An image classification method and system based on a multi-layer end-to-end dictionary learning network

Abstract

The application relates to an image classification method and system based on a multi-layer inner-rolled dictionary learning network, and belongs to the technical field of image processing. The method comprises the following steps: processing an obtained image dataset to obtain a processed dataset; performing feature extraction on image samples of the processed dataset to obtain feature samples of each image sample; linearly representing each feature sample by using a sub-dictionary to obtain a dictionary matrix; constructing an optimization mathematical model according to the dictionary matrix, and obtaining a representation coefficient matrix by solving the optimization mathematical model; calculating residual values of each channel sub-dictionary according to the feature samples, the dictionary matrix and the representation coefficient matrix, predicting labels of the image samples according to the residual values, and then realizing image classification. The application solves the information loss problem caused by a non-end-to-end architecture, and optimizes the insufficient classification performance caused by insufficient consideration of the specific difference between categories.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of image processing, and particularly relates to an image classification method and system based on a multi-layer inner convolution dictionary learning network. BACKGROUND

[0002] Image classification technology is an image processing technology that distinguishes different categories of images according to the characteristic information reflected by the images. In the field of image classification, single-label classification problems are particularly common, that is, each image is only assigned one label, and the output of the model is a specific class label. Usually, a Softmax activation function is used to normalize the predicted probability distribution to determine the most likely one of multiple categories. The main application scenarios of this technology include cat and dog classification, handwritten digit recognition, and object recognition.

[0003] In the past few decades, various single-label image classification algorithms have been proposed, among which dictionary learning-based image classification algorithms are particularly classic. Dictionary learning technology originally extracts a set of sparse signals from raw data that can linearly represent all original signals, and is later widely used in image denoising and image classification fields. However, traditional dictionary learning algorithms are limited by the size and dimension of the data set, and it is difficult to effectively process high-dimensional data such as images, speech, and videos. With the development of deep learning technology, it has been gradually integrated into dictionary learning. In the prior art, various methods combining dictionary learning and deep learning have been proposed. For example, Deep Micro-Dictionary Learning and Coding Network (DDLCN) combines feature extraction layers with multiple dictionary learning layers, and integrates dictionary pair learning layers into a deep learning architecture, using the learned dictionary as a classifier, although it is limited to single-layer use. In addition, there is a Multi-layer Discriminative Dictionary Learning (MDDL) architecture that uses multi-layer dictionaries for feature representation learning and combines a multi-class support vector machine (SVM) as a classifier. There is also a Multi-layer Attention Dictionary Pair Learning Network (MAPDL-net) that is specifically used for image recognition tasks.

[0004] Although these methods effectively fuse dictionary learning and deep learning, they are not end-to-end architectures, which can cause loss of information in the transmission process. To overcome this challenge, a multi-layer attention dictionary pair learning network (MAPDL-net) is proposed in the prior art, which is an end-to-end architecture that converts traditional dictionary learning methods into a dictionary learning network, effectively improving the speed of processing large-scale data sets and maintaining the integrity of feature information. However, the dictionary layer in MAPDL-net is designed as a single shared layer, without fully considering the specificity between different categories, which can lead to poor results when processing classification tasks with similar category information. SUMMARY

[0005] To this end, the technical problem to be solved by the present application is to overcome the information loss caused by the non-end-to-end architecture in the prior art, and the poor classification results caused by not fully considering the specificity between different categories.

[0006] In a first aspect, to solve the above technical problems, the present application provides an image classification method based on a multi-layer inner convolution dictionary learning network, comprising:

[0007] processing the obtained image data set to obtain a processed data set;

[0008] performing feature extraction on the image samples of the processed data set to obtain feature samples for each image sample;

[0009] linearly representing each feature sample with a sub-dictionary to obtain a dictionary matrix;

[0010] constructing an optimization mathematical model according to the dictionary matrix, and obtaining a representation coefficient matrix by solving the optimization mathematical model; wherein the optimization mathematical model is:

[0011]

[0012] wherein D is the dictionary matrix, X i is the total representation coefficient matrix, y′ i is the feature sample, D k is the kth channel sub-dictionary, is the representation coefficient matrix of the kth channel sub-dictionary pair y i , y i is the ith image sample, concat(·) is a concatenation function, and M is the total number of categories;

[0013] calculating the residual value of each channel sub-dictionary according to the feature sample, the dictionary matrix and the representation coefficient matrix, and predicting the label of the image sample according to the residual value.

[0014] In one embodiment of the present application, the formula of the residual value is:

[0015]

[0016] wherein e k is the residual of the kth class, E(c k-1 : c k : k) is the residual matrix of the kth class, and M is the total number of classes.

[0017] In one embodiment of the present application, when the minimum value of the residual value is e j = min{e1, e2, …, e M}, the label of the image sample is j; wherein M is the total number of classes.

[0018] In one embodiment of the present application, the process of feature extraction of the image sample of the processed data set comprises channel increasing processing of the image sample by using a convolution layer.

[0019] In one embodiment of the present application, the convolution layer comprises a first convolution layer, and the first convolution layer comprises a 2D convolution function, a batch normalization function, a ReLU activation function and a 2D max pooling function.

[0020] In one embodiment of the present application, the convolution layer further comprises a plurality of second convolution layers, and the second convolution layers comprise a plurality of 2D convolution functions and batch normalization functions.

[0021] In one embodiment of the present application, the process of obtaining the dictionary matrix comprises grouping the feature samples, setting a shared convolution kernel for each group, and convolving the channel features in the group by using the shared convolution kernel; wherein the number of the groups is the number of classes of the classification task.

[0022] The second aspect is to solve the above technical problems, and the present application provides an image classification system based on a multi-layer inner convolution dictionary learning network, comprising:

[0023] A data feature extraction module is configured to process the obtained image data set to obtain a processed data set, and to perform feature extraction on the image samples of the processed data set to obtain feature samples of each image sample.

[0024] An inner convolution dictionary learning module is configured to linearly represent each feature sample by a sub-dictionary to obtain a dictionary matrix, and to construct an optimization mathematical model according to the dictionary matrix to obtain a representation coefficient matrix by solving the optimization mathematical model; wherein the optimization mathematical model is:

[0025]

[0026] Where D is the dictionary matrix, X i Let y' be the total coefficient matrix. i For feature samples, D k For the k-th class of channel sub-dictionary, For the k-th class of channel sub-dictionary pairs y i The coefficient matrix y represents i Let be the i-th image sample, concat(·) be the concatenation function, and M be the total number of categories;

[0027] The classification module is used to calculate the residual value of each class channel sub-dictionary based on the feature samples, the dictionary matrix, and the representation coefficient matrix, and to predict the label of the image sample based on the residual value; wherein the minimum value of the residual value is the label of the image sample.

[0028] In one embodiment of the present invention, the involution dictionary learning module includes multiple involution layers and a dictionary layer. The involution layer is used to extract the category information of the image sample and pass it to the dictionary layer. The dictionary layer is used to integrate different category information and pass it to the next involution layer.

[0029] Thirdly, to solve the above-mentioned technical problems, the present invention provides a computer program product, which includes a computer program that, when the computer program is run, causes any of the methods described above to be executed.

[0030] Compared with the prior art, the above-described technical solution of the present invention has the following advantages:

[0031] This invention presents an image classification method and system based on a multi-layer involute dictionary learning network. Through deep feature extraction technology, it accurately captures the core features of images, laying a solid information foundation for image classification tasks. By employing a sub-dictionary linear representation strategy, a dictionary matrix is ​​constructed for each feature sample, which not only improves the sparsity of the representation but also enhances the model's ability to distinguish features from different categories. Furthermore, by constructing and solving an optimized mathematical model, the representation coefficient matrix can be accurately obtained, better adapting to the characteristics of the feature samples, thereby significantly improving the model's accuracy and robustness. In addition, based on residual values, image sample labels are predicted, achieving automated image classification, reducing manual intervention, and improving processing efficiency. This design can adapt to various image datasets and classification tasks, demonstrating good versatility and applicability, meeting the needs of image processing in different fields. This invention effectively solves the problem of information loss that may be caused by non-end-to-end architectures and the problem of classification performance being affected by insufficient consideration of the specific differences between different categories, providing an innovative solution for the field of image classification. Attached Figure Description

[0032] For the purpose of making the content of the present application more easily and clearly understood, the present application is further described in detail below according to specific embodiments of the present application and in conjunction with the accompanying drawings, in which

[0033] Figure 1 A flow chart of an image classification method based on a multi-layer inner-rolling dictionary learning network in a preferred embodiment of the present application;

[0034] Figure 2 A network model diagram in a preferred embodiment of the present application;

[0035] Figure 3 A system framework diagram of an image classification system based on a multi-layer inner-rolling dictionary learning network in a preferred embodiment of the present application. DETAILED DESCRIPTION

[0036] The present application is further described below in conjunction with the accompanying drawings and specific embodiments, so that those skilled in the art can better understand the present application and implement it. However, the embodiments are not intended to limit the present application.

[0037] Embodiment One

[0038] Referring to Figure 1 , the present embodiment provides an image classification method based on a multi-layer inner-rolling dictionary learning network, which comprises:

[0039] processing the obtained image data set to obtain a processed data set;

[0040] performing feature extraction on the image samples of the processed data set to obtain feature samples of each image sample;

[0041] linearly representing each feature sample with a sub-dictionary to obtain a dictionary matrix;

[0042] constructing an optimization mathematical model according to the dictionary matrix, and obtaining a representation coefficient matrix by solving the optimization mathematical model; wherein the optimization mathematical model is:

[0043]

[0044] wherein D is the dictionary matrix, X i is the total representation coefficient matrix, y′ i is the feature sample, D k is the kth channel sub-dictionary, is the representation coefficient matrix of the kth channel sub-dictionary for y i , y i is the ith image sample, concat(·) is a concatenation function, and M is the total number of categories;

[0045] Residual values of each channel sub-dictionary are calculated according to the feature sample, the dictionary matrix and the representation coefficient matrix, the label of the image sample is predicted according to the residual values, and the image classification is realized according to the label.

[0046] The embodiment of the application provides an image classification method based on a multi-layer inner convolution dictionary learning network, which can effectively capture the key features of the image by performing feature extraction on the image sample, thereby providing a rich information basis for subsequent classification tasks. The method of linear representation of sub-dictionary can construct a dictionary matrix for each feature sample, which not only improves the sparsity of representation, but also enhances the ability of the model to distinguish different types of features. Further, by constructing an optimization mathematical model and solving it, the representation coefficient matrix can be accurately obtained, thereby better adapting to the characteristics of the feature sample. This optimization process helps to improve the accuracy and robustness of the model. In addition, the label of the image sample is predicted according to the residual value, and the automatic image classification is realized, which reduces the manual intervention and improves the processing efficiency. The method can adapt to various image data sets and classification tasks, has good universality and applicability, and meets the image processing needs of different fields. The embodiment of the application effectively solves the problem of information loss caused by the non-end-to-end architecture, and the problem of affecting the classification effect due to insufficient consideration of the specific differences between different categories, providing an innovative solution for the image classification field.

[0047] Specifically, the embodiment of the application assumes that the image data set obtained is The label set is Where the i-th image sample is y i ∈R C×H×W , C is the number of channels of the image sample, H is the height of the image sample, W is the width of the image sample, L i ∈{1,…,M} is the label of y i , M is the total number of categories, and n is the total number of samples in the image data set. After standardization and normalization of the image data set, it is assumed that the image sample y i is mapped to the feature sample y′ i after data feature extraction, where y′ i ∈R C ′ ×H×W , C′ is the new number of channels after the original channel number C of the image sample is mapped after feature extraction.

[0048] Specifically, the process of feature extraction on the image sample of the processing data set includes increasing the number of channels of the image sample by using a convolution layer. Referring to Figure 2The convolutional layer includes a first convolutional layer and a plurality of second convolutional layers. The first convolutional layer includes a 2D convolution function, a batch normalization function, a ReLU activation function, and a 2D max-pooling function. The second convolutional layer includes a plurality of 2D convolution functions and a batch normalization function. The convolutional layer in the embodiment preferably includes one first convolutional layer and four second convolutional layers. The main purpose of the first convolutional layer is to perform channel augmentation processing on the input data, maintain feature information, and facilitate subsequent feature extraction. The four second convolutional layers are all stacked by a plurality of 2D convolution functions and a batch normalization function, but there is a difference in the number of convolution function channels. In the embodiment, the number of their channels gradually expands, preferably 32, 64, 256, and 512, and the step is preferably 1, 2, 1, and 2. The selection of the number of convolution function channels and the step should be based on the specific application requirements and data characteristics to achieve the best performance and effect.

[0049] The first convolutional layer integrates the arrangement of the 2D convolution function, the batch normalization function, the ReLU activation function, and the 2D max-pooling function, which has the advantage of effectively extracting the features of the input data while increasing the nonlinear expression ability of the network. The 2D convolution function is used to capture local features, the batch normalization function helps to reduce overfitting and speed up the training process, the ReLU activation function introduces nonlinearity, enabling the network to learn more complex features, and the 2D max-pooling function is used to reduce the spatial dimension of the features while maintaining feature invariance. This comprehensive design not only enhances the expressiveness of the features, but also improves the robustness of the model to input changes and lays the foundation for in-depth feature extraction by subsequent convolutional layers.

[0050] Specifically, the linear representation of each feature sample by the sub-dictionary includes grouping the data features of the previous layer, and setting a shared convolution kernel for each group to perform convolution on the channel features in the group. The number of groups is equal to the number of classes of the classification task. Different groups use different convolutions for channel convolution. This intra-channel convolution method embeds a channel attention mechanism, which can efficiently allocate information of different categories to different groups of dictionary channels, thereby enhancing the model's ability to distinguish features.

[0051] Specifically, for a feature sample y′ i The embodiment assumes that the feature sample y′ i can be linearly represented by all sub-dictionaries, and the mathematical expression is:

[0052]

[0053] where the concat(·) function represents concatenating matrices in the channel dimension, represents multiplying corresponding elements of a matrix, is the kth class channel sub-dictionary representation of y ia representation coefficient matrix, is a k-th channel sub-dictionary, C k is the number of atoms in the k-th channel sub-dictionary, and the dimension of each dictionary atom is HxW, k=1,…,M. All the sub-dictionaries form a dictionary matrix D in the channel dimension, and the expression of D is:

[0054]

[0055] An optimization mathematical model is constructed according to the dictionary matrix (the construction of the optimization mathematical model can also be referred to as the construction of an optimization problem), wherein the expression of the total representation coefficient matrix X i is:

[0056]

[0057] When the input image data y is obtained after data feature extraction, the feature sample is y', and the representation coefficient matrix k∈{1,…,M} is obtained, the label of the input sample is predicted according to the representation coefficient matrix.

[0058] Specifically, let the residual matrix and E(c k-1 :c k ,:,:) represents the residual matrix of the k-th class, wherein The calculation formula of the residual value of the k-th class is:

[0059]

[0060] wherein e k is the residual of the k-th class, and M is the total number of classes.

[0061] Further, when the minimum value of the residual value is e j =min{e1,e2,…,e M}, the label of the image sample is j.

[0062] Further, after the predicted label of the image is obtained, the image classification task can be realized.

[0063] In order to more intuitively show the superiority of the embodiment of the present application compared with the prior art, a set of comparative experiments is provided.

[0064] Specifically, the method provided by the embodiment of the present application can independently acquire an image data set by using an image acquisition technology, or directly adopt an existing image data set. The comparative experiment is tested on a well-known image classification data set, Fashion-MNIST. Fashion-MNIST, as a fashion variant of the MNIST handwritten digit set, is provided by the research department of Zalando, a German fashion technology company, and contains 70,000 front-facing product images from 10 clothing categories. Each image is a 28x28-pixel single-channel grayscale image, and the label numbers 0 to 9 represent T-shirt, pants, pullover, dress, coat, sandals, shirt, sneaker, bag and boots, respectively. In the experiment, the Fashion-MNIST data set is divided into a training set containing 60,000 samples and a test set containing 10,000 samples. After 350 training cycles (epochs) on the training set, the optimized model parameters are obtained. Subsequently, the trained model parameters (in particular, the dictionary weight-based model) are used to perform prediction classification on the test set, and the accuracy is evaluated. In addition, the parameter values corresponding to the Fashion-MNIST data set are as follows: n = 70,000, M = 10, B = 512, C = 1, H = 28, and W = 28. Among them, B represents the batch quantity, that is, the number of samples input at a time.

[0065] The classification accuracy of the method provided by the embodiment of the present application on various data sets is compared with the performance of existing algorithms, such as discriminative K-SVD (DKSVD), dictionary learning based on Fisher discriminant criterion (FDDL), projection dictionary pair learning (DPL), and multi-layer dictionary learning network (MADPL-net), on the same data set.

[0066] Table 1 Classification accuracy of different classification models on the Fashion-MNIST data set

[0067] Method DKSVD FDDL DPL MADPL-net The present invention Accuracy (%) 76.25 81.23 84.65 91.50 93.92

[0068] The specific comparison results are shown in Table 1. As can be seen from the data in the table, the classification performance of the embodiment of the present application exhibits higher accuracy compared to other algorithms, which fully proves the effectiveness and superiority of the present application in image classification tasks.

[0069] Embodiment Two

[0070] Based on the same inventive concept, the embodiment provides an image classification system based on a multi-layer intravolume dictionary learning network, which has a similar problem-solving principle to the image classification method based on a multi-layer intravolume dictionary learning network provided in Embodiment One, and the repeated parts will not be described again.

[0071] Reference Figure 3As shown, the embodiment provides an image classification system based on a multi-layer end-to-end dictionary learning network, comprising:

[0072] A data feature extraction module is configured to process the obtained image dataset to obtain a processed dataset, and extract features of image samples in the processed dataset to obtain feature samples of each image sample.

[0073] An end-to-end dictionary learning module is configured to linearly represent each feature sample by a sub-dictionary to obtain a dictionary matrix, and construct an optimization mathematical model according to the dictionary matrix to obtain a representation coefficient matrix by solving the optimization mathematical model; wherein the optimization mathematical model is as follows:

[0074]

[0075] wherein D is the dictionary matrix, X i is the total representation coefficient matrix, y′ i is the feature sample, D k is the kth channel sub-dictionary, is the representation coefficient matrix of the kth channel sub-dictionary for y i , y i is the ith image sample, concat(·) is a concatenation function, and M is the total number of categories.

[0076] A classification module is configured to calculate residual values of each channel sub-dictionary according to the feature sample, the dictionary matrix and the representation coefficient matrix, and predict a label of the image sample according to the residual values; wherein a minimum value of the residual values is the label of the image sample.

[0077] Specifically, the data feature extraction module is composed of 5 convolutional layers, wherein the first convolutional layer includes a 2D convolution function, a batch normalization function, a ReLU activation function and a 2D max pooling function. The main purpose of this convolutional layer is to perform channel increment processing on the input data, maintain feature information and facilitate subsequent feature extraction. The last four convolutional layers are composed of multiple 2D convolution functions and batch normalization functions. The difference lies in that the number of convolution function channels of each convolutional layer gradually expands from 32, 64, 256, 512, and the step lengths are 1, 2, 1 and 2, respectively.

[0078] Specifically, the purpose of the inner-convolution dictionary learning module is to obtain a dictionary matrix D and process the feature samples obtained by the data feature extraction module to obtain a representation coefficient matrix. The module includes multiple inner-convolution layers and dictionary layers, wherein the inner-convolution layers are used to extract the class information of the image samples and pass the class information to the dictionary layers; and the dictionary layers are used to integrate different class information and pass the integrated class information to the next inner-convolution layer. Specifically, the function of the inner-convolution layer is to group the data features of the previous layer, and the number of groups matches the number of classes in the classification task. Each group uses a shared convolution kernel to perform convolution operation on the channel features in the group, and different groups use different convolution kernels for channel convolution. This channel-based inner-convolution strategy embeds a channel attention mechanism, which can effectively map the information of different classes to different groups of dictionary channels.

[0079] Embodiment three

[0080] The embodiment provides a computer program product, which comprises a computer program, when the computer program is executed, causes the method in the embodiment one to be executed.

[0081] Those skilled in the art will understand that the embodiments of the present application can be provided as a method, a system, or a computer program product. Therefore, the present application can take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can take the form of a computer program product implemented on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0082] The present application is described with reference to flowcharts and / or block diagrams of the method, device (system), and computer program product according to the embodiments of the present application. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, and the combination of the flows and / or blocks in the flowcharts and / or block diagrams can be implemented by computer program instructions. These computer program instructions can be provided to a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to produce a machine, so that the instructions executed by the computer or other programmable data processing devices generate a device implemented in the flowcharts and / or block diagrams. Figure 1 The device that implements the function specified in one flow or multiple flows and / or blocks Figure 1 The device that implements the function specified in one flow or multiple flows and / or blocks

[0083] These computer program instructions can also be stored in a computer-readable memory that can guide the computer or other programmable data processing devices to work in a specific way, so that the instructions stored in the computer-readable memory produce a manufactured product including instruction devices that implement the flowcharts and / or block diagrams. Figure 1 The device that implements the function specified in one flow or multiple flows and / or blocksFigure 1 the function specified in the one or more blocks.

[0084] These computer program instructions can also be loaded into computer or other programmable data processing devices, so that a series of operation steps are performed on the computer or other programmable data processing devices to generate computer-implemented processes, thus the instructions executed on the computer or other programmable data processing devices provide processes for implementing the flows Figure 1 the flows or processes and / or blocks Figure 1 the steps of the function specified in the one or more blocks.

[0085] Obviously, the above embodiments are only examples for clearly illustrating the present application, and are not intended to limit the implementation modes. Based on the above description, other different forms of changes or variations can also be made by those skilled in the art. Here, all the implementation modes are not required or can not be exhausted. The obvious changes or variations derived therefrom are still within the protection scope of the present application.

Claims

1. An image classification method based on a multi-layer involution dictionary learning network, characterized in that, include: The acquired image dataset is processed to obtain a processed dataset. Feature extraction is performed on the image samples of the processing dataset to obtain feature samples for each image sample; Each feature sample is linearly represented using a sub-dictionary to obtain a dictionary matrix; An optimized mathematical model is constructed based on the dictionary matrix, and the representation coefficient matrix is ​​obtained by solving the optimized mathematical model; wherein, the optimized mathematical model is: Where D is the dictionary matrix, X i Let y' be the total coefficient matrix. i For feature samples, D k For the k-th class of channel sub-dictionary, For the k-th class of channel sub-dictionary pairs y i The coefficient matrix y represents i Let be the i-th image sample, concat(·) be the concatenation function, and M be the total number of categories; The residual value of each channel sub-dictionary is calculated based on the feature samples, the dictionary matrix, and the representation coefficient matrix, and the label of the image sample is predicted based on the residual value.

2. The image classification method based on a multi-layer involution dictionary learning network according to claim 1, characterized in that, The formula for calculating the residual value is: Among them, e k For the residual of class k, E(c) k-1 :c k ,:,:) is the residual matrix of the k-th class, and M is the total number of classes.

3. The image classification method based on a multi-layer involution dictionary learning network according to claim 1, characterized in that, When the minimum value of the residual is e j =min{e1,e2,…,e M When}, the label of the image sample is j; where M is the total number of categories.

4. The image classification method based on a multi-layer involution dictionary learning network according to claim 1, characterized in that, The process of feature extraction from image samples in the processing dataset includes using convolutional layers to increase the number of channels in the image samples.

5. The image classification method based on a multi-layer involution dictionary learning network according to claim 4, characterized in that, The convolutional layer includes a first convolutional layer, which includes a 2D convolutional function, a batch regularization function, a ReLU activation function, and a 2D max pooling function.

6. The image classification method based on a multi-layer involution dictionary learning network according to claim 4, characterized in that... The convolutional layer further includes multiple second convolutional layers, each of which includes multiple 2D convolutional functions and batch regularization functions.

7. The image classification method based on a multi-layer involution dictionary learning network according to claim 1, characterized in that, The process of obtaining the dictionary matrix includes grouping the feature samples, setting a shared convolution kernel for each group, and using the shared convolution kernel to convolve the channel features within the group; wherein the number of groups is the number of categories in the classification task.

8. An image classification system based on a multi-layer involution dictionary learning network, characterized in that, include: The data feature extraction module is used to process the acquired image dataset to obtain a processed dataset; and to extract features from the image samples in the processed dataset to obtain feature samples for each image sample. The involution dictionary learning module is used to linearly represent each feature sample using a sub-dictionary to obtain a dictionary matrix; construct an optimized mathematical model based on the dictionary matrix, and obtain the representation coefficient matrix by solving the optimized mathematical model; wherein, the optimized mathematical model is: Where D is the dictionary matrix, X i Let y' be the total coefficient matrix. i For feature samples, D k For the k-th class of channel sub-dictionary, For the k-th class of channel sub-dictionary pairs y i The coefficient matrix y represents i Let be the i-th image sample, concat(·) be the concatenation function, and M be the total number of categories; The classification module is used to calculate the residual value of each class channel sub-dictionary based on the feature samples, the dictionary matrix, and the representation coefficient matrix, and to predict the label of the image sample based on the residual value; wherein the minimum value of the residual value is the label of the image sample.

9. An image classification system based on a multi-layer involution dictionary learning network according to claim 8, characterized in that, The involution dictionary learning module includes multiple involution layers and a dictionary layer. The involution layer is used to extract the category information of the image samples and pass it to the dictionary layer. The dictionary layer is used to integrate different category information and pass it to the next involution layer.

10. A computer program product, characterized in that, The computer program product includes a computer program that, when executed, causes the method as described in any one of claims 1 to 7 to be performed.

Images