Image noise mark feature selection method and system, storage medium and computer

Abstract

The invention provides an image noise mark feature selection method and system, a storage medium and a computer. The method comprises the steps of obtaining a to-be-processed image noise mark data set; embedding a sample set in the image noise mark data set into a multi-granularity fuzzy cluster to construct a dynamic fuzzy membership evaluation matrix; dynamically evolving a multi-level high-precision granular ball cluster; obtaining mark distribution with high identification degree; constructing a rough perception feature evaluation framework based on granular ball topology driving, extracting decision equivalence classes by combining rough set upper and lower approximation and extended positive domain models, and determining and measuring the contribution degree of each feature to a decision system by fusing multi-granular-ball decision boundary information based on a dependency degree quantitative model; a particle and ball structure consistency verification mechanism is introduced, and multi-level evaluation is carried out on the importance of the features through dependency and consistency. According to the method, the optimal feature subset with strong anti-noise performance and high discrimination capability is obtained, and stable and efficient input support is provided for a subsequent image noise mark learning model.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, in particular to an image noise label feature selection method, system, storage medium and computer. BACKGROUND

[0002] With the rapid development of large-scale multimedia data and computer vision technology, automatic annotation of image data has become a key task in practical applications. Due to the dependence on crowd-sourced annotation, online scraping or weakly supervised information to generate labels, there are often problems such as omission, error or redundancy in the labels of image data, forming a significant noise label phenomenon. Especially in the task of bird image recognition and classification, due to the large number of bird species, high similarity in appearance and complex habitat, the bird image labels obtained through the crowd-sourcing platform often have noise problems such as species mislabeling, subspecies confusion and juvenile bird label errors. These noise labels can interfere with the training process of classification, retrieval and other models, and affect the accuracy of applications such as bird protection, ecological monitoring and intelligent bird watching. Therefore, how to extract accurate and effective label information from noisy bird image data is a technical problem that needs to be solved in the field of weakly supervised learning.

[0003] To reduce the negative impact of noise labels on bird image learning tasks, one of the key technical problems is to perform effective feature selection on noisy bird image data. By extracting high-dimensional features from the original bird image data and then selecting the most relevant discriminative features to the true semantics using a feature selection method, the performance of subsequent visual tasks such as classification, retrieval or target detection can be significantly improved. However, existing bird image feature selection methods often rely on a single correlation measure or assume that the labels are completely accurate, making it difficult to adapt to the feature-label dependence distortion and label confusion problems caused by label noise.

[0004] At the same time, image data itself usually has complex nonlinear distribution characteristics and multi-level local structure characteristics. Traditional feature selection methods often ignore the local consistency and structure information between samples, resulting in insufficient resistance of the selected features to noise labels, thereby affecting the generalization ability of the model. Therefore, how to combine multi-level granular structure evolution and fuzzy label representation, fully utilize the topological properties and local stability of image data, and perform robust feature selection on noisy bird image samples has become a core technical difficulty that needs to be broken through in the field of image noise label learning. SUMMARY

[0005] Therefore, the purpose of the present application is to provide an image noise label feature selection method, system, storage medium and computer to at least solve the deficiencies in the above technical field.

[0006] The present application provides an image noise label feature selection method, which comprises: S1: Obtain the image noise labeled dataset to be processed, wherein the image noise labeled dataset contains both feature space and label space; S2: An adaptive multi-core fusion and dynamic parameter optimization mechanism is adopted to embed the sample set in the image noise labeling dataset into a multi-granularity fuzzy cluster to construct a dynamic fuzzy membership evaluation matrix and determine the multi-dimensional fuzzy mapping relationship between quantified sample-cluster and sample-class. S3: Based on the global spectrum feature mapping mechanism, a macroscopic coarse-grained particle skeleton is constructed. For particles with insufficient label isomorphism or structural entropy exceeding the threshold, a quality measurement model is integrated for joint evaluation. Through a progressive self-organizing fission algorithm, multi-level high-precision sub-particle clusters are dynamically evolved. S4: Within each fine sphere, weight-driven sphere calculation and fuzzy inference are integrated. Weights are automatically assigned based on the sample-sphere center distance, and a category-cluster coupling association matrix is ​​constructed. Through fuzzy synthesis operation, the cluster membership is optimized into category membership, and finally a highly recognizable label distribution is obtained. S5: Construct a rough perception feature evaluation framework based on particle topology-driven, define adaptive neighborhoods with particle geometric characteristics, dynamically divide cohesive and non-overlapping local neighborhoods, combine rough set upper and lower approximations and extended positive domain models to extract decision equivalence classes, and based on the dependency metric model, fuse multi-particle decision boundary information to determine the contribution of each feature to the decision system. S6: Introduces a particle-sphere structure consistency verification mechanism to measure the expression stability of features in the local domain of particles from the dimension of structure preservation, and to evaluate the importance of features at multiple levels through dependency and consistency.

[0007] Furthermore, step S1 includes: Using ResNet to extract features from noise-labeled image data, a dataset is obtained. ,in, For the first The feature vector of each sample For the first A set of candidate labels for each sample, for each candidate label element ,when When, it indicates a mark. For the sample Candidate tags; when When, it indicates a mark. With sample Irrelevant.

[0008] Furthermore, in step S2, a combination of random perturbation initialization and nondeterministic exploration strategies is used to initially partition the sample set in the image noise-labeled dataset. Step S2 includes: S21: In a multi-kernel framework, each kernel function is assigned dynamic weights, and an entropy regularization term is introduced into the objective function to constrain the weight distribution. ; in, Let represent the overall objective function of multi-kernel fuzzy clustering, used to simultaneously optimize clustering quality and the rationality of kernel weight allocation, where Represents the membership matrix. Represents the cluster center matrix. Represents the kernel weight allocation vector; This represents the preset number of clusters; This represents the total number of samples in the dataset. This represents the total number of kernel functions in a multi-kernel framework; in the fuzzy clustering process, Represents the original sample For the first The fuzzy membership degree of each cluster, which is between 0 and 1 and satisfies the normalization constraint. , The fuzziness index, typically greater than 1, is used to control the degree of fuzziness of cluster boundaries; in multi-kernel mapping mechanisms, Representing the The dynamic weight coefficients of each kernel function, the weights satisfying and Constraints; Indicates the first Each kernel function will convert the original sample The feature vector obtained after mapping to a high-dimensional feature space; Indicates the first The clustering in the th case is at the th case. Cluster centers in the feature space induced by each kernel function; distance term Calculate the squared Euclidean distance between a sample and its corresponding cluster center in a specific kernel space to measure the similarity between the sample and the cluster center; The regularization parameter controls the uniformity of the kernel weight distribution; the entropy term... As a constraint, the more uniform the distribution of kernel weights, the larger the negative value of this term. S22: By dynamically adjusting the degree to which samples belong to each cluster, the rationality of fuzzy partitioning is ensured. ; in, Indicates the first The clustering in the th case is at the th case. Cluster centers in the feature space induced by each kernel function; S23: weights The assignment is based on the contribution of the current kernel function to the clustering, and the regularization parameter Controlling the uniformity of weight distribution: ; in, Indicates the first The clustering in the th case is at the th case. S24: Update the cluster centers in the feature space induced by the kernel function; update the cluster centers using the mean of the membership-weighted kernel mapping data to ensure the feasibility of the algorithm in the feature space. ; S25: After the update is completed, calculate the deviation between the result of this iteration and the result of the previous iteration. If the deviation is less than a preset threshold... If the condition is met, the algorithm terminates; otherwise, return to step S21 and repeat until the stopping condition is met. or ; in, and Indicates the first The set of all cluster centers generated in the next iteration and Indicates the first The set of all cluster centers generated in the next iteration The pre-set convergence threshold.

[0009] Furthermore, step S3 rapidly generates initial macroscopic particles based on full data spectral information. For particles with complex distributions or purity constraints, a dual-mode evaluation mechanism is embedded for real-time screening. Then, a progressive self-organizing fission algorithm is used to decompose them into finer sub-particle clusters. Specific steps include: S31: Transfer the dataset As the initial macroscopic particles, in the dataset The binary clustering algorithm is executed to obtain two initial spheres. and Each sphere The formula for calculating the center is: ; in, This represents the number of samples within the granule; S32: Each pellet The formula for calculating the radius is: ; S33: Check whether the above-mentioned particles have reached the threshold. Require: ; In the formula, ,in, Indicates granules the number of samples in the sample set, a candidate label set, representing the sample a label set of the sample; representing the information entropy of the granule , used to measure the uncertainty of the internal labels of the granule, wherein, representing the probability of the label appearing in the granule ; by adjusting the weight ratio of purity and entropy by the parameter, when the value is lower than the preset threshold, it indicates that the label consistency of the granule is insufficient or the topological structure is too complex; If the purity of the granule is lower than the preset threshold, perform binary clustering to split it into two sub-granules with higher consistency, recursively until the purity of all granules meets the standard, and finally complete the construction of the granule.

[0010] Further, the step S4, according to the geometric situation of the granule, fuses the distance measurement strategy triggered by the multi-source weight within each hierarchical granule, constructs a cluster-class tensor mapping framework, realizes adaptive mapping and fusion of sample-cluster membership to sample-class membership, and finally outputs a highly recognized semantic disambiguation distribution. The specific steps include: S41: Construct a hybrid distance measurement algorithm that fuses local geometric structure and dynamic feature importance to obtain the weight information of the sample: ; wherein, is a dynamic feature weight function, is the mean value of the th feature, is a sensitivity coefficient, represents the value of the sample in the th feature, is the reference center value of the th feature, is the standard deviation of the th feature within the granule, which eliminates the influence of dimension difference; The function is a hyperbolic tangent function, and its value range is [0, 1]. The parameter is an adjustment parameter used to control the smoothness of the density value to weight mapping, is a local density estimation based on nearest neighbors; S42: On the basis of the obtained fuzzy membership matrix and the sample weight , construct the class-cluster association matrix : ; matrix elements in Indicates the first The category and the first The association strength of each cluster is determined using fuzzy logic reasoning to correlate the association matrix with the samples. Membership degree of each cluster Perform fuzzy synthesis operation Thus achieving sample The membership degree of a cluster is transformed into the membership degree of a category.

[0011] Furthermore, step S5 constructs a coarse-perceptual feature evaluation framework based on particle-sphere topology-driven methods. It adaptively partitions the label distribution space into decision classes through particle-sphere computation and measures feature dependencies using a rough set of particle-sphere neighborhoods. Specific steps include: S51: For any sample Define it in the feature subset The adaptive neighborhood is as follows: ; in, and It is a pellet. The center and radius, For feature subset The distance below; S52: Construct a dynamic labeled spheroid group set, and use an adaptive spheroid splitting algorithm to target the entire real labeled domain. Perform multi-scale nonlinear spatial reconstruction: ; Each This represents a single marked sphere that collectively covers the entire marked space, with each sphere achieving optimal purity and coverage. S53: In a given decision system In the middle, decision attributes Divide the samples into equivalence classes, using feature subsets The generated set of spheres Define upper and lower approximate sets: , ; in, Through granules Equivalence Class The non-empty overlapping relationship determines the coverage boundary of the decision class; ; Through granules Equivalence Class extract the core support domain of the decision class; S54: Construct an extended positive region dependency quantification model for the image noise label decision system: ; wherein, is the sample that belongs to both and , and represents a sample set in but not in , is the distance from the sample to the center of the granule ball, is the distance from the sample to the center of the granule ball; S55: Through a dynamic multi-granularity collaborative feature importance quantification algorithm, the rough set theory and the granule ball calculation framework are fused to realize accurate evaluation of the features: .

[0012] Further, the step S6 is robust in feature selection from a bimodal perspective, and the specific steps include: S61: Measure the similarity of granule ball samples based on a similarity quantification algorithm of the internal sample structure consistency of the granule ball: ; wherein, represents the cardinality of the granule ball set; S62: Based on an additional feature importance evaluation mechanism, the discriminative contribution of the features is quantified by measuring the degree of change in the granule ball structure before and after the features are removed: ; S63: By constructing a feature dependency matrix and a sample consistency matrix within the granule ball, a method of jointly evaluating feature weights from multiple sources and bimodal dependencies-consistency is adopted to realize fine quantification and sorting of key features in high-dimensional data: .

[0013] The application also provides an image noise label feature selection system, which comprises: a data acquisition and processing module configured to acquire an image noise label data set to be processed, wherein the image noise label data set contains a feature space and a label space; a fuzzy clustering dynamic division module configured to embed a sample set in the image noise label data set into a multi-granularity fuzzy cluster group by using an adaptive multi-kernel fusion and dynamic parameter optimization mechanism, so as to construct a dynamic fuzzy membership evaluation matrix and determine a multi-dimensional fuzzy mapping relationship between the sample-cluster and the sample-class; A multi-scale granule level construction module is used for constructing a macro coarse-grained granule skeleton based on a global spectrum feature mapping mechanism, and for jointly evaluating granules with insufficient label isomorphism or structural entropy exceeding a threshold by integrating a quality measurement model, and dynamically evolving multi-level high-precision sub-granule clusters through a progressive self-organizing fission algorithm. A coupled granule label disambiguation module is used for fusing weight-driven granule calculation and fuzzy reasoning in each fine granule, automatically assigning weights according to sample-granule center distance, and constructing a class-cluster coupled correlation matrix, and finally obtaining a high-resolution label distribution by optimizing cluster group membership to class membership through fuzzy synthesis operation. A granule-rough set collaborative feature evaluation module is used for constructing a rough set perception feature evaluation framework based on granule topology driving, defining an adaptive neighborhood based on granule geometric characteristics, dynamically dividing local neighborhoods that are cohesive and do not overlap, combining rough set upper and lower approximation and extended positive region models to extract decision equivalence classes, and determining the contribution of each feature to the decision system based on a dependency quantification model and multi-granule decision boundary information. A multi-granularity feature stability verification module is used for introducing a granule structure consistency verification mechanism to measure the expression stability of features in the local domain of the granule from the perspective of structural preservation, and to perform multi-level evaluation of the importance of features through dependency and consistency.

[0014] The application further provides a storage medium having a computer program stored thereon, and the program is executed by a processor to implement the image noise label feature selection method.

[0015] The application further provides a computer comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, and the processor implements the image noise label feature selection method when executing the computer program.

[0016] The image noise label feature selection method, system, storage medium and computer in the application adopt a sample soft division mechanism driven by fuzzy membership, initialize fuzzy clusters, and on this basis introduce a quality-driven granular ball evolution strategy, construct a multi-granularity granular ball system from coarse to fine, realize dynamic disambiguation and noise interference suppression of the label space through progressive self-organizing fission; fuse the granular ball geometric topology and rough set approximation theory, define an adaptive neighborhood range, depict the boundary of the decision equivalence class by means of the upper and lower approximation set and the extended positive domain model, and realize high-precision quantitative evaluation of feature importance based on the dependence measurement model of multi-granular ball boundary fusion; construct a granular ball structure stability verification mechanism, evaluate the expression consistency and generalization ability of the selected features at different granularity levels from the two dimensions of sample similarity in the granular ball and feature disturbance response, and improve the robustness and reliability of feature selection; design a hierarchical weight integration strategy, weight and fuse the feature ranking results under multi-granularity, and finally obtain an optimal feature subset with strong noise resistance and high discrimination ability, providing stable and efficient input support for subsequent image noise label learning models. BRIEF DESCRIPTION OF DRAWINGS

[0017] Figure 1 The flowchart of the image noise label feature selection method in the first embodiment of the application; Figure 2 The framework schematic diagram of the image noise label feature selection method in the first embodiment of the application; Figure 3 The structural block diagram of the image noise label feature selection system in the second embodiment of the application; Figure 4 The structural block diagram of the computer in the third embodiment of the application.

[0018] The following specific embodiments will further illustrate the application in conjunction with the above drawings. DETAILED DESCRIPTION

[0019] In order to facilitate the understanding of the application, the application will be described more fully below with reference to the accompanying drawings. The drawings show several embodiments of the application. However, the application can be realized in many different forms and is not limited to the embodiments described herein. On the contrary, the purpose of providing these embodiments is to make the disclosure of the application more thorough and comprehensive.

[0020] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the application belongs. The terminology used in the specification of the application herein is only for the purpose of describing specific embodiments and is not intended to limit the application. The term "and / or" used herein includes any and all combinations of one or more of the associated listed items.

[0021] Embodiment one Please refer to Figure 1 , which is an image noise label feature selection method in the first embodiment of the application, and the method specifically comprises steps S1 to S6: S1: Obtain an image noise label data set to be processed, wherein the image noise label data set contains both a feature space and a label space; In specific implementation, the framework diagram of the bird image noise label feature selection method provided in the embodiment is as shown in Figure 2 The pre-trained ResNet network is used to extract deep features of the bird image, convert the image data into a structured multi-dimensional feature vector, and form a data set covering both the image feature space and the candidate label space; Specifically, let The dimensional feature space composed of samples is denoted as , and the label space containing possible labels is denoted as , wherein is the embedding vector of the sample in the feature space, is the candidate label set associated with the sample. The candidate label set may mix true labels and noise labels, and has the typical feature of label uncertainty. For any candidate label , when , it means that the label is a candidate label of the sample ; when , it means that the label is irrelevant to the sample .

[0022] S2: Use the adaptive multi-kernel fusion and dynamic parameter optimization mechanism to embed the sample set in the image noise label data set into multi-granularity fuzzy cluster groups, to construct a dynamic fuzzy membership evaluation matrix, and determine the multi-dimensional fuzzy mapping relationship between the quantized sample-cluster and sample-category; In specific implementation, the step is oriented to the distribution characteristics of the bird image features (such as feather texture details, feather color texture details, beak shape contour, background environment, etc.), and utilizes an adaptive multi-kernel fusion and dynamic parameter optimization mechanism to perform prior subspace division on the sample set, thereby constructing multiple original semantic clustering centers. The mechanism can automatically adjust the kernel function parameters according to the nonlinear distribution of the bird images in the feature space, and effectively capture the feature differences caused by light changes, posture differences, occlusions and background diversity. Subsequently, by fusing the fuzzy set theory, a soft attachment relationship between the samples and the clustering centers is realized, so that the gradual aggregation of the true species semantics can still be maintained in the presence of label noise. The step S2 specifically includes: Step S21: In the multi-kernel framework, a dynamic weight is given to each kernel function, and an entropy regularization term is introduced into the objective function to constrain the weight distribution: ; Wherein, represents the overall objective function of the multi-kernel fuzzy clustering, which is used to optimize the clustering quality and the rationality of the kernel weight distribution, wherein, represents the membership matrix, represents the clustering center matrix, represents the kernel weight distribution vector; represents the preset number of clustering clusters; is the total number of samples in the data set; represents the total number of kernel functions in the multi-kernel framework; in the fuzzy clustering process, represents the original sample (bird image sample) the fuzzy membership degree of the first clustering, the value is between 0 and 1, and satisfies the normalization constraint condition , is a fuzzification index, usually greater than 1, used to control the fuzziness of the clustering boundary; In the multi-kernel mapping mechanism, represents the dynamic weight coefficient of the first kernel function, and the weight satisfies and constraint conditions; represents the feature vector obtained after the first kernel function maps the original sample (bird image sample) to the high-dimensional feature space; represents the clustering center of the first clustering in the feature space induced by the first kernel function; the distance term calculates the Euclidean distance square of the sample in the specific kernel space and the corresponding clustering center, used to measure the similarity between the sample and the clustering center; In the noisy-labeled bird images, different kernel functions are designed to capture different types of visual features: color kernel extracts features such as feather color distribution and eye circle color, which are used to distinguish species with significant color differences; texture kernel captures information such as feather texture and wing spot details, which are used to deal with cases where feather colors are similar but textures are significantly different; shape kernel analyzes features such as beak shape geometry, body contour, and wing shape, which are used to distinguish species with large differences in body shape characteristics; multi-scale kernel captures detailed features at different shooting distances, poses, and scales.

[0023] The second part of the objective function introduces an entropy regularization mechanism, where is the regularization parameter, used to control the uniformity of the kernel weight distribution; the entropy term acts as a constraint term, and the negative value of this term becomes larger when the kernel weight distribution is more uniform, thereby preventing the algorithm from degenerating into a single kernel function and ensuring the effectiveness of multi-kernel fusion. The entire objective function is optimized by minimizing the weighted clustering error and maximizing the entropy constraint term, achieving the optimization of sample clustering in the multi-kernel space and the adaptive adjustment of kernel weights; Step S22: By dynamically adjusting the degree of sample belonging to each cluster, the rationality of fuzzy division is ensured: ; where, is the cluster center of the th cluster in the feature space induced by the th kernel function; Step S23: The kernel weight update formula dynamically allocates the importance of different kernel functions through the entropy regularization term. The weight is allocated based on the current kernel function's contribution to clustering, and the regularization parameter controls the uniformity of the weight distribution: ; where, is the cluster center of the th cluster in the feature space induced by the th kernel function; Step S24: In the kernel space, the cluster center cannot be explicitly expressed and needs to be implicitly calculated through the kernel matrix. The cluster center is updated using the membership-weighted kernel mapping data mean, ensuring the feasibility of the algorithm in the feature space: ; This process can make the clustering centers of bird images in multiple feature modalities closer to the visual feature distribution of real species; Step S25: After updating, the new membership matrix and the deviation between the previous membership matrix and the new center are calculated in turn, and if the deviation is less than a preset threshold If the condition is met, the algorithm terminates; otherwise, return to step S21 and repeat until the stopping condition is met. or ; in, and Indicates the first The set of all cluster centers generated in the next iteration and Indicates the first The set of all cluster centers generated in the next iteration The pre-defined convergence threshold is set based on the complexity of the bird image feature distribution (such as species diversity and background interference intensity).

[0024] S3: Based on the global spectrum feature mapping mechanism, a macroscopic coarse-grained particle skeleton is constructed. For particles with insufficient label isomorphism or structural entropy exceeding the threshold, a quality measurement model is integrated for joint evaluation. Through a progressive self-organizing fission algorithm, multi-level high-precision sub-particle clusters are dynamically evolved. In practical implementation, this step targets the distribution characteristics of bird image features in high-dimensional space (e.g., appearance differences of the same species under different postures, lighting conditions, and habitat environments). Based on full-data spectral information, initial macroscopic spheres are rapidly generated. For spheres containing high-noise markers or mixed species, a dual-mode evaluation mechanism is embedded for real-time filtering. Through a progressive self-organizing fission algorithm, complex spheres are gradually decomposed into high-purity subspheres capable of distinguishing fine-grained species characteristics, thereby improving the ability to identify subspecies and sex differences. Specific steps include: Step S31: Transfer the dataset As the initial macroscopic particles, in the dataset The binary clustering algorithm is executed to obtain two initial spheres. and Each sphere The formula for calculating the center is: ; in, This represents the number of samples within the granule; Step S32: Each ball The formula for calculating the radius is: ; Step S33: Check whether the above-mentioned particles have reached the threshold. Require: ; In the formula, ,in, Indicates granules The number of samples in a candidate label set, a label set representing the sample , the formula measures the concentration degree of the label distribution by calculating the maximum frequency of each label in the sample, the higher the purity value represents the more consistent the label distribution in the sample within the granule, for example, most images in the same granule are labeled as "white-headed bulbul" and the feather characteristics are consistent, and the purity is higher; information entropy of the granule , used to measure the uncertainty of the labels within the granule, wherein, the probability of the label appearing in the granule , the formula measures the uncertainty degree of the label distribution within the granule, such as different species (gray-headed bulbul and white-headed bulbul) with similar feather colors mixed in the same granule, the entropy value is higher, and the higher the entropy value represents the more complex and chaotic the label distribution within the granule; by adjusting the weight ratio of purity and entropy, when the value is lower than the preset threshold, it indicates that the label consistency of the granule is insufficient or the topological structure is too complex, and further splitting is needed to improve the quality of the granule; If the purity of the granule is lower than the preset threshold, perform binary clustering to split it into two sub-granules with higher consistency, recursively until all granules meet the purity threshold, and finally complete the granule construction. Iterate this process until the purity of all granules meets the threshold requirement, then terminate the splitting process and complete the adaptive granule construction S4: Within each fine granule, fuse weight-driven granule calculation and fuzzy reasoning, automatically assign weights according to sample-granule center distance, and construct class-cluster coupling association matrix, optimize cluster membership to class membership through fuzzy synthesis operation, and finally obtain high-resolution label distribution; In specific implementation, within each level granule, according to the granule geometric situation of bird image features, fuse multi-source weight triggered distance measurement strategies (such as color difference, texture difference and shape difference), construct cluster-class tensor mapping framework, realize adaptive mapping and fusion of sample-cluster membership to sample-class membership, and thus optimize label disambiguation in species recognition, the step S4 specifically includes: Step S41: Based on the core idea of multi-source weight triggering, combine manifold learning and adaptive feature weighting, construct a hybrid distance measurement algorithm that fuses local geometric structure and dynamic feature importance, obtain the weight information of the sample, and significantly improve the scene adaptability of distance measurement by introducing a third-order nonlinear regulation mechanism: ; wherein, is a dynamic feature weight function, is the the mean value of the i-th feature, is the sensitivity coefficient, denotes the sample In the first the value of the i-th feature on the sample, is the reference center value of the i-th feature, is the standard deviation of the i-th feature in the granule, eliminating the influence of dimensional difference; is the hyperbolic tangent function, whose value range is [0, 1], and the parameter is an adjustment parameter, used to control the smoothness of the density value to weight mapping, is the local density estimation based on the k-nearest neighbors; Step S42: On the basis of having obtained the fuzzy membership matrix and the sample weight , a class-cluster association matrix is constructed: ; Wherein, the initial value of the matrix A is a zero matrix, and the dimension is , is the number of classes, is the number of clusters, and the element in the matrix represents the association strength of the i-th class and the j-th cluster. By using fuzzy logic reasoning method, the association matrix is subjected to fuzzy composition operation with the membership of the sample to each cluster , so as to realize the conversion of the membership of the sample to the cluster to the membership to the class, thereby reducing the label confusion between the species with similar colors. S5: Construct a rough perception feature evaluation framework based on granule topology driving, define an adaptive neighborhood based on the geometric characteristics of the granule, dynamically divide local neighborhoods that are cohesive and do not overlap each other, extract decision equivalence classes by combining the upper and lower approximation of rough set and the extended positive region model, determine the contribution of each feature to the decision system based on the dependence quantification model, and fuse the decision boundary information of multiple granules.

[0025] S5: Construct a rough perception feature evaluation framework based on granule topology driving, define an adaptive neighborhood based on the geometric characteristics of the granule, dynamically divide local neighborhoods that are cohesive and do not overlap each other, extract decision equivalence classes by combining the upper and lower approximation of rough set and the extended positive region model, determine the contribution of each feature to the decision system based on the dependence quantification model, and fuse the decision boundary information of multiple granules. ​​​​​In specific implementation, a rough perception feature evaluation framework based on granule topology driving is constructed, relying on the geometric topology information of granules, the adaptive neighborhood of each sample is dynamically defined, the multi-dimensional division of the label distribution space is realized, the fine description of the decision equivalence class is completed by using the extended upper and lower approximation boundary and positive region expansion model, the dependence of each feature in different decision classes is calculated in real time based on the dynamic dependence measurement model, and then the discriminative features are accurately identified and sorted, and the step S5 specifically includes: Step S51: let the sample set be , the feature set be , and the attribute subset be . The granule calculation is applied to to obtain the granule set under the feature space . For any sample , the adaptive neighborhood of the sample under the feature subset is defined as: ; wherein and are the center and radius of the granule , and is the distance under the feature subset . The definition dynamically divides the samples into “cohesive” and “non-overlapping” local fields through the center and radius of the granule, avoids the rigidity of the fixed radius neighborhood, and ensures that the samples in the neighborhood are highly consistent in species characteristics; Step S52: construct a dynamic label granular sphere set, perform multi-scale nonlinear space reconstruction on the real label universe by using the adaptive granule splitting algorithm, and form the optimal granular topology partition with optimal coverage of the universe: ; Each represents a label granule, collectively covers the entire label space, and each granule reaches the optimal purity and coverage; Step S53: in the given decision system , the decision attribute divides the samples into equivalence classes, the granule set generated by the feature subset is used to define the upper and lower approximation sets: , ; wherein , the non-empty intersection relationship between the granule and the equivalence class is used to determine the coverage boundary of the decision class; ; By granules and equivalence classes , the core support domain of the decision class is extracted. In the bird application, the upper approximation set usually contains typical samples with completely consistent feather color, texture and morphological characteristics, such as high-definition photos of adult male birds; the lower approximation set contains boundary samples that are easily confused with other species, such as juvenile birds, intermediate individuals with feather color, or photos with partial occlusion; Step S54: Because the label space of the multi-label data set itself has inherent ambiguity, it is difficult to clearly segment the decision boundary, and the lower approximation set alone cannot fully reflect the contribution of boundary samples to the decision. Therefore, samples outside the lower approximation set but still have "weak association" or "partial overlap" with a certain decision class are also included in the measurement of the positive domain, so as to more comprehensively characterize the multi-label decision boundary. An extended positive region dependency measurement model for image noise label decision system is constructed: ; Wherein, is the sample that belongs to and , represents the sample set in but not in , is the distance from the sample to the center of the granule, is the distance from the sample to the center of the granule; Step S55: Through the dynamic multi-granularity collaborative feature importance quantification algorithm, the rough set theory and the granule calculation framework are fused to realize the accurate evaluation of the features: .

[0026] Specifically, the multi-granularity decision boundary information is fused to output the contribution degree ranking of each feature to bird classification. In actual deployment, the ranking result can be used to preferentially select the features with the highest contribution degree to species discrimination and the features that are not sensitive to environmental changes, which are used for model training and updating of the field bird automatic recognition system. S6: The particle ball structure consistency verification mechanism is introduced to measure the expression stability of the features in the local domain of the particle ball from the structure preservation dimension, and the importance of the features is evaluated at multiple levels through dependency and consistency.

[0027] In specific implementation, for the conditional attribute subset , the granule set generated from the training sample set , a similarity measurement method based on the structure consistency of the internal samples of the granule is introduced to measure the similarity of the granule samples, which is as follows: Step S61: measure the granule sample similarity based on the similarity quantization algorithm of the granule internal sample structure consistency: ; wherein, denotes the cardinality of the granule set, The greater the value, the more consistent the granule internal sample feature distribution, the higher the mutual similarity, and the clearer the boundary between different granules, proving that the current feature subset has strong discrimination ability and expression power; Step S62: based on the additional feature importance evaluation mechanism, the discrimination contribution of the feature is quantified by measuring the degree of change of the granule structure before and after the feature is removed: ; Step S63: by constructing the feature dependency matrix and the granule internal sample consistency matrix, the method of evaluating the feature weight based on the joint evaluation of multi-source and double modal dependency-consistency is adopted to realize the fine quantization and sorting of the key features in the high-dimensional data: .

[0028] In summary, the image noise label feature selection method in the above embodiments of the application adopts a sample soft partition mechanism driven by fuzzy membership to initialize fuzzy clusters, and introduces a quality-driven granule evolution strategy based on this to construct a multi-granularity granule system from coarse to fine, and realizes dynamic disambiguation and noise interference suppression of the label space through progressive self-organizing fission; the granule geometric topology and rough set approximation theory are fused to define an adaptive neighborhood range, the boundaries of decision equivalence classes are described by means of upper and lower approximation sets and extended positive domain models, and based on the dependency measure model of multi-granule boundary fusion, high-precision quantitative evaluation of feature importance is realized; a granule structure stability verification mechanism is constructed to evaluate the expression consistency and generalization ability of the selected features at different granular levels from the two dimensions of granule internal sample similarity and feature disturbance response, thereby improving the robustness and reliability of feature selection; a hierarchical weight integration strategy is designed to weight and fuse the feature ranking results at different granular levels, and finally an optimal feature subset with strong noise resistance and high discrimination ability is obtained, which provides stable and efficient input support for subsequent image noise label learning models.

[0029] Embodiment Two Another aspect of the application also proposes an image noise label feature selection system, please refer to Figure 3 , which is an image noise label feature selection system in the second embodiment of the application, the system comprises: A data acquisition and processing module 201 is used to acquire an image noise label data set to be processed, wherein the image noise label data set contains a feature space and a label space at the same time; The fuzzy clustering dynamic division module 202 is configured to embed sample sets in the image noise label data set into multi-granularity fuzzy clusters by using an adaptive multi-core fusion and dynamic parameter optimization mechanism, to construct a dynamic fuzzy membership evaluation matrix, and to determine a multi-dimensional fuzzy mapping relationship between quantized sample-clusters and sample-categories. The multi-scale granule layer construction module 203 is configured to construct a macroscopic coarse-granularity granule skeleton based on a global spectrum feature mapping mechanism, to fuse quality measurement models for joint evaluation of granules with insufficient label isomorphism or structural entropy exceeding a threshold, and to dynamically evolve multi-layer high-precision sub-granule clusters by using a progressive self-organizing fission algorithm. The coupled granule label disambiguation module 204 is configured to fuse granule calculation and fuzzy reasoning driven by weights within each fine granule, to automatically assign weights according to sample-granule center distances, to construct a category-cluster coupled correlation matrix, to optimize cluster membership to category membership by using fuzzy synthesis operation, and to finally obtain a high-resolution label distribution. The granule-rough set collaborative feature evaluation module 205 is configured to construct a rough set perception feature evaluation framework driven by granule topology, to define an adaptive neighborhood based on granule geometric characteristics, to dynamically divide local neighborhoods that are cohesive and do not overlap with each other, to extract decision equivalence classes in combination with rough set upper and lower approximation and extended positive region models, to determine the contribution of each feature to the decision system based on a dependency quantization model, and to fuse multi-granule decision boundary information. The multi-granularity feature stability verification module 206 is configured to introduce a granule structure consistency verification mechanism, to measure the expression stability of features in the local domain of the granule from the perspective of structural preservation, and to perform multi-level evaluation of the importance of features by using dependency and consistency.

[0030] The functions or operation steps realized when the above modules and units are executed are substantially the same as those of the above method embodiments, and thus will not be described here again.

[0031] The image noise label feature selection system provided in the embodiments of the present application has the same implementation principle and technical effects as the above method embodiments, and for brevity, the system embodiments not mentioned in the above will be described with reference to the corresponding contents in the above method embodiments.

[0032] Embodiment Three The present application also provides a computer, please refer to Figure 4 , which is a computer in the third embodiment of the present application, comprising a memory 10, a processor 20, and a computer program 30 stored in the memory 10 and executable on the processor 20, wherein the processor 20 implements the above-mentioned image noise label feature selection method when executing the computer program 30.

[0033] The memory 10 includes at least one type of storage medium including a flash memory, a hard disk, a multimedia card, a card-type memory (e.g., SD or DX memory, etc.), a magnetic memory, a magnetic disk, an optical disk, etc. The memory 10 can be an internal storage unit of a computer, e.g., a hard disk of the computer, in some embodiments. The memory 10 can also be an external storage device, e.g., a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) card, a Flash Card, etc., in other embodiments. Further, the memory 10 can include both an internal storage unit of a computer and an external storage device. The memory 10 can be used to store application software installed in the computer and various data, and to temporarily store data that has been output or is to be output.

[0034] The processor 20 can be an Electronic Control Unit (ECU), a Central Processing Unit (CPU), a controller, a microcontroller, a microprocessor, or other data processing chip, in some embodiments, for executing program codes or processing data stored in the memory 10, e.g., executing an access restriction program.

[0035] It is noted that, Figure 4 The illustrated structure does not constitute a limitation on the computer, which can include fewer or more components than illustrated, or combine certain components, or have different arrangement of components, in other embodiments.

[0036] The embodiments of the present application also propose a storage medium having a computer program stored thereon, which, when executed by a processor, implements the image noise marking feature selection method as described above.

[0037] Those skilled in the art can understand that the logic and / or steps represented in the flowcharts or otherwise described herein, for example, can be considered as a list of executable instructions for implementing the logic function, which can be embodied in any computer readable medium for use by or in connection with an instruction execution system, apparatus or device, such as a computer-based system, a system including a processor or other system that can fetch and execute instructions from an instruction execution system, apparatus or device. For the present specification, the "computer readable medium" can be any device that can contain, store, communicate, propagate or transport a program for use by or in connection with an instruction execution system, apparatus or device, or in conjunction with these instruction execution systems, apparatus or devices.

[0038] More specific examples (a non-exhaustive list) of the computer-readable medium include the following: an electrical connection (electronic) having one or more wires, a portable computer diskette (magnetic), a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), an optical fiber, and a portable compact disc read-only memory (CDROM). Additionally, the computer-readable medium can also be paper or another suitable medium upon which the program is printed, as the program can be electronically captured, for example, via an optical scanner, then compiled, interpreted, or otherwise processed, and stored in a computer memory in a form that is then reproducible into a computer readable medium.

[0039] It should be understood that aspects of the application can be implemented in hardware, software, firmware or combinations thereof. In the above embodiments, various steps or methods can be implemented in software or firmware that is stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, any of the following technologies, known in the art, or combinations thereof, can be employed: a discrete logic circuit having logic gates for implementing logic functions upon data signals, an application specific integrated circuit having appropriate combinational logic gates, a programmable gate array (PGA), a field programmable gate array (FPGA), or the like.

[0040] The technical features of the above-described embodiments can be combined in any manner, and in order to make the description concise, not all possible combinations of the technical features in the above-described embodiments are described, however, as long as the combinations of the technical features do not contradict each other, they should be considered within the scope of the present disclosure.

[0041] The above-described embodiments only express several implementation manners of the present application, and the description is relatively specific and detailed, but it should not be understood as a limitation on the scope of the patent. It should be noted that for those skilled in the art, without departing from the concept of the present application, a number of modifications and improvements can be made, which are within the scope of the present application. Therefore, the scope of protection of the patent of the present application should be subject to the appended claims.

Claims

1. A method for selecting image noise marker features, characterized in that, include: S1: Obtain the image noise labeled dataset to be processed, wherein the image noise labeled dataset contains both feature space and label space; S2: An adaptive multi-core fusion and dynamic parameter optimization mechanism is adopted to embed the sample set in the image noise labeling dataset into a multi-granularity fuzzy cluster to construct a dynamic fuzzy membership evaluation matrix and determine the multi-dimensional fuzzy mapping relationship between quantified sample-cluster and sample-class. S3: Based on the global spectrum feature mapping mechanism, a macroscopic coarse-grained particle skeleton is constructed. For particles with insufficient label isomorphism or structural entropy exceeding the threshold, a quality measurement model is integrated for joint evaluation. Through a progressive self-organizing fission algorithm, multi-level high-precision sub-particle clusters are dynamically evolved. S4: Within each fine sphere, weight-driven sphere calculation and fuzzy inference are integrated. Weights are automatically assigned based on the sample-sphere center distance, and a category-cluster coupling association matrix is ​​constructed. Through fuzzy synthesis operation, the cluster membership is optimized into category membership, and finally a highly recognizable label distribution is obtained. S5: Construct a rough perception feature evaluation framework based on particle topology-driven, define adaptive neighborhoods with particle geometric characteristics, dynamically divide cohesive and non-overlapping local neighborhoods, combine rough set upper and lower approximations and extended positive domain models to extract decision equivalence classes, and based on the dependency metric model, fuse multi-particle decision boundary information to determine the contribution of each feature to the decision system. S6: Introduces a particle-sphere structure consistency verification mechanism to measure the expression stability of features in the local domain of particles from the dimension of structure preservation, and to evaluate the importance of features at multiple levels through dependency and consistency.

2. The image noise marker feature selection method according to claim 1, characterized in that, Step S1 includes: Using ResNet to extract features from noise-labeled image data, a dataset is obtained. ,in, For the first The feature vector of each sample For the first A set of candidate labels for each sample, for each candidate label element ,when When, it indicates a mark. For the sample Candidate tags; when When, it indicates a mark. With sample Irrelevant.

3. The image noise marker feature selection method according to claim 2, characterized in that, In step S2, a combination of random perturbation initialization and nondeterministic exploration strategies is used to initially partition the sample set in the image noise-labeled dataset. Step S2 includes: S21: In a multi-kernel framework, each kernel function is assigned dynamic weights, and an entropy regularization term is introduced into the objective function to constrain the weight distribution. ； in, Let represent the overall objective function of multi-kernel fuzzy clustering, used to simultaneously optimize clustering quality and the rationality of kernel weight allocation, where Represents the membership matrix. Represents the cluster center matrix. Represents the kernel weight allocation vector; This represents the preset number of clusters; This represents the total number of samples in the dataset. This represents the total number of kernel functions in a multi-kernel framework; in the fuzzy clustering process, Represents the original sample For the first The fuzzy membership degree of each cluster, which is between 0 and 1 and satisfies the normalization constraint. , The fuzziness index, typically greater than 1, is used to control the degree of fuzziness of cluster boundaries; in multi-kernel mapping mechanisms, Representing the The dynamic weight coefficients of each kernel function, the weights satisfying and Constraints; Indicates the first Each kernel function will convert the original sample The feature vector obtained after mapping to a high-dimensional feature space; Indicates the first The clustering in the th case is at the th case. Cluster centers in the feature space induced by each kernel function; distance term Calculate the squared Euclidean distance between a sample and its corresponding cluster center in a specific kernel space to measure the similarity between the sample and the cluster center; The regularization parameter controls the uniformity of the kernel weight distribution; the entropy term... As a constraint, the more uniform the distribution of kernel weights, the larger the negative value of this term. S22: By dynamically adjusting the degree to which samples belong to each cluster, the rationality of fuzzy partitioning is ensured. ； in, Indicates the first The clustering in the th case is at the th case. Cluster centers in the feature space induced by each kernel function; S23: Weight The assignment is based on the contribution of the current kernel function to the clustering, and the regularization parameter Controlling the uniformity of weight distribution: ； in, Indicates the first The clustering in the th case is at the th case. Cluster centers in the feature space induced by each kernel function; S24: Update the cluster centers using the mean of the kernel mapping data weighted by membership degree to ensure the feasibility of the algorithm in the feature space: ； S25: After the update is completed, calculate the deviation between the result of this iteration and the result of the previous iteration. If the deviation is less than a preset threshold... If the condition is met, the algorithm terminates; otherwise, return to step S21 and repeat until the stopping condition is met. or ; in, and Indicates the first The set of all cluster centers generated in the next iteration and Indicates the first The set of all cluster centers generated in the next iteration The pre-set convergence threshold.

4. The image noise marker feature selection method according to claim 3, characterized in that, Step S3 rapidly generates initial macroscopic particles based on full data spectral information. For particles with complex distributions or purity constraints, a dual-mode evaluation mechanism is embedded for real-time screening. Furthermore, a progressive self-organizing fission algorithm is used to decompose them into finer sub-particle clusters. Specific steps include: S31: Transfer the dataset As the initial macroscopic particles, in the dataset The binary clustering algorithm is executed to obtain two initial spheres. and , where each sphere The formula for calculating the center is: ； in, This represents the number of samples within the granule; S32: Each pellet The formula for calculating the radius is: ； S33: Check whether the above-mentioned particles have reached the threshold. Require: ； In the formula, ,in, Indicates granules The number of samples in For the candidate tag set, Indicates sample A set of tags; Indicates granules The information entropy is used to measure the uncertainty of the markings inside the granule, where... Indicates a marker In granules The probability of occurrence in; through The parameter adjusts the weight ratio of purity and entropy. When the value is lower than the preset threshold, it indicates that the labeling consistency of the particle is insufficient or the topology is too complex. If the purity of the spheres is lower than the preset threshold, binary clustering is performed to split them into two more consistent sub-spheres. This process is repeated until the purity of all spheres meets the threshold, and finally the sphere construction is completed.

5. The image noise marker feature selection method according to claim 4, characterized in that, Step S4 involves constructing a cluster-class tensor mapping framework within each level of the granular sphere, based on the sphere's geometric shape and by integrating a distance metric strategy triggered by multi-source weights. This framework enables adaptive mapping and fusion of sample-cluster membership to sample-class membership, ultimately outputting a highly discriminative semantic disambiguation distribution. Specific steps include: S41: Construct a hybrid distance metric algorithm that integrates local geometric structure and dynamic feature importance to obtain sample weight information: ； in, , is the dynamic feature weight function. For the first dimensional feature mean, This is the sensitivity coefficient. Indicates sample In the Values ​​on the dimensional features, For the first Reference center value of dimensional feature For the first granule Standard deviation of dimensional characteristics to eliminate the influence of dimensional differences; The function is the hyperbolic tangent function, its range is [0,1], and the parameter is... The adjustable parameters control the smoothness of the mapping from density values ​​to weights. Based on Local density estimation of nearest neighbors; S42: With the fuzzy membership matrix already obtained and sample weights Based on this, construct a category-cluster association matrix. : ； matrix elements in Indicates the first The category and the first The association strength of each cluster is determined using fuzzy logic reasoning to correlate the association matrix with the samples. Membership degree of each cluster Perform fuzzy synthesis operation Thus achieving sample The membership degree of a cluster is transformed into the membership degree of a category.

6. The image noise marker feature selection method according to claim 5, characterized in that, Step S5 constructs a coarse-aware feature evaluation framework based on particle-sphere topology-driven methods. It adaptively partitions the label distribution space into decision classes through particle-sphere computation and measures feature dependencies using a rough set of particle-sphere neighborhoods. Specific steps include: S51: For any sample Define it in the feature subset The adaptive neighborhood is as follows: ； in, and It is a pellet. The center and radius, For feature subset The distance below; S52: Construct a dynamic labeled spheroid group set, and use an adaptive spheroid splitting algorithm to target the entire real labeled domain. Perform multi-scale nonlinear spatial reconstruction: ； Each This represents a single marked sphere that collectively covers the entire marked space, with each sphere achieving optimal purity and coverage. S53: In a given decision system In the middle, decision attributes Divide the samples into equivalence classes, using feature subsets The generated set of spheres Define upper and lower approximate sets: ， ； in, Through granules Equivalence Class The non-empty overlapping relationship determines the coverage boundary of the decision class; ； Through granules Equivalence Class The full inclusion relationship is used to extract the core support domain of the decision class; S54: Constructing an extended positive region dependency quantification model for image noise labeling decision systems: ； in, For both belong to It also belongs to The samples in Indicates in In but not in The sample set in For the sample Distance to the center of the grain For the sample Distance to the center of the grain; S55: Achieves accurate feature evaluation by integrating rough set theory and the granular sphere computation framework through a dynamic multi-granularity collaborative feature importance quantification algorithm. 。 7. The image noise marker feature selection method according to claim 6, characterized in that, Step S6, from a dual-modal perspective, addresses the robustness of feature selection and includes the following specific steps: S61: A similarity quantification algorithm based on the consistency of the internal sample structure of spheres measures the similarity of sphere samples: ； in, The cardinality of the set of particles; S62: Based on an additional feature importance evaluation mechanism, the discriminative contribution of a feature is quantified by measuring the degree of change in granular structure before and after feature removal. ； S63: By constructing a feature dependency matrix and an intrasphere sample consistency matrix, a multi-source bimodal dependency-consistency joint evaluation method for feature weights is adopted to achieve refined quantification and ranking of key features in high-dimensional data. 。 8. An image noise marker feature selection system, characterized in that, include: The data acquisition and processing module is used to acquire the image noise-labeled dataset to be processed, wherein the image noise-labeled dataset contains both a feature space and a label space; The fuzzy clustering dynamic partitioning module is used to embed the sample set in the image noise labeling dataset into a multi-granularity fuzzy cluster using an adaptive multi-kernel fusion and dynamic parameter optimization mechanism, so as to construct a dynamic fuzzy membership evaluation matrix and determine the multi-dimensional fuzzy mapping relationship between quantified sample-cluster and sample-class. The multi-scale particle-sphere hierarchical construction module is used to construct a macroscopic coarse-grained particle-sphere skeleton based on the global spectrum feature mapping mechanism. For particles with insufficient label isomorphism or structural entropy exceeding the threshold, it is integrated into the quality measurement model for joint evaluation, and dynamically evolves into multi-level high-precision sub-particle clusters through a progressive self-organizing fission algorithm. The coupled particle sphere labeling and disambiguation module is used to fuse weight-driven particle sphere calculation and fuzzy inference within each fine particle sphere. It automatically assigns weights based on the sample-particle center distance and constructs a category-cluster coupling association matrix. Through fuzzy synthesis operation, it optimizes the cluster membership degree into the category membership degree, and finally obtains a highly recognizable label distribution. The particle-rough set collaborative feature evaluation module is used to construct a rough perception feature evaluation framework based on particle topology. It defines an adaptive neighborhood based on the geometric characteristics of the particle, dynamically divides cohesive and non-overlapping local neighborhoods, and extracts decision equivalence classes by combining the upper and lower approximations of rough sets and the extended positive domain model. Based on the dependency metric model, it integrates multi-particle decision boundary information to determine the contribution of each feature to the decision system. The multi-granularity feature stability verification module is used to introduce a particle-sphere structure consistency verification mechanism. From the dimension of structure preservation, it measures the expression stability of features in the local domain of particles and spheres, and evaluates the importance of features at multiple levels through dependency and consistency.

9. A storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the image noise marker feature selection method as described in any one of claims 1 to 7.

10. A computer comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the image noise marker feature selection method as described in any one of claims 1 to 7.

Images