Feeder line fault monitoring and alarming method and system based on machine vision

Abstract

The invention discloses a feeder line fault monitoring and warning method and system based on machine vision, and relates to the technical field of warning devices.The method comprises the following steps that an enhanced image is generated based on a feeder line equipment image, and quality optimization is conducted on the enhanced image under the typical weather condition through a self-adaptive illumination compensation model; obtaining standardized feeder line equipment image data; performing feature extraction by using a deep learning model, establishing a visual feature vector of the feeder equipment according to the extracted texture features, color features and shape features, and analyzing the change trend of the visual feature vector through a time sequence analysis algorithm to obtain an aging degree evaluation index of the feeder equipment; time sequence prediction and risk assessment are carried out, and a fuzzy inference system is combined with an expert knowledge base to generate graded alarm information. According to the invention, comprehensive monitoring of the feeder equipment is realized through a machine vision technology, and continuity and reliability of state monitoring of the feeder equipment are ensured.

Description

Technical Field

[0001] The present invention relates to the technical field of alarm devices, and in particular to a feeder fault monitoring and alarm method and system based on machine vision. Background Art

[0002] As power systems continue to expand in size and complexity, feeders, as key components of power transmission, are crucial for the reliability of the entire power grid. Feeder failures can not only cause localized power outages but can also trigger chain reactions, leading to widespread blackouts. Therefore, effective monitoring and timely warning of feeder equipment have become crucial tasks in power system operations and maintenance.

[0003] Currently, feeder fault monitoring primarily relies on regular manual inspections and traditional electrical parameter measurement methods. These methods include using infrared thermal imagers to detect feeder temperature anomalies, using partial discharge detectors to monitor insulation performance, and analyzing feeder operating conditions using electrical parameters such as current and voltage. With technological advancements, automated monitoring systems are gradually being implemented, such as remote monitoring based on Supervisory Control and Data Acquisition (SCADA) systems and distributed temperature monitoring using fiber optic sensors.

[0004] However, existing feeder fault monitoring technology still has some shortcomings. First, manual inspections are limited by human resources and inspection frequency, making it difficult to achieve real-time, continuous monitoring of feeder equipment. Second, while traditional electrical parameter measurement methods can partially reflect the operating status of the feeder, they cannot fully capture subtle changes in the appearance of feeder equipment, such as discoloration of the insulation layer, rust on metal components, and other early signs of degradation. Finally, existing automated monitoring systems mainly focus on detecting and locating feeder faults when they occur, and lack continuous tracking and early warning mechanisms for the progressive aging characteristics of feeders during long-term use. These shortcomings limit the early detection of feeder faults and the implementation of preventive maintenance.

[0005] Currently, no effective solutions have been proposed for the problems in related technologies. Summary of the Invention

[0006] In response to the problems in the related art, the present invention proposes a feeder fault monitoring and alarm method and system based on machine vision, which has the advantages of realizing comprehensive monitoring of feeder equipment through machine vision technology, ensuring the continuity and reliability of feeder equipment status monitoring, and thus solving the problems in the existing technology that it is difficult to realize real-time and continuous monitoring of feeder equipment, it is difficult to fully capture the slight appearance changes of feeder equipment, and there is a lack of continuous tracking and early warning mechanism for the progressive aging characteristics of feeders during long-term use.

[0007] To this end, the specific technical solutions adopted in the present invention are as follows:

[0008] According to one aspect of the present invention, a feeder fault monitoring and alarm method based on machine vision is provided. The feeder fault monitoring and alarm method based on machine vision comprises the following steps:

[0009] S1. Based on the feeder equipment images captured by industrial cameras, an image fusion algorithm is used to generate enhanced images. The quality of the enhanced images under typical weather conditions is optimized using an adaptive illumination compensation model to obtain standardized feeder equipment image data.

[0010] S2. Use a deep learning model to extract features from standardized feeder equipment image data, establish a visual feature vector for the feeder equipment based on the extracted texture features, color features, and shape features, and analyze the changing trend of the visual feature vector using a time series analysis algorithm to obtain an aging assessment index for the feeder equipment;

[0011] S3. Perform time series prediction and risk assessment based on the aging evaluation indicators of feeder equipment, and use the fuzzy reasoning system combined with the expert knowledge base to generate graded alarm information.

[0012] Furthermore, the feeder equipment image includes the feeder terminal body image, the aerial plug interface image and the opening and closing position indication device image;

[0013] The enhanced images under typical weather conditions include feeder equipment enhanced images on sunny days, feeder equipment enhanced images on rainy days, feeder equipment enhanced images on foggy days, and feeder equipment enhanced images at night.

[0014] Furthermore, based on the feeder equipment images captured by the industrial camera, an enhanced image is generated using an image fusion algorithm, and the quality of the enhanced image under typical weather conditions is optimized using an adaptive illumination compensation model. Obtaining standardized feeder equipment image data includes the following steps:

[0015] S11. Establish an original image dataset based on the feeder terminal body image, the aerial plug interface image, and the opening and closing position indication device image captured by the industrial camera;

[0016] S12. Perform feature-level fusion on the original image dataset using an image fusion algorithm, and combine it with real-time meteorological data to generate enhanced images of feeder equipment on sunny days, enhanced images of feeder equipment on rainy days, enhanced images of feeder equipment on foggy days, and enhanced images of feeder equipment at night;

[0017] S13. Optimize the quality of the feeder equipment enhanced image on a sunny day, the feeder equipment enhanced image on a rainy day, the feeder equipment enhanced image on a foggy day, and the feeder equipment enhanced image at night using an adaptive illumination compensation model to obtain standardized feeder equipment image data.

[0018] Furthermore, the original image dataset is fused at the feature level by an image fusion algorithm, and combined with real-time meteorological data, to generate enhanced images of feeder equipment on sunny days, enhanced images of feeder equipment on rainy days, enhanced images of feeder equipment on foggy days, and enhanced images of feeder equipment at night, respectively, including the following steps:

[0019] S121. Perform feature-level fusion on the original image dataset based on a multi-scale decomposition and weighted average fusion method, and combine it with clear-day meteorological data to generate a clear-day feeder equipment enhanced image.

[0020] S122. Processing the original image dataset using a rain line detection and removal algorithm, and combining it with rainy day meteorological data to generate an enhanced image of the feeder equipment on a rainy day;

[0021] S123, optimizing the original image dataset using an improved dark channel prior defogging algorithm, and combining it with foggy meteorological data to generate an enhanced image of the feeder equipment in foggy weather;

[0022] S124. Process the original image data set using an adaptive brightness enhancement and noise suppression algorithm, and combine it with nighttime meteorological data to generate a nighttime feeder equipment enhanced image.

[0023] Furthermore, the expression of the adaptive illumination compensation model is:

[0024] ;

[0025] Where x is the horizontal coordinate of the pixel in the image, y is the vertical coordinate of the pixel in the image, and I c (x, y) is the compensated feeder device image, R(x, y) is the reflection component, L'(x, y) is the adjusted illumination component, I(x, y) is the pixel value of the original feeder device image at coordinate (x, y), L(x, y) is the estimated original illumination component, α is the illumination intensity adjustment coefficient, and β is the illumination offset adjustment parameter.

[0026] Furthermore, a deep learning model is used to extract features from the standardized feeder equipment image data. A visual feature vector of the feeder equipment is established based on the extracted texture features, color features, and shape features. The changing trend of the visual feature vector is analyzed by a time series analysis algorithm. The aging degree assessment index of the feeder equipment is obtained, which includes the following steps:

[0027] S21. Extract features from the standardized feeder equipment image data based on a pre-trained deep convolutional neural network to obtain a multi-level feature map;

[0028] S22. Establishing a visual feature vector of the feeder device based on the multi-level feature map; wherein the visual feature vector of the feeder device includes texture features, color features, and shape features;

[0029] S23. Use a time series analysis algorithm to perform trend analysis on the visual feature vectors at consecutive time points to generate an aging evaluation index for the feeder equipment.

[0030] Furthermore, establishing a visual feature vector of the feeder device based on the multi-level feature map includes the following steps:

[0031] S221. Obtain texture features of the feeder device image using a gray-level co-occurrence matrix and a local binary pattern algorithm;

[0032] S222. Extracting color features of the feeder device image based on color moment and color histogram methods;

[0033] S223. Generate shape features of the feeder device image through contour analysis and morphological operations;

[0034] S224: Construct a visual feature vector of the feeder device based on the texture features, color features, and shape features of the feeder device image.

[0035] Furthermore, the expression of the aging evaluation index of the feeder equipment is:

[0036] ;

[0037] ;

[0038] ;

[0039] ;

[0040] Where, I is the aging evaluation index of feeder equipment, T i (t), C i (t) and S i (t) are the i-th component of the texture feature, color feature and shape feature of the feeder equipment image at time point t, N is the dimension of the feature vector, ΔT(t), ΔC(t) and ΔS(t) are the feature change rates of the texture feature, color feature and shape feature of the feeder equipment image at T consecutive time points, respectively. t 、w c and w s are texture feature weights, color feature weights, and shape feature weights, respectively, and w t +w c +w s =1.

[0041] Furthermore, based on the aging evaluation index of the feeder equipment, time series prediction and risk assessment are performed, and the fuzzy inference system is combined with the expert knowledge base to generate graded alarm information, including the following steps:

[0042] S31. Use a long short-term memory network to perform time series prediction on the aging evaluation index of the feeder equipment to obtain an aging trend prediction result;

[0043] S32. Based on the historical fault data of the feeder equipment and the aging evaluation index of the feeder equipment, a feeder fault risk assessment model is constructed using support vector machine technology to generate a fault risk probability;

[0044] S33. Combine the aging trend prediction results and the failure risk probability, conduct a comprehensive analysis through the fuzzy reasoning system, match them with the pre-established expert knowledge base, and output graded alarm information.

[0045] According to another aspect of the present invention, a feeder fault monitoring and alarm system based on machine vision is further provided. The feeder fault monitoring and alarm system based on machine vision includes:

[0046] The image enhancement and optimization processing unit is used to generate enhanced images based on the feeder equipment images captured by industrial cameras using an image fusion algorithm. The image quality of the enhanced images under typical weather conditions is optimized using an adaptive illumination compensation model to obtain standardized feeder equipment image data.

[0047] The feature extraction and aging assessment unit is used to extract features from standardized feeder equipment image data using a deep learning model, establish visual feature vectors of the feeder equipment based on the extracted texture features, color features, and shape features, and analyze the changing trend of the visual feature vectors using a time series analysis algorithm to obtain an aging assessment index for the feeder equipment;

[0048] The multi-level warning information generation unit is used to perform time series prediction and risk assessment based on the aging evaluation indicators of the feeder equipment, and use the fuzzy reasoning system combined with the expert knowledge base to generate graded alarm information.

[0049] Compared with the prior art, the present invention has the following beneficial effects:

[0050] (1) The present invention realizes comprehensive monitoring of feeder equipment through machine vision technology, uses industrial cameras to collect images of the feeder terminal body, aerial plug interface and opening and closing position indication device, constructs a complete original image data set, and uses image fusion algorithm and adaptive illumination compensation model to enhance and optimize images under different weather conditions to obtain standardized feeder equipment image data; this image acquisition and processing method based on machine vision effectively solves the problem of unstable image quality of traditional visual detection methods in complex environments, and significantly improves the accuracy of subsequent feature extraction and analysis; especially in adverse weather conditions such as rainy days, foggy days and nighttime, the present invention can maintain high image quality, ensure the continuity and reliability of feeder equipment status monitoring, and thus provide a strong guarantee for the safe and stable operation of the power system.

[0051] (2) The present invention combines deep learning with machine vision technology, uses a deep learning model to extract features from standardized feeder equipment images, obtains multi-level feature maps, and extracts comprehensive texture, color and shape features through machine vision algorithms to construct visual feature vectors of feeder equipment. In addition, the aging degree assessment index of feeder equipment is obtained by using a time series analysis algorithm, combined with a long short-term memory network for time series prediction and a fault risk assessment model constructed by a support vector machine, and finally outputs graded alarm information through a fuzzy reasoning system and an expert knowledge base. This intelligent analysis method based on machine vision can not only accurately assess the current status of feeder equipment, but also predict its future aging trend and potential fault risks, greatly improving the preventive maintenance capability of the power system, effectively reducing the incidence of feeder failures, and extending the service life of the equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] In order to more clearly illustrate the embodiments of the present invention 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 of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0053] Figure 1 1 is a flow chart of a feeder fault monitoring and alarm method based on machine vision according to an embodiment of the present invention;

[0054] Figure 2 The figure is a principle block diagram of a feeder fault monitoring and alarm system based on machine vision according to an embodiment of the present invention. DETAILED DESCRIPTION

[0055] To further illustrate each embodiment, the present invention provides drawings, which are part of the disclosure of the present invention. They are mainly used to illustrate the embodiments and can be used in conjunction with the relevant descriptions in the specification to explain the operating principles of the embodiments. By referring to these contents, ordinary technicians in this field should be able to understand other possible implementation methods and advantages of the present invention. The components in the figures are not drawn to scale, and similar component symbols are generally used to represent similar components.

[0056] According to an embodiment of the present invention, a feeder fault monitoring and alarm method and system based on machine vision are provided.

[0057] The present invention will now be further described with reference to the accompanying drawings and specific embodiments. Figure 1 As shown, according to one embodiment of the present invention, a feeder fault monitoring and alarm method based on machine vision is provided, and the feeder fault monitoring and alarm method based on machine vision includes the following steps:

[0058] S1. Based on the feeder equipment images captured by industrial cameras, an image fusion algorithm is used to generate enhanced images. The quality of the enhanced images under typical weather conditions is optimized using an adaptive illumination compensation model to obtain standardized feeder equipment image data.

[0059] S2. Use a deep learning model to extract features from standardized feeder equipment image data, establish a visual feature vector for the feeder equipment based on the extracted texture features, color features, and shape features, and analyze the changing trend of the visual feature vector using a time series analysis algorithm to obtain an aging assessment index for the feeder equipment;

[0060] S3. Perform time series prediction and risk assessment based on the aging evaluation indicators of feeder equipment, and use the fuzzy reasoning system combined with the expert knowledge base to generate graded alarm information.

[0061] In one embodiment, the feeder equipment image includes an image of the feeder terminal body, an image of the aerial plug interface, and an image of the opening and closing position indicating device;

[0062] The enhanced images under typical weather conditions include feeder equipment enhanced images on sunny days, feeder equipment enhanced images on rainy days, feeder equipment enhanced images on foggy days, and feeder equipment enhanced images at night.

[0063] In one embodiment, based on feeder equipment images captured by industrial cameras, an enhanced image is generated using an image fusion algorithm, and the quality of the enhanced image under typical weather conditions is optimized using an adaptive illumination compensation model to obtain standardized feeder equipment image data, including the following steps:

[0064] S11. Establish an original image dataset based on the feeder terminal body image, the aerial plug interface image, and the opening and closing position indication device image captured by the industrial camera;

[0065] S12. Perform feature-level fusion on the original image dataset using an image fusion algorithm, and combine it with real-time meteorological data to generate enhanced images of feeder equipment on sunny days, enhanced images of feeder equipment on rainy days, enhanced images of feeder equipment on foggy days, and enhanced images of feeder equipment at night;

[0066] S13. Optimize the quality of the feeder equipment enhanced image on a sunny day, the feeder equipment enhanced image on a rainy day, the feeder equipment enhanced image on a foggy day, and the feeder equipment enhanced image at night using an adaptive illumination compensation model to obtain standardized feeder equipment image data.

[0067] In one embodiment, the image fusion algorithm is used to perform feature-level fusion on the original image dataset, and combined with real-time meteorological data, to generate enhanced images of feeder equipment on sunny days, enhanced images of feeder equipment on rainy days, enhanced images of feeder equipment on foggy days, and enhanced images of feeder equipment at night, respectively, including the following steps:

[0068] S121. Perform feature-level fusion on the original image dataset based on a multi-scale decomposition and weighted average fusion method, and combine it with clear-day meteorological data to generate a clear-day feeder equipment enhanced image.

[0069] Specifically, the multi-scale decomposition and weighted average fusion method in S121 utilizes the Non-Subsampled Contourlet Transform (NSCT) technique. An adaptive weighted average method based on regional energy is used for low-frequency subbands, while a selective fusion rule based on local Laplacian energy is employed for high-frequency subbands. An inverse NSCT transform is used to generate an enhanced image of the clear-sky feeder equipment, effectively improving image clarity and contrast.

[0070] S122. Processing the original image dataset using a rain line detection and removal algorithm, and combining it with rainy day meteorological data to generate an enhanced image of the feeder equipment on a rainy day;

[0071] Specifically, the rain line detection and removal algorithm in S122 combines an improved Sobel operator with a deep learning network. It first detects rain line features, then performs rain removal processing. Finally, a guided filter is applied to restore details, generating an enhanced image of the feeder equipment on rainy days. This effectively removes rain line interference while preserving the original structural information.

[0072] S123, optimizing the original image dataset using an improved dark channel prior defogging algorithm, and combining it with foggy meteorological data to generate an enhanced image of the feeder equipment in foggy weather;

[0073] Specifically, the improved dark channel prior dehazing algorithm in S123 combines adaptive gamma correction and local contrast enhancement. Guided filtering is used to optimize transmittance estimation, adaptive gamma correction is applied to adjust global brightness, and finally, a local contrast enhancement algorithm (CLAHE) is used to generate an enhanced image of the feeder equipment in foggy weather, effectively removing the effects of haze and avoiding over-enhancement.

[0074] S124. Process the original image data set using an adaptive brightness enhancement and noise suppression algorithm, and combine it with nighttime meteorological data to generate a nighttime feeder equipment enhanced image.

[0075] Specifically, the S124's adaptive brightness enhancement and noise suppression algorithms are based on a multi-scale approach based on Retinex theory. Nonlinear mapping is performed on the separated illumination components, and a modified bilateral filter is applied to the reflectance components to suppress noise. These components are then recombined and color restored to generate enhanced images of nighttime feeder equipment, effectively improving nighttime image quality and suppressing noise.

[0076] In one embodiment, the adaptive illumination compensation model is expressed as:

[0077] ;

[0078] Where x is the horizontal coordinate of the pixel in the image, y is the vertical coordinate of the pixel in the image, and I c (x, y) is the compensated feeder device image, R(x, y) is the reflection component, L'(x, y) is the adjusted illumination component, I(x, y) is the pixel value of the original feeder device image at coordinate (x, y), L(x, y) is the estimated original illumination component, α is the illumination intensity adjustment coefficient, and β is the illumination offset adjustment parameter.

[0079] In one embodiment, a deep learning model is used to extract features from standardized feeder equipment image data, a visual feature vector of the feeder equipment is established based on the extracted texture features, color features, and shape features, and a time series analysis algorithm is used to analyze the changing trend of the visual feature vector to obtain an aging assessment index for the feeder equipment. The following steps are included:

[0080] S21. Extract features from the standardized feeder equipment image data based on a pre-trained deep convolutional neural network to obtain a multi-level feature map;

[0081] S22. Establishing a visual feature vector of the feeder device based on the multi-level feature map; wherein the visual feature vector of the feeder device includes texture features, color features, and shape features;

[0082] S23. Use a time series analysis algorithm to perform trend analysis on the visual feature vectors at consecutive time points to generate an aging evaluation index for the feeder equipment.

[0083] Specifically, the time series analysis algorithm uses a sliding window autoregressive integrated moving average (ARIMA) model. First, the visual feature vectors at consecutive time points are differentiated to stabilize the time series. Then, the order of the ARIMA model is determined through autocorrelation function (ACF) and partial autocorrelation function (PACF) analysis. Model parameters are estimated using maximum likelihood estimation, and the optimal model is selected using the Akaike Information Criterion (AIC). Finally, the trained ARIMA model is used to predict feature vectors at several future time points. By comparing the deviation between the predicted and actual values, the aging trend of the feeder equipment is determined.

[0084] In one embodiment, establishing a visual feature vector of a feeder device based on a multi-level feature map includes the following steps:

[0085] S221. Obtain texture features of the feeder device image using a gray-level co-occurrence matrix and a local binary pattern algorithm;

[0086] Specifically, texture feature extraction in the S221 utilizes the Gray Level Co-occurrence Matrix (GLCM) and Local Binary Pattern (LBP) algorithms from machine vision. GLCM calculates statistics such as energy, contrast, entropy, and correlation to capture global texture information. LBP compares the grayscale values ​​of a central pixel with those of its neighbors to generate a binary code that describes the local texture structure. Combining these two methods can comprehensively characterize subtle texture variations on the surface of feeder equipment, such as fine lines caused by aging insulation or minor corrosion marks on metal components.

[0087] S222. Extracting color features of the feeder device image based on color moment and color histogram methods;

[0088] Specifically, color feature extraction in S222 utilizes color moment and color histogram methods from machine vision. Color moments calculate the first-order moment (mean), second-order moment (variance), and third-order moment (skewness) of an image in the RGB or HSV color space, describing the overall characteristics of the color distribution. Color histograms count the number of pixels for each color component, reflecting the local distribution of color. These features effectively capture color variations on the surface of feeder equipment, such as fading insulation or hue changes caused by metal oxidation.

[0089] S223. Generate shape features of the feeder device image through contour analysis and morphological operations;

[0090] Specifically, shape feature generation in S223 utilizes contour analysis and morphological operations in machine vision. The Canny edge detection algorithm is first used to extract the outline of the feeder equipment. Geometric features such as perimeter, area, and circularity are then calculated. Morphological operations such as opening and closing are used to remove noise and fill small holes. These techniques accurately describe geometric variations in the feeder equipment, such as connector deformation or insulation expansion.

[0091] S224. Construct a visual feature vector of the feeder device according to the texture features, color features, and shape features of the feeder device image.

[0092] Specifically, the visual feature vector constructed in S224 is constructed by normalizing and concatenating the features obtained in the previous three steps. Specifically, texture features (GLCM statistics and LBP histograms), color features (color moments and color histograms), and shape features (geometric parameters) are dimensionally normalized and then concatenated into a high-dimensional vector. This combined visual feature vector comprehensively characterizes the appearance of the feeder equipment, providing a reliable data foundation for subsequent condition assessment and fault warning.

[0093] In one embodiment, the expression for the aging evaluation index of the feeder equipment is:

[0094] ;

[0095] ;

[0096] ;

[0097] ;

[0098] Where, I is the aging evaluation index of feeder equipment, T i (t), C i (t) and S i (t) are the i-th component of the texture feature, color feature and shape feature of the feeder equipment image at time point t, N is the dimension of the feature vector, ΔT(t), ΔC(t) and ΔS(t) are the feature change rates of the texture feature, color feature and shape feature of the feeder equipment image at T consecutive time points, respectively. t 、w c and w s are texture feature weights, color feature weights, and shape feature weights, respectively, and w t +w c +w s =1.

[0099] In one embodiment, performing time series prediction and risk assessment based on the aging evaluation index of feeder equipment and utilizing a fuzzy inference system combined with an expert knowledge base to generate graded alarm information includes the following steps:

[0100] S31. Use a long short-term memory network to perform time series prediction on the aging evaluation index of the feeder equipment to obtain an aging trend prediction result;

[0101] Specifically, the LSTM time series prediction model in S31 employs a multi-layered architecture, consisting of an input layer, multiple LSTM hidden layers, and a fully connected output layer. The model is trained using the Adam optimizer and the Mean Separation Error (MSE) loss function, and uses a sliding window method to predict aging indicators for M future time points. Exponential smoothing techniques are applied post-processing to improve prediction stability and effectively capture long-term trends and cyclical variations in feeder equipment aging.

[0102] S32. Based on the historical fault data of the feeder equipment and the aging evaluation index of the feeder equipment, a feeder fault risk assessment model is constructed using support vector machine technology to generate a fault risk probability;

[0103] Specifically, the SVM fault risk assessment model in S32 uses the RBF kernel function. Parameters are optimized through grid search and cross-validation, and the SMOTE algorithm is used to address sample imbalance. Platt scaling is used to convert the SVM output into probabilities, and SHAP values ​​are introduced to analyze feature importance. This method accurately assesses the fault risk of feeder equipment and provides a quantitative analysis of risk factors.

[0104] S33. Combine the aging trend prediction results and the failure risk probability, conduct a comprehensive analysis through the fuzzy reasoning system, match them with the pre-established expert knowledge base, and output graded alarm information.

[0105] Specifically, the fuzzy inference system in S33 employs the Mamdani inference model. First, the aging trend prediction results and failure risk probability are used as input variables to define corresponding fuzzy sets and membership functions. Then, a fuzzy rule base is designed based on expert knowledge, such as "If the aging trend is rapid and the failure risk is high, the alarm level is urgent." Fuzzy inference is performed using the min-max synthesis method to obtain fuzzy outputs, which are then defuzzified using the center of gravity method to determine the alarm level. Combined with an expert knowledge base containing typical failure modes and treatment recommendations, detailed alarm information is generated through semantic matching.

[0106] like Figure 2 According to another embodiment of the present invention, a feeder fault monitoring and alarm system based on machine vision is provided. The feeder fault monitoring and alarm system based on machine vision includes:

[0107] The image enhancement and optimization processing unit 1 is used to generate an enhanced image based on the feeder equipment image captured by the industrial camera using an image fusion algorithm, and optimize the quality of the enhanced image under typical weather conditions using an adaptive illumination compensation model to obtain standardized feeder equipment image data;

[0108] Feature extraction and aging assessment unit 2, used to extract features from standardized feeder equipment image data using a deep learning model, establish a visual feature vector of the feeder equipment based on the extracted texture features, color features, and shape features, and analyze the changing trend of the visual feature vector using a time series analysis algorithm to obtain an aging assessment index for the feeder equipment;

[0109] The multi-level warning information generating unit 3 is used to perform time series prediction and risk assessment based on the aging evaluation index of the feeder equipment, and generate graded warning information by using a fuzzy reasoning system combined with an expert knowledge base.

[0110] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A feeder fault monitoring and alarm method based on machine vision, characterized in that: The following steps are involved: S1. Based on the feeder equipment images captured by industrial cameras, an image fusion algorithm is used to generate enhanced images. The quality of the enhanced images under typical weather conditions is optimized using an adaptive illumination compensation model to obtain standardized feeder equipment image data. S2. Use a deep learning model to extract features from standardized feeder equipment image data, establish a visual feature vector for the feeder equipment based on the extracted texture features, color features, and shape features, and analyze the changing trend of the visual feature vector using a time series analysis algorithm to obtain an aging assessment index for the feeder equipment; S3. Perform time series prediction and risk assessment based on the aging evaluation indicators of feeder equipment, and use the fuzzy reasoning system combined with the expert knowledge base to generate graded alarm information.

2. A feeder fault monitoring and alarm method based on machine vision according to claim 1, characterized in that: The feeder equipment image includes the feeder terminal body image, the aerial plug interface image and the opening and closing position indication device image; The enhanced images under typical weather conditions include a clear day feeder equipment enhanced image, a rainy day feeder equipment enhanced image, a foggy day feeder equipment enhanced image, and a nighttime feeder equipment enhanced image.

3. A feeder fault monitoring and alarm method based on machine vision according to claim 2, characterized in that: The method of generating an enhanced image based on the feeder device image captured by the industrial camera using an image fusion algorithm, and optimizing the quality of the enhanced image under typical weather conditions using an adaptive illumination compensation model to obtain standardized feeder device image data includes the following steps: S11. Establish an original image dataset based on the feeder terminal body image, the aerial plug interface image, and the opening and closing position indication device image captured by the industrial camera; S12. Perform feature-level fusion on the original image dataset using an image fusion algorithm, and combine it with real-time meteorological data to generate enhanced images of feeder equipment on sunny days, enhanced images of feeder equipment on rainy days, enhanced images of feeder equipment on foggy days, and enhanced images of feeder equipment at night; S13. Optimize the quality of the feeder equipment enhanced image on a sunny day, the feeder equipment enhanced image on a rainy day, the feeder equipment enhanced image on a foggy day, and the feeder equipment enhanced image at night using an adaptive illumination compensation model to obtain standardized feeder equipment image data.

4. A feeder fault monitoring and alarm method based on machine vision according to claim 3, characterized in that: The feature-level fusion of the original image data set by the image fusion algorithm and the combination with real-time meteorological data to generate a clear day feeder equipment enhanced image, a rainy day feeder equipment enhanced image, a foggy day feeder equipment enhanced image, and a nighttime feeder equipment enhanced image respectively include the following steps: S121. Perform feature-level fusion on the original image dataset based on a multi-scale decomposition and weighted average fusion method, and combine it with clear-day meteorological data to generate a clear-day feeder equipment enhanced image. S122. Processing the original image dataset using a rain line detection and removal algorithm, and combining it with rainy day meteorological data to generate an enhanced image of the feeder equipment on a rainy day; S123, optimizing the original image dataset using an improved dark channel prior defogging algorithm, and combining it with foggy meteorological data to generate an enhanced image of the feeder equipment in foggy weather; S124. Process the original image data set using an adaptive brightness enhancement and noise suppression algorithm, and combine it with nighttime meteorological data to generate a nighttime feeder equipment enhanced image.

5. The feeder fault monitoring and alarm method based on machine vision according to claim 4 is characterized in that: The expression of the adaptive illumination compensation model is: ； Where x is the horizontal coordinate of the pixel in the image, y is the vertical coordinate of the pixel in the image, and I c (x, y) is the compensated feeder device image, R(x, y) is the reflection component, L'(x, y) is the adjusted illumination component, I(x, y) is the pixel value of the original feeder device image at coordinate (x, y), L(x, y) is the estimated original illumination component, α is the illumination intensity adjustment coefficient, and β is the illumination offset adjustment parameter.

6. The feeder fault monitoring and alarm method based on machine vision according to claim 1 is characterized in that: The method of extracting features from standardized feeder equipment image data using a deep learning model, establishing a visual feature vector of the feeder equipment based on the extracted texture features, color features, and shape features, and analyzing the changing trend of the visual feature vector using a time series analysis algorithm to obtain an aging evaluation index for the feeder equipment includes the following steps: S21. Extract features from the standardized feeder equipment image data based on a pre-trained deep convolutional neural network to obtain a multi-level feature map; S22. Establishing a visual feature vector of the feeder device based on the multi-level feature map; wherein the visual feature vector of the feeder device includes texture features, color features, and shape features; S23. Use a time series analysis algorithm to perform trend analysis on the visual feature vectors at consecutive time points to generate an aging evaluation index for the feeder equipment.

7. The feeder fault monitoring and alarm method based on machine vision according to claim 6 is characterized in that: The method of establishing a visual feature vector of a feeder device according to a multi-level feature map comprises the following steps: S221. Obtain texture features of the feeder device image using a gray-level co-occurrence matrix and a local binary pattern algorithm; S222. Extracting color features of the feeder device image based on color moment and color histogram methods; S223. Generate shape features of the feeder device image through contour analysis and morphological operations; S224: Construct a visual feature vector of the feeder device based on the texture features, color features, and shape features of the feeder device image.

8. The feeder fault monitoring and alarm method based on machine vision according to claim 6 is characterized in that: The expression of the aging evaluation index of the feeder equipment is: ； ； ； ； Where, I is the aging evaluation index of feeder equipment, T i (t), C i (t) and S i (t) are the i-th component of the texture feature, color feature and shape feature of the feeder equipment image at time point t, N is the dimension of the feature vector, ΔT(t), ΔC(t) and ΔS(t) are the feature change rates of the texture feature, color feature and shape feature of the feeder equipment image at T consecutive time points, respectively. t 、w c and w s are texture feature weights, color feature weights, and shape feature weights, respectively, and w t +w c +w s =1.

9. The feeder fault monitoring and alarm method based on machine vision according to claim 1, characterized in that: The method of performing time series prediction and risk assessment based on the aging evaluation index of the feeder equipment and generating graded alarm information by using a fuzzy inference system combined with an expert knowledge base includes the following steps: S31. Use a long short-term memory network to perform time series prediction on the aging evaluation index of the feeder equipment to obtain an aging trend prediction result; S32. Based on the historical fault data of the feeder equipment and the aging evaluation index of the feeder equipment, a feeder fault risk assessment model is constructed using support vector machine technology to generate a fault risk probability; S33. Combine the aging trend prediction results and the failure risk probability, conduct a comprehensive analysis through the fuzzy reasoning system, match them with the pre-established expert knowledge base, and output graded alarm information.

10. A feeder fault monitoring and alarm system based on machine vision, used to implement the feeder fault monitoring and alarm method based on machine vision according to any one of claims 1 to 9, characterized in that: include: The image enhancement and optimization processing unit is used to generate enhanced images based on the feeder equipment images captured by industrial cameras using an image fusion algorithm. The image quality of the enhanced images under typical weather conditions is optimized using an adaptive illumination compensation model to obtain standardized feeder equipment image data. The feature extraction and aging assessment unit is used to extract features from standardized feeder equipment image data using a deep learning model, establish visual feature vectors of the feeder equipment based on the extracted texture features, color features, and shape features, and analyze the changing trend of the visual feature vectors using a time series analysis algorithm to obtain an aging assessment index for the feeder equipment; The multi-level warning information generation unit is used to perform time series prediction and risk assessment based on the aging evaluation indicators of the feeder equipment, and use the fuzzy reasoning system combined with the expert knowledge base to generate graded alarm information.

Images