminal device to obtain an initial permission verification feature, includes:

[0094] Step S110: performing a biometric feature extraction operation on the vital sign collection image of the terminal device through a pre-calibrated and converged authority verification neural network to obtain an initial authority verification feature.

[0095] When the remote verification system for access rights of a terminal device receives a vital sign collection image of the terminal device, it inputs it into a previously calibrated and converged permission verification neural network, such as a convolutional neural network or any other feasible neural network. Taking a facial image as an example, the remote verification system for access rights of a terminal device first preprocesses the vital sign collection image of the terminal device. This may include normalization of the image, that is, mapping the pixel values ​​of the image to a specific range, such as [0, 1], to eliminate differences between different images due to factors such as brightness and contrast. Then, the normalized image is input into the permission verification neural network. Each neuron and layer within the network will gradually analyze and extract features from the image. For example, in the first convolution layer of the network, multiple different convolution kernels are used to scan the image. These convolution kernels can be regarded as some small filters, each of which has its own specific weight parameters. When the convolution kernel slides on the image, it performs a dot product operation with the local area of ​​the image, thereby extracting low-level features such as edges and lines in the image. For example, a vertical convolution kernel can detect vertical edge information in an image.

[0096] As the network layers deepen, subsequent layers will further extract higher-level and more abstract features based on the low-level features extracted by the previous layers. For example, after multiple layers of convolution and pooling operations, the network can extract feature information of facial organs such as eyes, nose, and mouth. These feature information will be gradually integrated and processed to eventually form the initial permission verification feature. For example, the initial permission verification feature may be a set of feature vectors of key parts of the face, which can accurately represent the biometric information of the face.

[0097] As an implementation manner, in the step S110, before the biometric feature extraction operation is performed on the vital sign collection image of the terminal device by using the pre-calibrated converged authority verification neural network to obtain the initial authority verification feature, the method further includes a network training process:

[0098] Step S101: obtaining an initial permission verification neural network, wherein the initial permission verification neural network is a pre-debugging permission verification neural network calibrated according to vital sign acquisition image samples and first-order priori permission verification features corresponding to the vital sign acquisition image samples, the vital sign acquisition image samples correspond to second-order priori permission verification features and first-order priori permission verification features, the second-order priori permission verification features are image features of first precision corresponding to the vital sign acquisition image samples, and the first-order priori permission verification features are image features of second precision corresponding to the vital sign acquisition image samples.

[0099] In step S101, the vital sign acquisition image samples correspond to second-order prior authority verification features and first-order prior authority verification features. The second-order prior authority verification features are image features of first precision corresponding to the vital sign acquisition image samples, and the first-order prior authority verification features are image features of second precision corresponding to the vital sign acquisition image samples, wherein the first precision is weak precision and the second precision is strong precision, and the specific precision division threshold can be set according to actual conditions.

[0100] Taking facial sign collection as an example, the remote verification system of terminal device access rights collects a large number of facial images of different individuals as samples of sign collection images. These samples cover a variety of facial features, such as different face shapes (round, square, oval), facial features ratios (big eyes, small eyes, high nose bridge, collapsed nose bridge, etc.) and skin color differences. For each facial image sample, its second-order priori permission verification features and first-order priori permission verification features are determined by combining specific implementation methods and manual annotation. The second-order priori permission verification features may be relatively broad and basic facial feature descriptions, such as the approximate contour range of the face in the image, the approximate shape category of the eyes (such as the approximate judgment of single eyelids or double eyelids). The accuracy of such features is relatively low and can only provide some basic image information pointing. The first-order priori permission verification features are more accurate and detailed, such as accurate facial contour coordinate data, specific shape parameters of the eyes (such as the angle of the corner of the eye, the length of the palpebral fissure, etc.), which can more accurately depict the true features of the face.

[0101] When obtaining the initial permission verification neural network, the remote verification system for terminal device access rights uses the pre-debugging permission verification neural network. This pre-debugging neural network has undergone preliminary design and some basic training, and has certain feature extraction capabilities, but it still needs to be further optimized to adapt to specific physical sign collection image samples and verification requirements. The remote verification system for terminal device access rights inputs the physical sign collection image samples and their corresponding first-order prior permission verification features into the pre-debugging permission verification neural network for adjustment. During the adjustment process, by adjusting the weight parameters and bias parameters of the network, the network can better extract feature information that matches the first-order prior permission verification features from the physical sign collection image samples. For example, for facial image samples, the network gradually learns how to accurately identify key features such as the precise shape of the eyes and the specific height of the nose from the image during the adjustment process, and by continuously adjusting the weights and biases, the features output by the network are closer to the first-order prior permission verification features. For example, the remote verification system for terminal device access rights can use a stochastic gradient descent (SGD) algorithm to adjust the weight parameters and bias parameters of the network.

[0102] Step S102: generating a proposed adjustment extraction network according to the network weight parameters and bias parameters of the initial permission verification neural network, wherein the network composition architecture of the proposed adjustment extraction network is the same as that of the pre-debugged permission verification neural network, and the network weight parameters and bias parameters of the proposed adjustment extraction network are the same as those of the initial permission verification neural network;

[0103] Step S103: loading the vital sign acquisition image sample into the extraction network to be adjusted, obtaining a first prediction feature output by the extraction network to be adjusted; and determining a first training error value according to the first prediction feature and the first-order prior authority verification feature;

[0104] Step S104: loading the vital sign acquisition image sample into the initial authority verification neural network to obtain a second prediction feature output by the initial authority verification neural network; and determining a second training error value based on the second prediction feature, the first prediction feature, the second-order prior authority verification feature, and the first-order prior authority verification feature;

[0105] Step S105: determining a comprehensive training error value according to the first training error value and the second training error value;

[0106] Step S106: According to the comprehensive training error value, iterate the network weight parameters and bias parameters of the extraction network to be adjusted until the preset iteration stop condition is met, and obtain the pre-adjusted and converged extraction neural network, and the pre-adjusted and converged extraction neural network is the pre-adjusted and converged authority verification neural network.

[0107] Step S102 requires the remote verification system of the terminal device access rights to generate a proposed adjustment extraction network based on the network weight parameters and bias parameters of the initial permission verification neural network. The network composition architecture of the proposed adjustment extraction network is the same as the permission verification neural network debugged in advance, and its network weight parameters and bias parameters are also the same as the initial permission verification neural network. The purpose of this step is to provide a copy network for subsequent network training and optimization, so as to make various adjustments and experiments without affecting the original initial permission verification neural network.

[0108] Taking the neural network of facial recognition as an example, the initial permission verification neural network has a certain feature extraction capability after being adjusted in step S101. Assume that the neural network has multiple convolutional layers, pooling layers and fully connected layers. The convolutional layer is used to extract local features in the facial image, such as edges and textures in different directions; the pooling layer downsamples the features extracted by the convolutional layer to reduce the amount of data and retain the main features; the fully connected layer integrates and classifies the features extracted by the previous layer. The remote verification system for access rights of terminal devices accurately replicates the structure of this neural network, including the number of neurons in each layer, the connection method, etc. At the same time, the weight parameters (i.e., the weight value of each neuron connection) and the bias parameters (bias value of each neuron) of the initial permission verification neural network are copied intact to the extraction network to be adjusted. In this way, the extraction network to be adjusted has exactly the same feature extraction capability and behavior performance as the initial permission verification neural network in the initial state.

[0109] Step S103 requires the remote verification system of the terminal device access rights to load the vital sign acquisition image samples into the proposed calibration extraction network to obtain the first prediction feature output by the proposed calibration extraction network; and determine the first training error value based on the first prediction feature and the first-order priori authority verification feature. This step is a preliminary test of the proposed calibration extraction network to evaluate the gap between its feature extraction capability for the vital sign acquisition image samples in the current state and the real features.

[0110] Continuing with facial recognition as an example, the remote verification system for terminal device access rights sequentially inputs a large number of facial feature collection image samples into the extraction network to be adjusted. The network processes the image according to its internal structure and parameters, extracts various feature information from the input facial image, and finally outputs the first predicted feature. For example, the first predicted feature may be a set of feature vectors of various parts of the face, which attempt to describe information such as the shape of the face and the proportion of the facial features. Then, the remote verification system for terminal device access rights compares these first predicted features with the corresponding first-order priori permission verification features. The first-order priori permission verification features are precisely labeled information that represents the real features of the face. By calculating the difference between the two, the remote verification system for terminal device access rights can determine the first training error value.

[0111] Step S104 requires the remote verification system of the terminal device access rights to load the vital sign collection image sample into the initial permission verification neural network to obtain the second prediction feature output by the initial permission verification neural network; and determine the second training error value based on the second prediction feature, the first prediction feature, the second-order prior permission verification feature and the first-order prior permission verification feature. This step evaluates the performance and error of the network more comprehensively by extracting features of the same sample on the initial permission verification neural network and combining more levels of prior feature information.

[0112] Still taking facial recognition as an example, the remote verification system for terminal device access rights inputs the same facial feature collection image sample into the initial permission verification neural network. After processing, the network outputs the second prediction feature, which is also a set of information describing facial features. At this time, the remote verification system for terminal device access rights comprehensively considers multiple factors to determine the second training error value. The second-order prior permission verification feature, as a relatively broad and basic facial feature description, is compared and analyzed with the second prediction feature and the first prediction feature together with the first-order prior permission verification feature (precise feature).

[0113] The process of calculating the second training error value is relatively complex, and it needs to consider the difference relationship between multiple features. One possible calculation method is to comprehensively consider the weighted differences between different features. For example, for different parts of facial features (such as eyes, nose, mouth, etc.), different weights are assigned according to their importance for identity recognition. Assume that the weight of the eye feature is , the nose feature weight is , the mouth feature weight is For each sample, calculate the error of the eye part separately ,in is the eye feature in the first-order prior authority verification feature, is the eye feature in the first prediction feature, is the eye feature in the second prediction feature; the error of the nose part ; Error in the mouth Then, the second training error value In this way, the prior features at different levels and the predicted features of the two networks are comprehensively considered, which more comprehensively reflects the error of the initial authority verification neural network in feature extraction.

[0114] Step S105 requires the remote verification system of the terminal device access rights to determine the comprehensive training error value based on the first training error value and the second training error value. This step integrates the error values ​​obtained in the previous two steps to obtain a single error indicator that can fully reflect the current training state and performance of the network, providing a unified basis for subsequent parameter iterations. A variety of methods can be used to determine the comprehensive training error value. One method is to perform a weighted summation of the first training error value and the second training error value.

[0115] Step S106 requires the remote verification system of the terminal device access rights to iteratively adjust the network weight parameters and bias parameters of the extraction network according to the comprehensive training error value until the preset iteration stop condition is met, and the extraction neural network that has been adjusted and converged in advance is obtained. This network is the permission verification neural network that has been adjusted and converged in advance. This step is the core part of network training. By continuously adjusting the parameters of the network, the feature extraction capability of the network is continuously optimized, and finally reaches a stable and accurate state.

[0116] The remote verification system for access rights of terminal devices uses an optimization algorithm to iteratively adjust the parameters of the extraction network. One optimization algorithm is stochastic gradient descent (SGD) and its variants, such as Adagrad, Adadelta, Adam, etc. Taking the stochastic gradient descent algorithm as an example, in each iteration, the remote verification system for access rights of terminal devices calculates the gradient according to the comprehensive training error value.

[0117] During the iteration process, the remote verification system for the access rights of the terminal device repeatedly calculates the comprehensive training error value, gradient, and updates the parameters. After each update, it will check whether the preset iteration stop condition is met. The preset iteration stop condition can be in various forms, such as the comprehensive training error value is less than a specific threshold, such as 0.01. When the comprehensive training error value drops below this threshold, the remote verification system for the access rights of the terminal device believes that the network has reached a sufficiently accurate state and stops iterating. Alternatively, an upper limit on the number of iterations can be set, for example, a maximum of 1,000 iterations. When this upper limit on the number of iterations is reached, the iteration is stopped regardless of whether the comprehensive training error value reaches the threshold.

[0118] Through a series of operations from step S102 to step S106, the remote verification system for terminal device access rights has comprehensively optimized and adjusted the initial permission verification neural network. From generating the extraction network to be adjusted, to evaluating the errors of the two networks respectively, to determining the comprehensive training error value and optimizing the parameters through iteration, finally obtaining the permission verification neural network that has been adjusted and converged in advance.

[0119] As an implementation manner, before obtaining the initial permission verification neural network in step S101, the method further includes:

[0120] Step S10: performing a biometric feature extraction operation on the target vital sign collection image sample to obtain an initial sample authority verification feature; wherein the target vital sign collection image sample is any one of the plurality of vital sign collection image samples;

[0121] Step S20: performing a feature adjustment operation on the initial sample authority verification feature to obtain a debugging transition image feature;

[0122] Step S30: Determine the first-order priori authority verification feature corresponding to the target vital sign acquisition image sample based on the debugging transition image feature; the second-order priori authority verification feature corresponding to the target vital sign acquisition image sample is the initial sample authority verification feature or the debugging transition image feature.

[0123] Step S10 requires the remote verification system of the terminal device access rights to perform a biometric extraction operation on the target vital sign collection image sample to obtain the initial sample access verification feature. The target vital sign collection image sample is a representative sample selected from a large number of vital sign collection images for subsequent analysis and processing. The remote verification system of the terminal device access rights uses a special biometric extraction technology to process these samples.

[0124] Taking facial vital signs collection as an example, the target vital signs collection image sample obtained by the remote verification system of terminal device access rights may be a clear face photo. In order to extract the biometric features, the remote verification system of terminal device access rights can use convolutional neural network (CNN) technology based on deep learning. First, the image is preprocessed, such as normalization, to map the pixel values ​​of the image to a specific range, such as [0, 1], to eliminate differences in brightness, contrast, etc. between different images.

[0125] Next, the normalized image is input into CNN. CNN contains multiple convolutional layers, pooling layers, and fully connected layers. In the convolutional layer, different convolution kernels scan the image to extract local features. For example, some convolution kernels are specifically used to extract edge features of the face. By performing dot product operations with local areas of the image, the contour edges of parts such as eyes, nose, and mouth can be detected. After being processed by multiple convolutional layers, the image is gradually converted into a feature map containing rich semantic information. The pooling layer downsamples the feature map to reduce the amount of data while retaining key features. Finally, the fully connected layer integrates these features and outputs the initial sample authority verification features. These features may be a vector containing information about key parts of the face, such as the shape of the eyes, the position of the nose, etc.

[0126] Step S20 requires the remote verification system of the terminal device access rights to perform feature adjustment operations on the initial sample permission verification features to obtain debugging transition image features. This step aims to optimize and improve the initially extracted features so that they can more accurately reflect the true situation of the target signs.

[0127] Still taking facial signs as an example, the remote verification system for terminal device access rights further processes the initial sample permission verification features containing information about key facial parts. First, the initial sample permission verification features are feature decomposed and split into multiple more targeted initial sample verification sub-features. For example, for eye features, they can be decomposed into initial sample verification sub-features such as eye shape, size, and color; for the nose, they can be decomposed into sub-features such as nose bridge height and nose tip shape.

[0128] Then, based on the target sign collection image samples, feature extraction error adjustment is performed on each initial sample verification sub-feature. The remote verification system for terminal device access rights can use a dedicated error adjustment model, which learns a large amount of facial image data to understand the error patterns that may occur in the extraction process of different facial features. For example, if the eye shape extraction in the initial sample verification sub-feature is inaccurate, the error adjustment model will adjust the initial sample verification sub-feature of the eye shape based on the actual shape information of the eye in the target sign collection image sample and the learned adjustment rules to obtain a more accurate sample verification adjustment sub-feature.

[0129] Based on multiple sample verification adjustment sub-features, the remote verification system of the terminal device access rights determines the sample verification adjustment feature. For example, the adjusted sample verification adjustment sub-features of the eyes, nose, mouth and other parts are integrated to form a comprehensive sample verification adjustment feature. Then, image samples are collected according to the target signs, and the sample verification adjustment feature is improved. The remote verification system of the terminal device access rights can use the generative adversarial network (GAN) technology, which consists of a generator and a discriminator. The generator generates more complete features based on the sample verification adjustment feature and the information of the target sign collected image sample, and the discriminator determines whether the generated features are true and accurate. Through the adversarial training of the two, the features generated by the generator are constantly approaching the real facial features, thereby obtaining the sample permission verification improvement feature. The sample verification adjustment feature and the sample permission verification improvement feature together constitute the debugging transition image feature.

[0130] Step S30 requires the remote verification system of the terminal device access rights to determine the first-order priori permission verification feature corresponding to the target vital sign acquisition image sample based on the debugging transition image feature. At the same time, the second-order priori permission verification feature corresponding to the target vital sign acquisition image sample is any one of the initial sample permission verification feature or the debugging transition image feature.

[0131] Continuing with the example of facial features, the remote verification system for terminal device access rights further analyzes the sample permission verification and improvement features. First, the sample permission verification and improvement features are feature decomposed to obtain multiple sample permission verification and improvement sub-features. For example, the sample permission verification and improvement features of the face are decomposed into more detailed feature sub-elements of the eyes, nose, mouth and other parts.

[0132] Then, according to the target vital signs image samples, feature extraction error adjustment is performed on each sample permission verification and improvement sub-feature. The remote verification system of the terminal device access rights uses the error adjustment model again, combined with the detailed information of the target vital signs image samples, to accurately adjust each sub-feature and obtain the corresponding sample permission verification and improvement adjustment sub-feature.

[0133] Finally, based on multiple sample authority verification and improvement adjustment sub-features, the first-order prior authority verification feature corresponding to the target sign acquisition image sample is determined. For example, the adjusted sample authority verification and improvement adjustment sub-features of the eyes, nose, mouth and other parts are comprehensively considered and integrated to form a first-order prior authority verification feature that can accurately represent the real facial features. The second-order prior authority verification feature can choose the initial sample authority verification feature, which is relatively broad and basic and can provide a rough feature range; it can also choose the debugging transition image feature, which has been optimized and improved to a certain extent and can better reflect the actual situation of facial features.

[0134] Through steps S10-S30, starting from the target vital sign collection image sample, after a series of operations such as biometric feature extraction, feature adjustment and prior permission verification feature determination, accurate and reliable prior knowledge is provided for the subsequent acquisition of the initial permission verification neural network. These prior permission verification features serve as an important basis for network training, and can help the initial permission verification neural network learn more quickly and accurately how to extract effective biometric features from the terminal device vital sign collection image, thereby improving the accuracy and stability of the entire terminal device access right remote verification method, ensuring that only legitimate users can obtain access rights to the terminal device, and effectively protecting the security of the remote operation and maintenance terminals of government and enterprise customers.

[0135] As an implementation mode, the debugging transition image feature includes a sample verification adjustment feature and a sample authority verification improvement feature; based on this, the step S20, performing a feature adjustment operation on the initial sample authority verification feature to obtain the debugging transition image feature, includes:

[0136] Step S21: Decomposing the initial sample authority verification feature to obtain a plurality of initial sample verification sub-features;

[0137] Step S22: collecting image samples according to the target physical sign, performing feature extraction error adjustment on each of the initial sample verification sub-features, and obtaining corresponding sample verification adjustment sub-features;

[0138] Step S23: determining a sample verification adjustment feature according to a plurality of the sample verification adjustment sub-features;

[0139] Step S24: collecting image samples according to the target vital signs, performing feature improvement on the sample verification adjustment features, and obtaining sample authority verification improvement features.

[0140] Step S21 requires the remote verification system of the terminal device access rights to decompose the initial sample access rights verification features to obtain multiple initial sample verification sub-features. The initial sample access rights verification features are extracted from the target vital signs acquisition image samples by the remote verification system of the terminal device access rights in step S10. It is a comprehensive feature set that contains multiple aspects of target vital signs information. In order to process and optimize these features more carefully, the remote verification system of the terminal device access rights needs to decompose them into multiple more targeted initial sample verification sub-features.

[0141] Taking facial signs as an example, assuming that the initial sample permission verification feature is a vector containing overall facial feature information, it may integrate information such as facial contour, facial features, and skin color. The remote verification system for terminal device access rights performs feature decomposition based on different parts of the face and feature dimensions. For facial contours, features such as face shape (such as round, square, oval, etc.) and curvature of facial lines are extracted to form initial sample verification sub-features about facial contours. For example, through specific algorithms and models, the pixel distribution of contour lines in facial images is analyzed, and the curvature change of contours is calculated to determine the specific type of face shape. This process can use edge detection algorithms, such as the Canny edge detection algorithm, to first extract facial edge information, and then determine the initial sample verification sub-features related to face shape by fitting and analyzing edge curves.

[0142] For example, a remote verification system for access rights of terminal devices can use a feature extraction network based on deep learning to perform feature decomposition. For example, a convolutional neural network (CNN) model can be constructed, and the structure of the model is designed to be able to extract features for different feature areas and dimensions. For the initial sample permission verification features corresponding to facial images, through the design of different convolution kernels, some convolution kernels are specifically used to extract facial contour features, while other convolution kernels focus on the extraction of facial features. When these convolution kernels perform convolution operations on the initial sample permission verification features, they will extract the corresponding feature information according to their weight parameters and separate it into different initial sample verification sub-features. For the initial sample permission verification features of fingerprint images, different types of features such as fingerprint patterns, breakpoints, and bifurcation points can also be extracted through carefully designed convolution kernels to achieve feature decomposition.

[0143] Step S22 requires the remote verification system of the terminal device access rights to collect image samples based on the target vital signs, adjust the feature extraction error for each initial sample verification sub-feature, and obtain the corresponding sample verification adjustment sub-feature. Although the biometric feature extraction has been performed in step S10, due to factors such as the image acquisition environment, device accuracy, and limitations of the extraction algorithm, the initial sample verification sub-feature may have a certain extraction error. The remote verification system of the terminal device access rights needs to correct and optimize these sub-features based on the complete information contained in the image samples collected by the target vital signs.

[0144] Still taking facial sign verification as an example, for the initial sample verification sub-feature of eye shape, it is assumed that the eye shape extracted by the previous step is "circular", but in the actual target sign acquisition image sample, it is found that the true shape of the eye is closer to "almond shape", and there is a certain shape extraction error. The remote verification system of the terminal device access rights inputs the target sign acquisition image sample and the initial sample verification sub-feature of the eye shape into a special error adjustment model. This error adjustment model can be a neural network based on deep learning, which learns the accurate features of the eye shape in a large number of different facial images and feasible error patterns during the training process.

[0145] The model analyzes the actual shape information of the eye in the target sign acquisition image sample and compares it with the current initial sample verification sub-feature. By calculating the difference between the two, for example, the Euclidean distance formula can be used to measure the distance between the two shape feature vectors. According to this distance and the learned adjustment rule, the initial sample verification sub-feature of the eye shape is adjusted to be corrected to an "almond shape" that is closer to the actual situation, thereby obtaining the adjusted eye shape sample verification adjustment sub-feature. For example, the error adjustment model can adopt a residual network (ResNet) structure. The residual network can effectively solve the problem of gradient disappearance in deep learning by introducing jump connections, so that the model can learn more complex error adjustment modes. During the training process, a large number of target sign acquisition image samples with accurate labels and the corresponding initial sample verification sub-features are input into the residual network, and the network outputs the adjusted sample verification adjustment sub-feature. By calculating the loss function between the prediction result and the true label, such as the mean square error loss function. Then, the back propagation algorithm is used to update the weight parameters of the network according to the gradient of the loss function, so that the network can learn accurate error adjustment capabilities.

[0146] Step S23 requires the remote verification system of the terminal device access rights to determine the sample verification adjustment feature based on multiple sample verification adjustment sub-features. This step is to integrate the multiple sub-features after error adjustment to form a comprehensive sample verification adjustment feature that can more accurately reflect a certain aspect of the target physical sign. In facial sign verification, the remote verification system of the terminal device access rights fuses the adjusted eyes, nose, mouth, facial contour and other sample verification adjustment sub-features. A feasible fusion method is weighted fusion. The remote verification system of the terminal device access rights assigns corresponding weights to each sub-feature according to its importance to the description of facial features. For example, the facial contour feature plays an important role in the description of the overall facial structure, and may be assigned a weight of 0.3; the eye feature may have a weight of 0.3 due to its uniqueness and key role in facial recognition; the nose and mouth feature weights are 0.15 respectively; the skin color feature weight is 0.1. Through the weighted fusion formula: sample verification adjustment feature = facial contour sample verification adjustment sub-feature × 0.3 + eye sample verification adjustment sub-feature × 0.3 + nose sample verification adjustment sub-feature × 0.15 + mouth sample verification adjustment sub-feature × 0.15 + skin color sample verification adjustment sub-feature × 0.1, the sub-features are integrated to obtain the final facial sample verification adjustment feature. This sample verification adjustment feature combines the adjusted feature information of multiple facial parts and describes the facial features more comprehensively and accurately. In addition to weighted fusion, the remote verification system for terminal device access rights can also use feature splicing and convolution fusion technology in deep learning. First, multiple sample verification adjustment sub-features are spliced ​​in the feature dimension. For example, the sample verification adjustment sub-features of various parts of the face are connected together in a certain order to form a longer feature vector. Then, the spliced ​​feature vector is processed by a convolutional neural network. The convolution layer in the convolutional neural network uses a specific convolution kernel to perform a convolution operation on this long feature vector to extract more advanced and abstract feature information. After multiple layers of convolution operations, the remote verification system for terminal device access rights finally obtains a fused and refined sample verification adjustment feature.

[0147] Step S24 requires the remote verification system of the terminal device access rights to collect image samples based on the target vital signs, perform feature improvement on the sample verification adjustment features, and obtain the sample access rights verification improvement features. Although the sample verification adjustment features have been error adjusted and initially fused, in order to make them more comprehensively and accurately reflect the true features of the target vital signs, the remote verification system of the terminal device access rights needs to collect image samples based on the target vital signs and perform further feature improvement on them.

[0148] The remote verification system of terminal device access rights can use the generative adversarial network (GAN) technology to improve the sample verification adjustment features. The generative adversarial network consists of a generator and a discriminator. The function of the generator is to generate more complete features based on the sample verification adjustment features and the target signs to collect information of the image sample; the discriminator is responsible for judging whether the generated features are true and accurate.

[0149] In facial sign verification, the generator will generate facial features that are more in line with the actual situation based on the facial sample verification adjustment features and other information in the target sign acquisition image samples, such as facial expressions, lighting and other factors. For example, the generator can generate facial features with corresponding expression details based on the basic information of the facial contour and facial features in the sample verification adjustment features, combined with the facial expression information in the target sign acquisition image samples. The discriminator will judge the generated features to see if they are consistent with the real facial features. If the discriminator judges that the generated features are not real, the generator will adjust them according to the feedback of the discriminator and continuously improve the generated features. Through repeated adversarial training of the generator and the discriminator, the generator can eventually generate more accurate and complete facial sample authority verification and complete features.

[0150] In addition, the remote verification system for terminal device access rights can also use reinforcement learning algorithms to improve the sample verification adjustment features. The reinforcement learning algorithm interacts with the environment through the agent, and optimizes its own behavior based on the reward signal fed back by the environment. In the feature improvement process, the agent can be regarded as a module that operates the sample verification adjustment feature, and the environment is the target sign collection image sample. The agent generates a tentative feature based on the sample verification adjustment feature, and then obtains a reward signal based on the feedback provided by the target sign collection image sample (for example, the similarity between the generated feature and the real feature). The agent adjusts its operation strategy based on the reward signal, and continuously generates sample permission verification improvement features that are closer to the real feature.

[0151] As an implementation manner, the step S30, determining the first-order priori authority verification feature corresponding to the target vital sign acquisition image sample according to the debugging transition image feature, includes:

[0152] Step S31: Decomposing the sample authority verification and improvement feature to obtain a plurality of sample authority verification and improvement sub-features;

[0153] Step S32: collecting image samples according to the target vital signs, performing feature extraction error adjustment on each of the sample authority verification and improvement sub-features, and obtaining corresponding sample authority verification and improvement adjustment sub-features;

[0154] Step S33: Determine the first-order prior authority verification feature corresponding to the target vital sign acquisition image sample based on the multiple sample authority verification and improvement adjustment sub-features; the second-order prior authority verification feature corresponding to the target vital sign acquisition image sample is any one of the following features: the initial sample authority verification feature, the sample verification adjustment feature, and the sample authority verification improvement feature.

[0155] Step S31 requires the remote verification system of the terminal device access rights to perform feature decomposition on the sample permission verification improvement feature to obtain multiple sample permission verification improvement sub-features. The sample permission verification improvement feature is a comprehensive feature set obtained after feature adjustment and improvement in the previous steps, which more comprehensively reflects the relevant information of the target vital signs. However, in order to analyze and utilize these features more deeply, the remote verification system of the terminal device access rights decomposes them into multiple more targeted and detailed sub-features.

[0156] For example, the remote verification system of terminal device access rights can use deep convolutional neural network (DCNN) for feature decomposition. DCNN has powerful feature extraction capabilities. By designing different convolution kernels and network hierarchical structures, it can accurately extract features for different feature areas and dimensions. For example, for the sample permission verification and improvement features corresponding to facial images, the remote verification system of terminal device access rights can build a DCNN with a multi-branch structure. Different branches are responsible for extracting features of different parts of the face, and each branch contains multiple convolution layers and pooling layers. By carefully designing the parameters and sizes of the convolution kernels, some convolution kernels are specifically used to extract detailed features of facial contours, while other convolution kernels focus on extracting microscopic features of facial features. When these convolution kernels perform convolution operations on the sample permission verification and improvement features, they will extract the corresponding feature information according to their weight parameters and separate it into different sample permission verification and improvement sub-features. For the sample permission verification and improvement features of fingerprint images, it is also possible to use a customized DCNN structure to use different convolution kernels to extract different types of detailed features such as fingerprint lines, breakpoints, and bifurcation points, so as to achieve accurate feature decomposition.

[0157] Step S32 requires the remote verification system of the terminal device access rights to collect image samples based on the target vital signs, perform feature extraction error adjustment on each sample permission verification and improvement sub-feature, and obtain the corresponding sample permission verification and improvement adjustment sub-feature. Although the sample permission verification and improvement sub-feature has undergone the previous feature adjustment and improvement steps, due to various uncertain factors in the image acquisition process (such as illumination changes, equipment accuracy limitations) and the inherent limitations of the feature extraction algorithm, there may still be certain extraction errors. The remote verification system of the terminal device access rights needs to perform fine error adjustment on these sub-features based on the complete information contained in the image samples collected by the target vital signs.

[0158] Continuing with the facial sign verification as an example, for the sample permission verification and improvement sub-feature of the eye shape, it is assumed that the eye shape parameters extracted through the previous steps deviate from the actual situation in the target sign acquisition image sample in some details. For example, there is a certain error between the extracted Danfeng eye angle parameter and the actual angle. The remote verification system for the access rights of the terminal device inputs the target sign acquisition image sample and the sample permission verification and improvement sub-feature of the eye shape into a special error adjustment model. This error adjustment model can be a residual network (ResNet) based on deep learning. By introducing jump connections, ResNet can effectively learn complex error adjustment patterns and solve the gradient disappearance problem in deep learning.

[0159] The model will first perform a deep analysis of the eye area in the target vital sign acquisition image sample to extract the true eye shape feature. Then, it will be compared with the current sample authority verification and improvement sub-feature. By calculating the difference between the two, for example, the mean square error (MSE) formula can be used to measure the degree of difference, and the sample authority verification and improvement sub-feature of the eye shape is adjusted according to the calculated difference and the adjustment rules learned by the model during training. For example, if it is found that the angle of the Danfeng eye in the current sub-feature is larger than the true angle, the model will reduce the angle parameter accordingly according to the adjustment rule, thereby obtaining the adjusted eye shape sample authority verification and improvement adjustment sub-feature.

[0160] Step S33 requires the remote verification system of the terminal device access rights to verify and adjust the sub-features based on multiple sample permissions, and determine the first-order priori permission verification feature corresponding to the target vital sign acquisition image sample. This step is to organically integrate the multiple sub-features after error adjustment to form a first-order priori permission verification feature that can most accurately and comprehensively reflect the true situation of the target vital sign.

[0161] In facial sign verification, the remote verification system for terminal device access rights will fuse multiple sample permission verification and adjustment sub-features such as eyes, nose, mouth, and facial contours after adjustment. A feasible and effective fusion method is weighted fusion. The remote verification system for terminal device access rights assigns corresponding weights to each sub-feature according to its importance to the overall facial feature description. For example, facial contour features play a key role in determining the overall structure and identity of the face, and may be assigned a weight of 0.3; eye features may have a weight of 0.3 due to their uniqueness and high recognition in facial recognition; although the recognition of nose and mouth features is slightly lower than that of eyes and facial contours, they also make important contributions to the overall facial features, with weights of 0.15 respectively; other detail features (such as skin color, facial texture, etc.) have a weight of 0.1. Through the weighted fusion formula: first-order prior authority verification feature = facial contour sample authority verification and improvement adjustment sub-feature × 0.3 + eye sample authority verification and improvement adjustment sub-feature × 0.3 + nose sample authority verification and improvement adjustment sub-feature × 0.15 + mouth sample authority verification and improvement adjustment sub-feature × 0.15 + other detail sample authority verification and improvement adjustment sub-feature × 0.1, each sub-feature is integrated to obtain the final facial first-order prior authority verification feature. This first-order prior authority verification feature combines the feature information of multiple facial parts after fine adjustment, and can highly accurately represent the true feature situation of the face.

[0162] In addition, the remote verification system for terminal device access rights can also use feature splicing and convolution fusion technology in deep learning to determine the first-order priori permission verification features. First, multiple sample permission verification and improvement adjustment sub-features are spliced ​​in the feature dimension. For example, the sample permission verification and improvement adjustment sub-features of various parts of the face are connected together in a certain logical order to form a longer and richer feature vector. Then, the spliced ​​feature vector is processed by a convolutional neural network. The convolution layer in the convolutional neural network uses carefully designed convolution kernels to perform convolution operations on this long feature vector. These convolution kernels can capture the associations and interactions between different sub-features and extract higher-level and more abstract feature information. After multiple layers of convolution operations, the remote verification system for terminal device access rights finally obtains a first-order priori permission verification feature that has been deeply fused and refined.

[0163] At the same time, the second-order priori permission verification features corresponding to the target vital sign acquisition image samples can be selected from the initial sample permission verification features, sample verification adjustment features, and sample permission verification improvement features. The initial sample permission verification features are relatively broad and basic, and can provide a rough feature range. As a second-order priori permission verification feature, it can provide a basic reference framework for subsequent network training; the sample verification adjustment feature has undergone a certain error adjustment and is relatively more accurate. Selecting it as a second-order priori permission verification feature can balance the accuracy and generalization of the feature to a certain extent; the sample permission verification improvement feature is a feature that has been adjusted and improved many times, and is the most accurate, but may be slightly weaker in generalization. The remote verification system for terminal device access rights can flexibly select appropriate features as second-order priori permission verification features based on specific verification requirements and data characteristics.

[0164] As an implementation manner, the initial permission verification feature includes multiple verification feature elements, and the multiple verification feature elements include the first verification feature element; based on this, the step S110, performing a biometric feature extraction operation on the vital sign collection image of the terminal device through a pre-adjusted converged permission verification neural network, to obtain the initial permission verification feature, includes:

[0165] Step S111: vectorizing the vital sign acquisition image of the terminal device to obtain an image matrix representation;

[0166] Step S112: Mapping the image matrix representation to a high-dimensional space to obtain an image tensor representation;

[0167] Step S113: performing a feature weight focusing operation on the image tensor representation to obtain a focused image tensor representation;

[0168] Step S114: performing position extraction on the focused image tensor representation to obtain a position extraction result;

[0169] Step S115: obtaining a position code corresponding to the position extraction result in a preset feature library, wherein the position code corresponding to the position extraction result in the preset feature library is a first position code;

[0170] Step S116: restore and map the first position code to obtain the first verification feature element of the initial authority verification feature.

[0171] Step S111 requires the remote verification system of the terminal device access rights to vectorize the terminal device vital sign collection image to obtain an image matrix representation. The terminal device vital sign collection image is raw data containing vital sign information, and the vectorization operation is intended to convert it into a matrix form that is easy for the remote verification system of the terminal device access rights to process and analyze. Taking the facial vital sign collection image as an example, it is assumed that a color image with a resolution of m×n is collected. A color image is usually composed of three color channels (such as RGB channels), and each pixel has a value ranging from 0 to 255 on each channel, representing the intensity of the pixel on the color channel. The remote verification system of the terminal device access rights arranges the RGB values ​​of each pixel of this image in a certain order to form a three-dimensional matrix. For example, the R value, G value, and B value of the first row and first column of the image are used as the first element group of the matrix, and the RGB value of the second column of the pixel is used as the second element group, and so on, until the RGB value of the last row and last column of the pixel, and finally an image matrix of size m×n×3 is obtained. This matrix completely stores the color and spatial information of the image and provides the basic data structure for subsequent processing.

[0172] Step S112 requires the remote verification system of the terminal device access rights to map the image matrix representation to a high-dimensional space to obtain an image tensor representation. The purpose of mapping the image matrix to a high-dimensional space is to enable the remote verification system of the terminal device access rights to capture more complex and abstract feature information in the image. In deep learning, a tensor is a general data structure used to represent high-dimensional data. Taking the m×n×3 matrix of a facial image as an example, the remote verification system of the terminal device access rights uses some layer structures in the neural network, such as a fully connected layer or a convolution layer, to perform linear and nonlinear transformations on the image matrix. For example, through the convolution operation of the convolution layer, the image matrix is ​​convolved with multiple convolution kernels. The convolution kernel is a small matrix that slides on the image matrix and performs a dot product operation with the local area of ​​the image matrix to generate a new feature matrix. Assuming that the convolution kernel size is k×k, it slides on the image matrix with a step size of s. For each sliding position, the dot product between the convolution kernel and the local area of ​​the image matrix is ​​calculated to obtain a new value, and these new values ​​form a new matrix. After convolution operations with multiple convolution kernels, multiple new matrices will be obtained. Combining these new matrices in a certain way will form a higher-dimensional tensor. For example, after a series of convolution operations, an image tensor of size m'×n'×c may be obtained from the original m×n×3 image matrix, where m' and n' are the spatial dimensions after the convolution operation, and c is the number of convolution kernels, that is, the number of newly generated feature channels.

[0173] Step S113 requires the remote verification system of the terminal device access rights to perform a feature weight focusing operation (such as performing attention) on the image tensor representation to obtain a focused image tensor representation. The purpose of the feature weight focusing operation is to enable the remote verification system of the terminal device access rights to pay more attention to the important parts of the image tensor related to the biometric features and ignore some unimportant information, thereby improving the accuracy and efficiency of feature extraction. Taking the facial image tensor as an example, the remote verification system of the terminal device access rights can adopt an attention mechanism. The attention mechanism determines which parts need to be focused on by calculating the importance weight of each element in the image tensor. For example, an attention network is used to analyze the image tensor and calculate the importance score of each element for representing the facial biometric features. The importance score can be obtained by calculating the similarity between the feature vectors at different positions in the image tensor and a learnable query vector. Assume that the image tensor is T and the query vector is q. For each element in the image tensor , calculate the similarity score (Here represents the transpose of the query vector q). Then, these scores are normalized by the softmax function to obtain the weight of each element , where N is the total number of elements in the image tensor. These weights represent the importance of each element in feature extraction.

[0174] Next, the remote verification system for the terminal device's access rights multiplies each element in the image tensor by its corresponding weight to focus on important features. For example, the focused image tensor element After such an operation, the part of the image tensor related to the key biometric features of the face is enhanced, while some unimportant background information or noise information is weakened, thus obtaining a focused image tensor representation.

[0175] Step S114 requires the remote verification system of the terminal device access rights to perform position extraction on the focused image tensor representation to obtain a position extraction result. The position extraction operation is intended to determine key position information related to the biometric feature from the focused image tensor. Taking the focused image tensor of a facial image as an example, the remote verification system of the terminal device access rights can use some algorithms or models based on spatial positions to extract key positions. For example, through a target detection algorithm, such as the Faster R-CNN algorithm based on a convolutional neural network.

[0176] Step S115 requires the remote verification system of the terminal device access rights to obtain the position code corresponding to the position extraction result in the preset feature library, wherein the position code corresponding to the position extraction result in the preset feature library is the first position code. The preset feature library is a database containing various biometric position information pre-built by the remote verification system of the terminal device access rights, and the position code is a digital representation of these position information. Taking facial features as an example, the preset feature library may contain standard position information of key facial features (such as eyes, nose, and mouth) of different facial expressions and different individuals. The remote verification system of the terminal device access rights matches the position information of the key facial features extracted in step S114 with the preset feature library. For example, for the extracted eye position information, the remote verification system of the terminal device access rights searches for the eye position template that best matches it in the preset feature library. Assuming that the eye position template in the preset feature library is represented by a coordinate range and a relative proportional relationship, the remote verification system of the terminal device access rights calculates the similarity between the extracted eye position and these templates. The similarity can be calculated using a metric such as Euclidean distance or cosine similarity.

[0177] After finding the best matching template, the remote verification system for the access rights of the terminal device obtains the position code corresponding to the template as the first position code. This position code is a digitally processed feature representation, which can be a vector or a specific numerical sequence, used to uniquely identify the position information. In the processing of fingerprint features, the preset feature library contains the standard position information of the key features (such as breakpoints and bifurcation points) of different types of fingerprints (such as dustpan patterns and bucket patterns). The remote verification system for the access rights of the terminal device matches the extracted fingerprint key feature position with the preset feature library, calculates the similarity, finds the best matching template and obtains its corresponding position code as the first position code. The implementation method for realizing this operation mainly relies on database query and matching algorithms, and obtains the corresponding position code by quickly and accurately matching the extracted position information with the templates in the preset feature library.

[0178] Step S116 requires the remote verification system of the terminal device access rights to restore the first position code to obtain the first verification feature element of the initial permission verification feature. The restoration mapping operation is to convert the position code back to information with actual biometric significance as the first verification feature element of the initial permission verification feature. Taking facial features as an example, it is assumed that the first position code is a digitized vector that represents the position information of the eyes in the facial image. The remote verification system of the terminal device access rights uses a pre-trained restoration mapping model that learns the mapping relationship between the position code and the actual facial features. This model can be a neural network that takes the first position code as input and outputs a facial feature information with actual significance, such as the specific shape and size of the eyes, as the first verification feature element of the initial permission verification feature through a series of neuron calculations and nonlinear transformations. For example, the model may determine whether the eyes are single eyelids or double eyelids, as well as information such as the approximate size range of the eyes based on the position code.

[0179] As an implementation method, the initial permission verification feature includes multiple verification feature elements, and the multiple verification feature elements also include verification feature elements other than the first one, and the verification feature elements other than the first one are any verification feature elements other than the first verification feature element. Based on this, the step S110, through the pre-calibrated converged permission verification neural network, performs a biometric feature extraction operation on the vital sign collection image of the terminal device to obtain the initial permission verification feature, including:

[0180] Step S117: Map the position code corresponding to the previous verification feature element to a high-dimensional space to obtain a position code tensor representation;

[0181] Step S118: performing a feature weight focusing operation on the position coding tensor representation to obtain a focused position coding tensor representation;

[0182] Step S119: obtaining the position code corresponding to the preset feature library of the focus position code tensor representation, wherein the position code corresponding to the preset feature library of the focus position code tensor representation is a second position code;

[0183] Step S1110: restore and map the second position code to obtain verification feature elements other than the first one of the initial authority verification features.

[0184] In step S117, the remote verification system of the terminal device access rights obtains the position code corresponding to the previous verification feature element. Position coding is a coding method for digitally representing position information, which records the position information of the feature in a specific space. In practical applications, such as when processing the vital signs collected by the terminal device, different vital signs parts in the image have different spatial positions, and the position code is used to identify these positions. For example, assuming that the vital signs collected by the terminal device is an image containing a face, different parts such as the eyes, nose, and mouth have their own positions in the image, and the position code is a digital code used to accurately describe the positions of these parts.

[0185] Next, the remote verification system of the terminal device access rights maps the location code to a high-dimensional space. The mapping to a high-dimensional space is to be able to express and process the location information more richly in a higher-dimensional vector space. This mapping can usually be achieved using mathematical methods such as linear transformation. For example, assuming that the location code is a one-dimensional vector [x], the remote verification system of the terminal device access rights can map it to a high-dimensional space through a matrix multiplication, such as using an m×1 matrix W, and by calculating Y=W×[x] (where Y is the mapped high-dimensional vector), the position code tensor representation is obtained. Here, m represents the dimension of the high-dimensional space, and m is usually greater than 1, for example, m=10. After mapping, the original one-dimensional location code becomes a ten-dimensional vector, which can carry more information in the high-dimensional space.

[0186] In step S118, the remote verification system for the access rights of the terminal device performs a feature weight focusing operation on the obtained position coding tensor representation, which is similar to executing an attention mechanism. The core idea of ​​the attention mechanism is to enable the remote verification system for the access rights of the terminal device to automatically focus on the important parts of the data. In this step, the remote verification system for the access rights of the terminal device performs weighted processing on each dimension in the position coding tensor representation according to different feature weights. For example, assuming that the position coding tensor representation is a ten-dimensional vector [y1, y2, y3, y4, y5, y6, y7,y8, y9, y10], the remote verification system for the access rights of the terminal device assigns a weight value to each dimension, such as [w1,w2, w3, w4, w5, w6, w7, w8, w9, w10], and these weight values ​​represent the importance of each dimension in the current task. Through weighted calculation, that is, the new vector element zi=wi×yi (i=1 to 10), a new vector [z1,z2, z3, z4, z5, z6, z7, z8, z9, z10] is obtained. This new vector is the focused position encoding tensor representation after the feature weight focusing operation. The purpose of this is to highlight the more important information in the position encoding tensor representation, so that subsequent processing can focus more on key features.

[0187] In step S119, the remote verification system for the access rights of the terminal device matches the focus position coding tensor representation obtained through the feature weight focusing operation with the preset feature library. The preset feature library is a pre-established database containing various features and their corresponding position codes. For example, in the processing scenario of the terminal device vital sign acquisition image, the preset feature library may contain standard position coding information of different facial features, fingerprint features, etc. at different positions. The remote verification system for the access rights of the terminal device searches the preset feature library for the position code that best matches the focus position coding tensor representation, and the found position code is the second position code. The specific search process can use a similarity calculation method, such as cosine similarity. Assume that the focus position coding tensor is represented by vector A, and each position coding vector in the preset feature library is Bi (i=1 to n, n is the number of position coding vectors in the preset feature library). By calculating the cosine similarity Sim(A, Bi)=(A·Bi) / (||A||×||Bi||) (where A·Bi is the vector dot product, ||A|| and ||Bi|| are the moduli of vectors A and Bi respectively), find the Bi with the highest similarity, which is the second position coding. For example, when processing face images, the focus position coding tensor may correspond to the position information of a local feature in the image. By searching in the preset feature library, the standard position coding that best matches it is found. This standard position coding is the second position coding, which can more accurately reflect the position of the local feature in the overall feature system.

[0188] In step S1110, the remote verification system for the access rights of the terminal device performs a restoration mapping on the second position code found. The restoration mapping is the opposite operation of the previous mapping to the high-dimensional space, and the purpose is to restore the position code processed in the high-dimensional space to a suitable dimension to obtain the verification feature element. For example, the one-dimensional position code was previously mapped to the high-dimensional space by matrix multiplication W×[x], so the restoration mapping may be performed by calculating (in is the inverse matrix of matrix W, Y' is the second position encoding vector, and [x'] is the restored vector). After restoration mapping, the result is the verification feature element other than the first one of the initial permission verification feature. This verification feature element further enriches the content of the initial permission verification feature and provides more feature information for more accurate subsequent permission verification. For example, when processing facial images for permission verification, the verification feature element obtained by restoration mapping the second position encoding may be a detailed feature description of a specific area of ​​the face, such as the texture features of the cheek, etc. These features together with other verification feature elements constitute a more comprehensive and accurate initial permission verification feature, which helps to improve the accuracy and reliability of permission verification.

[0189] Figure 2 A hardware entity diagram of a remote verification system for access rights of a terminal device provided by an embodiment of the present invention, such as Figure 2 As shown, the hardware entity of the remote verification system 1000 for access rights of the terminal device includes: a processor 1001 and a memory 1002, wherein the memory 1002 stores a computer program that can be run on the processor 1001, and the processor 1001 implements the steps in the method of any of the above embodiments when executing the program.

[0190] The memory 1002 stores a computer program that can be run on the processor. The memory 1002 is configured to store instructions and applications executable by the processor 1001. It can also cache the processor 1001 and the data to be processed or processed by each module in the remote verification system 1000 of the terminal device access rights (for example, image data, audio data, voice communication data and video communication data), which can be implemented through flash memory (FLASH) or random access memory (Random Access Memory, RAM).

[0191] When the processor 1001 executes the program, the steps of any of the above-mentioned methods for remote verification of terminal device access rights are implemented. The processor 1001 generally controls the overall operation of the system 1000 for remote verification of terminal device access rights.

[0192] The above description is only an implementation mode of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A remote verification method for access rights of a terminal device, characterized in that: The method comprises: Perform biometric feature extraction on the terminal device’s vital sign collection image to obtain initial permission verification features; Decomposing the initial authority verification feature to obtain a plurality of initial verification sub-features; According to the terminal device vital sign acquisition image, adjusting the feature extraction error for each of the initial verification sub-features to obtain a corresponding verification adjustment sub-feature; Determining a verification adjustment feature based on a plurality of the verification adjustment sub-features; According to the terminal device's vital sign collection image, the verification adjustment feature is improved to obtain the permission verification improvement feature; Determine the target permission verification feature based on the permission verification improvement feature; Identification is performed based on the target permission verification feature to obtain a verification result of the terminal device access permission.

2. The method according to claim 1, characterized in that The step of determining a target permission verification feature based on the permission verification improvement feature includes: Decomposing the permission verification improvement feature to obtain a plurality of permission verification improvement sub-features; According to the terminal device vital sign acquisition image, feature extraction error adjustment is performed on each of the permission verification improvement sub-features to obtain a corresponding permission verification improvement adjustment sub-feature; The target permission verification feature is determined based on a plurality of the permission verification improvement adjustment sub-features.

3. The method according to claim 1, characterized in that The step of performing feature extraction error adjustment on each of the initial verification sub-features based on the terminal device vital sign acquisition image to obtain a corresponding verification adjustment sub-feature includes: The terminal device vital sign acquisition image and the proposed processing initial verification sub-feature are embedded into a feature extraction error adjustment indication framework to obtain a proposed adjustment image feature, wherein the proposed adjustment image feature is used to instruct the image processing network to perform feature extraction error adjustment on the proposed processing initial verification sub-feature, and the proposed processing initial verification sub-feature is any one of the multiple initial verification sub-features; Loading the image feature to be adjusted into the image processing network, obtaining an image feature analysis result of the image feature to be adjusted output by the image processing network, wherein the image feature analysis result includes a verification adjustment sub-feature corresponding to the initial verification sub-feature to be processed and an extraction error category, wherein the extraction error category identifies a type corresponding to the extraction feature error; The step of collecting the image based on the terminal device's vital signs and improving the verification and adjustment features to obtain the permission verification and improvement features includes: Embedding the terminal device vital sign acquisition image and the verification adjustment feature into a feature improvement indication framework to obtain a proposed improvement image feature, wherein the proposed improvement image feature is used to instruct an image processing network to perform feature improvement on the verification adjustment feature; The to-be-perfected image features are loaded into the image processing network, and a to-be-perfected image feature analysis result output by the image processing network for the to-be-perfected image features is obtained, wherein the to-be-perfected image feature analysis result includes the permission verification perfection feature.

4. The method according to claim 1, characterized in that The biometric feature extraction operation is performed on the terminal device vital sign collection image to obtain the initial permission verification feature, including: The biometric feature extraction operation is performed on the vital sign collection image of the terminal device by using the permission verification neural network that has been adjusted and converged in advance to obtain the initial permission verification feature.

5. The method according to claim 4, characterized in that Before performing a biometric feature extraction operation on the vital sign collection image of the terminal device through the pre-adjusted and converged authority verification neural network to obtain an initial authority verification feature, the method further includes: Acquire an initial permission verification neural network, wherein the initial permission verification neural network is obtained by adjusting the pre-debugging permission verification neural network according to the vital sign acquisition image sample and the first-order priori permission verification feature corresponding to the vital sign acquisition image sample, the vital sign acquisition image sample corresponds to a second-order priori permission verification feature and a first-order priori permission verification feature, the second-order priori permission verification feature is an image feature of a first precision corresponding to the vital sign acquisition image sample, and the first-order priori permission verification feature is an image feature of a second precision corresponding to the vital sign acquisition image sample; Generate a proposed adjustment extraction network according to the network weight parameters and bias parameters of the initial permission verification neural network, wherein the network composition architecture of the proposed adjustment extraction network is the same as that of the pre-debugging permission verification neural network, and the network weight parameters and bias parameters of the proposed adjustment extraction network are the same as those of the initial permission verification neural network; Loading the vital sign acquisition image sample into the proposed calibration extraction network to obtain a first prediction feature output by the proposed calibration extraction network; and determining a first training error value based on the first prediction feature and the first-order prior authority verification feature; Loading the vital sign acquisition image sample into the initial authority verification neural network to obtain a second prediction feature output by the initial authority verification neural network; and determining a second training error value based on the second prediction feature, the first prediction feature, the second-order prior authority verification feature, and the first-order prior authority verification feature; Determining a comprehensive training error value according to the first training error value and the second training error value; According to the comprehensive training error value, the network weight parameters and bias parameters of the extraction network to be adjusted are iterated until the preset iteration stop condition is met, so as to obtain the extraction neural network that has been previously adjusted and converged. The extraction neural network that has been previously adjusted and converged is the authority verification neural network that has been previously adjusted and converged.

6. The method according to claim 5, characterized in that Before obtaining the initial permission verification neural network, the method further includes: Performing a biometric feature extraction operation on the target vital sign collection image sample to obtain an initial sample authority verification feature; wherein the target vital sign collection image sample is any one of the plurality of vital sign collection image samples; Performing a feature adjustment operation on the initial sample authority verification feature to obtain a debugging transition image feature; Based on the debugging transition image feature, the first-order priori authority verification feature corresponding to the target vital sign acquisition image sample is determined; the second-order priori authority verification feature corresponding to the target vital sign acquisition image sample is the initial sample authority verification feature or the debugging transition image feature.

7. The method according to claim 6, characterized in that The debugging transition image features include sample verification adjustment features and sample authority verification improvement features; The performing a feature adjustment operation on the initial sample authority verification feature to obtain a debugging transition image feature includes: Decomposing the initial sample authority verification feature to obtain a plurality of initial sample verification sub-features; Collecting image samples according to the target physical sign, performing feature extraction error adjustment on each of the initial sample verification sub-features, and obtaining corresponding sample verification adjustment sub-features; Determining a sample validation regulation feature based on a plurality of the sample validation regulation sub-features; Image samples are collected according to the target vital signs, and feature improvement is performed on the sample verification and adjustment features to obtain sample authority verification and improvement features.

8. The method according to claim 7, characterized in that Determining the first-order priori authority verification feature corresponding to the target vital sign acquisition image sample according to the debugging transition image feature includes: Decomposing the sample authority verification and improvement feature to obtain a plurality of sample authority verification and improvement sub-features; Collect image samples according to the target vital signs, perform feature extraction error adjustment on each of the sample authority verification and improvement sub-features, and obtain corresponding sample authority verification and improvement adjustment sub-features; Based on multiple sample authority verification and improvement adjustment sub-features, the first-order prior authority verification feature corresponding to the target vital sign acquisition image sample is determined; the second-order prior authority verification feature corresponding to the target vital sign acquisition image sample is any one of the following features: the initial sample authority verification feature, the sample verification adjustment feature, and the sample authority verification improvement feature.

9. The method according to claim 4, characterized in that The initial permission verification feature includes a plurality of verification feature elements, wherein the plurality of verification feature elements include a first verification feature element; the biometric feature extraction operation is performed on the vital sign collection image of the terminal device by the permission verification neural network that has been adjusted and converged in advance to obtain the initial permission verification feature, including: Vectorizing the vital sign acquisition image of the terminal device to obtain an image matrix representation; Mapping the image matrix representation to a high-dimensional space to obtain an image tensor representation; Performing a feature weight focusing operation on the image tensor representation to obtain a focused image tensor representation; Performing position extraction on the focused image tensor representation to obtain a position extraction result; Obtaining a position code corresponding to the position extraction result in a preset feature library, wherein the position code corresponding to the position extraction result in the preset feature library is a first position code; Performing restoration mapping on the first position code to obtain a first verification feature element of the initial authority verification feature; The multiple verification feature elements also include verification feature elements other than the first one, and the verification feature elements other than the first one are any verification feature elements other than the first verification feature element; the biometric feature extraction operation is performed on the vital sign collection image of the terminal device by the pre-adjusted and converged authority verification neural network to obtain the initial authority verification feature, including: Map the position code corresponding to the previous verification feature element to a high-dimensional space to obtain a position code tensor representation; Performing a feature weight focusing operation on the position coding tensor representation to obtain a focused position coding tensor representation; Obtaining a position code corresponding to a preset feature library in the focus position code tensor representation, wherein the position code corresponding to the preset feature library in the focus position code tensor representation is a second position code; The second position code is restored and mapped to obtain verification feature elements other than the first one of the initial authority verification features.

10. A remote verification system for access rights of a terminal device, comprising a memory and a processor, wherein the memory stores a computer program that can be run on the processor, characterized in that: When the processor executes the program, the steps in the method according to any one of claims 1 to 9 are implemented.

Images