Power equipment state evaluation method and system based on gray probability distribution

Abstract

The invention discloses a power equipment state evaluation method and system based on gray scale probability distribution, and belongs to the technical field of power equipment monitoring, and the method comprises the steps: evaluating the feature accuracy through introducing a local gray scale consistency factor and an edge structure integrity factor, and constructing an accuracy scoring function to carry out the quantitative judgment of the image quality; and when the score is lower than a threshold value, an adaptive entropy-oriented diffusion mapping gray correction algorithm is automatically triggered to intelligently repair an abnormal region, and then reliable features are re-extracted from a corrected image and state evaluation is executed, so that the accuracy and robustness of an evaluation system in a complex environment are remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of power equipment monitoring, and in particular to a method and system for evaluating the state of power equipment based on grayscale probability distribution. Background Art

[0002] Grayscale probability distribution refers to the probability distribution of each grayscale level (typically between 0 and 255) within an image. It reflects the distribution of image pixels at different grayscale values ​​and is used to describe the overall brightness characteristics of the image. By counting the frequency of each grayscale level and dividing it by the total number of pixels, a normalized grayscale probability distribution can be obtained, which is useful for image processing tasks such as image enhancement, segmentation, and feature extraction.

[0003] The existing technology has the following shortcomings:

[0004] Existing technologies for evaluating the status of power equipment based on the extraction of statistical features based on grayscale probability distribution may be highly sensitive to non-uniform illumination and local highlight artifacts in images, which can easily lead to distortion of statistical features (such as entropy, skewness, and kurtosis). For example, in infrared images, local bright spots caused by sunlight reflection or uneven equipment surfaces may be mistaken for thermal anomalies, causing the assessment model to misjudge. This problem stems from the fact that grayscale probability distribution lacks the ability to recognize the spatial structure of the image and the true temperature characteristics, and is easily misled by external interference factors, affecting the accuracy of equipment status judgment. Summary of the Invention

[0005] The purpose of the present invention is to provide a method and system for evaluating the state of power equipment based on grayscale probability distribution to address the shortcomings of the background technology.

[0006] In order to achieve the above object, the present invention provides the following technical solution: a method for evaluating the state of power equipment based on grayscale probability distribution, comprising:

[0007] Collect infrared images or visible light images of power equipment and pre-process the images;

[0008] Perform pixel grayscale statistics on the preprocessed image, construct a grayscale histogram, and calculate the grayscale probability distribution;

[0009] Extracting a plurality of statistical characteristic parameters of the image based on the grayscale probability distribution;

[0010] Based on the image content, feature accuracy analysis parameters are further extracted, including the local grayscale consistency factor LGC and the edge structure integrity factor ESI;

[0011] Constructing an accuracy function, taking the LGC and ESI as input, outputting an accuracy score value, and dividing the accuracy level according to a preset score threshold;

[0012] When the accuracy score is lower than a preset threshold, performing a grayscale correction scheme of an adaptive entropy-guided diffusion mapping;

[0013] Statistical features are re-extracted from the corrected image, and the operating status of the power equipment is evaluated based on the feature values ​​and the training model or threshold rules.

[0014] Preferably, the calculation of the grayscale probability distribution includes:

[0015] The grayscale value of each pixel in the image is counted to generate a grayscale histogram H[i], where i∈[0,255]; the number of pixels at each grayscale level is divided by the total number of pixels to obtain the normalized grayscale probability distribution p(i); the grayscale probability distribution result is output in the form of an array or tensor.

[0016] Preferably, the statistical characteristic parameters include: grayscale mean, variance, entropy, skewness and kurtosis of the image; each statistical parameter is obtained by weighted calculation of the normalized grayscale probability distribution p(i).

[0017] Preferably, the method for extracting the local grayscale consistency factor LGC is:

[0018] Input the preprocessed grayscale image I, with a size of M×N and a pixel grayscale value range of [0, 255]; construct the image pyramid block and set L pyramid levels;

[0019] Each level l divides the image into 2 l ×2 l sub-regions of equal size; for the j-th region of the l-th layer, extract its pixel subset R l,j ; Calculate the mean grayscale value μ of pixels in the area l,j and variance

[0020] For each level l, find the mean variance V of all sub-regions l , the expression is: Weighted combination of variance values ​​at different levels, setting the weight w of each level l , calculate the final local gray consistency factor LGC:

[0021] Preferably, the method for obtaining the edge structure integrity factor ESI is:

[0022] Input image I, use edge detection algorithm to extract edge map E, and obtain binary edge map E(x,y)∈{0,1}; divide the image into m×n grid blocks of equal size; each grid block is denoted as B i,j , where i∈[1,m], j∈[1,n]; for each block Bi,j , count the number of edge pixels e i,j :e i,j =∑(x,y)∈B i,j E(x,y); calculate edge density d i,j :

[0023] Divide the sparse area and the dense area, set the edge density threshold δ; if d i,j <δ, then area B i,j Recorded as sparse area; otherwise recorded as dense area; let the total number of areas be N b =m×n, the number of sparse regions is N s ; Define ESI as the proportion of non-sparse areas, the expression is:

[0024] Preferably, an accuracy function is constructed, which takes the LGC and ESI as input, outputs an accuracy score value, and divides the accuracy level according to a preset score threshold, specifically including:

[0025] The local grayscale consistency factor LGC and the edge structure integrity factor ESI are normalized so that they are both between [0, 1]. The accuracy score value is calculated based on the normalized local grayscale consistency factor LGC and the edge structure integrity factor ESI.

[0026] Preferably, the image feature accuracy is divided into three levels according to the scoring results:

[0027] An accuracy score of ≥0.8 is classified as high accuracy, indicating good image quality, and features are directly used for evaluation;

[0028] 0.5≤Accuracy score<0.8 is classified as medium accuracy, indicating that the image has slight interference and needs selective correction;

[0029] An accuracy score value <0.5 is classified as low accuracy, indicating that the image interference is serious and a correction mechanism is implemented.

[0030] Preferably, the grayscale correction scheme of the adaptive entropy-guided diffusion mapping includes:

[0031] Construct a local entropy map of the image and calculate the information entropy value for each pixel neighborhood;

[0032] Generate abnormal area masks based on the entropy map to identify areas that need repair;

[0033] Construct a graph structure model with pixels as nodes and edge weights that fuse grayscale difference and entropy difference;

[0034] Based on the graph Laplace diffusion model, the grayscale value is propagated from the credible area to the abnormal area, and the corrected image is output.

[0035] The present invention also provides a power equipment status assessment system based on grayscale probability distribution, which includes an image acquisition and preprocessing module, a grayscale probability analysis module, a feature extraction module, an accuracy analysis module, an accuracy assessment module, an image correction module, and a status assessment and judgment module;

[0036] Image acquisition and preprocessing module: collects infrared images or visible light images of power equipment and preprocesses the images;

[0037] Grayscale probability analysis module: performs pixel grayscale statistics on the preprocessed image, constructs a grayscale histogram, and calculates the grayscale probability distribution;

[0038] Feature extraction module: extracts several statistical feature parameters of the image based on the grayscale probability distribution;

[0039] Accuracy analysis module: Based on the image content, it further extracts feature accuracy analysis parameters, including local grayscale consistency factor LGC and edge structure integrity factor ESI;

[0040] Accuracy evaluation module: constructs an accuracy function, takes the LGC and ESI as input, outputs an accuracy score value, and divides the accuracy level according to a preset score threshold;

[0041] Image correction module: when the accuracy score is lower than a preset threshold, executing a grayscale correction scheme of adaptive entropy-guided diffusion mapping;

[0042] State assessment and judgment module: re-extract statistical features from the corrected image, and evaluate the operating status of the power equipment based on the feature values ​​and the training model or threshold rules.

[0043] In the above technical solution, the technical effects and advantages provided by the present invention are:

[0044] 1. This invention significantly improves image quality assessment capabilities by introducing a local grayscale consistency factor and an edge structure integrity factor to construct a feature accuracy assessment mechanism. This solves the problem of feature distortion and device status misjudgment in traditional grayscale probability distribution-based methods when dealing with complex scenarios such as non-uniform lighting and local highlight artifacts. An accuracy scoring function dynamically determines the reliability of an image and selects whether to perform corrections based on the score, establishing an automated, closed-loop feedback image quality control process.

[0045] 2. This invention innovatively employs an adaptive entropy-guided diffusion mapping algorithm to correct grayscale in interference areas, ensuring the continuity of image structure while achieving smooth transfer of thermal information and effectively restoring the authenticity of image features. Ultimately, the system re-extracts statistical features from the corrected image and performs state assessment, improving the accuracy and robustness of power equipment anomaly detection and enhancing the system's intelligent response capabilities in complex environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments described in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0047] Figure 1 This is a mind map of the method of the present invention.

[0048] Figure 2 This is a mind map of the system modules of the present invention. DETAILED DESCRIPTION

[0049] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0050] Example 1, please refer to Figure 1 As shown, the method for evaluating the state of electric power equipment based on grayscale probability distribution described in this embodiment includes:

[0051] Collect infrared images or visible light images of power equipment and pre-process the images;

[0052] Perform pixel grayscale statistics on the preprocessed image, construct a grayscale histogram, and calculate the grayscale probability distribution;

[0053] Extracting a plurality of statistical characteristic parameters of the image based on the grayscale probability distribution;

[0054] Based on the image content, feature accuracy analysis parameters are further extracted, including the local grayscale consistency factor LGC and the edge structure integrity factor ESI;

[0055] Constructing an accuracy function, taking the LGC and ESI as input, outputting an accuracy score value, and dividing the accuracy level according to a preset score threshold;

[0056] When the accuracy score is lower than a preset threshold, performing a grayscale correction scheme of an adaptive entropy-guided diffusion mapping;

[0057] Statistical features are re-extracted from the corrected image, and the operating status of the power equipment is evaluated based on the feature values ​​and the training model or threshold rules.

[0058] The image acquisition and preprocessing steps specifically include:

[0059] Use an infrared thermal imager to obtain thermal distribution images of power equipment in working condition, which is suitable for detecting faults such as overheating and poor contact; use a visible light camera to obtain the appearance structure diagram of the equipment to assist in locating abnormal areas in the infrared image.

[0060] Infrared image acquisition temperature range: generally set to -20℃ ~ 150℃, covering the operating temperature of common power equipment; resolution is not less than 320×240 pixels, preferably 640×480 or higher to obtain sufficient spatial information; use real-time continuous acquisition (frame rate not less than 9fps) or timed single-frame shooting to obtain clear images.

[0061] Collection should be carried out during the period with minimal interference from natural light, avoiding direct sunlight or reflections; the image collection angle should be as perpendicular to the device surface as possible to reduce angular distortion; if it is an outdoor device, external interference such as obstructions, wind, and rain should be considered.

[0062] The collected images are processed in the following multi-stage process to improve the accuracy of subsequent grayscale feature extraction:

[0063] Image denoising: Removes thermal noise, environmental interference, etc. generated during image acquisition. Method selection: Gaussian filtering: Smoothes the image and removes random noise; Median filtering: Suitable for removing salt and pepper noise while preserving edges; Bilateral filtering: Removes noise while maintaining edge clarity. Adjustable parameters: The filter window size is generally set to 3×3 or 5×5.

[0064] Image grayscale conversion: Convert the image uniformly into a single-channel grayscale image to facilitate subsequent grayscale probability distribution and statistical feature analysis; Implementation: Apply the weighted average grayscale conversion formula for RGB images: G = 0.299·R + 0.587·G + 0.114·B; Infrared images are usually grayscale images themselves and can be used directly or normalized (0 to 255). Image enhancement: Improve image contrast and detail resolution, highlighting abnormal areas of the device; Enhancement methods include: Histogram equalization: Enhances global contrast and stretches the grayscale dynamic range, often used for low-contrast infrared images;

[0065] Adaptive histogram equalization: maintains local details and avoids over-enhancement of the entire image. It is suitable for images with alternating bright and dark areas. Gradient enhancement (such as the Sobel operator): strengthens edge features, which helps with subsequent structural integrity analysis (ESI). Gamma correction: adjusts the image brightness distribution to improve visual perception. The parameter γ is usually between 0.4 and 1.2.

[0066] Output high-quality grayscale images in a unified format, providing clean, contrasty input for subsequent grayscale probability distribution analysis; retain preprocessing parameters and intermediate images for debugging and tracing.

[0067] The global statistical features of pixel grayscale values ​​are extracted from the preprocessed image, and the grayscale probability distribution is formed by constructing a grayscale histogram and normalizing it, providing a basic data structure for subsequent statistical feature calculations (such as entropy, variance, kurtosis, etc.).

[0068] The input is a single-channel image that has been grayscaled, with pixel values ​​ranging from 0 to 255, and the distribution represents different brightness levels;

[0069] The image size is M×N, with a total of P=M×N pixels.

[0070] Initialize the grayscale histogram array H[i], where i∈[0,255], with a total of 256 gray levels;

[0071] Traverse all pixels in the image, and for each pixel grayscale value g, increase the corresponding count H[g] by 1;

[0072] The generated one-dimensional grayscale histogram H = [H0, H1, ..., H255]; each element represents the number of pixels of that grayscale level in the image.

[0073] Divide the frequency of each gray level by the total number of pixels P in the image to obtain the probability value p(i) of each gray level, i∈[0,255];

[0074] Generate a probability distribution array and obtain the normalized grayscale probability distribution vector: P gray =[p(0),p(1),...,p(255)], and

[0075] If some grayscale values ​​in the image do not exist at all (such as an image that is too dark or too bright), the corresponding p(i) = 0, and the mathematical anomaly of log(0) must be processed in the subsequent entropy calculation, for example, by adding a small constant ε for smoothing.

[0076] Output grayscale histogram H[i] and grayscale probability distribution p(i); store them as a callable data structure, such as array, vector or tensor form, for the next step of statistical calculation such as entropy and variance; and generate probability distribution graphs for visual analysis.

[0077] Based on the grayscale probability distribution p(i), a set of statistical feature parameters that reflect the image's grayscale structure and thermal distribution characteristics are extracted as key inputs to the power equipment condition assessment model. These features reveal the image's brightness level, contrast, complexity, and skewness, helping to determine whether there are thermal anomalies, defects, or operational abnormalities.

[0078] The following common grayscale statistical features are extracted, each of which is obtained by calculating the grayscale probability distribution p(i):

[0079] Mean: Reflects the overall brightness level of the image, similar to the average temperature or reflection intensity of the device surface.

[0080] Variance: measures the degree of discreteness of grayscale distribution and reflects whether the contrast or thermal distribution of the image is uniform.

[0081] Entropy: Indicates the amount of information and complexity of the grayscale distribution of an image. A higher value indicates a more complex thermal distribution. Abnormal hot spots often correspond to increased entropy values.

[0082] Skewness: measures the symmetry of grayscale distribution and is used to detect grayscale shift trends in images, such as abnormal bright or dark area shifts.

[0083] Kurtosis: measures the sharpness of the grayscale distribution and reflects whether local anomalies are concentrated, such as whether there are prominent hot spots.

[0084] Based on the image content, feature accuracy analysis parameters are further extracted, including the local grayscale consistency factor LGC and the edge structure integrity factor ESI, specifically:

[0085] The extraction method of local gray consistency factor LGC is:

[0086] Input the preprocessed grayscale image I, with a size of M×N and a pixel grayscale value range of [0, 255]; construct the image pyramid block and set L pyramid levels, for example, L = 3;

[0087] Each level l divides the image into 2 l ×2 l Sub-regions of equal size (i.e. the total number of regions is N l =4 l ); For the jth region of the lth layer, extract its pixel subset R l,j ; Calculate the mean grayscale value μ of pixels in the area l,j and variance

[0088] For each level l, find the mean variance V of all sub-regionsl , the expression is: Weighted combination of variance values ​​at different levels, setting the weight w of each level l , calculate the final local gray consistency factor LGC:

[0089] The larger the LGC, the more obvious the grayscale fluctuation in the local area of ​​the image, which may be caused by non-uniform lighting, reflection, highlight interference, etc. The smaller the LGC, the more consistent the local grayscale distribution of the image, the more stable the overall lighting and structure, and the higher the feature credibility.

[0090] The method for obtaining the edge structure integrity factor ESI is:

[0091] Input image I (grayscale image) and use edge detection algorithm to extract edge map E; edge detection algorithm includes: HED (Holistically-Nested Edge Detection) or Canny operator; obtain binary edge map E(x,y)∈{0,1}.

[0092] Divide the image into m×n grid blocks of equal size (recommended value: m=n=8 or 10);

[0093] Each grid block is denoted as B i,j , where i∈[1,m], j∈[1,n].

[0094] For each block B i,j , count the number of edge pixels e i,j :e i,j =∑(x,y)∈B i,j E(x,y); calculate edge density d i,j :

[0095] Divide the sparse area and the dense area, set the edge density threshold δ (recommended value: 0.02~0.05); if d i,j <δ, then area B i,j It is recorded as a sparse area; otherwise it is recorded as a dense area.

[0096] Let the total number of regions be N b =m×n, the number of sparse regions is N s ; ESI is defined as the proportion of non-sparse areas (i.e., structural integrity), and the expression is:

[0097] ESI=1 indicates that the structure is highly intact and the edges are continuous; ESI<0.6 indicates that there may be abnormalities such as structural fracture, reflection, and highlight interference. It can be used to assist in determining whether the image is affected by factors such as occlusion and hot spot overflow, which affect the status assessment results.

[0098] Construct an accuracy function that takes the LGC and ESI as input, outputs an accuracy score value, and divides the accuracy level according to a preset score threshold, specifically including:

[0099] The local grayscale consistency factor LGC and the edge structure integrity factor ESI are normalized so that they are both between [0, 1]. The accuracy score value is calculated based on the normalized local grayscale consistency factor LGC and the edge structure integrity factor ESI.

[0100] For example, the present invention can use the following formula to calculate the accuracy score value, and the calculation expression is: Where B is the accuracy score, a1 and a2 are the weight coefficients of the local grayscale consistency factor LGC and the edge structure integrity factor ESI (which can be optimized based on experimental experience or machine learning), and a1 and a2 are both greater than 0.

[0101] Based on the scoring results, the image feature accuracy is divided into three levels:

[0102] An accuracy score of ≥0.8 is classified as high accuracy, indicating good image quality, and features are directly used for evaluation;

[0103] 0.5≤Accuracy score<0.8 is classified as medium accuracy, indicating that the image has slight interference and needs selective correction;

[0104] An accuracy score value <0.5 is classified as low accuracy, indicating that the image interference is serious and a correction mechanism is implemented.

[0105] If the score is in the medium-to-low accuracy range, the score value is passed as an input signal to the subsequent image correction module; it can be used to control the correction intensity or dynamically adjust the processing path; if it is high accuracy, it directly enters the state evaluation stage.

[0106] When the accuracy score of an image is lower than a preset threshold (such as 0.5), it is determined that the current image has serious grayscale distribution abnormality or structural interference, and a grayscale correction operation needs to be performed.

[0107] A local entropy map is constructed to identify disturbed areas in the image, and the information stability is judged by the change of local entropy.

[0108] For each pixel (x, y) in the input image I, calculate the local entropy value H(x, y) within the neighborhood window (such as 5×5 or 7×7): H(x, y) = -∑ i p i ·log2(p i +ε); where p iis the probability of a pixel with grayscale value i in the neighborhood, and ε is a small constant. The output is a local entropy map Hlocal∈[0,1], which represents the information complexity of different regions of the image.

[0109] Abnormal area identification: perform threshold segmentation on Hlocal to identify high entropy areas (possibly noise or bright spots); at the same time, combine grayscale gradient or edge features (such as Sobel) to enhance boundary positioning; output the abnormal mask area M(x,y)∈{0,1}, where M=1 indicates the area that needs to be corrected.

[0110] Construct a graph structure and diffusion weight matrix, and model the image as an undirected graph G = (V, E) composed of pixels, where V is all pixels in the image; E is the edge between pixels (such as 4-neighborhood or 8-neighborhood connection);

[0111] The weight w of each edge uv Determined by grayscale difference and entropy similarity: Among them, σ and γ control the sensitivity of grayscale and entropy.

[0112] Perform diffusion mapping, select credible pixels in the non-abnormal area as source points; construct a diffusion matrix (random walk or Laplace matrix based on the graph); perform grayscale diffusion based on the graph structure, so that the credible grayscale value flows to the abnormal area and updates the regional pixel value I'(x,y). Using the graph Laplace diffusion model: I ′ =D -1 WI; where W is the graph weight matrix, D is the degree matrix, and I is the original image grayscale vector.

[0113] The pixel area where M(x,y)=1 in the image is replaced by the diffusion calculation result I'(x,y); the non-abnormal area keeps the original value; and the image Icorrected with continuous structure, consistent grayscale and suppressed artifacts is obtained.

[0114] Icorrected was input into the subsequent feature extraction process, and LGC, ESI, and accuracy scores were recalculated to confirm the correction effect.

[0115] After the image is corrected using the adaptive entropy-guided diffusion mapping method, the system uses the corrected image as a new input data source and re-executes the grayscale probability distribution statistics and statistical feature extraction steps to ensure that the features used for evaluation have higher accuracy and reliability. Specifically, the system reconstructs the grayscale histogram on the corrected image and normalizes it to generate a grayscale probability distribution. On this basis, it recalculates multiple grayscale statistical feature parameters of the image, including but not limited to mean, variance, entropy, skewness, and kurtosis. These feature parameters comprehensively reflect the brightness distribution, degree of thermal anomaly, and local structural characteristics of the image, and are important input for evaluating the operating status of power equipment.

[0116] The system then inputs these re-extracted feature values ​​into a pre-built state assessment module. This module can be implemented in two ways: one is based on statistical learning training models, such as support vector machines (SVMs), decision trees, and neural networks, which use classification models trained on a large number of labeled images to perform pattern recognition and classification judgments on the current features; the other is based on manually set judgment threshold rules, which divide the numerical distribution range of each feature value under normal and abnormal conditions to achieve rapid logical judgment.

[0117] Based on the output results of the model or rules, the system can classify the current operating status of the power equipment into levels such as "normal", "warning" or "fault", and can use the results to issue maintenance recommendations, alarm prompts or record them in the equipment operation and maintenance log, thereby realizing intelligent, controllable and highly robust real-time monitoring and evaluation of the power equipment status.

[0118] Example 2, please refer to Figure 2 As shown, the power equipment status assessment system based on grayscale probability distribution described in this embodiment includes an image acquisition and preprocessing module, a grayscale probability analysis module, a feature extraction module, an accuracy analysis module, an accuracy assessment module, an image correction module, and a status assessment and judgment module;

[0119] Image acquisition and preprocessing module: collects infrared images or visible light images of power equipment and preprocesses the images;

[0120] Grayscale probability analysis module: performs pixel grayscale statistics on the preprocessed image, constructs a grayscale histogram, and calculates the grayscale probability distribution;

[0121] Feature extraction module: extracts several statistical feature parameters of the image based on the grayscale probability distribution;

[0122] Accuracy analysis module: Based on the image content, it further extracts feature accuracy analysis parameters, including local grayscale consistency factor LGC and edge structure integrity factor ESI;

[0123] Accuracy evaluation module: constructs an accuracy function, takes the LGC and ESI as input, outputs an accuracy score value, and divides the accuracy level according to a preset score threshold;

[0124] Image correction module: when the accuracy score is lower than a preset threshold, executing a grayscale correction scheme of adaptive entropy-guided diffusion mapping;

[0125] State assessment and judgment module: re-extract statistical features from the corrected image, and evaluate the operating status of the power equipment based on the feature values ​​and the training model or threshold rules.

[0126] The above formulas are all dimensionless and numerical calculations. The formulas are obtained by collecting a large amount of data and performing software simulation to obtain the most recent real situation. The preset parameters in the formulas are set by technicians in this field according to actual conditions.

[0127] It should be understood that the term "and / or" as used herein simply describes a relationship between associated objects, indicating that three possible relationships exist. For example, "A and / or B" can represent: A alone, A and B together, or B alone. A and B can be singular or plural. Furthermore, the character " / " as used herein generally indicates an "or" relationship between the associated objects, but it may also indicate an "and / or" relationship. For specific understanding, please refer to the context.

[0128] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0129] The above is only a specific implementation method of the present application, but the scope of protection of the present application is not limited thereto. Any technician familiar with this technical field can easily think of changes or replacements within the technical scope disclosed in this application, which should be covered by the scope of protection of the present application.

Claims

1. A method for evaluating the state of power equipment based on grayscale probability distribution, characterized by: include: Collect infrared images or visible light images of power equipment and pre-process the images; Perform pixel grayscale statistics on the preprocessed image, construct a grayscale histogram, and calculate the grayscale probability distribution; Extracting a plurality of statistical characteristic parameters of the image based on the grayscale probability distribution; Based on the image content, feature accuracy analysis parameters are further extracted, including the local grayscale consistency factor LGC and the edge structure integrity factor ESI; Constructing an accuracy function, taking the LGC and ESI as input, outputting an accuracy score value, and dividing the accuracy level according to a preset score threshold; When the accuracy score is lower than a preset threshold, performing a grayscale correction scheme of an adaptive entropy-guided diffusion mapping; Statistical features are re-extracted from the corrected image, and the operating status of the power equipment is evaluated based on the feature values ​​and the training model or threshold rules.

2. The method for evaluating the state of electric power equipment based on grayscale probability distribution according to claim 1, wherein: The calculation of the grayscale probability distribution includes: The grayscale value of each pixel in the image is counted to generate a grayscale histogram H[i], where i∈[0,255]; the number of pixels at each grayscale level is divided by the total number of pixels to obtain the normalized grayscale probability distribution p(i); the grayscale probability distribution result is output in the form of an array or tensor.

3. The method for evaluating the state of electric power equipment based on grayscale probability distribution according to claim 1, wherein: The statistical characteristic parameters include: grayscale mean, variance, entropy, skewness and kurtosis of the image; each statistical parameter is obtained by weighted calculation of the normalized grayscale probability distribution p(i).

4. The method for evaluating the state of electric power equipment based on grayscale probability distribution according to claim 1, wherein: The extraction method of local gray consistency factor LGC is: Input the preprocessed grayscale image I, with a size of M×N and a pixel grayscale value range of [0, 255]; construct the image pyramid block and set L pyramid levels; Each level l divides the image into 2 l ×2 l sub-regions of equal size; for the j-th region of the l-th layer, extract its pixel subset R l,j ; Calculate the mean grayscale value μ of pixels in the area l,j and variance For each level l, find the mean variance V of all sub-regions l , the expression is: Weighted combination of variance values ​​at different levels, setting the weight w of each level l , calculate the final local gray consistency factor LGC:

5. The method for evaluating the state of electric power equipment based on grayscale probability distribution according to claim 4, characterized in that: The method for obtaining the edge structure integrity factor ESI is: Input image I, use edge detection algorithm to extract edge map E, and obtain binary edge map E(x,y)∈{0,1}; divide the image into m×n grid blocks of equal size; each grid block is denoted as B i,j , where i∈[1,m], j∈[1,n]; for each block B i,j , count the number of edge pixels e i,j : Calculate edge density d i,j : Divide the sparse area and the dense area, set the edge density threshold δ; if d i,j <δ, then area B i,j Recorded as sparse area; Otherwise it is recorded as a dense area; let the total number of areas be N b =m×n, the number of sparse regions is N s ; definition ESI is the proportion of non-sparse areas, and its expression is:

6. The method for evaluating the state of electric power equipment based on grayscale probability distribution according to claim 5, characterized in that: Construct an accuracy function that takes the LGC and ESI as input, outputs an accuracy score value, and divides the accuracy level according to a preset score threshold, specifically including: The local grayscale consistency factor LGC and the edge structure integrity factor ESI are normalized so that they are both between [0, 1]. The accuracy score value is calculated based on the normalized local grayscale consistency factor LGC and the edge structure integrity factor ESI.

7. The method for evaluating the state of electric power equipment based on grayscale probability distribution according to claim 6, characterized in that: Based on the scoring results, the image feature accuracy is divided into three levels: An accuracy score of ≥0.8 is classified as high accuracy, indicating good image quality, and features are directly used for evaluation; 0.5≤Accuracy score<0.8 is classified as medium accuracy, indicating that the image has slight interference and needs selective correction; An accuracy score value <0.5 is classified as low accuracy, indicating that the image interference is serious and a correction mechanism is implemented.

8. The method for evaluating the state of electric power equipment based on grayscale probability distribution according to claim 7, characterized in that: The grayscale correction scheme of the adaptive entropy-guided diffusion mapping includes: Construct a local entropy map of the image and calculate the information entropy value for each pixel neighborhood; Generate abnormal area masks based on the entropy map to identify areas that need repair; Construct a graph structure model with pixels as nodes and edge weights that fuse grayscale difference and entropy difference; Based on the graph Laplace diffusion model, the grayscale value is propagated from the credible area to the abnormal area, and the corrected image is output.

9. A power equipment status assessment system based on grayscale probability distribution, used to implement the power equipment status assessment method based on grayscale probability distribution according to any one of claims 1 to 8, characterized in that: It includes image acquisition and preprocessing module, grayscale probability analysis module, feature extraction module, accuracy analysis module, accuracy assessment module, image correction module and state assessment and judgment module; Image acquisition and preprocessing module: collects infrared images or visible light images of power equipment and preprocesses the images; Grayscale probability analysis module: performs pixel grayscale statistics on the preprocessed image, constructs a grayscale histogram, and calculates the grayscale probability distribution; Feature extraction module: extracts several statistical feature parameters of the image based on the grayscale probability distribution; Accuracy analysis module: Based on the image content, it further extracts feature accuracy analysis parameters, including local grayscale consistency factor LGC and edge structure integrity factor ESI; Accuracy evaluation module: constructs an accuracy function, takes the LGC and ESI as input, outputs an accuracy score value, and divides the accuracy level according to a preset score threshold; Image correction module: when the accuracy score is lower than a preset threshold, executing a grayscale correction scheme of adaptive entropy-guided diffusion mapping; State assessment and judgment module: re-extract statistical features from the corrected image, and evaluate the operating status of the power equipment based on the feature values ​​and the training model or threshold rules.

Images