A palm print multi-feature fusion recognition method and device based on multi-view learning

Abstract

The application provides a palmprint multi-feature fusion recognition method and device based on multi-view learning, comprising: extracting multiple palmprint view features from a palmprint region of interest image; for any one of the palmprint view features, using a linear discriminant analysis method to maximize the variance between different categories and minimize the variance between different samples of the same category to obtain the palmprint view feature after projection dimension reduction; using a deep generalized canonical correlation analysis method to extract the correlation features between the palmprint view features after projection dimension reduction as palmprint fusion features; and applying a collaborative representation classification method according to the palmprint fusion features to perform classification recognition. The palmprint features of multiple different views can be fused for recognition, and the stability and accuracy of palmprint feature recognition can be improved.

Description

Technical Field

[0001] This application belongs to the field of palmprint recognition technology, and in particular relates to a palmprint multi-feature fusion recognition method and device based on multi-view learning. Background Technology

[0002] Biometric technologies utilize biological or behavioral characteristics to automatically identify individuals and are considered one of the most reliable and effective solutions for identity verification. Examples include facial recognition, fingerprint recognition, iris recognition, and palmprint recognition. Biometric authentication has become a key component of secure access control systems. Palmprint features, due to their difficulty in forging and their uniqueness and near-lifelong immutability, have attracted increasing research attention.

[0003] Palmprints contain stable features, such as ridges, textures, and main lines, which enable palmprint recognition to achieve high accuracy. However, recognition accuracy decreases under varying environmental conditions or with limited sample sizes. Achieving stable and high recognition accuracy by fusing multiple palmprint features in complex environments is a current challenge in palmprint recognition applications. Summary of the Invention

[0004] The main objective of this invention is to provide a palmprint multi-feature fusion recognition method and apparatus based on multi-view learning, which can fuse palmprint features from multiple different views for recognition, thereby improving the stability and accuracy of palmprint feature recognition and solving the problem of unstable and low palmprint recognition accuracy.

[0005] Firstly, a palmprint multi-feature fusion recognition method based on multi-view learning is provided. The method includes: extracting multiple palmprint view features from a palmprint region of interest image; for any one of the palmprint view features, using linear discriminant analysis to maximize the variance between different categories while minimizing the variance between different samples of the same category, to obtain the palmprint view features after projection dimensionality reduction; using deep generalized canonical correlation analysis to extract the correlation features between the palmprint view features after projection dimensionality reduction, as palmprint fusion features; and applying a collaborative representation classification method to perform classification and recognition based on the palmprint fusion features.

[0006] In one possible implementation, the palmprint view features include palmprint texture features, palmprint orientation features, and palmprint depth convolution features. Extracting multiple palmprint view features from the palmprint region of interest (ROI) image includes: uniformly dividing each subspace of the ROI image into regional pixel blocks; applying an improved local binary mode algorithm to extract texture and position information from all regional pixel blocks; concatenating and fusing the texture and position information to construct palmprint texture features; applying 2D Gabor filters in 10 directions to obtain convolution responses with the ROI image in 10 directions; calculating the differences in convolution responses between adjacent directions through weighted fusion; and combining these to form palmprint orientation features; and extracting palmprint depth features from the ROI image using a Vanillanet lightweight neural network to obtain a feature matrix, which is then transformed to form the palmprint depth convolution features.

[0007] In another possible implementation, the improved local binary mode algorithm is used to extract the texture and position information of all region pixel blocks, including: for any region pixel block, taking the region pixel block as the center region pixel block, calculating the absolute value of the difference between the total average gray value of the eight adjacent region pixel blocks and the average gray value of the center region pixel block to obtain a first threshold; replacing the comparison of individual pixels in the original local binary mode algorithm with the comparison of the average gray values ​​of the region pixel blocks, and extracting the texture information of the center region pixel block based on the following relationship:

[0008]

[0009]

[0010] in, It is the average grayscale value of the pixel block in the central region. is the average gray value of adjacent pixel blocks, P is the number of adjacent pixel blocks, P=8, R is the radius of each pixel block, and LBP represents the Local Binary Mode Algorithm; among the average gray values ​​of adjacent pixel blocks, the maximum and minimum values ​​are determined, a coordinate axis is established in the central pixel block, the span of each pixel block is considered as 1, and the distance between them is calculated as the position information of the central pixel block.

[0011] In another possible implementation, calculating the distance between the two includes: applying the following relation to calculate the distance between the two:

[0012]

[0013] in, The coordinates of the central region block are... These are the coordinates of the pixel block with the highest average gray value within the adjacent region. The coordinates of the pixel block with the smallest average gray value in the adjacent area are given. LBP stands for Local Binary Mode Algorithm.

[0014] In another possible implementation, the step of using linear discriminant analysis to maximize the variance between different categories while minimizing the variance between different samples of the same category to obtain the palmprint view features after projection dimensionality reduction includes: projecting each sample in the palmprint view features; calculating the mean of data samples of different categories and the mean of all samples after projection mapping; calculating the inter-class scatter matrix and intra-class scatter matrix after projection mapping based on the mean of data samples and the mean of all samples; and learning the projection matrix within the palmprint view based on the inter-class scatter matrix and the intra-class scatter matrix using an optimization objective function, and performing projection dimensionality reduction on the palmprint view features to obtain the palmprint view features after projection dimensionality reduction.

[0015] In another possible implementation, the step of using deep generalized canonical correlation analysis to extract the correlation features between the palmprint view features after projection dimensionality reduction as palmprint fusion features includes: applying a multi-layer neural network to process each palmprint view feature after projection dimensionality reduction to obtain the nonlinear mapping features of each palmprint view feature; and applying a deep generalized canonical correlation analysis model to maximize the correlation features between each nonlinear mapping feature to obtain the palmprint fusion features.

[0016] In another possible implementation, the classification and recognition based on the palmprint fusion features using a collaborative representation method includes: constructing a training sample set A: , Indicates the first The first in the class Fusion features of individual palm print samples , , This represents the total number of palmprint sample categories. For the training sample set, the first The number of class samples, This represents the dimension of the fused feature vector for each palm print. This represents the total number of training feature samples; using the palmprint fusion features as test samples, the fusion feature vector of a test sample is linearly represented by the palmprint fusion features in the training sample set. : , , is the coefficient matrix. Indicates the first The coefficient vector of each class; the distance between the fused feature vector of the test sample and the fused feature vector of each class of training samples is used as the weight to calculate the coefficient matrix. The estimated value; calculate the fusion feature vector of the test sample. With the The matching error between classes is considered, and the fusion feature vector of the test sample is determined when the matching error is minimized. For the first kind.

[0017] Secondly, a palmprint multi-feature fusion recognition device based on multi-view learning is provided. The device includes: a data acquisition module for extracting multiple palmprint view features from a palmprint region of interest image; a projection dimensionality reduction module for maximizing the variance between different categories and minimizing the variance between different samples of the same category for any one of the palmprint view features, thereby obtaining the palmprint view features after projection dimensionality reduction; a feature fusion module for extracting the correlation features between the palmprint view features after projection dimensionality reduction using a deep generalized canonical correlation analysis method, as palmprint fusion features; and a classification and recognition module for performing classification and recognition based on the palmprint fusion features using a collaborative representation classification method.

[0018] Thirdly, an electronic device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements a palmprint multi-feature fusion recognition method based on multi-view learning as provided in the first aspect.

[0019] Fourthly, a non-transitory computer-readable storage medium is provided, on which a computer program is stored, which, when executed by a processor, implements a palmprint multi-feature fusion recognition method based on multi-view learning as provided in the first aspect. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments of this application will be briefly introduced below.

[0021] Figure 1 A flowchart illustrating a palmprint multi-feature fusion recognition method based on multi-view learning, provided as an embodiment of the present invention;

[0022] Figure 2 The diagram shows a structure of a palmprint multi-feature fusion recognition device based on multi-view learning, according to an embodiment of the present invention.

[0023] Figure 3 This is a schematic diagram of the physical structure of an electronic device provided by the present invention. Detailed Implementation

[0024] The embodiments of this application are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar modules or modules having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain this application, and should not be construed as limiting the invention.

[0025] Those skilled in the art will understand that, unless specifically stated otherwise, the singular forms “a,” “an,” “the,” and “the” used herein may also include the plural forms. It should be further understood that the term “comprising” as used in this application means the presence of the stated features, integers, steps, operations, modules, and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, modules, components, and / or groups thereof. It should be understood that when we say a module is “connected” or “coupled” to another module, it can be directly connected or coupled to the other module, or there may be an intermediate module. Furthermore, “connected” or “coupled” as used herein can include wireless connection or wireless coupling. The term “and / or” as used herein includes all or any of the modules and all combinations thereof of one or more associated listed items.

[0026] To make the objectives, technical solutions, and advantages of this application clearer, the implementation of this application will be described in further detail below with reference to the accompanying drawings.

[0027] The technical solutions of this application and how they solve the aforementioned technical problems will be described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments. The embodiments of this application will now be described with reference to the accompanying drawings.

[0028] like Figure 1 The diagram shows a flowchart of a palmprint multi-feature fusion recognition method based on multi-view learning according to an embodiment of the present invention. The method includes:

[0029] Step S11: Extract various palm print view features from the palm print region of interest image.

[0030] Specifically, three types of view features are extracted simultaneously from the palmprint region of interest image: the palmprint texture features are extracted from the palmprint region of interest image using an improved Local Binary Pattern (LBP) algorithm, the palmprint orientation features are extracted from the palmprint region of interest image using an improved Gabor filter, and the palmprint depth convolution features are extracted from the palmprint region of interest image using a Vanillanet lightweight neural network.

[0031] Step S12: For any of the palmprint view features, use linear discriminant analysis to maximize the variance between different categories while minimizing the variance between different samples of the same category, to obtain the palmprint view features after projection dimensionality reduction.

[0032] Specifically, for each view feature, Linear Discriminant Analysis (LDA) is used to maximize the variance between different categories while minimizing the variance between different samples within the same category, thereby effectively distinguishing samples from different categories.

[0033] Step S13: Use deep generalized canonical correlation analysis to extract the correlation features between the palmprint view features after projection dimensionality reduction, and use them as palmprint fusion features.

[0034] By using deep generalized canonical correlation analysis to extract correlation features among three palmprint view features as discriminative features, redundant information is eliminated, and more discriminative and representative palmprint fusion features are obtained.

[0035] Step S14: Classify and identify the palmprint fusion features using a collaborative representation classification method.

[0036] The classification method using collaborative representation is used to identify samples by constructing a training sample set using palm print fusion feature vectors to represent the relationships between samples and capture the inherent structure of the data. The distance between the fusion feature vector of the test sample and the fusion feature vector of each class of training samples is used as the weight to learn the linear representation of the test sample on the training sample set for classification and identification.

[0037] This invention presents a palmprint multi-feature fusion recognition based on multi-view learning, which can fuse palmprint features from multiple different views for recognition, thereby improving the stability and accuracy of palmprint feature recognition and solving the problem of unstable and low accuracy in palmprint recognition.

[0038] In this embodiment of the invention, in step S11, each subspace of the palm print region of interest image is uniformly divided into blocks to obtain region pixel blocks. An improved local binary mode algorithm is applied to extract the texture information and position information of all region pixel blocks. The texture information and the position information are then spliced ​​and fused to construct palm print texture features (view1).

[0039] Optionally, for any given region pixel block, taking that region pixel block as the center region pixel block, the absolute value of the difference between the total average gray value of the eight adjacent region pixel blocks and the average gray value of the center region pixel block is calculated to obtain a first threshold. During feature extraction, each subspace is uniformly divided into blocks. After block division, the comparison of the average gray value of the region blocks replaces the comparison of individual pixels in the original LBP. The absolute value of the difference between the total average gray value of the eight surrounding region pixel blocks and the average gray value of the center region pixel block is calculated and used as the first threshold. :

[0040] (1)

[0041] Then, the comparison of the average grayscale value of the region pixel block is applied instead of the comparison of individual pixels in the original local binary mode algorithm, and the texture information of the central region pixel block is extracted based on the following relationship:

[0042] (2)

[0043] (3)

[0044] in, It is the average grayscale value of the pixel block in the central region. is the average grayscale value of adjacent pixel blocks, P is the number of adjacent pixel blocks (P=8), R is the radius of each pixel block, and LBP stands for Local Binary Mode Algorithm. , The threshold is set to 3. Each pixel block is a 3×3 square region with R=3.

[0045] Among the average grayscale values ​​of adjacent pixel blocks, determine the maximum and minimum values. Establish a coordinate axis for the central pixel block, treating the span of each pixel block as 1. Calculate the distance between the two axes, which serves as the position information of the central pixel block. Specifically, the following formula is used to calculate the distance:

[0046] (4)

[0047] in, The coordinates of the central region block are... These are the coordinates of the pixel block with the highest average gray value within the adjacent region. The coordinates of the pixel block with the minimum average gray value among adjacent regions are given. LBP stands for Local Binary Mode Algorithm. If all pixel blocks in a region have the same pixel value, it is defined as 1.

[0048] The texture information of all pixels is extracted using Equation (2) for the region of interest image of the palm print; the position information is extracted using Equation (4), and the two information are spliced ​​and fused to construct the palm print texture feature.

[0049] In this embodiment of the invention, 2D Gabor filters in 10 directions are applied to obtain the convolutional responses of the palm print region of interest image in 10 directions, and the differences in convolutional responses between adjacent directions are calculated by weighted fusion and combined to form palm print directional features (view2). The 2D Gabor filter is shown in Equation (5).

[0050] (5)

[0051] in, This represents the standard deviation of the Gaussian function. , It is the frequency of a sine wave. This represents the direction of the Gabor filter. As shown in equation (6), 2D Gabor filters in 10 directions are selected and their responses are obtained, denoted as... , The corresponding direction is , It is the convolution response of the Gabor template with the palmprint region of interest image in 10 directions, which obtains information in multiple directions. I(x,y) represents the palmprint region of interest image.

[0052] (6)

[0053] in," "It is a convolution operator."

[0054] The difference in convolutional responses between adjacent directions is calculated by weighted fusion, as shown in equation (7):

[0055] (7)

[0056] Palm print direction features are formed by combining formula (8) (view2):

[0057] (8)

[0058] in, , This represents the corresponding weight, and and The sum is 1.

[0059] The palmprint depth features in the palmprint region of interest image are extracted using the Vanillanet lightweight neural network to obtain a feature matrix. After matrix transformation, the palmprint depth convolutional features are formed (view3). Specifically, the Vanillanet lightweight network is trained using the cosine annealing algorithm for feature extraction. Then, the trained Vanillanet network is used to extract the palmprint depth features in the palmprint region of interest image to obtain a feature matrix. After matrix transformation, the palmprint depth convolutional features are formed.

[0060] In step S12, optionally, projection mapping is performed on each sample in the palmprint view features; the mean of data samples of different categories after projection mapping and the mean of all samples after projection mapping are calculated; the inter-class scatter matrix and intra-class scatter matrix after projection mapping are calculated based on the mean of data samples and the mean of all samples; the projection matrix in the palmprint view is learned based on the inter-class scatter matrix and the intra-class scatter matrix using an optimization objective function, and the palmprint view features are reduced in dimensionality by projection to obtain the palmprint view features after dimensionality reduction by projection.

[0061] Taking palm print texture features as an example, For the texture view feature, the first The class of A sample, processed by equation (9) Perform projection mapping, calculate the mean of data samples of different categories after mapping according to equation (10), and calculate the mean of all samples after mapping according to equation (11); calculate the inter-class scatter matrix after mapping according to equation (12), and calculate the intra-class scatter matrix after mapping according to equation (13).

[0062] (9)

[0063] (10)

[0064] (11)

[0065] (12)

[0066] (13)

[0067] Let be the projection matrix within the palmprint view. Design the objective function as shown in Equation (14), learn the projection matrix within the view, and perform projection dimensionality reduction on the palmprint texture features to obtain the projection-reduced palmprint texture features. Then, perform projection dimensionality reduction on both the palmprint orientation features and the palmprint depth convolution features to obtain the projection-reduced palmprint orientation features and palmprint depth convolution features.

[0068] (14)

[0069] in, Represents trace operation.

[0070] In step S13, optionally, a multi-layer neural network is first applied to process each of the palmprint view features after projection dimensionality reduction, to obtain the nonlinear mapping features of each palmprint view feature. Each palmprint view feature after projection dimensionality reduction is then passed through a multi-layer neural network, allowing... Represents the input view features, and , Let the variable be... Indicates the data sample size. Indicates the first The dimension of the view feature. Let the first view feature be... The network of views has Layer, number The first in the view There are layers There are neurons, and the output layer has _____ neurons. The _ neuron, the _ ... The first in the view Layer output As shown in equation (15), where It is a non-linear activation function (ReLU). It is a weight matrix. It is the bias vector. The output of the last layer of the network is expressed using equation (16) to obtain the nonlinear mapping characteristics of this view. , This is a general summary of the function transformations in a multilayer neural network.

[0071] (15)

[0072] , (16)

[0073] Then, a deep generalized canonical correlation analysis model is applied to maximize the correlation features between the nonlinear mapping features to obtain palmprint fusion features. To fuse features, calculate With fusion features The reconstruction error is used as the objective function to construct a deep generalized canonical correlation analysis model. Let the variables be... Indicates the data sample size. For the first The projection change of the network output, variable express Dimensions and It is a penalty parameter.

[0074] (17)

[0075] in, It is an identity matrix. Under the conditions, the aim is to find Palmprint fusion features from different views .

[0076] In step S14, the training sample set A is constructed:

[0077] (18)

[0078] Indicates the first The first in the class Palmprint fusion features of individual palmprint samples. , , This represents the total number of palmprint sample categories. For the training sample set, the first The number of class samples, This represents the dimension of each palmprint fusion feature. This represents the total number of training feature samples.

[0079] Using the palmprint fusion features as test samples, the fusion feature vector of a test sample is linearly represented by the palmprint fusion features in the training sample set. : , , is the coefficient matrix. Represented as the first The coefficient vector of the class. If The fused feature vector representing a test sample can be linearly represented by the fused feature vectors in the training sample set. The coefficient matrix is ​​calculated using the distance between the fused feature vector of the test sample and the fused feature vector of each class of training samples as weights. The estimated value. Weight matrix. Determined by equation (19)

[0080] (19)

[0081] in, Adjust the parameters for the weights. Then, establish the relationship between the weights and the parameters according to equation (20). The objective function is denoted by λ, where λ is the regularization parameter and y is the feature of the given test sample. make sure Sparsity.

[0082] (20)

[0083] in, By differentiating formula (20), we obtain: .

[0084] Calculate the fusion feature vector of the test sample With the The matching error between classes is considered, and the fusion feature vector of the test sample is determined when the matching error is minimized. For the first Class. Specifically, formula (21) can be used to calculate the class residual. .in For vectors With the Matching error between classes In and category The associated terms are not zero, and the others are zero. Use equation (22) to calculate the category of the test sample.

[0085] (twenty one)

[0086] (twenty two)

[0087] in, For the training sample set Regarding the first A partial set vector of class samples. When When the minimum value is reached, the corresponding The category represented is used as the category of the test sample, and is denoted by z.

[0088] In summary, this embodiment of the invention extracts multiple palmprint view features from the palmprint region of interest image; for any one of the palmprint view features, linear discriminant analysis is used to maximize the variance between different categories while minimizing the variance between different samples of the same category, resulting in the palmprint view features after projection dimensionality reduction; deep generalized canonical correlation analysis is used to extract the correlation features between the palmprint view features after projection dimensionality reduction, which are used as palmprint fusion features; and a collaborative representation classification method is applied to classify and identify the palmprint fusion features. This allows for the fusion of palmprint features from multiple different views for identification, improving the stability and accuracy of palmprint feature recognition.

[0089] like Figure 2 The diagram shown is a structural diagram of a palmprint multi-feature fusion recognition device based on multi-view learning according to an embodiment of the present invention. The device includes:

[0090] The data acquisition module is used to extract various palmprint view features from the palmprint region of interest image;

[0091] The projection dimension reduction module is used to maximize the variance between different categories and minimize the variance between different samples of the same category for any of the palmprint view features, thereby obtaining the palmprint view features after projection dimension reduction.

[0092] The feature fusion module is used to extract the correlation features between the palmprint view features after projection dimensionality reduction using the deep generalized canonical correlation analysis method, and use them as palmprint fusion features.

[0093] The classification and recognition module is used to perform classification and recognition based on the palmprint fusion features by applying a collaborative representation classification method.

[0094] The apparatus of the above embodiments is applied to the corresponding methods in the foregoing embodiments and has the beneficial effects of the corresponding method embodiments, which will not be repeated here.

[0095] Figure 3 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 3 As shown, the electronic device may include: a processor 301, a communication interface 302, a memory 303, and a communication bus 304, wherein the processor, communication interface, and memory communicate with each other through the communication bus. The processor can call logical instructions in the memory to execute a palmprint multi-feature fusion recognition method based on multi-view learning. This method includes: extracting multiple palmprint view features from a palmprint region of interest image; for any one of the palmprint view features, using linear discriminant analysis to maximize the variance between different categories while minimizing the variance between different samples of the same category, to obtain the palmprint view features after projection dimensionality reduction; using deep generalized canonical correlation analysis to extract the correlation features between the palmprint view features after projection dimensionality reduction as palmprint fusion features; and applying a collaborative representation classification method to perform classification and recognition based on the palmprint fusion features.

[0096] Furthermore, the logical instructions in the aforementioned memory can be implemented as software functional units and sold or used as independent products, and can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0097] On the other hand, embodiments of the present invention also provide a computer program product, the computer program product including a computer program stored on a non-transitory computer-readable storage medium, the computer program including program instructions, when the program instructions are executed by a computer, the computer is able to execute a palmprint multi-feature fusion recognition method based on multi-view learning provided in the above-described method embodiments, the method including: extracting multiple palmprint view features from a palmprint region of interest image; for any one of the palmprint view features, using a linear discriminant analysis method to maximize the variance between different categories, while minimizing the variance between different samples of the same category, to obtain the palmprint view features after projection dimensionality reduction; using a deep generalized canonical correlation analysis method to extract the correlation features between the palmprint view features after projection dimensionality reduction, as palmprint fusion features; and applying a collaborative representation classification method to perform classification and recognition based on the palmprint fusion features.

[0098] In another aspect, embodiments of the present invention also provide a non-transitory computer-readable storage medium storing a computer program thereon. When executed by a processor, the computer program is implemented to perform a palmprint multi-feature fusion recognition method based on multi-view learning provided in the above embodiments. The method includes: extracting multiple palmprint view features from a palmprint region of interest image; for any one of the palmprint view features, using linear discriminant analysis to maximize the variance between different categories while minimizing the variance between different samples of the same category, to obtain the palmprint view features after projection dimensionality reduction; using deep generalized canonical correlation analysis to extract the correlation features between the palmprint view features after projection dimensionality reduction as palmprint fusion features; and applying a collaborative representation classification method to perform classification and recognition based on the palmprint fusion features.

[0099] It should be understood that although the steps in the flowcharts of the accompanying figures are shown sequentially as indicated by the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the accompanying figures may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the sub-steps or stages of other steps.

[0100] The above description is only a partial implementation of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A palmprint multi-feature fusion recognition method based on multi-view learning, characterized in that, The method includes: Extracting various palmprint view features from the palmprint region of interest image; For any of the palmprint view features, linear discriminant analysis is used to maximize the variance between different categories while minimizing the variance between different samples of the same category, to obtain the palmprint view features after projection dimensionality reduction. The correlation features between the palmprint view features after projection dimensionality reduction are extracted using the deep generalized canonical correlation analysis method and used as palmprint fusion features. Classification and recognition are performed using a collaborative representation classification method based on the palmprint fusion features; The method of extracting correlation features among the palmprint view features after projection dimensionality reduction using deep generalized canonical correlation analysis as palmprint fusion features includes: A multi-layer neural network is applied to process each palmprint view feature after projection dimensionality reduction to obtain the nonlinear mapping features of each palmprint view feature. By applying a deep generalized canonical correlation analysis model to maximize the correlation features among the nonlinear mapping features, palmprint fusion features are obtained. The palmprint view features include palmprint texture features, palmprint orientation features, and palmprint depth convolution features. The extraction of multiple palmprint view features from the palmprint region of interest image includes: uniformly dividing each subspace of the palmprint region of interest image into regional pixel blocks, applying an improved local binary mode algorithm to extract the texture information and position information of all regional pixel blocks, and concatenating and fusing the texture information and the position information to construct palmprint texture features. Ten 2D Gabor filters in 10 directions are applied to obtain the convolutional responses of the palmprint region of interest image in 10 directions. The differences in convolutional responses between adjacent directions are calculated by weighted fusion and combined to form palmprint directional features. The palmprint depth features are extracted from the palmprint region of interest image using the Vanillanet lightweight neural network to obtain a feature matrix, which is then transformed to form the palmprint depth convolution feature.

2. The method as described in claim 1, characterized in that, The improved local binary mode algorithm is used to extract texture and position information of all pixel blocks in the application, including: For any region pixel block, taking the region pixel block as the center region pixel block, calculate the absolute value of the difference between the total average gray value of the 8 adjacent region pixel blocks and the average gray value of the center region pixel block to obtain the first threshold. The comparison of the average grayscale value of the pixel block in the application region replaces the comparison of individual pixels in the local binary mode algorithm, and the texture information of the pixel block in the central region is extracted based on the following relationship: Among them, g c It is the average grayscale value of the pixel block in the central region, g w is the average grayscale value of adjacent pixel blocks, P is the number of adjacent pixel blocks (P=8), R is the radius of each pixel block, and LBP stands for Local Binary Mode Algorithm. X = g w -g c N is the first threshold; Among the average gray values ​​of adjacent pixel blocks, determine the maximum and minimum values, establish a coordinate axis in the central pixel block, consider the span of each pixel block as 1, calculate the distance between them as the position information of the central pixel block.

3. The method as described in claim 2, characterized in that, The calculation of the distance between the two includes: The distance between the two can be calculated using the following formula: Among them, (x c y c (x) represents the coordinates of the pixel block in the central region. max y max (x) represents the coordinates of the pixel block with the highest average gray value in the adjacent region. min y min ) represents the coordinates of the pixel block with the smallest average gray value in the adjacent region. LBP stands for Local Binary Mode Algorithm.

4. The method as described in claim 1, characterized in that, The method of using linear discriminant analysis to maximize the variance between different categories while minimizing the variance between different samples of the same category yields the palmprint view features after projection dimensionality reduction, including: Projection mapping is performed on each sample in the palmprint view features; Calculate the mean of data samples of different categories after projection mapping and the mean of all samples after projection mapping; Calculate the inter-class scatter matrix and intra-class scatter matrix after projection mapping based on the mean of the data samples and the mean of all samples; Based on the inter-class scatter matrix and the intra-class scatter matrix, the projection matrix within the palmprint view is learned using an optimization objective function. The palmprint view features are then subjected to projection dimensionality reduction to obtain the projection-reduced palmprint view features.

5. The method as described in claim 1, characterized in that, The classification and identification based on the palmprint fusion features using a collaborative representation method includes: Construct training sample set A: G i，j Let i represent the palmprint fusion feature of the j-th palmprint sample in the i-th class, i = 1, 2, ..., k, j = 1, 2, ..., n k k is the total number of palmprint sample types, n k Let r be the number of samples of class k in the training sample set, r be the dimension of each palmprint fusion feature, and N be the total number of training samples. Using the palmprint fusion features as test samples, the fusion feature vector y of a test sample is linearly represented by the palmprint fusion features in the training sample set: y≈Aα, α=[α1; α2; ...; α k ], where a is the coefficient matrix. k Represents the coefficient vector of the k-th class; The distance between the fused feature vector of the test sample and the fused feature vector of each class of training samples is used as the weight. Calculate the estimated value of the coefficient matrix a; Calculate the matching error between the fused feature vector y of the test sample and the i-th class. When the matching error is minimized, determine that the fused feature vector y of the test sample belongs to the i-th class.

6. A palmprint multi-feature fusion recognition device based on multi-view learning, using the method described in any one of claims 1-5, characterized in that, The device includes: The data acquisition module is used to extract various palmprint view features from the palmprint region of interest image; The projection dimension reduction module is used to maximize the variance between different categories and minimize the variance between different samples of the same category for any of the palmprint view features, thereby obtaining the palmprint view features after projection dimension reduction. The feature fusion module is used to extract the correlation features between the palmprint view features after projection dimensionality reduction using the deep generalized canonical correlation analysis method, and use them as palmprint fusion features. The classification and recognition module is used to perform classification and recognition based on the palmprint fusion features by applying a collaborative representation classification method.

7. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements a palmprint multi-feature fusion recognition method based on multi-view learning as described in any one of claims 1-5.

8. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements a palmprint multi-feature fusion recognition method based on multi-view learning as described in any one of claims 1-5.

Images