Face recognition method and device based on artificial intelligence, computer equipment and medium

Abstract

The invention belongs to the field of artificial intelligence and the field of financial science and technology, and relates to an artificial intelligence-based face recognition method, which comprises the steps of preprocessing initial face image data to obtain face image training data; performing feature extraction on the face image training data based on a feature extraction model to obtain a target feature vector; training a preset number of classification models based on the target feature vector to obtain a plurality of face recognition models; screening all the face recognition models based on the face image test data to obtain a target face recognition model; and recognizing the face image based on the target face recognition model and the feature extraction model to generate a target face recognition result. The invention further provides a human face recognition device based on artificial intelligence, computer equipment and a storage medium. In addition, the target face recognition result can be stored in the block chain. According to the invention, based on the use of the feature extraction model and the target face recognition model, the model training efficiency and recognition precision of face recognition are improved.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence development technology and financial technology, and in particular to artificial intelligence-based face recognition methods, devices, computer equipment and storage media. Background Art

[0002] With the development of artificial intelligence and computer vision technology, face recognition technology has been widely used in various fields as a fast and convenient new authentication method. Financial business systems usually use face recognition to achieve identity authentication when processing financial industry-related business. Financial business systems can be insurance systems, banking systems, transaction systems, and order systems.

[0003] At present, many financial business systems use PCANet, a deep learning network based on principal component analysis, to perform face recognition processing. PCANet is a simplified deep learning model that combines the ideas of principal component analysis and convolutional neural networks, providing an image classification method with a simple and fast training process that can adapt to different tasks. However, PCANet cannot process image samples in real time during the training of the convolution kernel, and the entire training set needs to be processed offline, resulting in low efficiency of PCANet network training. In addition, the recognition accuracy of face recognition processing using PCANet is generally not high. Summary of the invention

[0004] The purpose of the embodiments of the present application is to propose a face recognition method, device, computer equipment and storage medium based on artificial intelligence to solve the existing way of performing face recognition processing based on principal component analysis deep learning network PCANet. PCANet cannot process image samples in real time during the training of convolution kernels, and the entire training set needs to be processed offline, resulting in low efficiency of PCANet network training. In addition, the recognition accuracy of face recognition processing using PCANet is generally not high.

[0005] In order to solve the above technical problems, the embodiment of the present application provides a face recognition method based on artificial intelligence, which adopts the following technical solutions:

[0006] Acquire pre-constructed initial face image data; wherein the initial face image data is a data set that has been labeled with identity information;

[0007] Preprocessing the initial face image data to obtain corresponding face image training data;

[0008] Based on a preset feature extraction model, feature extraction processing is performed on the face image training data to obtain a corresponding target feature vector; wherein the feature extraction model is a deep subspace network model based on incremental row-column two-dimensional principal component analysis;

[0009] Based on the target feature vector, a preset number of classification models are trained to obtain a plurality of trained face recognition models;

[0010] Screening all the face recognition models based on preset face image test data to obtain a target face recognition model that meets preset evaluation conditions;

[0011] The face image to be processed is recognized and processed based on the target face recognition model and the feature extraction model to generate a target face recognition result corresponding to the face image to be processed.

[0012] Furthermore, the step of preprocessing the initial face image data to obtain corresponding face image training data specifically includes:

[0013] Normalizing the initial facial image data to obtain corresponding first facial image data;

[0014] Performing grayscale processing on the first facial image data to obtain corresponding second facial image data;

[0015] Performing histogram equalization processing on the second facial image data to obtain corresponding third facial image data;

[0016] The third facial image data is used as the facial image training data.

[0017] Furthermore, the step of performing feature extraction processing on the face image training data based on a preset feature extraction model to obtain a corresponding target feature vector specifically includes:

[0018] Dividing the face image training data into corresponding image blocks; wherein the number of the image blocks includes a plurality;

[0019] Performing principal component analysis on each row in the image block based on the feature extraction model to obtain a corresponding first feature matrix;

[0020] Performing principal component analysis on each column of the first feature matrix to obtain a corresponding second feature matrix;

[0021] Performing a convolution operation on the face image training data based on the second feature matrix to obtain a corresponding feature map;

[0022] Performing a preset operation on the feature map to obtain a corresponding first feature vector;

[0023] The target feature vector is generated based on the first feature vector.

[0024] Furthermore, the step of performing principal component analysis on each column of the first feature matrix to obtain a corresponding second feature matrix specifically includes:

[0025] Calculating a covariance matrix of the first feature matrix;

[0026] Performing eigendecomposition on the covariance matrix to obtain eigenvectors and eigenvalues ​​corresponding to the covariance matrix;

[0027] Constructing a corresponding column-wise projection matrix based on the eigenvectors and the eigenvalues;

[0028] Projecting each column of the first feature matrix based on the projection matrix in the column direction to obtain a specified feature matrix after dimension reduction in the column direction;

[0029] The designated characteristic matrix is ​​used as the second characteristic matrix.

[0030] Furthermore, the step of screening all the face recognition models based on the preset face image test data to obtain a target face recognition model that meets the preset evaluation conditions specifically includes:

[0031] Using the face image test data to perform capability evaluation processing on each of the face recognition models, to obtain capability index values ​​of each of the face recognition models corresponding to preset capability indexes;

[0032] Calculating the ability evaluation score of each face recognition model based on the ability indicator value of each face recognition model;

[0033] Selecting a first face recognition model with the highest ability evaluation score from all the face recognition models;

[0034] Determine whether the number of the first face recognition models is equal to 1;

[0035] If the number of the first face recognition models is equal to 1, taking the first face recognition model as the target face recognition model;

[0036] If the number of the first face recognition models is greater than 1, obtaining the accuracy value of each of the first face recognition models from the capability indicator value, and screening out the second face recognition model with the highest accuracy value from the first face recognition models;

[0037] The second face recognition model is used as the target face recognition model.

[0038] Furthermore, the capability index at least includes accuracy, processing time and recall rate; the step of calculating the capability evaluation score of each face recognition model based on the capability index value of each face recognition model specifically includes:

[0039] Obtaining a specified accuracy value, a specified processing time value, and a specified recall value from the capability indicator value of a third face recognition model; wherein the third face recognition model is any one of all the face recognition models;

[0040] Get the preset weight construction algorithm;

[0041] Generate a first weight corresponding to the accuracy rate, a second weight corresponding to the processing time, and a third weight corresponding to the recall rate based on the weight construction algorithm;

[0042] Call the preset score calculation formula;

[0043] Based on the score calculation formula, the specified accuracy value, the specified processing time value, the specified recall value, the first weight, the second weight, and the third weight are calculated to obtain a corresponding score result;

[0044] The score result is used as the capability evaluation score of the third face recognition model.

[0045] Furthermore, the step of performing recognition processing on the face image to be processed based on the target face recognition model and the feature extraction model to generate a target face recognition result corresponding to the face image to be processed includes:

[0046] Preprocessing the face image to be processed to obtain a corresponding designated face image;

[0047] Performing feature extraction processing on the designated face image based on the feature extraction model to obtain a corresponding second feature vector;

[0048] Performing recognition processing on the second feature vector based on the target face recognition model to obtain a corresponding designated face recognition result;

[0049] The designated face recognition result is used as the target face recognition result.

[0050] In order to solve the above technical problems, the embodiment of the present application also provides a face recognition device based on artificial intelligence, which adopts the following technical solution:

[0051] An acquisition module, used to acquire pre-constructed initial face image data; wherein the initial face image data is a data set that has been labeled with identity information;

[0052] A preprocessing module, used for preprocessing the initial face image data to obtain corresponding face image training data;

[0053] An extraction module, used to perform feature extraction processing on the face image training data based on a preset feature extraction model to obtain a corresponding target feature vector; wherein the feature extraction model is a deep subspace network model based on incremental row-column two-dimensional principal component analysis;

[0054] A training module, used to train a preset number of classification models based on the target feature vector to obtain a plurality of trained face recognition models;

[0055] A screening module, used to screen all the face recognition models based on preset face image test data to obtain a target face recognition model that meets preset evaluation conditions;

[0056] The recognition module is used to perform recognition processing on the face image to be processed based on the target face recognition model and the feature extraction model, and generate a target face recognition result corresponding to the face image to be processed.

[0057] In order to solve the above technical problems, the embodiment of the present application further provides a computer device, which adopts the following technical solution:

[0058] The computer device includes a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the steps of any one of the artificial intelligence-based face recognition methods proposed in the embodiments of the present application are implemented.

[0059] In order to solve the above technical problems, the embodiment of the present application further provides a computer-readable storage medium, which adopts the following technical solution:

[0060] The computer-readable storage medium stores a computer program, and when the computer program is executed by the processor, the steps of any one of the artificial intelligence-based face recognition methods proposed in the embodiments of the present application are implemented.

[0061] Compared with the prior art, the embodiments of the present application have the following beneficial effects:

[0062] The present application first obtains pre-constructed initial face image data; then pre-processes the initial face image data to obtain corresponding face image training data; then performs feature extraction processing on the face image training data based on a preset feature extraction model to obtain a corresponding target feature vector; wherein the feature extraction model is a deep subspace network model based on incremental row-column two-dimensional principal component analysis; subsequently, a preset number of classification models are trained based on the target feature vector to obtain a plurality of trained face recognition models; further, all the face recognition models are screened based on preset face image test data to obtain a target face recognition model that meets preset evaluation conditions; finally, the face image to be processed is recognized based on the target face recognition model and the feature extraction model to generate a target face recognition result corresponding to the face image to be processed. After preprocessing the pre-constructed initial face image data to obtain face image training data, the present application will use a deep subspace network model based on incremental row-column two-dimensional principal component analysis to perform feature extraction processing on the face image training data to obtain a target feature vector, and then train a preset number of classification models based on the target feature vector, and use the face image test data to screen all face recognition models to obtain a target face recognition model that meets the preset evaluation conditions, and finally use the target face recognition model and the feature extraction model to perform face recognition processing on the face image to be processed to generate the corresponding target face recognition result. The present application uses a deep subspace network model based on incremental row-column two-dimensional principal component analysis to process training samples in real time while training the convolution kernel. This real-time processing capability effectively improves the efficiency of face recognition model training. In addition, by screening out a target face recognition model that meets the preset evaluation conditions from a plurality of trained face recognition models to perform recognition processing on the face image to be processed, the recognition accuracy of face recognition can be effectively improved, and the data accuracy of the generated target face recognition result corresponding to the face image to be processed is guaranteed. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] In order to more clearly illustrate the scheme in the present application, a brief introduction is given below to the drawings required for use in the description of the embodiments of the present application. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0064] Figure 1 is an exemplary system architecture diagram to which the present application may be applied;

[0065] Figure 2 A flowchart of an embodiment of a face recognition method based on artificial intelligence according to the present application;

[0066] Figure 3 is a structural schematic diagram of an embodiment of an artificial intelligence-based face recognition device according to the present application;

[0067] Figure 4 It is a structural diagram of an embodiment of a computer device according to the present application. DETAILED DESCRIPTION

[0068] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by technicians in the technical field of the present application; the terms used in the specification of the application herein are only for the purpose of describing specific embodiments and are not intended to limit the present application; the terms "including" and "having" and any variations thereof in the specification and claims of the present application and the above-mentioned drawings are intended to cover non-exclusive inclusions. The terms "first", "second", etc. in the specification and claims of the present application or the above-mentioned drawings are used to distinguish different objects, not to describe a specific order.

[0069] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0070] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the accompanying drawings.

[0071] like Figure 1 As shown, the system architecture 100 may include terminal devices 101, 102, 103, a network 104 and a server 105. The network 104 is used to provide a medium for communication links between the terminal devices 101, 102, 103 and the server 105. The network 104 may include various connection types, such as wired, wireless communication links or optical fiber cables, etc.

[0072] Users can use terminal devices 101, 102, 103 to interact with server 105 through network 104 to receive or send messages, etc. Various communication client applications can be installed on terminal devices 101, 102, 103, such as web browser applications, shopping applications, search applications, instant messaging tools, email clients, social platform software, etc.

[0073] The terminal devices 101, 102, and 103 can be various electronic devices with display screens and supporting web browsing, including but not limited to smart phones, tablet computers, e-book readers, MP3 players (Moving Picture Experts Group Audio Layer III), MP4 (Moving Picture Experts Group Audio Layer IV) players, laptop computers, desktop computers, and the like.

[0074] The server 105 may be a server that provides various services, such as a background server that provides support for web pages displayed on the terminal devices 101 , 102 , and 103 .

[0075] It should be noted that the artificial intelligence-based face recognition method provided in the embodiments of the present application is generally executed by a server / terminal device, and accordingly, the artificial intelligence-based face recognition device is generally set in the server / terminal device.

[0076] It should be understood that Figure 1 The number of terminal devices, networks and servers in the embodiment is only for illustration. Any number of terminal devices, networks and servers may be provided according to implementation requirements.

[0077] Continue to refer Figure 2 , showing a flow chart of an embodiment of an artificial intelligence-based face recognition method according to the present application. According to different requirements, the order of the steps in the flow chart can be changed, and some steps can be omitted. The artificial intelligence-based face recognition method provided in the embodiment of the present application can be applied to any scenario where face recognition is required, and the artificial intelligence-based face recognition method can be applied to products in these scenarios, for example, face recognition in the process of identity verification of customers in the financial insurance field. The artificial intelligence-based face recognition method comprises the following steps:

[0078] Step S201, obtaining pre-constructed initial face image data; wherein the initial face image data is a data set that has been labeled with identity information.

[0079] In this embodiment, the face recognition method based on artificial intelligence is executed on the electronic device (for example Figure 1The server / terminal device shown in the figure) can obtain the initial facial image data through a wired connection or a wireless connection. It should be noted that the above-mentioned wireless connection method may include but is not limited to 3G / 4G / 5G connection, Wi-Fi connection, Bluetooth connection, Wi MAX connection, Zigbee connection, UWB (ultrawi deband) connection, and other wireless connection methods currently known or to be developed in the future. Among them, the above-mentioned initial facial image data is a data set containing facial images of known identities, and each facial image contained in the data set is associated with its corresponding identity tag. The above-mentioned labeling processing of identity information can be carried out by manual labeling or automatic machine labeling.

[0080] Step S202, preprocessing the initial face image data to obtain corresponding face image training data.

[0081] In this embodiment, the specific implementation process of preprocessing the initial facial image data to obtain the corresponding facial image training data will be further described in detail in subsequent specific embodiments of the present application and will not be elaborated on here.

[0082] Step S203, performing feature extraction processing on the face image training data based on a preset feature extraction model to obtain a corresponding target feature vector; wherein the feature extraction model is a deep subspace network model based on incremental row-column two-dimensional principal component analysis.

[0083] In the present embodiment, the above-mentioned feature extraction model based on the preset feature extraction model performs feature extraction processing on the face image training data to obtain the corresponding target feature vector. The present application will further describe this in detail in the subsequent specific embodiments, and will not be elaborated on here. Among them, the above-mentioned feature extraction model is a deep subspace network model based on incremental row-column two-dimensional principal component analysis (I RC2DPCA, i ncremental sequential row-column 2DPCA), which can also be called I RC2DPCA Net. The above-mentioned feature extraction model is based on the PCANet framework, and the method for obtaining the convolution kernel parameters is replaced by the I RC2DPCA algorithm. The I RC2DPCA method can process the training samples in real time during the filter training process, thereby improving the efficiency of network training and improving the efficiency of face recognition.

[0084] The overall structure of IRC2DPCANet and PCANet (Principal Component Analysis Network) is the same. Both of them first undergo two layers of convolution operations, and then perform binary hashing and histogram statistics to extract image feature vectors. The difference is that in IRC2DPCANet, the two layers of convolution kernels are processed by the IRC2DPCA method before the convolution operation. The PCA process requires the original image matrix to be rearranged into vectors, which will cause two problems. First, it will destroy the original spatial structure of the data. Second, the high-dimensional vectors obtained in the vectorization process will cause dimensionality disasters in the eigenvalue and eigenvector calculation process. In addition, PCA is a batch processing method, and feature vectors can only be extracted after all image samples are collected. In contrast, the IRC2DPCA algorithm directly treats the scanned image sub-blocks as matrices, avoiding the process of arranging them into vectors and retaining the spatial structure of the data. The incremental processing method makes it possible to continuously update the current principal component matrix by adding new samples, thereby avoiding the calculation of the covariance matrix of all samples.

[0085] Since traditional PCANet cannot process image samples in real time during the training of convolution kernels, the entire training set needs to be processed offline. I RC2DPCANet uses an incremental row-column two-dimensional principal component analysis method (I RC2DPCA), which can process training samples in real time while the convolution kernel is being trained. This real-time processing capability greatly improves the efficiency of network training, and is particularly suitable for large-scale data sets or scenarios that require fast training. In addition, since I RC2DPCANet can process training samples in real time, it can dynamically add new training samples during the training of convolution kernels without retraining previously trained samples. This incremental training method greatly reduces the time and computational cost of network training, and also improves the scalability of the network. In addition, I RC2DPCANet avoids the complexity of offline processing of the entire training set in traditional PCANet by introducing an incremental row-column two-dimensional principal component analysis method. This can reduce the complexity and number of parameters of the network, simplify the structure of the network model, and improve the efficiency and interpretability of the network.

[0086] Step S204: training a preset number of classification models based on the target feature vector to obtain a plurality of trained face recognition models.

[0087] In this embodiment, the above classification models may specifically include logistic regression, decision tree, random forest, support vector machine, K nearest neighbor algorithm, and naive Bayes. Correspondingly, the above preset number is 6. Each classification model may be trained respectively by using the target feature vector, so that each classification model can learn the mapping relationship between the input target feature vector and the target output (identity label), thereby obtaining a plurality of trained face recognition models.

[0088] Step S205, screening all the face recognition models based on preset face image test data to obtain a target face recognition model that meets preset evaluation conditions.

[0089] In this embodiment, the above-mentioned specific implementation process of screening and processing all the face recognition models based on the preset face image test data to obtain the target face recognition model that meets the preset evaluation conditions will be further described in detail in the subsequent specific embodiments of this application and will not be elaborated here.

[0090] Step S206, performing recognition processing on the face image to be processed based on the target face recognition model and the feature extraction model, and generating a target face recognition result corresponding to the face image to be processed.

[0091] In the present embodiment, the above-mentioned specific implementation process of performing recognition processing on the face image to be processed based on the target face recognition model and the feature extraction model to generate a target face recognition result corresponding to the face image to be processed will be further described in detail in the subsequent specific embodiments of the present application and will not be elaborated here.

[0092] After preprocessing the pre-constructed initial face image data to obtain face image training data, the present application will use a deep subspace network model based on incremental row-column two-dimensional principal component analysis to perform feature extraction processing on the face image training data to obtain a target feature vector, and then train a preset number of classification models based on the target feature vector, and use the face image test data to screen all face recognition models to obtain a target face recognition model that meets the preset evaluation conditions, and finally use the target face recognition model and the feature extraction model to perform face recognition processing on the face image to be processed to generate the corresponding target face recognition result. The present application uses a deep subspace network model based on incremental row-column two-dimensional principal component analysis to process training samples in real time while training the convolution kernel. This real-time processing capability effectively improves the efficiency of face recognition model training. In addition, by screening out a target face recognition model that meets the preset evaluation conditions from a plurality of trained face recognition models to perform recognition processing on the face image to be processed, the recognition accuracy of face recognition can be effectively improved, and the data accuracy of the generated target face recognition result corresponding to the face image to be processed is guaranteed.

[0093] In some optional implementations, step S202 includes the following steps:

[0094] The initial face image data is normalized to obtain corresponding first face image data.

[0095] In this embodiment, the initial facial image data is normalized to ensure that the size of the initial facial image data is uniform, and the first facial image data is obtained.

[0096] The first facial image data is gray-scaled to obtain corresponding second facial image data.

[0097] In this embodiment, the grayscale processing refers to converting the first facial image data into a grayscale image, that is, obtaining the second facial image data.

[0098] The second facial image data is subjected to histogram equalization processing to obtain corresponding third facial image data.

[0099] In this embodiment, the histogram equalization process refers to a process of enhancing the contrast of an image.

[0100] The third facial image data is used as the facial image training data.

[0101] The present application performs normalization, grayscale processing and histogram equalization processing on the initial face image data to accurately complete the preprocessing of the initial face image data. By preprocessing the initial face image data, key information in the initial face image data can be effectively extracted, and the performance impact of illumination and noise on the subsequent use of the model for face recognition processing can be reduced.

[0102] In some optional implementations of this embodiment, step S203 includes the following steps:

[0103] The face image training data is divided into corresponding image blocks; wherein the number of the image blocks includes multiple.

[0104] In this embodiment, the face image training data is processed into blocks to obtain a corresponding plurality of image blocks for subsequent principal component analysis.

[0105] Based on the feature extraction model, principal component analysis is performed on each row in the image block to obtain a corresponding first feature matrix.

[0106] In this embodiment, each row of each image block is subjected to principal component analysis by using a feature extraction model and the corresponding specified covariance matrix is ​​calculated; the specified covariance matrix is ​​then subjected to eigendecomposition to obtain a specified eigenvector and a corresponding specified eigenvalue; then the eigenvectors corresponding to the specified number of the largest specified eigenvalues ​​are selected to form a projection matrix in the row direction; the projection matrix in the row direction is subsequently used to project each row of the image block to obtain a feature matrix after dimension reduction in the row direction; finally, the feature matrix after dimension reduction in the row direction is used as the first feature matrix mentioned above. There is no specific limitation on the selection of the above specified number, and it can be set according to actual business usage requirements.

[0107] Perform principal component analysis on each column of the first feature matrix to obtain a corresponding second feature matrix.

[0108] In this embodiment, the above-mentioned principal component analysis is performed on each column of the first feature matrix to obtain the specific implementation process of the corresponding second feature matrix. This application will further describe this in detail in subsequent specific embodiments, which will not be elaborated on here. In addition, the feature extraction model also includes an incremental learning process. During the training process, with the arrival of new samples, the projection matrix in the row and column directions is gradually updated using an incremental PCA method, such as Rank-One Update or a more complex online PCA algorithm.

[0109] A convolution operation is performed on the facial image training data based on the second feature matrix to obtain a corresponding feature map.

[0110] In this embodiment, the convolution operation specifically includes: filter initialization: initializing the learned filter as a convolution kernel. Among them, the second feature matrix is ​​obtained by rearranging the second feature matrix into a shape suitable for the convolution operation (such as 3x3, 5x5, etc.) to meet the input requirements of the convolution operation. Boundary processing: In order to maintain the size of the feature map, the input image needs to be border-filled (such as zero-filled). Convolution operation: Slide each filter (convolution kernel) on the input image (i.e., face image training data), calculate the dot product between the filter and the local area of ​​the image, and obtain the value of the corresponding position on the feature map. Activation function: In order to increase nonlinearity, an activation function (such as ReLU, Sigmo id, etc.) can be applied after the convolution operation to perform a nonlinear transformation on the feature map. After the above convolution operation, each filter will generate a corresponding feature map. These feature maps together constitute the representation of the input image in the feature space learned by the deep subspace network model based on incremental row and column two-dimensional principal component analysis. Each feature map emphasizes a specific feature or pattern in the input image related to the corresponding filter.

[0111] A preset operation process is performed on the feature map to obtain a corresponding first feature vector.

[0112] In this embodiment, the above-mentioned preset operation processing may specifically include binary hash and block histogram processing.

[0113] The target feature vector is generated based on the first feature vector.

[0114] In this embodiment, the first feature vector obtained directly may be used as the target feature vector.

[0115] The present application completes the feature extraction processing of the face image training data by dividing the face image training data into corresponding image blocks, and then performs incremental row and column two-dimensional principal component analysis processing, convolution operation and preset operation processing on the image blocks based on the feature extraction model, so as to quickly and accurately extract the target feature vector from the face image training data. The present application uses the feature extraction model to perform secondary feature extraction of the rows and columns of the image blocks, which not only considers the variation of the face image in the column direction, but also considers the variation in the row direction. This dual feature extraction helps to retain more details and structural information in the face image, and can capture the information of the face image more comprehensively, which is conducive to improving the performance and accuracy of subsequent face recognition tasks of the face image.

[0116] In some optional implementations, performing principal component analysis on each column of the first feature matrix to obtain a corresponding second feature matrix includes the following steps:

[0117] The covariance matrix of the first feature matrix is ​​calculated.

[0118] In this embodiment, the covariance matrix of the first feature matrix may be calculated by using a general covariance matrix calculation formula.

[0119] Perform eigendecomposition on the covariance matrix to obtain eigenvectors and eigenvalues ​​corresponding to the covariance matrix.

[0120] In this embodiment, by performing eigendecomposition on the covariance matrix, an eigenvector and an eigenvalue corresponding to the eigenvector can be obtained.

[0121] A corresponding column-wise projection matrix is ​​constructed based on the eigenvectors and the eigenvalues.

[0122] In this embodiment, the projection matrix in the column direction can be constructed by selecting the eigenvector corresponding to the largest eigenvalue of the target number. There is no specific limitation on the selection of the above target number, and it can be set according to actual business usage requirements.

[0123] Projection processing is performed on each column of the first feature matrix based on the projection matrix in the column direction to obtain a designated feature matrix after dimension reduction in the column direction.

[0124] In this embodiment, by projecting the first feature matrix onto the selected column feature vector, that is, the projection matrix in the column direction, a designated feature matrix after dimension reduction in the column direction can be obtained. The designated feature matrix is ​​a representation of the original data in the new space, which contains the main information of the original data, but the dimension has been reduced.

[0125] The designated characteristic matrix is ​​used as the second characteristic matrix.

[0126] The present application calculates the covariance matrix of the first feature matrix; then performs eigendecomposition on the covariance matrix to obtain eigenvectors and eigenvalues ​​corresponding to the covariance matrix; then constructs a corresponding column-wise projection matrix based on the eigenvectors and the eigenvalues; subsequently, projects each column of the first feature matrix based on the column-wise projection matrix to obtain a designated feature matrix after dimensionality reduction in the column direction, and uses the designated feature matrix as the second feature matrix, so as to quickly and accurately complete principal component analysis of each column of the first feature matrix, thereby ensuring the data accuracy of the obtained second feature matrix.

[0127] In some optional implementations, step S205 includes the following steps:

[0128] The facial image test data is used to perform capability evaluation processing on each of the facial recognition models to obtain capability indicator values ​​of each of the facial recognition models corresponding to preset capability indicators.

[0129] In this embodiment, the facial image test data is input into each of the facial recognition models respectively, so that each of the facial recognition models performs recognition processing on the facial image test data respectively, and then the ability of each facial recognition model is evaluated according to the recognition results output by each facial recognition model, and the ability index values ​​of each facial recognition model are counted. The above-mentioned ability index values ​​may include specific values ​​corresponding to the ability indicators such as accuracy, processing time, and recall rate calculated for each facial recognition model to evaluate the model ability. Among them, the processing time refers to the total time taken by the facial recognition model to complete all recognition processing of the facial image test data.

[0130] The ability evaluation score of each face recognition model is calculated based on the ability indicator value of each face recognition model.

[0131] In this embodiment, the specific implementation process of calculating the ability evaluation score of each face recognition model based on the ability indicator value of each face recognition model will be described in further details in subsequent specific embodiments of the present application and will not be elaborated here.

[0132] A first face recognition model with the highest ability evaluation score is selected from all the face recognition models.

[0133] In this embodiment, the capability evaluation scores of the various face recognition models may be numerically compared, and then the first face recognition model with the highest capability evaluation score may be screened out from all the face recognition models according to the obtained numerical comparison results.

[0134] Determine whether the number of the first face recognition models is equal to 1.

[0135] In this embodiment, the number of the first face recognition models may be 1 or greater than 1.

[0136] If the number of the first face recognition models is equal to 1, the first face recognition model is used as the target face recognition model.

[0137] In this embodiment, if the number of first face recognition models is equal to 1, the first face recognition model with the highest capability evaluation score and the optimal processing capability is directly used as the above-mentioned target face recognition model for ultimately identifying the face image to be processed, thereby effectively ensuring the accuracy of the determination of the target face recognition model.

[0138] If the number of the first face recognition models is greater than 1, the accuracy value of each of the first face recognition models is obtained from the capability indicator value, and the second face recognition model with the highest accuracy value is screened out from the first face recognition models.

[0139] In this embodiment, a numerical comparison can be performed on the accuracy values ​​of each first face recognition model obtained from the capability indicator value, and then the second face recognition model with the highest accuracy value can be screened out from all the first face recognition models based on the obtained numerical comparison results.

[0140] The second face recognition model is used as the target face recognition model.

[0141] In this embodiment, when there are multiple first face recognition models with optimal processing capabilities, the second face recognition model with the optimal processing capabilities and the highest accuracy value will be intelligently selected as the target face recognition model, which can further ensure the accuracy of the target face recognition model.

[0142] This application uses facial image test data to perform capability evaluation processing on all the facial recognition models to screen out the first facial recognition model with the highest capability evaluation score from all the facial recognition models. When it is detected that there is only one first facial recognition model, the first facial recognition model with the highest capability evaluation score and the best processing capability will be directly used as the above-mentioned target facial recognition model for the final recognition of the facial image to be processed, effectively ensuring the accuracy of the target facial recognition model. In addition, when it is detected that there are multiple first facial recognition models with the best processing capability, the second facial recognition model with the best processing capability and the highest accuracy value will be intelligently used as the target facial recognition model, thereby further ensuring the accuracy of the obtained target facial recognition model, which is conducive to the subsequent use of the target facial recognition model to recognize the facial image to be processed, and can effectively improve the data accuracy of the generated target facial recognition result corresponding to the facial image to be processed.

[0143] In some optional implementations of this embodiment, the capability indicators include at least accuracy, processing time and recall rate; and the step of calculating the capability evaluation score of each face recognition model based on the capability indicator values ​​of each face recognition model includes the following steps:

[0144] Gets a specified accuracy value, a specified processing time value, and a specified recall value from the capability indicator value of the third face recognition model.

[0145] In this embodiment, the third face recognition model is any one of all the face recognition models.

[0146] Get the preset weight construction algorithm.

[0147] In this embodiment, an algorithm can be randomly selected from a plurality of initial weight construction algorithms in advance as the above-mentioned weight construction algorithm. Among them, the initial weight construction algorithm may include at least: principal component analysis, entropy method, factor analysis, and hierarchical analysis method. In addition, the algorithm with the highest comprehensive effect of processing efficiency and processing evaluation can be selected from all the initial weight construction algorithms as the above-mentioned weight construction algorithm, so that when the weight construction algorithm is subsequently used to generate weights of various indicators, the generation efficiency and accuracy of weight generation can be effectively improved. The above-mentioned processing efficiency and processing evaluation can be obtained by querying the open source test data results of various initial weight construction algorithms.

[0148] Based on the weight construction algorithm, a first weight corresponding to the accuracy, a second weight corresponding to the processing time, and a third weight corresponding to the recall rate are generated.

[0149] In this embodiment, the accuracy, processing time and recall rate can be weighted according to the operation steps of constructing indicator weights corresponding to the weight construction algorithm to obtain the corresponding first weight, second weight and third weight.

[0150] Call the preset score calculation formula.

[0151] In this embodiment, the above score calculation formula specifically includes: G=a*X+b / Y+c*Z, wherein G is the capability assessment score, X is the accuracy value, a is the first weight, Y is the processing time value, b is the second weight, Z is the recall rate value, and c is the third weight.

[0152] Based on the score calculation formula, the specified accuracy value, the specified processing time value, the specified recall value, the first weight, the second weight and the third weight are calculated and processed to obtain a corresponding score result.

[0153] In this embodiment, the corresponding score result can be obtained by substituting the specified accuracy value, the specified processing time value, the specified recall value, the first weight, the second weight and the third weight into the corresponding positions in the above score calculation formula in sequence.

[0154] The score result is used as the capability evaluation score of the third face recognition model.

[0155] The present application uses a weight construction algorithm to generate a first weight corresponding to the accuracy rate, a second weight corresponding to the processing time, and a third weight corresponding to the recall rate, and then calls the score calculation formula to perform calculation processing on the specified accuracy rate value, the specified processing time value, and the specified recall rate value of the obtained third face recognition model based on the corresponding weights. This can quickly and accurately calculate the capability evaluation scores of each of the face recognition models, which is conducive to the subsequent analysis and processing of the obtained capability evaluation scores to accurately screen out the target face recognition models that meet the evaluation conditions.

[0156] In some optional implementations of this embodiment, step S206 includes the following steps:

[0157] The face image to be processed is preprocessed to obtain a corresponding designated face image.

[0158] In this embodiment, the implementation process of preprocessing the face image to be processed can refer to the aforementioned step of preprocessing the initial face image data, and will not be described in detail here.

[0159] Perform feature extraction processing on the designated face image based on the feature extraction model to obtain a corresponding second feature vector.

[0160] In this embodiment, the implementation process of performing feature extraction processing on the specified facial image based on the feature extraction model can refer to the aforementioned steps of performing feature extraction processing on the facial image training data based on the feature extraction model, and will not be elaborated on here.

[0161] The second feature vector is processed based on the target face recognition model to obtain a corresponding designated face recognition result.

[0162] In this embodiment, by using a target face recognition model that meets preset evaluation conditions selected from multiple trained face recognition models to perform recognition processing on the second feature vector, the processing efficiency and accuracy of the recognition processing can be effectively improved, thereby ensuring the accuracy of the generated specified face recognition result.

[0163] The designated face recognition result is used as the target face recognition result.

[0164] The present application obtains a corresponding designated face image by preprocessing the face image to be processed; and performs feature extraction processing on the designated face image based on the feature extraction model to obtain a corresponding second feature vector; and then uses a target face recognition model that meets preset evaluation conditions and is screened out from multiple trained face recognition models to perform recognition processing on the second feature vector, which can effectively improve the processing efficiency and accuracy of the recognition processing, thereby improving the accuracy of the generated designated face recognition result.

[0165] It should be understood that the order of execution of the steps in the above embodiment does not necessarily mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiment of the present invention.

[0166] It should be emphasized that in order to further ensure the privacy and security of the above-mentioned target face recognition results, the above-mentioned target face recognition results can also be stored in a blockchain node.

[0167] The blockchain referred to in this application is a new application model of computer technologies such as distributed data storage, peer-to-peer transmission, consensus mechanism, encryption algorithm, etc. Blockchain is essentially a decentralized database, a string of data blocks generated by cryptographic methods. Each data block contains a batch of network transaction information, which is used to verify the validity of its information (anti-counterfeiting) and generate the next block. Blockchain can include the underlying blockchain platform, platform product service layer, and application service layer.

[0168] The embodiments of the present application can acquire and process relevant data based on artificial intelligence technology. Artificial Intelligence (AI) is the theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results.

[0169] AI basic technologies generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing technology, operation / interaction systems, mechatronics, etc. AI software technologies mainly include computer vision technology, robotics technology, biometrics technology, speech processing technology, natural language processing technology, and machine learning / deep learning.

[0170] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through computer-readable instructions, and the computer-readable instructions can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, the aforementioned storage medium can be a non-volatile storage medium such as a disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM).

[0171] It should be understood that, although the steps in the flowchart of the accompanying drawings are displayed in sequence as indicated by the arrows, these steps are not necessarily executed in sequence in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least a part of the steps in the flowchart of the accompanying drawings may include multiple sub-steps or multiple stages, and these sub-steps or stages are not necessarily executed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be executed in turn or alternately with other steps or at least a part of the sub-steps or stages of other steps.

[0172] Further references Figure 3 , as a response to the above Figure 2 The present application provides an embodiment of a face recognition device based on artificial intelligence, and the device embodiment is similar to Figure 2 Corresponding to the method embodiment shown, the device can be specifically applied to various electronic devices.

[0173] like Figure 3 As shown, the artificial intelligence-based face recognition device 300 described in this embodiment includes: an acquisition module 301, a preprocessing module 302, an extraction module 303, a training module 304, a screening module 305 and a recognition module 306. Among them:

[0174] The acquisition module 301 is used to acquire pre-constructed initial face image data; wherein the initial face image data is a data set that has been labeled with identity information;

[0175] A preprocessing module 302 is used to preprocess the initial face image data to obtain corresponding face image training data;

[0176] An extraction module 303 is used to perform feature extraction processing on the face image training data based on a preset feature extraction model to obtain a corresponding target feature vector; wherein the feature extraction model is a deep subspace network model based on incremental row-column two-dimensional principal component analysis;

[0177] A training module 304 is used to train a preset number of classification models based on the target feature vector to obtain a plurality of trained face recognition models;

[0178] A screening module 305 is used to screen all the face recognition models based on preset face image test data to obtain a target face recognition model that meets preset evaluation conditions;

[0179] The recognition module 306 is used to perform recognition processing on the face image to be processed based on the target face recognition model and the feature extraction model, and generate a target face recognition result corresponding to the face image to be processed.

[0180] In some optional implementations of this embodiment, the preprocessing module 302 includes:

[0181] A first processing submodule, configured to perform normalization processing on the initial face image data to obtain corresponding first face image data;

[0182] A second processing submodule, configured to perform grayscale processing on the first facial image data to obtain corresponding second facial image data;

[0183] a third processing submodule, configured to perform histogram equalization processing on the second facial image data to obtain corresponding third facial image data;

[0184] The first determining submodule is configured to use the third facial image data as the facial image training data.

[0185] In some optional implementations of this embodiment, the extraction module 303 includes:

[0186] A division submodule, used for dividing the face image training data into corresponding image blocks; wherein the number of the image blocks includes a plurality;

[0187] A first analysis submodule, configured to perform principal component analysis on each row in the image block based on the feature extraction model to obtain a corresponding first feature matrix;

[0188] A second analysis submodule is used to perform principal component analysis on each column of the first feature matrix to obtain a corresponding second feature matrix;

[0189] A convolution submodule, configured to perform a convolution operation on the face image training data based on the second feature matrix to obtain a corresponding feature map;

[0190] A first extraction submodule, used for performing a preset operation on the feature map to obtain a corresponding first feature vector;

[0191] A generating submodule is used to generate the target feature vector based on the first feature vector.

[0192] In some optional implementations of this embodiment, the second analysis submodule includes:

[0193] A first calculation unit, used for calculating the covariance matrix of the first feature matrix;

[0194] A decomposition unit, used for performing eigendecomposition on the covariance matrix to obtain eigenvectors and eigenvalues ​​corresponding to the covariance matrix;

[0195] A construction unit, configured to construct a corresponding column-wise projection matrix based on the eigenvector and the eigenvalue;

[0196] A projection unit, configured to perform a projection process on each column of the first feature matrix based on the projection matrix in the column direction to obtain a specified feature matrix after dimension reduction in the column direction;

[0197] The first determining unit is configured to use the designated characteristic matrix as the second characteristic matrix.

[0198] In some optional implementations of this embodiment, the screening module 305 includes:

[0199] An evaluation submodule, used to perform capability evaluation processing on each of the face recognition models using the face image test data, and obtain capability index values ​​of each of the face recognition models corresponding to preset capability indexes;

[0200] A calculation submodule, used to calculate the ability evaluation score of each face recognition model based on the ability indicator value of each face recognition model;

[0201] A first screening submodule, used to screen out a first face recognition model with the highest ability evaluation score from all the face recognition models;

[0202] A judging submodule, used to judge whether the number of the first face recognition models is equal to 1;

[0203] a second determining submodule, configured to use the first face recognition model as the target face recognition model if the number of the first face recognition models is equal to 1;

[0204] a second screening submodule, configured to obtain the accuracy value of each of the first face recognition models from the capability indicator value if the number of the first face recognition models is greater than 1, and screen out the second face recognition model with the highest accuracy value from the first face recognition models;

[0205] The third determination submodule is used to use the second face recognition model as the target face recognition model.

[0206] In some optional implementations of this embodiment, the capability index includes at least accuracy, processing time and recall rate; the calculation submodule includes:

[0207] A first acquisition unit is used to acquire a specified accuracy value, a specified processing time value, and a specified recall value from the capability indicator value of a third face recognition model; wherein the third face recognition model is any one of all the face recognition models;

[0208] A second acquisition unit, used to acquire a preset weight construction algorithm;

[0209] A generating unit, configured to generate, based on the weight building algorithm, a first weight corresponding to the accuracy rate, a second weight corresponding to the processing time, and a third weight corresponding to the recall rate;

[0210] A calling unit, used to call a preset score calculation formula;

[0211] A second calculation unit is used to calculate the specified accuracy value, the specified processing time value, the specified recall value, the first weight, the second weight and the third weight based on the score calculation formula to obtain a corresponding score result;

[0212] The second determining unit is used to use the score result as the ability evaluation score of the third face recognition model.

[0213] In some optional implementations of this embodiment, the identification module 306 includes:

[0214] A preprocessing submodule, used for preprocessing the face image to be processed to obtain a corresponding designated face image;

[0215] A second extraction submodule, configured to perform feature extraction processing on the specified face image based on the feature extraction model to obtain a corresponding second feature vector;

[0216] A recognition submodule, configured to perform recognition processing on the second feature vector based on the target face recognition model to obtain a corresponding designated face recognition result;

[0217] The fourth determination submodule is used to use the designated face recognition result as the target face recognition result.

[0218] To solve the above technical problems, the present application also provides a computer device. Figure 4 , Figure 4This is a basic structural block diagram of the computer device in this embodiment.

[0219] The computer device 4 includes a memory 41, a processor 42, and a network interface 43 that are interconnected through a system bus. It should be noted that the figure only shows a computer device 4 with components 41-43, but it should be understood that it is not required to implement all the components shown, and more or fewer components can be implemented instead. Among them, those skilled in the art can understand that the computer device here is a device that can automatically perform numerical calculations and / or information processing according to pre-set or stored instructions, and its hardware includes but is not limited to microprocessors, application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), digital processors (DSPs), embedded devices, etc.

[0220] The computer device may be a computing device such as a desktop computer, a notebook, a PDA, a cloud server, etc. The computer device may interact with a user through a keyboard, a mouse, a remote controller, a touch pad, or a voice control device.

[0221] The memory 41 includes at least one type of readable storage medium, and the readable storage medium includes flash memory, hard disk, multimedia card, card-type memory (for example, SD or DX memory, etc.), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, disk, optical disk, etc. In some embodiments, the memory 41 can be an internal storage unit of the computer device 4, such as a hard disk or memory of the computer device 4. In other embodiments, the memory 41 can also be an external storage device of the computer device 4, such as a plug-in hard disk, a smart memory card (Smart Med ia Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), etc. equipped on the computer device 4. Of course, the memory 41 can also include both the internal storage unit of the computer device 4 and its external storage device. In this embodiment, the memory 41 is generally used to store the operating system and various application software installed on the computer device 4, such as computer-readable instructions of a face recognition method based on artificial intelligence, etc. In addition, the memory 41 can also be used to temporarily store various types of data that have been output or are to be output.

[0222] The processor 42 may be a central processing unit (CPU), a controller, a microcontroller, a microprocessor, or other data processing chips in some embodiments. The processor 42 is generally used to control the overall operation of the computer device 4. In this embodiment, the processor 42 is used to run the computer-readable instructions stored in the memory 41 or process data, such as running the computer-readable instructions of the artificial intelligence-based face recognition method.

[0223] The network interface 43 may include a wireless network interface or a wired network interface. The network interface 43 is generally used to establish a communication connection between the computer device 4 and other electronic devices.

[0224] The present application also provides another implementation, namely, providing a computer-readable storage medium, wherein the computer-readable storage medium stores computer-readable instructions, and the computer-readable instructions can be executed by at least one processor to enable the at least one processor to perform the steps of the above-mentioned artificial intelligence-based face recognition method.

[0225] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes a number of instructions for a terminal device (which can be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in each embodiment of the present application.

[0226] Obviously, the embodiments described above are only some embodiments of the present application, rather than all embodiments. The preferred embodiments of the present application are given in the accompanying drawings, but they do not limit the patent scope of the present application. The present application can be implemented in many different forms. On the contrary, the purpose of providing these embodiments is to make the understanding of the disclosure of the present application more thorough and comprehensive. Although the present application is described in detail with reference to the aforementioned embodiments, for those skilled in the art, it is still possible to modify the technical solutions recorded in the aforementioned specific implementation methods, or to perform equivalent replacement of some of the technical features therein. Any equivalent structure made using the contents of the specification and drawings of this application, directly or indirectly used in other related technical fields, is similarly within the scope of patent protection of this application.

Claims

1. A face recognition method based on artificial intelligence, characterized in that: The steps include: Acquire pre-constructed initial face image data; wherein the initial face image data is a data set that has been labeled with identity information; Preprocessing the initial face image data to obtain corresponding face image training data; Based on a preset feature extraction model, feature extraction processing is performed on the face image training data to obtain a corresponding target feature vector; wherein the feature extraction model is a deep subspace network model based on incremental row-column two-dimensional principal component analysis; Based on the target feature vector, a preset number of classification models are trained to obtain a plurality of trained face recognition models; Screening all the face recognition models based on preset face image test data to obtain a target face recognition model that meets preset evaluation conditions; The face image to be processed is recognized and processed based on the target face recognition model and the feature extraction model to generate a target face recognition result corresponding to the face image to be processed.

2. The artificial intelligence-based face recognition method according to claim 1, characterized in that: The step of preprocessing the initial face image data to obtain corresponding face image training data specifically includes: Normalizing the initial facial image data to obtain corresponding first facial image data; Performing grayscale processing on the first facial image data to obtain corresponding second facial image data; Performing histogram equalization processing on the second facial image data to obtain corresponding third facial image data; The third facial image data is used as the facial image training data.

3. The face recognition method based on artificial intelligence according to claim 1, characterized in that: The step of performing feature extraction processing on the face image training data based on a preset feature extraction model to obtain a corresponding target feature vector specifically includes: Dividing the face image training data into corresponding image blocks; wherein the number of the image blocks includes a plurality; Performing principal component analysis on each row in the image block based on the feature extraction model to obtain a corresponding first feature matrix; Performing principal component analysis on each column of the first feature matrix to obtain a corresponding second feature matrix; Performing a convolution operation on the face image training data based on the second feature matrix to obtain a corresponding feature map; Performing a preset operation on the feature map to obtain a corresponding first feature vector; The target feature vector is generated based on the first feature vector.

4. The artificial intelligence-based face recognition method according to claim 3, characterized in that: The step of performing principal component analysis on each column of the first feature matrix to obtain a corresponding second feature matrix specifically includes: Calculating a covariance matrix of the first feature matrix; Performing eigendecomposition on the covariance matrix to obtain eigenvectors and eigenvalues ​​corresponding to the covariance matrix; Constructing a corresponding column-wise projection matrix based on the eigenvectors and the eigenvalues; Projecting each column of the first feature matrix based on the projection matrix in the column direction to obtain a specified feature matrix after dimension reduction in the column direction; The designated characteristic matrix is ​​used as the second characteristic matrix.

5. The face recognition method based on artificial intelligence according to claim 1, characterized in that: The step of screening all the face recognition models based on the preset face image test data to obtain a target face recognition model that meets the preset evaluation conditions specifically includes: Using the face image test data to perform capability evaluation processing on each of the face recognition models, to obtain capability index values ​​of each of the face recognition models corresponding to preset capability indexes; Calculating the ability evaluation score of each face recognition model based on the ability indicator value of each face recognition model; Selecting a first face recognition model with the highest ability evaluation score from all the face recognition models; Determine whether the number of the first face recognition models is equal to 1; If the number of the first face recognition models is equal to 1, taking the first face recognition model as the target face recognition model; If the number of the first face recognition models is greater than 1, obtaining the accuracy value of each of the first face recognition models from the capability indicator value, and screening out the second face recognition model with the highest accuracy value from the first face recognition models; The second face recognition model is used as the target face recognition model.

6. The artificial intelligence-based face recognition method according to claim 5, characterized in that: The capability indicators at least include accuracy, processing time and recall rate; the step of calculating the capability evaluation score of each face recognition model based on the capability indicator values ​​of each face recognition model specifically includes: Obtaining a specified accuracy value, a specified processing time value, and a specified recall value from the capability indicator value of a third face recognition model; wherein the third face recognition model is any one of all the face recognition models; Get the preset weight construction algorithm; Generate a first weight corresponding to the accuracy rate, a second weight corresponding to the processing time, and a third weight corresponding to the recall rate based on the weight construction algorithm; Call the preset score calculation formula; Based on the score calculation formula, the specified accuracy value, the specified processing time value, the specified recall value, the first weight, the second weight, and the third weight are calculated to obtain a corresponding score result; The score result is used as the capability evaluation score of the third face recognition model.

7. The face recognition method based on artificial intelligence according to claim 1, characterized in that: The step of performing recognition processing on the face image to be processed based on the target face recognition model and the feature extraction model to generate a target face recognition result corresponding to the face image to be processed comprises: Preprocessing the face image to be processed to obtain a corresponding designated face image; Performing feature extraction processing on the designated face image based on the feature extraction model to obtain a corresponding second feature vector; Performing recognition processing on the second feature vector based on the target face recognition model to obtain a corresponding designated face recognition result; The designated face recognition result is used as the target face recognition result.

8. A face recognition device based on artificial intelligence, characterized in that: include: An acquisition module, used to acquire pre-constructed initial face image data; wherein the initial face image data is a data set that has been labeled with identity information; A preprocessing module, used for preprocessing the initial face image data to obtain corresponding face image training data; An extraction module, used to perform feature extraction processing on the face image training data based on a preset feature extraction model to obtain a corresponding target feature vector; wherein the feature extraction model is a deep subspace network model based on incremental row-column two-dimensional principal component analysis; A training module, used to train a preset number of classification models based on the target feature vector to obtain a plurality of trained face recognition models; A screening module, used to screen all the face recognition models based on preset face image test data to obtain a target face recognition model that meets preset evaluation conditions; The recognition module is used to perform recognition processing on the face image to be processed based on the target face recognition model and the feature extraction model, and generate a target face recognition result corresponding to the face image to be processed.

9. A computer device comprising a memory and a processor, wherein the memory stores computer-readable instructions, and when the processor executes the computer-readable instructions, the steps of the artificial intelligence-based face recognition method as described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-readable instructions, which, when executed by a processor, implement the steps of the artificial intelligence-based face recognition method as described in any one of claims 1 to 7.

Images