An incomplete multi-view image segmentation method, system, device and storage medium

Abstract

The application discloses an incomplete multi-view image segmentation method, system, device and storage medium, comprising the following processes: acquiring an incomplete multi-view feature matrix of an image to be segmented; defining an anchor point matrix and an anchor point graph, and minimizing the error between the representation of the anchor point matrix and the anchor point graph and the incomplete multi-view feature matrix by using an adaptive norm; in an anchor point graph self-learning framework, different learnable weights are given to the anchor point graph self-learning framework in different views; the anchor point graph is subjected to one-sided orthogonal non-negative matrix factorization, integrated into a target framework, and the target framework is subjected to iterative convergence to obtain an updated coefficient matrix; and a segmentation result of the image to be segmented is extracted from the coefficient matrix. The efficiency and performance of a large-scale multi-view image segmentation task are greatly improved.

Description

TECHNICAL FIELD

[0001] The present application belongs to the field of image processing, and relates to an incomplete multi-view image segmentation method, system, device and storage medium. BACKGROUND

[0002] In recent years, image segmentation algorithms based on multi-view clustering have received extensive attention due to their ability to integrate more abundant image information, and have been successfully applied to different fields such as remote sensing analysis, autonomous driving and medical diagnosis. However, the multi-view features are often incomplete due to uncontrollable factors during image acquisition, making it difficult to apply existing segmentation methods based on complete multi-view clustering. At the same time, with the rapid increase in the number of images to be processed in the real world, multi-view data gradually exhibits large-scale and high-dimensional characteristics. Traditional multi-view clustering methods are difficult to effectively process these large-scale and high-dimensional multi-view data due to high time and space overhead. How to deal with incompleteness while improving the performance and efficiency of large-scale incomplete multi-view image segmentation methods has become an important research problem in the field of information technology. In the present application, we mainly focus on incomplete multi-view data of images, and based on the theory of incomplete multi-view clustering, we propose an efficient incomplete multi-view image segmentation method and system.

[0003] Current segmentation methods for large-scale incomplete multi-view image data mainly fall into two categories: the first category is based on anchor graph, that is, after obtaining multi-view features from images, a small number of anchors are generated, an anchor graph is constructed between the original data and the anchors, and then spectral clustering is performed. This method has two different ways of generating anchors. One is pre-defined anchor generation, that is, a small number of anchors are generated by using methods such as mean value, and then an anchor graph is constructed based on this. However, this method often causes a two-stage mismatch between anchor generation and anchor graph construction, resulting in a decrease in segmentation accuracy. The second way is to simultaneously update the anchors and the anchor graph through self-learning, which solves the two-stage mismatch problem and improves the effectiveness of segmentation. The second category of methods for large-scale incomplete data is based on accelerated matrix decomposition, which operates on the original data, decomposes it into the product of a small-scale coefficient matrix and a basis matrix, and then performs mean clustering on the obtained coefficient matrix to obtain the segmentation result. Compared with traditional multi-view image segmentation methods, these methods have significantly reduced computational complexity, improving the efficiency of multi-view image segmentation and enabling them to be used for large-scale tasks.

[0004] ​​However, the efficiency of these two approaches in large-scale image segmentation tasks still needs to be improved. For the first type of anchor graph-based methods, constructing an anchor graph instead of the traditional full-sample graph reduces computational complexity, but the time and space overhead of spectral clustering on the anchor graph remains high. For the second type of matrix decomposition-based methods, directly decomposing the raw data becomes time-consuming when faced with high-dimensional data, meaning that segmentation efficiency is constrained by large-scale matrix multiplications. Summary of the Invention

[0005] The purpose of the present invention is to overcome the shortcomings of the above-mentioned prior art and provide an incomplete multi-view image segmentation method, system, device and storage medium, which greatly improves the efficiency and performance of large-scale multi-view image segmentation tasks.

[0006] In order to achieve the above object, the present invention adopts the following technical solutions:

[0007] An incomplete multi-view image segmentation method includes the following steps:

[0008] S01, obtaining an incomplete multi-view feature matrix of the image to be segmented;

[0009] S02, define the anchor matrix and anchor graph, use the adaptive norm to minimize the error between the representation of the anchor matrix and anchor graph and the incomplete multi-view feature matrix, and obtain the anchor graph self-learning framework;

[0010] S03, assigning weights of different perspectives to the anchor graph self-learning framework to obtain a weighted anchor graph self-learning framework;

[0011] S04, performing unilateral orthogonal non-negative matrix factorization on the anchor graph to obtain an anchor graph decomposition framework, integrating the weighted anchor graph self-learning framework and the anchor graph self-learning framework into a target framework, iteratively converging the target framework to obtain an updated coefficient matrix;

[0012] S05, extracting the segmentation result of the image to be segmented from the coefficient matrix.

[0013] Preferably, in S1, the image to be segmented is in the form of an incomplete multi-view image data set, a missing sample indication matrix is ​​defined, the incomplete multi-view image data set is multiplied by the missing indication matrix, and the missing rows or columns caused by the incomplete multi-view image data set are removed to obtain an ordered incomplete multi-view feature matrix.

[0014] Preferably, in S2, an anchor matrix independent of each perspective and an anchor graph matrix consistent with all perspectives are defined, and an anchor graph consistent with all perspectives is defined. Under the constraint of an adaptive norm, each perspective independent anchor point and an anchor graph consistent with the perspective are obtained by self-learning. The anchor matrix, the anchor graph and the missing indicator matrix are multiplied to obtain the representation of the anchor matrix and the anchor graph.

[0015] Preferably, in S3, a self-learning weight and a power parameter are added to the anchor graph self-learning framework for different view angles.

[0016] Preferably, in S4, a single-sided orthogonal non-negative matrix factorization is used to reconstruct the anchor graph by using a norm constraint decomposition.

[0017] Preferably, in S4, each variable in the target framework is updated one by one by an iterative optimization algorithm until the value of the target framework converges when the change is less than a threshold value, and an updated coefficient matrix is obtained.

[0018] Preferably, in S5, a mean clustering is used to extract a segmentation result of the multi-view image data set from the coefficient matrix.

[0019] An incomplete multi-view image segmentation system, comprising:

[0020] A feature acquisition module is configured to acquire an incomplete multi-view feature matrix of an image to be segmented.

[0021] An anchor graph learning module is configured to define an anchor matrix and an anchor graph, minimize the error between the representation of the anchor matrix and the anchor graph and the incomplete multi-view feature matrix by using an adaptive norm, and obtain an anchor graph self-learning framework.

[0022] A weight assignment module is configured to assign a weight to the anchor graph self-learning framework for different view angles, and obtain a weighted anchor graph self-learning framework.

[0023] A target framework module is configured to perform a single-sided orthogonal non-negative matrix factorization on the anchor graph to obtain an anchor graph decomposition framework, integrate the weighted anchor graph self-learning framework and the anchor graph self-learning framework into a target framework, and perform an iterative convergence on the target framework to obtain an updated coefficient matrix.

[0024] A segmentation result acquisition module is configured to extract a segmentation result of the image to be segmented from the coefficient matrix.

[0025] A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the incomplete multi-view image segmentation method when executing the computer program.

[0026] A computer-readable storage medium, storing a computer program, wherein the computer program is executable by a processor to implement the steps of the incomplete multi-view image segmentation method.

[0027] Compared with the prior art, the present application has the following beneficial effects: ​​

[0028] The application integrates anchor point learning based on adaptive norm and anchor point graph decomposition into a unified clustering framework, which can effectively improve the clustering efficiency and give the model the characteristics of global anchor point adaptive adjustment and overall robustness with the help of adaptive norm; single-sided orthogonal non-negative matrix decomposition is used to decompose the anchor point graph to obtain the coefficient matrix, which can effectively reduce the size of the matrix and further improve the framework performance, and can be robustly applied to the actual noisy image segmentation task, greatly improving the efficiency and performance of large-scale multi-view image segmentation task. BRIEF DESCRIPTION OF DRAWINGS

[0029] Figure 1 Flow chart of the incomplete multi-view image segmentation method of the application;

[0030] Figures 2a-2c The figure is the change of each segmentation index with the parameter of the application.

[0031] Figures 3a-3c The figure is the convergence of the three incomplete multi-view image data sets to be clustered after iterative optimization of the application;

[0032] Figures 4a-4c The figure is the visualization of the segmentation results of the application with the number of iterations increasing in turn;

[0033] Figures 5a-5i The figure is the accuracy comparison of the method of the application and other image segmentation methods;

[0034] Figures 6a-6c The figure is the efficiency comparison of the method of the application and other image segmentation methods. DETAILED DESCRIPTION

[0035] The embodiments of the application are described in detail below, and examples of the embodiments are shown in the drawings, wherein the same or similar reference numerals represent the same or similar elements or elements having the same or similar functions throughout. The embodiments described below by referring to the drawings are exemplary and are only used to explain the application, and cannot be understood as a limitation of the application.

[0036] The following disclosure provides many different embodiments or examples for implementing different structures of the application. In order to simplify the disclosure of the application, the components and settings of specific examples are described below. Of course, they are only examples and the purpose is not to limit the application. In addition, the application can repeatedly refer to numerals and / or reference letters in different examples, and such repetition is for the purpose of simplification and clarity, which itself does not indicate the relationship between the various embodiments and / or settings discussed. In addition, the application provides examples of various specific processes and materials, but those skilled in the art can realize the application of other processes and / or the use of other materials.

[0037] As Figure 1As shown in FIG, the incomplete multi-view image segmentation method based on global adaptive anchor graph decomposition according to the present invention includes the following steps:

[0038] Step S01: Establish a machine learning model to obtain incomplete multi-view features of the image to be segmented, where the image to be segmented is in the form of an incomplete multi-view image dataset.

[0039] Feeding an incomplete multi-view image dataset into a machine learning model , in order to deal with the missing part of the data matrix caused by incomplete features, for the input incomplete multi-view image dataset ,definition Is a given indicator vector containing the index of the existing sample of the v-th perspective. On this basis, the missing sample indicator matrix is ​​defined ,in represents the number of samples, It is The number of instances where the viewpoints are not missing, The specific definition is as follows:

[0040]

[0041] Incomplete multi-view image dataset and missing indicator matrix Multiply to get , we can get the processed ordered incomplete multi-view feature matrix, which contains incomplete multi-view features. By introducing the missing indicator matrix To remove the missing rows or columns caused by incomplete multi-view image datasets.

[0042] Step S02: Define the anchor matrix and anchor graph, multiply the anchor matrix, anchor graph and missing indicator matrix to obtain the representation of the anchor matrix and anchor graph, use the adaptive norm to minimize the error between the representation of the anchor matrix and anchor graph and the incomplete multi-view feature matrix, and under the constraint of the adaptive norm, self-learn to obtain the independent anchor points of each view and the anchor graph with consistent view, thereby improving the robustness of the machine learning model.

[0043] In some embodiments, a per-view independent anchor matrix and an anchor matrix consistent across all views are defined, where the anchor matrix is ​​defined as: ,in Indicates the The number of dimensions of the perspective data, Expressed as the number of anchor points; define the anchor point map with consistent viewpoints , the anchor matrix, anchor graph and missing indicator matrix are multiplied to obtain the representation of the anchor matrix and anchor graph. The anchor graph self-learning framework with adaptive norm constraint can be expressed as:

[0044]

[0045]

[0046] in represents the adaptive norm, is the parameter of the adaptive norm, Represents a vector whose elements are all 1.

[0047] Step S03: Assign weights to the anchor graph self-learning framework at different perspectives, and assign different importance to anchor learning at different perspectives.

[0048] Add self-learning weights and power parameters to anchor graphs of different perspectives. The self-learning framework of anchor graphs with weights is expressed as follows:

[0049]

[0050]

[0051] in , indicating the The anchor graph learning weights of each perspective are used to measure the different weights of different perspectives in the consistent anchor graph, and to give different importance to the anchor learning of different perspectives. is the power parameter of the weight, fixed to 2.

[0052] Step S04: performing unilateral orthogonal non-negative matrix factorization on the anchor graph to obtain a coefficient matrix for characterizing the sample category structure.

[0053] In some embodiments, a one-sided orthogonal non-negative matrix factorization is employed, using Norm Constrained Anchor Graph Matrix And the decomposed matrix is ​​used to reconstruct the anchor graph, reducing the number of matrix multiplication operations. The anchor graph decomposition framework is expressed as:

[0054]

[0055]

[0056] in express norm, is the coefficient matrix, is an orthogonal matrix.

[0057] The weighted anchor graph self-learning framework and the anchor graph decomposition framework are integrated into the target framework of the machine learning model, that is, the incomplete multi-view image segmentation framework. The complete target equation can be written as:

[0058]

[0059]

[0060] in is a hyperparameter that balances the self-learning anchor term and the anchor graph factorization term. The relationship between the changes in the three parameters and the changes in image segmentation indicators and segmentation time is as follows: Figures 2a-2c As shown in Figure 2, the line graph shows the impact of the three hyperparameters in this method on the effectiveness and efficiency of clustering. The hyperparameters are the number of anchor points, ( Figure 2a ), balance parameters ( Figure 2b ) and the adaptive norm parameter ( Figure 2c Clustering effectiveness is expressed by ACC, NMI, and Purity (%), and efficiency is expressed by time (s).

[0061] Each variable is updated one by one through the iterative optimization algorithm until the value of the target equation changes less than the threshold value and the model converges to obtain the updated coefficient matrix .

[0062] Step S05: Use Mean clustering,extracts segmentation results for incomplete multi-view image datasets from the coefficient matrix.

[0063] In some embodiments, the coefficient matrix that can characterize the sample category structure obtained by decomposing the full-view consistent anchor point graph is ,use The mean clustering method extracts the discrete categories corresponding to the samples and obtains the image segmentation results.

[0064] like Figures 3a-3c As shown in Figure 3, the three incomplete multi-view image datasets to be clustered have all been effectively converged after iterative optimization, proving that good convergence results are achieved for different incomplete multi-view image datasets to be clustered. Figures 4a-4c As shown in Figure 3, as the number of iterations increases, the segmentation effect becomes better and better.

[0065] The embodiment of the present application can effectively improve the efficiency of incomplete multi-view image segmentation and maintain high segmentation effectiveness, improve the multi-view image segmentation result under the premise of good robustness. Compared with other incomplete multi-view segmentation methods for image segmentation in the related art, the image segmentation effectiveness of the embodiment of the present application is higher, as shown in Figures 5a-5i . Figures 5a-5i The segmentation performance comparison of the embodiment of the present application and other advanced methods on different types, different sizes (conventional data sets and large-scale data sets) of real multi-view image data sets is shown, the indicators of methods that cannot complete segmentation within a specified time or run out of memory are represented as 0 in the column chart, and the effectiveness comparison of the present method and four comparison methods on three different missing rates corresponding to five image data sets is shown by a 3D column chart. Among them Figures 5a-5c , Figures 5d-5f and Figures 5g-5i Three groups represent 10%, 30%, and 50% sample missing rates, respectively, Figures 5a-5c , Figures 5d-5f and Figures 5g-5i Three groups show ACC, NMI and Purity indicators in turn, each image shows HCP-IMSC, MIMB, SEC-IMVC and HCLS as comparison methods, and MSRC, Yale, Mnist4, STL-10 and Animal as image data sets. Among different indicators, the segmentation effect of the embodiment of the present application is the highest.

[0066] The embodiment of the present application can significantly improve the efficiency of incomplete multi-view image segmentation, compared with other incomplete multi-view segmentation methods for image segmentation in the related art, the image segmentation efficiency of the embodiment of the present application is significantly improved, as shown in Figures 6a-6c . Figures 6a-6c The running time comparison of the embodiment of the present application and other advanced methods on different types, different sizes (conventional data sets and large-scale data sets) of real multi-view image data sets is shown, Figures 6a-6c the efficiency comparison of the present method and four comparison methods on MSRC, Yale and Caltech101-7 image data sets is shown in turn. Each image shows the running time (s) of each method under the condition of 10%, 30%, and 50% sample missing rate of the data set. The comparison methods are: HCP-IMSC, MIMB, SEC-IMVC and HCLS. Through the running time comparison, the segmentation efficiency of the embodiment of the present application is the highest.

[0067] It can be seen that the application is an incomplete multi-view image segmentation method based on global adaptive anchor point graph decomposition, anchor point graph decomposition is performed on the anchor point graph obtained by learning under the constraint of adaptive norm, the data size is reduced, the calculation complexity is reduced, and the clustering efficiency is greatly improved; meanwhile, a new self-learning anchor point graph framework based on adaptive norm constraint is provided, the robustness and global adaptive adjustment capability brought by adaptive norm are utilized to improve the effectiveness of clustering.

[0068] The following is a device embodiment of the application, which can be used to execute the method embodiment of the application. For details not covered in the device embodiment, please refer to the method embodiment of the application.

[0069] In another embodiment of the application, an incomplete multi-view image segmentation system is provided, which can be used to implement the above incomplete multi-view image segmentation method. Specifically, the incomplete multi-view image segmentation system comprises a feature acquisition module, an anchor point graph learning module, a weight assignment module, a target framework module and a segmentation result acquisition module.

[0070] The feature acquisition module is configured to acquire an incomplete multi-view feature matrix of an image to be segmented.

[0071] The anchor point graph learning module is configured to define an anchor point matrix and an anchor point graph, minimize the error between the representation of the anchor point matrix and the anchor point graph and the incomplete multi-view feature matrix by using adaptive norm, and obtain an anchor point graph self-learning framework.

[0072] The weight assignment module is configured to assign weights of different view features in the anchor point graph in the anchor point graph self-learning framework, and obtain a weighted anchor point graph self-learning framework.

[0073] The target framework module is configured to perform one-sided orthogonal non-negative matrix factorization on the anchor point graph, obtain an anchor point graph decomposition framework, integrate the weighted anchor point graph self-learning framework and the anchor point graph self-learning framework into a target framework, perform iterative convergence on the target framework, and obtain an updated coefficient matrix.

[0074] The segmentation result acquisition module is configured to extract a segmentation result of the image to be segmented from the coefficient matrix.

[0075] In still another embodiment of the present application, a terminal device is provided, which comprises a processor and a memory, the memory being configured to store a computer program, the computer program comprising program instructions, and the processor being configured to execute the program instructions stored in the computer storage medium. The processor can be a central processing unit (CPU), and can also be other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic, discrete hardware components, etc., which are the computing core and control core of the terminal, and are suitable for implementing one or more instructions, and are particularly suitable for loading and executing one or more instructions to implement a corresponding method flow or a corresponding function. The processor in the embodiments of the present application can be used for the operation of the incomplete multi-view image segmentation method, which comprises the following steps: S01, obtaining an incomplete multi-view feature matrix of an image to be segmented; S02, defining an anchor point matrix and an anchor point graph, and minimizing the error between the representation of the anchor point matrix and the anchor point graph and the incomplete multi-view feature matrix by using an adaptive norm to obtain an anchor point graph self-learning framework; S03, in the anchor point graph self-learning framework, assigning weights to different view features in the anchor point graph to obtain a weighted anchor point graph self-learning framework; S04, performing one-sided orthogonal non-negative matrix factorization on the anchor point graph to obtain an anchor point graph decomposition framework, integrating the weighted anchor point graph self-learning framework and the anchor point graph self-learning framework into a target framework, and performing iterative convergence on the target framework to obtain an updated coefficient matrix; and S05, extracting a segmentation result of the image to be segmented from the coefficient matrix.

[0076] In still another embodiment of the present application, a computer readable storage medium (Memory) is also provided, which is a memory device in the terminal device, and is configured to store programs and data. It can be understood that the computer readable storage medium herein can include an internal storage medium of the terminal device, and of course can also include an expansion storage medium supported by the terminal device. The computer readable storage medium provides a storage space, which stores an operating system of the terminal. Moreover, one or more instructions suitable for being loaded and executed by the processor are also stored in the storage space, and the instructions can be one or more computer programs (including program codes). It should be noted that the computer readable storage medium herein can be a high-speed RAM memory, or a non-volatile memory such as at least one disk memory.

[0077] The one or more instructions stored in the computer readable storage medium can be loaded and executed by the processor to implement the corresponding steps of the incomplete multi-view image segmentation method in the above embodiments; the one or more instructions stored in the computer readable storage medium are loaded and executed by the processor to perform the following steps: S01, obtaining an incomplete multi-view feature matrix of an image to be segmented; S02, defining an anchor point matrix and an anchor point graph, minimizing the error between the representation of the anchor point matrix and the anchor point graph and the incomplete multi-view feature matrix by using an adaptive norm, and obtaining an anchor point graph self-learning framework; S03, in the anchor point graph self-learning framework, assigning weights to different view features in the anchor point graph, and obtaining a weighted anchor point graph self-learning framework; S04, performing one-sided orthogonal non-negative matrix factorization on the anchor point graph, obtaining an anchor point graph decomposition framework, integrating the weighted anchor point graph self-learning framework and the anchor point graph self-learning framework into a target framework, and performing iterative convergence on the target framework to obtain an updated coefficient matrix; and S05, extracting a segmentation result of the image to be segmented from the coefficient matrix.

[0078] Those skilled in the art should 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-ROMs, optical storage, etc.) containing computer-usable program code.

[0079] The present application is described with reference to flowcharts and / or block diagrams of the methods, devices (systems), and computer program products 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, as well as combinations of 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 processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing apparatus to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing apparatus generate a means for implementing the functions specified in the flowcharts and / or block diagrams. Figure One The functions specified in one or more flows and / or blocks Figure One The functions specified in one or more flows and / or blocks

[0080] These computer program instructions can also be stored in a computer readable storage medium that can direct the computer or other programmable data processing apparatus to work in a specific manner, so that the instructions stored in the computer readable storage medium produce a product including an instruction apparatus that implements the functions specified in the flowcharts and / or block diagrams. Figure One The functions specified in one or more flows and / or blocks Figure One The functions specified in one or more flows and / or blocks

[0081] These computer program instructions can also be loaded into a computer or other programmable data processing devices, so that a series of operation steps are performed on the computer or other programmable devices to generate computer-implemented processes, thus the instructions executed on the computer or other programmable devices provide a process for implementing the functions specified in the flowchart Figure One one flow or multiple flows and / or one block or multiple blocks. Figure One the functions specified in the flowchart

[0082] The above-mentioned sequence numbers of the embodiments of the present application are only for description, and do not represent advantages or disadvantages of the embodiments.

[0083] In the above-described embodiments of the present application, the description of each embodiment has its own emphasis, and the parts not described in detail in a certain embodiment can be referred to the related description of other embodiments.

[0084] The above only describes the preferred embodiments of the present application, and it should be pointed out that, for ordinary skilled in the art, without departing from the principles of the present application, a number of improvements and refinements can be made, which should be considered as the protection scope of the present application.

[0085] It should be understood that the above description is for illustration only and not for limitation. Many embodiments and many applications other than those provided in the foregoing description will be apparent to those skilled in the art from the foregoing description. The scope of the patent should therefore not be determined with reference to the foregoing description, but instead should be determined with reference to the appended claims along with their full scope of equivalents. For purposes of completeness, all articles and references, including patents and patent documents, are incorporated herein by reference in their entirety. The disclosure of any patent, publication, or other disclosure provided herein is for the purpose of presenting the combined teachings of the present application as well as possible claims to the patent, publication or other disclosure as it existed at the time of the filing of this patent application, and should not be interpreted, in any way, to limit the further broad scope of features described herein, including any claims.

Claims

1. An incomplete multi-view image segmentation method, characterized in that: The following processes are included: S1, obtain the incomplete multi-view feature matrix of the image to be segmented; S2, define the anchor matrix and anchor graph, use the adaptive norm to minimize the error between the representation of the anchor matrix and anchor graph and the incomplete multi-view feature matrix, and obtain the anchor graph self-learning framework; Define an independent anchor matrix for each view and an anchor matrix consistent for all views, where the anchor matrix is ​​defined as: ,in Indicates the The number of dimensions of the perspective data, Expressed as the number of anchor points; define the anchor point map with consistent viewpoints , the anchor matrix, anchor graph and missing indicator matrix are multiplied to obtain the representation of the anchor matrix and anchor graph. The anchor graph self-learning framework with adaptive norm constraint can be expressed as: in represents the adaptive norm, is the parameter of the adaptive norm, represents a vector whose elements are all 1; S3, assigns different learnable weights to the anchor graph self-learning framework in different perspectives; Add self-learning weights and power parameters to anchor graphs of different perspectives. The self-learning framework of anchor graphs with weights is expressed as follows: in , indicating the The anchor graph learning weights of each perspective are used to measure the different weights of different perspectives in the consistent anchor graph, and to give different importance to the anchor learning of different perspectives. is the power parameter of the weight, fixed to 2; S4, performing unilateral orthogonal non-negative matrix factorization on the anchor graph to obtain the anchor graph decomposition framework; Using unilateral orthogonal non-negative matrix factorization, using Norm Constrained Anchor Graph Matrix And the decomposed matrix is ​​used to reconstruct the anchor graph, reducing the number of matrix multiplication operations. The anchor graph decomposition framework is expressed as: in express norm, is the coefficient matrix, is an orthogonal matrix; The weighted anchor graph self-learning framework and the anchor graph decomposition framework are integrated into the target framework, and the target framework is iteratively updated until convergence to obtain the updated coefficient matrix; S5, extracting the segmentation result of the image to be segmented from the coefficient matrix.

2. The incomplete multi-view image segmentation method according to claim 1, characterized in that: In S1, the image to be segmented is in the form of an incomplete multi-view image dataset. A missing sample indicator matrix is ​​defined, and the incomplete multi-view image dataset is multiplied by the missing indicator matrix. The missing rows or columns caused by the incomplete multi-view image dataset are removed to obtain an ordered incomplete multi-view feature matrix.

3. The incomplete multi-view image segmentation method according to claim 1, characterized in that: In S2, an anchor matrix independent of each view and an anchor map matrix consistent with all views are defined, and an anchor map consistent with all views is defined. Under the adaptive norm constraint, each view-independent anchor point and the anchor map consistent with the view are obtained by self-learning. The anchor matrix, anchor map and missing indicator matrix are multiplied to obtain the representation of the anchor matrix and anchor map.

4. The incomplete multi-view image segmentation method according to claim 1, characterized in that: In S3, self-learning weights and power parameters are added to the anchor graph self-learning framework.

5. The incomplete multi-view image segmentation method according to claim 1, characterized in that: In S4, a unilateral orthogonal non-negative matrix factorization is used, using Norm-constrained decomposition,performs low-rank reconstruction on anchor graphs.

6. The incomplete multi-view image segmentation method according to claim 1, characterized in that: In S4, the learnable variables in the target frame are updated one by one through the iterative optimization algorithm until the value change of the target frame is less than the threshold and converges to obtain the updated coefficient matrix.

7. The incomplete multi-view image segmentation method according to claim 1, wherein: In S5, Mean clustering,extracts segmentation results for multi-view image datasets from the coefficient matrix.

8. An incomplete multi-view image segmentation system, characterized in that: include: A feature acquisition module is used to obtain an incomplete multi-view feature matrix of the image to be segmented; Anchor graph learning module, which is used to define the anchor matrix and anchor graph, and uses adaptive norm to minimize the error between the representation of the anchor matrix and anchor graph and the incomplete multi-view feature matrix, thus obtaining the anchor graph self-learning framework; Define an independent anchor matrix for each view and an anchor matrix consistent for all views, where the anchor matrix is ​​defined as: ,in Indicates the The number of dimensions of the perspective data, Expressed as the number of anchor points; define the anchor point map with consistent viewpoints , the anchor matrix, anchor graph and missing indicator matrix are multiplied to obtain the representation of the anchor matrix and anchor graph. The anchor graph self-learning framework with adaptive norm constraint can be expressed as: in represents the adaptive norm, is the parameter of the adaptive norm, represents a vector whose elements are all 1; The weight assignment module is used to assign weights to the anchor graph self-learning framework at different perspectives to obtain a weighted anchor graph self-learning framework; Add self-learning weights and power parameters to anchor graphs of different perspectives. The self-learning framework of anchor graphs with weights is expressed as follows: in , indicating the The anchor graph learning weights of each perspective are used to measure the different weights of different perspectives in the consistent anchor graph, and to give different importance to the anchor learning of different perspectives. is the power parameter of the weight, fixed to 2; The target framework module is used to perform unilateral orthogonal non-negative matrix factorization on the anchor graph to obtain the anchor graph decomposition framework; Using unilateral orthogonal non-negative matrix factorization, using Norm Constrained Anchor Graph Matrix And the decomposed matrix is ​​used to reconstruct the anchor graph, reducing the number of matrix multiplication operations. The anchor graph decomposition framework is expressed as: in express norm, is the coefficient matrix, is an orthogonal matrix; The weighted anchor graph self-learning framework and the anchor graph self-learning framework are integrated into the target framework, and the target framework is iteratively converged to obtain the updated coefficient matrix; The segmentation result acquisition module is used to extract the segmentation result of the image to be segmented from the coefficient matrix.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the steps of the incomplete multi-view image segmentation method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the incomplete multi-view image segmentation method according to any one of claims 1 to 7 are implemented.

Images