Image classification method and system based on improved robust sparse linear discriminant analysis

Abstract

The present invention relates to an image classification method and system based on improved robust sparse linear discriminant analysis. The classification method comprises the following steps: inputting an image data set into a data input module and storing it in a database; extracting an image data set in the database, and extracting image features from the image data set in a data feature extraction module; inputting the extracted image features into a data analysis module, and using the RSLDA+ method to perform data analysis and processing; and inputting the results of the analysis and processing into an image classification module, and performing image classification in the image classification module. The classification system comprises a data input module, a data feature extraction module, a data analysis module and an image classification module. Aiming at the problem that the image classification method has poor noise resistance, the present invention proposes the RSLDA+ method, which effectively improves the interpretability of the image feature matrix and can significantly improve the accuracy of image classification.

Description

Technical Field

[0001] The present invention relates to the field of image processing technology, and in particular to an image classification method and system based on improved robust sparse linear discriminant analysis. The present invention can be applied to the fields of image recognition, image classification, etc. Background Art

[0002] In application fields such as image recognition and image classification, we often encounter a lot of image data. Since image data is generally high-dimensional data, if we directly process high-dimensional data, the requirements for computer hardware are high and the recognition rate is low. Therefore, before classifying, recognizing or clustering images, it is generally necessary to reduce the dimensionality of the image, and feature extraction is one of the most common dimensionality reduction methods. Feature extraction is a technology that finds a small number of features that best represent the data from a large amount of raw data. It is usually used as a data preprocessing method to overcome problems such as the "curse of dimensionality".

[0003] At present, there are many image feature extraction methods, such as principal component analysis (PCA), linear discriminant analysis (LDA), locality preserving projection (LPP), etc. However, all of these methods use the L2 norm, which is very sensitive to noise and heterogeneous data, to construct the objective function, which makes these methods sensitive to noise and not robust enough.

[0004] Sparsity is a very important property. Sparse modeling can automatically find relevant features from the training data while ignoring other features. It not only improves the generalization ability of the algorithm, but also increases the interpretability of the results.

[0005] The problems and defects of existing image classification technology are:

[0006] 1. The existing method simplifies the calculation of the image classification algorithm and performs convex approximation smoothing on the original data features, which loses some of the authenticity of the image and cannot perfectly classify the data features.

[0007] 2. Existing methods only use the row vector norm of the data set for regularization, which makes it difficult for image classification algorithms to obtain ideal sparse solutions in some cases.

[0008] 3. Existing methods perform poorly when processing high-complexity data sets, are sensitive to noise, and are not robust enough. Summary of the invention

[0009] In view of some or all of the problems in the prior art, the present invention provides an image classification method based on improved robust sparse linear discriminant analysis, the method comprising the following steps:

[0010] Input the image data set in the data input module and store it in the database;

[0011] Extracting an image data set from a database, and extracting image features from the image data set in a data feature extraction module;

[0012] Inputting the extracted image features into a data analysis module, and using the RSLDA+ method to perform data analysis and processing; and

[0013] The result of the analysis process is input into an image classification module, and the image classification is performed in the image classification module.

[0014] Furthermore, extracting image features from the image data set in the data feature extraction module includes: using principal component analysis to decompose image information to decompose main image data features and noise items.

[0015] Furthermore, the data analysis and processing using the RSLDA+ method includes:

[0016] The image data processing problem constructed by the RSLDA+ method is converted into the problem of solving the optimal values ​​of the three parameter matrices P, Q, and E. The three parameter matrices P, Q, and E satisfy the following formulas:

[0017]

[0018]

[0019] Among them, min represents the minimum value of the function, st represents the constraint, I represents the identity matrix, Tr represents the trace of the matrix, T represents the matrix transpose, the P matrix is ​​the linear transformation matrix of the original data set, and E represents the noise term matrix; ||Q|| 2，0 ,||Q||0,||E||0 represents the L0 norm,||Q|| 2，0 represents the number of all non-zero rows in the matrix Q, ||Q||0 and ||E||0 represent the number of all non-zero elements in the matrices Q and E, respectively, and ||Q|| 2，0 Used to preserve row / structure sparsity, ||Q||0, ||E||0 are used to preserve element sparsity; weighting parameters λ1, λ2, λ3>0, X=PQ T X+E represents the PCA decomposition of the data set matrix X;

[0020] Among them, Q represents the projection matrix, which satisfies the following formula:

[0021]

[0022] Among them, S satisfies the following formula,

[0023] S=(S W -μS b )

[0024] Among them, the parameter μ>0, S b represents the inter-class scatter matrix, S W represents the within-class scatter matrix.

[0025] Furthermore, the data analysis and processing using the RSLDA+ method also includes:

[0026] Use the alternating direction multiplication method to introduce the auxiliary term matrix F, and rewrite the equations satisfied by the three parameter matrices P, Q, and E into equations satisfied by the four matrices P, Q, E, and F.

[0027]

[0028]

[0029] Among them, ||F||0 represents the L0 norm, which indicates the number of all non-zero elements in the matrix F.

[0030] Furthermore, the data analysis and processing using the RSLDA+ method also includes:

[0031] Based on the equations satisfied by the four matrices P, Q, E, and F, the augmented Lagrangian function is constructed. The function is expressed as follows:

[0032]

[0033] Among them, Λ1Λ2 is the augmented Lagrange multiplier, the penalty parameters β1, β2>0, <·> represents the matrix inner product, Represents the square of the F-norm of the matrix.

[0034] Furthermore, the data analysis and processing using the RSLDA+ method also includes:

[0035] For the augmented Lagrangian function, the alternating direction multiplier method is used to design a program for solving it, and the calculation iterations are performed in sequence to update the values ​​of matrices P, Q, E, F, Λ1, and Λ2; the criterion for stopping the calculation iteration is that the relative error of the results of two calculation iterations is less than 10^-3 or the number of iterations reaches 100 times.

[0036] Further, performing classification of the image in the image classification module includes:

[0037] The KNN classifier is used to classify the results of the analysis and processing to obtain a classification result.

[0038] The present invention also provides an image classification system based on improved robust sparse linear discriminant analysis, the system comprising the following modules:

[0039] A data input module is configured to input an image data set and store it in a database;

[0040] A data feature extraction module is configured to extract an image data set from a database, wherein image features are extracted from the image data set;

[0041] A data analysis module, configured to input the extracted image features and perform data analysis using the RSLDA+ method; and

[0042] The image classification module is configured to input the result of the analysis process and perform classification of the image therein.

[0043] The present invention also provides a computer system, comprising:

[0044] a processor configured to execute machine-readable instructions; and

[0045] The memory stores machine-readable instructions, which, when executed by the processor, perform the steps of the classification method.

[0046] The present invention also provides a computer-readable storage medium on which machine-readable instructions are stored. When the machine-readable instructions are executed by a processor, the steps of the classification method are executed.

[0047] Compared with the prior art, the present invention has the following advantages:

[0048] 1. Based on the mathematical model of extracting image features, the present invention starts from the fundamental sparse features and combines the linear discriminant analysis method to realize a new non-convex and non-smooth processing method for image classification, which not only extracts the global features of the image but also retains the local structure of the image, making the image classification processing more accurate.

[0049] 2. The present invention provides a feasible solution with better robustness and better sparsity for the classification of image data by obtaining a local optimal closed solution. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] To further illustrate the above and other advantages and features of various embodiments of the present invention, a more specific description of various embodiments of the present invention will be presented with reference to the accompanying drawings. It will be understood that these drawings only depict typical embodiments of the present invention and are therefore not to be considered as limiting the scope thereof. In the accompanying drawings, for clarity, identical or corresponding parts will be represented by identical or similar reference numerals.

[0051] Figure 1A schematic diagram of a flow chart of an image classification method according to an embodiment of the present invention;

[0052] Figure 2 A schematic diagram of a process of solving an augmented Lagrangian function using an alternating direction multiplier method according to an embodiment of the present invention;

[0053] Figure 3 is a schematic diagram of a first set of image data sets according to an embodiment of the present invention;

[0054] Figure 4 is a schematic diagram of a second set of image data sets according to an embodiment of the present invention; and

[0055] Figure 5 FIG. 4 is a schematic diagram of an image classification system according to an embodiment of the present invention. DETAILED DESCRIPTION

[0056] In the following description, the present invention is described with reference to various embodiments. However, those skilled in the art will recognize that various embodiments may be implemented without one or more specific details or with other replacement and / or additional methods or components. In other cases, well-known structures or operations are not shown or described in detail to avoid obscuring the inventive point of the present invention. Similarly, for the purpose of explanation, specific numbers and configurations are set forth to provide a comprehensive understanding of embodiments of the present invention. However, the present invention is not limited to these specific details.

[0057] In this specification, reference to "one embodiment" or "the embodiment" means that a particular feature, structure, or characteristic described in conjunction with the embodiment is included in at least one embodiment of the present invention. The phrase "in one embodiment" appearing in various places in this specification does not necessarily all refer to the same embodiment.

[0058] It should be noted that the embodiments of the present invention describe the method steps in a specific order, but this is only for the purpose of illustrating the specific embodiment, rather than limiting the order of the steps. On the contrary, in different embodiments of the present invention, the order of the steps can be adjusted according to actual needs.

[0059] In the present invention, each module of the system according to the present invention can be implemented using software, hardware, firmware or a combination thereof. When the module is implemented using software, the function of the module can be implemented by a computer program flow, for example, the module can be implemented by a code segment (such as a code segment of a language such as C, C++) stored in a storage device (such as a hard disk, a memory, etc.), wherein the corresponding function of the module can be implemented when the code segment is executed by a processor. When the module is implemented using hardware, the function of the module can be implemented by setting a corresponding hardware structure, for example, the function of the module can be implemented by hardware programming a programmable device such as a field programmable gate array (FPGA), or the function of the module can be implemented by designing an application-specific integrated circuit (ASIC) including electronic devices such as a plurality of transistors, resistors and capacitors. When the module is implemented using firmware, the function of the module can be written into a read-only memory such as an EPROM or EEPROM of the device in the form of a program code, and the corresponding function of the module can be implemented when the program code is executed by a processor. In addition, certain functions of the module may need to be implemented by separate hardware or by collaboration with the hardware, for example, the detection function is implemented by corresponding sensors (such as proximity sensors, acceleration sensors, gyroscopes, etc.), the signal transmission function is implemented by corresponding communication devices (such as Bluetooth devices, infrared communication devices, baseband communication devices, Wi-Fi communication devices, etc.), the output function is implemented by corresponding output devices (such as displays, speakers, etc.), and so on.

[0060] The most critical issue in image classification is to reduce the dimensionality of the image data matrix to extract the most representative features and use a small number of the most useful features to characterize different types of images to complete the image classification.

[0061] The RSLDA+ (improved robust sparse linear discriminant analysis) method proposed in the present invention is an improvement on the RSLDA (Robust Sparse Linear Discriminant Analysis) method. RSLDA targets the regularized model of the L1 norm, i.e., the approximate problem; while the RSLDA+ proposed in the present invention targets the L0 norm, i.e., the original problem. Compared with the approximate method, the L0 norm can more accurately characterize the sparsity and has a better feature extraction effect.

[0062] The RSLDA+ method proposed in the present invention is based on the Linear Discriminant Analysis (LDA) algorithm, and comprehensively constrains the element sparsity and row sparsity of the data set and noise items, which can make the best sparse classification of data sample features and maximize the shielding of noise items. Therefore, in the case of high-complexity and strong interference image classification, the classification algorithm can have stronger anti-noise ability, i.e. high robustness, more effective image classification accuracy, i.e. good sparsity, and faster image classification processing, i.e. fast running speed.

[0063] The technical solutions in the embodiments of the present invention are described clearly and completely below in conjunction with the accompanying drawings in the embodiments of the present invention. Figure 1 FIG. 1 is a flow chart of an image classification method according to an embodiment of the present invention. Figure 1 As shown, the image classification method based on improved robust sparse linear discriminant analysis includes the following steps:

[0064] First, the image data set is input into the data input module and stored in the database.

[0065] Next, extract the image data set from the database, and extract image features from the image data set in the data feature extraction module. Extracting image features from the image data set in the data feature extraction module includes: using principal component analysis to decompose image information, decomposing the main data features and noise items of the image. PCA (Principal Component Analysis) is a commonly used linear dimensionality reduction method, which aims to map high-dimensional data to a low-dimensional space for expression through linear mapping, and expects the variance of the data to be maximized in the projected dimension, thereby retaining more characteristics of the original data using fewer data dimensions.

[0066] Next, the extracted image features are input into the data analysis module, and the RSLDA+ method is used to analyze and process the data. The RSLDA+ method proposed in the present invention is based on the LDA algorithm. Linear discriminant analysis LDA (Linear Discriminant Analysis) is also called Fisher linear discriminant, which is widely used in the field of pattern recognition (such as face recognition, ship recognition and other graphic image recognition fields).

[0067] The core of the LDA algorithm is the Fisher criterion, that is, the classification of samples is to find a projection matrix Q so that the inter-class distance of all samples after projection is maximized and the intra-class distance is minimized. This means that the data is projected onto a low dimension, and the projection points of the same category data are as close as possible after projection, and the center points of the projection points of different category data are as far as possible. The formula satisfied by the Fisher criterion is as follows,

[0068]

[0069]

[0070] Among them, min represents the minimum value of the function, st represents the constraint, I represents the identity matrix, Tr represents the trace of the matrix, T represents the matrix transpose, Q represents the projection matrix, parameter μ>0, S b represents the inter-class scatter matrix, S W represents the within-class scatter matrix.

[0071] The inter-class scatter matrix S b and the intra-class scatter matrix S W It is expressed by the following formula:

[0072]

[0073]

[0074] in, represents the i-th sample in the j-th class, represents the average value of the sample vector in the jth class, represents the average value of all sample vectors, n represents the number of samples in the X matrix, c represents the total number of classes of samples, n j represents the number of samples in the jth class.

[0075] and It is expressed by the following formula:

[0076]

[0077]

[0078] Where n represents the number of samples in the X matrix, and each sample has m features, that is, the X matrix is ​​a data set with m rows and n columns. n can be expressed as follows:

[0079]

[0080] Introduce parameter S, which is composed of the inter-class scatter matrix S b and the intra-class scatter matrix S W express,

[0081] S=(S W -μS b )

[0082] The RSLDA+ method proposed in the present invention is a further optimization and improvement of the LDA classification method. The LDA method and derived methods such as SLDA, RSLDA, etc. all introduce the L1 norm for regularization, while the RSLDA+ method is an innovation in the regularization term. The RSLDA+ method constructs the most original image classification feature extraction mathematical model by using the L0 norm regularization term, maintains element sparsity and structural sparsity respectively, and represents and restores image features from both local and global aspects. The RSLDA+ method effectively improves the interpretability of the image feature matrix, can significantly improve the accuracy of image classification, makes the desired classification results more sparse and robust, and can eliminate the influence of noise and a small number of outliers to the greatest extent.

[0083] The image data processing problem constructed by the RSLDA+ method is converted into the problem of solving the optimal values ​​of the three parameter matrices P, Q, and E. The three parameter matrices P, Q, and E satisfy the following formulas:

[0084]

[0085]

[0086] Among them, min represents the minimum value of the function, st represents the constraint, I represents the identity matrix, Tr represents the trace of the matrix, T represents the matrix transpose, the P matrix is ​​the linear transformation matrix of the original data set, Q represents the projection matrix, and E represents the noise term matrix; ||Q|| 2,0 ,||Q||0,||E||0 represents the L0 norm,||Q|| 2,0 represents the number of all non-zero rows in the matrix Q, ||Q||0 and ||E||0 represent the number of all non-zero elements in the matrices Q and E, respectively, and ||Q|| 2,0 Used to preserve row / structure sparsity, ||Q||0, ||E||0 are used to preserve element sparsity; weighting parameters λ1, λ2, λ3>0, X=PQ T X+E represents the PCA decomposition of the data set matrix X.

[0087] By using the Alternating Direction Method of Multipliers (ADMM) to introduce the auxiliary term matrix F, the equations satisfied by the three parameter matrices P, Q, and E can be rewritten as equations satisfied by the four matrices P, Q, E, and F.

[0088]

[0089]

[0090] Among them, ||F||0 represents the L0 norm, which indicates the number of all non-zero elements in the matrix F.

[0091] Based on the equations satisfied by the four matrices P, Q, E, and F, the augmented Lagrangian function is constructed. The function is expressed as follows:

[0092]

[0093] Among them, Λ1 and Λ2 are augmented Lagrange multipliers, penalty parameters β1 and β2>0, <·> represents the inner product of the matrix, Represents the square of the F-norm of the matrix.

[0094] Figure 2 FIG. 1 is a flow chart of solving the augmented Lagrangian function using the alternating direction multiplier method according to an embodiment of the present invention. Figure 2 As shown, for the augmented Lagrangian function, the alternating direction multiplier method is used to design a program for solving, and the iterative calculations are performed in sequence to update the values ​​of matrices P, Q, E, F, Λ1, and Λ2; the criterion for stopping the calculation iteration is that the relative error of the results of two calculation iterations is less than 10^-3 or the number of iterations reaches 100 times.

[0095] Finally, the result of the analysis process is input into the image classification module, and the image classification is performed in the image classification module. The result of the analysis process is classified using a KNN (k-nearest neighbor) classifier to obtain a classification result.

[0096] The effect of the present invention can be further illustrated by the following experimental results of image classification.

[0097] The following example illustrates the effect of the image classification method based on improved robust sparse linear discriminant analysis provided by the present invention on the accuracy of image classification. Figure 3 FIG. 1 is a schematic diagram of a first set of image data sets according to an embodiment of the present invention. Figure 3 As shown, the first set of image datasets contains a total of 9146 objects from all over the world divided into 101 categories. Figure 4 FIG. 1 is a schematic diagram of a second set of image data sets according to an embodiment of the present invention. Figure 4 As shown, the second set of image data sets is a data set of 2000 images of 10 different types of cars with various complex backgrounds. For the classification experiments of the two sets of image data sets, the existing LDA, SLDA, SULDA, RSLDA, RSMDA, RSLDA+IIKC, SUWLDA methods and the RSLDA+ method proposed in the present invention are used for image classification, and the image classification experimental results of different methods are obtained.

[0098] Table 1 shows the classification accuracy (%) of different methods for the first set of image data sets. As shown in Table 1, the image classification experiment randomly selected 10, 15, 20, and 25 images from the first set of image data sets for classification using different image classification methods. As shown in Table 1, when classifying the same number of images, the RSLDA+ method proposed in the present invention has a higher classification accuracy than the existing LDA, SLDA, SULDA, RSLDA, RSMDA, RSLDA+IIKC, and SUWLDA methods.

[0099] Table 1 Classification accuracy for the first set of image datasets (%)

[0100]

[0101] Table 2 shows the classification accuracy (%) of different methods for the second set of image data sets. As shown in Table 2, the image classification experiment randomly selected 10, 15, 20, and 25 images from the second set of image data sets for classification using different image classification methods. As shown in Table 2, when classifying the same number of images, the RSLDA+ method proposed in the present invention has a higher classification accuracy than the existing LDA, SLDA, SULDA, RSLDA, RSMDA, RSLDA+IIKC, and SUWLDA methods.

[0102] Table 2 Classification accuracy for the second set of image data sets (%)

[0103]

[0104] Table 3 shows the classification accuracy (%) of different methods after adding pepper noise to the second group of image data sets. As shown in Table 3, the image classification experiment randomly selected 25, 50, 75, and 100 pictures from the second group of image data sets for classification using different image classification methods, and added different degrees (10%, 30%, 50%, and 70%) of pepper noise. As shown in Table 3, when classifying the same number of pictures under the same degree of pepper noise, the RSLDA+ method proposed in the present invention has a higher classification accuracy than the existing LDA, SLDA, SULDA, RSLDA, RSMDA, RSLDA+IIKC, and SUWLDA methods.

[0105] Table 3 Classification accuracy after adding pepper noise to the second set of image data sets (%)

[0106]

[0107] The experimental results of the image data set classification in Table 1, Table 2 and Table 3 show that when facing high-dimensional, strong-noise, and multi-complexity image classification problems, the RSLDA+ method proposed in the present invention has a more obvious classification accuracy advantage than other traditional linear discriminant analysis methods.

[0108] The RSLDA+ method proposed in the present invention further optimizes and improves the LDA classification method. By using the L0 norm regularization term, the most original image classification feature extraction mathematical model is constructed, and the element sparsity and structural sparsity are maintained respectively, and the image features are restored from both local and global aspects. The RSLDA+ method proposed in the present invention effectively improves the interpretability of the image feature matrix, can significantly improve the accuracy of image classification, makes the required classification results more sparse and robust, and can eliminate the influence of noise and a small number of outliers to the maximum extent.

[0109] The invention also provides an image classification system based on improved robust sparse linear discriminant analysis. Figure 5 FIG. 1 is a schematic diagram of an image classification system according to an embodiment of the present invention. Figure 5 As shown, the system includes the following modules:

[0110] A data input module is configured to input an image data set and store it in a database;

[0111] A data feature extraction module is configured to extract an image data set from a database, wherein image features are extracted from the image data set;

[0112] A data analysis module, configured to input the extracted image features and perform data analysis using the RSLDA+ method; and

[0113] The image classification module is configured to input the result of the analysis process and perform classification of the image therein.

[0114] In one embodiment, a computer system is also provided, including a memory and a processor, the memory storing machine-readable instructions, and the processor implementing the following processing steps when executing the machine-readable instructions: inputting an image data set into a data input module and storing it in a database; extracting the image data set in the database, and extracting image features from the image data set in a data feature extraction module; inputting the extracted image features into a data analysis module, and performing data analysis and processing using the RSLDA+ method; and inputting the results of the analysis and processing into an image classification module, and performing image classification in the image classification module.

[0115] The memory includes: a USB flash drive, a read-only memory (ROM), a random access memory (RAM), a mobile hard disk, a magnetic disk or an optical disk, and other media that can store machine-readable instructions.

[0116] It can be understood that in addition to the memory and processor mentioned above, the above-mentioned computer system also includes other software and hardware components not listed in this specification. The specific components can be determined according to the model of specific data processing equipment in different application scenarios, and this specification will not list them one by one in detail.

[0117] In one embodiment, a computer-readable storage medium is also provided, on which machine-readable instructions are stored. When the machine-readable instructions are executed by a processor, the following processing steps are implemented: inputting an image data set into a data input module and storing it in a database; extracting an image data set from the database, and extracting image features from the image data set in a data feature extraction module; inputting the extracted image features into a data analysis module, and performing data analysis and processing using the RSLDA+ method; and inputting the results of the analysis and processing into an image classification module, and performing image classification in the image classification module.

[0118] Although various embodiments of the present invention are described above, it should be understood that they are presented as examples only and not as limitations. It is obvious to those skilled in the relevant art that various combinations, modifications and changes can be made thereto without departing from the spirit and scope of the present invention. Therefore, the breadth and scope of the present invention disclosed herein should not be limited by the exemplary embodiments disclosed above, but should only be defined according to the attached claims and their equivalents.

Claims

1. An image classification method based on improved robust sparse linear discriminant analysis, characterized in that: The steps include: Input the image data set in the data input module and store it in the database; Extracting an image data set from a database, and extracting image features from the image data set in a data feature extraction module; Input the extracted image features into a data analysis module, and use the RSLDA+ method to perform data analysis and processing; as well as Inputting the result of the analysis process into an image classification module, and performing image classification in the image classification module; The data analysis and processing using RSLDA+ method includes: The image data processing problem constructed by the RSLDA+ method is converted into the problem of solving the optimal values ​​of the three parameter matrices P, Q, and E. The three parameter matrices P, Q, and E satisfy the following formulas: Where min represents the minimum value of the function, st represents the constraint, I represents the identity matrix, Tr represents the trace of the matrix, T represents the matrix transpose, P matrix is ​​the linear transformation matrix of the original data set, and E represents the noise term matrix; ||Q||2,0, ||Q||0, ||E||0 represent the L0 norm, ||Q||2,0 represents the number of all non-zero rows in the matrix Q, ||Q||0, ||E||0 represent the number of all non-zero elements in the matrices Q and E respectively, ||Q||1,0 is used to preserve row / structure sparsity, ||Q||0, ||E||0 are used to preserve element sparsity; weighting parameters λ1, λ2, λ3>0, X=PQ T X+E represents the PCA decomposition of the data set matrix X; Among them, Q represents the projection matrix, which satisfies the following formula: Among them, S satisfies the following formula, S=(S W -μS b ) Among them, the parameter μ>0, S b represents the inter-class scatter matrix, S W represents the within-class scatter matrix; The data analysis and processing using the RSLDA+ method also includes: Use the alternating direction multiplication method to introduce the auxiliary term matrix F, and rewrite the equations satisfied by the three parameter matrices P, Q, and E into equations satisfied by the four matrices P, Q, E, and F. Among them, ||F||0 represents the L0 norm, which indicates the number of all non-zero elements in the matrix F; The data analysis and processing using the RSLDA+ method also includes: Based on the equations satisfied by the four matrices P, Q, E, and F, the augmented Lagrangian function is constructed. The function is expressed as follows: Among them, Λ1 and Λ2 are augmented Lagrange multipliers, penalty parameters β1 and β2>0, <·> represents the inner product of the matrix, Represents the square of the Frobenius norm of the matrix.

2. The image classification method based on improved robust sparse linear discriminant analysis according to claim 1, characterized in that: Extracting image features from the image data set in the data feature extraction module includes: The principal component analysis method is used to decompose the image information and decompose the main data features and noise items of the image.

3. The image classification method based on improved robust sparse linear discriminant analysis according to claim 1, characterized in that: The data analysis and processing using the RSLDA+ method also includes: For the augmented Lagrangian function, the alternating direction multiplier method is used to design a program for solving it, and the calculation iterations are performed in sequence to update the values ​​of matrices P, Q, E, F, Λ1, and Λ2; the criterion for stopping the calculation iteration is that the relative error of the results of two calculation iterations is less than 10^-3 or the number of iterations reaches 100 times.

4. The image classification method based on improved robust sparse linear discriminant analysis according to claim 1, characterized in that: The image classification module performs image classification including: The KNN classifier is used to classify the results of the analysis and processing to obtain a classification result.

5. A system for the image classification method based on improved robust sparse linear discriminant analysis according to any one of claims 1 to 4, characterized in that: Includes the following modules: A data input module is configured to input an image data set and store it in a database; A data feature extraction module is configured to extract an image data set from a database, wherein image features are extracted from the image data set; A data analysis module is configured to input the extracted image features and perform data analysis and processing using the RSLDA+ method; as well as The image classification module is configured to input the result of the analysis process and perform classification of the image therein.

6. A computer system, characterized in that: include: a processor configured to execute machine-readable instructions; and A memory storing machine-readable instructions, which, when executed by a processor, perform the steps of the method according to any one of claims 1 to 4.

7. A computer-readable storage medium, characterized in that: Machine readable instructions are stored thereon, and when the machine readable instructions are executed by a processor, the steps of the method according to any one of claims 1 to 4 are performed.

Images