A method, system and device for synergistic detection of three-light-one-field and hydrophobicity of distribution network insulators, and a storage medium

Abstract

The application discloses a kind of coordination detection methods, systems, equipment and storage medium of three light one field and hydrophobicity of distribution network insulator, it is related to power transmission and distribution network equipment state monitoring and fault diagnosis technical field, method includes: collecting the visible light, infrared, ultraviolet, space electric field sequence and hydrophobicity water drop image of insulator, aligning electric field and umbrella skirt profile and registering multi-source data to target umbrella skirt position, extracting surface, temperature, discharge, electric field feature and water drop contact angle, construct compensation feature group after contact angle is corrected using temperature and electric field, generate feature space distance matrix by distance calculation, classify mapping output insulator defect classification result;The application fuses multi-source representation and physical law, can improve the accuracy of insulator complex deterioration diagnosis;Objective empowerment avoids subjective bias;Mahalanobis distance eliminates dimension and correlation influence, truly reflects defect attribution.

Description

Technical Field

[0001] This invention relates to the field of power transmission and distribution network equipment condition monitoring and fault diagnosis technology, and in particular to a method, system, equipment and storage medium for the coordinated detection of three-light field and hydrophobicity of distribution network insulators. Background Technology

[0002] Insulators are core equipment in power transmission and distribution systems that ensure electrical insulation and mechanical support. During long-term outdoor operation, they are susceptible to defects such as material aging and insulation degradation caused by the combined effects of environmental surface contamination and electrical stress. Currently, visible light inspection, infrared thermometry, or hydrophobicity testing are commonly used to obtain characterization parameters such as surface morphology, temperature distribution, and water droplet contact angle, in order to assess the condition of the equipment and diagnose defects.

[0003] Most existing diagnostic methods rely on independent analysis of single features, without fully considering the physical effects of local temperature rise and dielectric constant perturbation on the hydrophobicity and surface condition of insulators during actual operation. Due to the lack of effective integration of multi-source external characterization with internal physical laws, existing technologies are easily affected by environmental interference under complex operating conditions and the characterization is incomplete, resulting in the need to further improve the accuracy of insulator deterioration diagnosis. Summary of the Invention

[0004] In view of the above-mentioned problems, the present invention provides a method, system, device and storage medium for the coordinated detection of three-light field and hydrophobicity of distribution network insulators.

[0005] Therefore, the technical problem solved by this invention is that most existing diagnostic methods rely on independent analysis of a single feature and do not fully consider the physical effects of local temperature rise and dielectric constant perturbation on the hydrophobicity and surface condition of insulators during actual operation.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: In a first aspect, the present invention provides a method for the coordinated detection of three-light field and hydrophobicity of distribution network insulators, including: Acquire visible light, infrared, and ultraviolet images, axial spatial electric field sequences, and hydrophobic water droplet images of the target insulator; Extract the visible light umbrella skirt contour to generate a structural reference sequence, align the electric field sequence with the structural reference sequence to generate an aligned electric field sequence; Based on the visible light image, the infrared and ultraviolet images, the alignment electric field sequence, and the hydrophobic water droplet image are registered to the umbrella skirt position to construct a registration space model; Based on the registration spatial model, surface, temperature, discharge, electric field features and water droplet contact angle are extracted to generate a multi-source feature vector set; The contact angle of water droplets is corrected by using the temperature and electric field features in the multi-source feature vector set, and a compensation feature vector set is generated. Distance calculation is performed on the compensated feature vector group to generate a feature space distance matrix; Based on the feature space distance matrix classification mapping, the insulator defect classification results are output.

[0007] As a preferred scheme for a synergistic detection method of three-light field and hydrophobicity of distribution network insulators, wherein: The step of extracting the visible light umbrella skirt contour to generate a structural reference sequence, aligning the electric field sequence with the structural reference sequence to generate an aligned electric field sequence includes: Extract the edge contour of the umbrella skirt from the visible light image, and generate a structural reference sequence by projection integration along the insulator axis; Local peaks and valleys are identified from the spatial electric field intensity sequence based on differential gradients and then merged to generate an electric field extremum sequence. After normalizing the structural reference sequence and the electric field extremum sequence, dynamic time warping is used to calculate the cumulative distance cost matrix, and a dynamic alignment path is generated by reverse search. The original electric field sequence is resampled by piecewise linear interpolation based on the dynamic alignment path, and the output is an aligned electric field sequence of the same length as the structural reference sequence.

[0008] As a preferred scheme for a synergistic detection method of three-light field and hydrophobicity of distribution network insulators, wherein: The process of registering infrared and ultraviolet images, alignment electric field sequences, and hydrophobic water droplet images to the umbrella skirt position, based on visible light images, to construct a registration space model includes: Based on the visible light image, the corresponding feature points between the visible light image, the infrared thermal image, and the ultraviolet discharge image are extracted, the spatial transformation matrix is ​​calculated, and the infrared and ultraviolet images are mapped and superimposed onto the pixel coordinate system of the visible light image to generate a multispectral fusion image. The axial center coordinates of the target skirt position in the multispectral fusion image are extracted, and the data nodes in the aligned electric field sequence are mapped and assigned to the axial center coordinates according to their positions to generate an electric field distribution topology map. Acquire the acquisition pose parameters of the hydrophobic water droplet image, perform scaling, rotation and translation transformations on the hydrophobic image, and anchor it to the corresponding target umbrella skirt position in the multispectral fusion image; The multispectral fused image, the electric field distribution topology map, and the anchored hydrophobic water droplet image are combined into a coordinate system to output a registration space model.

[0009] As a preferred scheme for a synergistic detection method of three-light field and hydrophobicity of distribution network insulators, wherein: The process involves extracting surface, temperature, discharge, electric field features, and water droplet contact angle based on a registration spatial model to generate a multi-source feature vector set, including: Gray-scale parameters and texture parameters are extracted from the visible light image channel at the target umbrella skirt location as surface features; The difference between the local maximum temperature and the ambient reference temperature is extracted from the infrared thermal image channel as a temperature feature; The number of effective discharge photons per unit area is extracted from the ultraviolet discharge image channel as the discharge feature. The electric field amplitude distortion rate within the target umbrella skirt location is extracted as an electric field feature from the electric field distribution topology map in the registration space model. The water droplet edge contour is extracted from the hydrophobic water droplet image after being anchored in the registration space model. The tangent angle of the water droplet edge contour at the gas-liquid-solid three-phase interface is calculated by fitting the edge function and used as the water droplet contact angle. Based on the historical prior distribution interval of each feature, the numerical values ​​of each feature are nonlinearly stretched and normalized to generate a normalized feature subset. The system acquires the operating parameters of the insulator, dynamically calculates the weight coefficients of each feature dimension, performs weighted modulation on the normalized feature subsets, and splices them in a preset order to output a multi-source feature vector group.

[0010] As a preferred scheme for a synergistic detection method of three-light field and hydrophobicity of distribution network insulators, wherein: The step of using temperature and electric field features from a multi-source feature vector set to correct the water droplet contact angle and generate a compensation feature vector set includes: Analyze the multi-source feature vector set to extract local temperature rise parameters and electric field distortion amplitude; Thermodynamic contact angle deformation is calculated based on local temperature rise parameters, and electrodynamic contact angle offset is calculated based on electric field distortion amplitude and surface features. The corrected water droplet contact angle is obtained by subtracting the thermodynamic deformation and electrodynamic offset from the original water droplet contact angle. The original contact angle in the multi-source feature vector set is updated to the corrected contact angle, and the compensated feature vector set is output.

[0011] As a preferred scheme for a synergistic detection method of three-light field and hydrophobicity of distribution network insulators, wherein: The step of calculating the distance to the compensated feature vector group and generating a feature space distance matrix includes: The coordinate positions of the compensation feature vector group in the feature dimension space are extracted as test sample points, and multiple preset standard defect center clusters are called. Each standard defect center cluster corresponds to a typical deterioration state or normal state of an insulator. The eigenvector difference between the test sample point and the cluster center mean vector of each standard defect center cluster is calculated. The eigenvector difference is weighted by the pre-set covariance matrix of each feature dimension. The correlation effect and dimensional difference between feature dimensions are eliminated by the Mahalanobis distance calculation formula to obtain the Mahalanobis distance of the test sample point relative to each standard defect center cluster. The Mahalanobis distances are combined according to the category order of the corresponding standard defect center clusters to generate a feature space distance matrix.

[0012] As a preferred scheme for a synergistic detection method of three-light field and hydrophobicity of distribution network insulators, wherein: The classification mapping based on the feature space distance matrix outputs the insulator defect classification results, including: Analyze the feature space distance matrix and extract the Mahalanobis distance elements of the test sample points relative to each standard defect center cluster; Traverse all Mahalanobis distance values, use a minimization optimization strategy to determine the minimum value, and extract the category index number of the standard defect center cluster corresponding to the minimum value. The system calls a preset defect status lookup table, uses the extracted category index number as the input primary key, performs matching and addressing in the defect status lookup table, and maps and outputs the final defect classification result of the target insulator. The defect classification result includes normal operation status, surface area contamination and deterioration, aging of the skirt material, or internal insulation breakdown.

[0013] Secondly, the present invention provides a synergistic detection system for the three-light field and hydrophobicity of distribution network insulators, comprising: The insulator global sensing module is used to acquire visible light, infrared, and ultraviolet images, axial spatial electric field sequences, and hydrophobic water droplet images of the target insulator. The electric field-skirt dynamic warping module is used to extract the visible light skirt outline to generate a structural reference sequence, align the electric field sequence with the structural reference sequence, and generate an aligned electric field sequence. The multi-source image fusion and registration module is used to register infrared, ultraviolet, aligned electric field sequences and hydrophobic water droplet images to the umbrella skirt position based on visible light images, and to build a registration space model. The composite feature vector generation module is used to extract surface, temperature, discharge, electric field features and water droplet contact angle based on the registration space model, and generate a multi-source feature vector set. The hydrophobic contact angle compensation module is used to correct the water droplet contact angle by utilizing the temperature and electric field characteristics in the multi-source feature vector set, and to generate a compensation feature vector set. The Mahalanobis distance evaluation module is used to calculate the distance to the compensated feature vector group and generate a feature space distance matrix. The defect status output module is used to output the insulator defect classification results based on the feature space distance matrix classification mapping.

[0014] Thirdly, the present invention provides a computer device, comprising: Memory and processor; The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the computer-executable instructions are executed by the processor, they implement the steps of the method for the coordinated detection of the three-light field and hydrophobicity of distribution network insulators.

[0015] Fourthly, the present invention provides a computer-readable storage medium storing computer-executable instructions that, when executed by a processor, implement the steps of a method for the coordinated detection of the three light fields and hydrophobicity of distribution network insulators.

[0016] The beneficial effects of this invention are as follows: This invention extracts multi-dimensional operational features of insulators and combines local temperature rise and dielectric constant perturbation to physically correct the original water droplet contact angle. It integrates multi-source external characterization with internal physical laws, effectively overcoming the limitations of traditional single-feature detection, which is easily interfered with or has incomplete characterization, and improving the accuracy of diagnosing complex insulator degradation states. In the feature fusion stage, this invention performs adaptive weighting on multi-dimensional features including the corrected water droplet contact angle, avoiding subjective bias from manual experience-based assignment and highlighting the key feature dimensions most sensitive to the current degradation state of the insulator. When calculating feature space distance, this invention uses covariance matrix for non-independent weighting and calculates Mahalanobis distance, fully considering the non-independent correlation law between multi-dimensional physical features, eliminating the bias caused by feature dimension differences and correlation overlap, and enabling distance measurement to more realistically reflect the defect category of the target insulator. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is an overall flowchart of a method for the coordinated detection of three-light field and hydrophobicity of distribution network insulators provided by the present invention.

[0019] Figure 2 This is a decision flowchart of a collaborative detection method for the three light fields and hydrophobicity of distribution network insulators provided by the present invention.

[0020] Figure 3This is a flowchart of a method for dynamically aligning the spatial electric field sequence with the insulator skirt outline and judging the effectiveness of a collaborative detection method for three-light field and hydrophobicity of distribution network insulators provided by the present invention. Detailed Implementation

[0021] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.

[0022] Example 1, referring to Figure 1 This is the first embodiment of the present invention, which provides a method for the coordinated detection of three-light field and hydrophobicity of distribution network insulators, including: S1: Acquire visible light, infrared, and ultraviolet images of the target insulator, axial spatial electric field sequence, and images of hydrophobic water droplets; S2: Extract the visible light umbrella skirt contour to generate a structural reference sequence, align the electric field sequence with the structural reference sequence to generate an aligned electric field sequence; S3: Using the visible light image as a reference, register the infrared and ultraviolet images, the alignment electric field sequence, and the hydrophobic water droplet image to the umbrella skirt position to construct a registration space model; S4: Based on the registration spatial model, extract surface, temperature, discharge, electric field features and water droplet contact angle to generate a multi-source feature vector set; S5: Use the temperature and electric field features in the multi-source feature vector set to correct the water droplet contact angle and generate a compensation feature vector set; S6: Calculate the distance to the compensated feature vector group and generate the feature space distance matrix; S7: Based on the feature space distance matrix classification mapping, output the insulator defect classification results.

[0023] It should be noted that through steps S1-S7, the collaborative acquisition, spatial registration, and physical fusion of multispectral images, spatial electric fields, and hydrophobic characteristics of insulators are realized. The contact angle is corrected by mechanism using temperature and electric field, and the defect classification results are output based on feature spatial distance calculation and classification mapping, which effectively improves the accuracy and robustness of insulator degradation state diagnosis under complex working conditions.

[0024] Example 2, refer to Figures 1-3 As an embodiment of the present invention, based on the previous embodiment, a method for the coordinated detection of three-light field and hydrophobicity of distribution network insulators is provided, comprising: In this embodiment, the acquisition of visible light, infrared, and ultraviolet images, axial spatial electric field sequences, and hydrophobic water droplet images of the target insulator in step S1 above includes: A drone equipped with a multi-sensor integrated pod flies along the power distribution line, remaining relatively stationary at a preset observation distance from the target insulator. A visible light camera captures the surface structure and appearance of the insulator, outputting visible light images; an infrared thermal imager detects the surface temperature distribution corresponding to the physical location of the insulator, outputting infrared thermal images; and an ultraviolet imager captures ultraviolet photons generated by partial discharge on the insulator surface, outputting ultraviolet discharge images.

[0025] To eliminate background stray light noise caused by complex outdoor environments, a spatial adaptive filtering algorithm is used to preprocess the ultraviolet pixel matrix. The calculation formula is as follows: ; In the formula: This represents the effective discharge photon intensity at pixel coordinates x and y in the output ultraviolet discharge image; This represents a local filtering window centered on the current pixel. and This represents the relative coordinate offset within the local filtering window; This indicates that the summation is performed on all pixels within the local filtering window. This represents the original photon count value in the relative offset coordinates; Represents the spatial weight coefficients within the local filtering window calculated based on the spatial center distance; This represents the background stray light threshold, which is dynamically calculated based on the ambient brightness of the visible light environment.

[0026] During the same frequency period when images are acquired, a miniature electric field sensor array mounted on a UAV is used to measure the spatial electric field around the insulator. The electric field sensor continuously samples the electric field intensity along a spatial trajectory parallel to the insulator axis. To eliminate abnormal fluctuation noise in the raw measurement data caused by UAV flight attitude yaw or mechanical jitter, a spatial trajectory compensation algorithm is used to correct the raw electric field data. The calculation formula is as follows: ; In the formula: This represents the corrected spatial electric field strength at the i-th sampling point along the axial direction of the target insulator; This indicates the original electric field strength value measured by the sensor at the sampling point location; This indicates the spatial yaw angle between the sensor's measuring axis and the insulator's actual physical axis at the sampling point. This represents the background environmental electric field calibration constant obtained through pre-calibration.

[0027] The compensated electric field strength values ​​are arranged sequentially according to the sampling spatial coordinates to generate a spatial electric field strength sequence distributed along the axis of the target insulator. Then, the electric water spraying device mounted on the UAV is controlled to quantitatively spray atomized deionized water onto the target skirt area of ​​the insulator. To ensure that the collected water droplets accurately reflect the static hydrophobic angle characteristics, a frame difference energy assessment algorithm is used to determine the dynamic-static transition process of the water droplets on the insulator surface. The calculation formula is as follows: ; In the formula: This represents the energy of the water droplet morphology change at time t; This indicates the distribution area of ​​water droplets on the target umbrella skirt in a macro image; This indicates that the coordinates of all pixels within the water droplet distribution area are summed. and These represent the grayscale values ​​of the corresponding pixel coordinates in the macro image of the current frame and the previous frame, respectively; This represents the preset scaling factor for the surface state of the insulating medium.

[0028] when If the water droplets remain below the stability threshold for an extended period within a preset time window, it is determined that the water droplets have reached a static equilibrium state on the insulating medium. At this point, a high-resolution telephoto macro lens is triggered to capture a magnified image of the umbrella skirt surface. This image contains the morphology and edge contours of multiple static water droplets attached to the insulating medium, and is output as an image of hydrophobic water droplets.

[0029] It should be noted that by acquiring multidimensional physical parameter data of the target insulator and eliminating interference from ambient stray light, attitude jitter, and dynamic changes in water droplets on the underlying data through algorithms, the objectivity and accuracy of the collected data are ensured.

[0030] In this embodiment, step S2 above involves extracting the visible light umbrella skirt contour to generate a structural reference sequence, aligning the electric field sequence with the structural reference sequence, and generating an aligned electric field sequence, including: like Figure 3 As shown, this embodiment achieves precise alignment between the electric field sequence and the umbrella skirt outline through extreme value extraction, dynamic time warping, and alignment path validity judgment.

[0031] Specifically, local peaks and troughs are extracted from the spatial electric field intensity sequence and merged to generate an electric field extremum sequence; Calculate the distance cost matrix between the electric field extremum sequence and the structural reference sequence, and search for coordinate correspondences that meet the preset minimum conditions based on the distance cost matrix to generate a dynamic alignment path; The spatial electric field intensity sequence is resampled based on the dynamic alignment path, and the aligned electric field sequence with the same length as the structural reference sequence is output.

[0032] Furthermore, the visible light image and the spatial electric field intensity sequence distributed along the axis of the target insulator are obtained. An edge detection operator is used to extract the skirt contour of the target insulator from the visible light image. The extracted binarized edge pixels are then horizontally projected and integrated along the physical axis of the target insulator, calculated using the following formula: ; In the formula: The contour feature projection value represents the pixel coordinate of the j-th axis in the structural reference sequence; This represents the horizontal pixel width of a visible light image; Represents the horizontal pixel coordinates of a visible light image; This represents the coordinate of the j-th vertical pixel along the physical axis of the insulator in the visible light image; This indicates the visible light image at pixel coordinates x and y. Binarized edge feature values ​​at the location; Represents the current vertical pixel coordinate Iterate through and sum all the corresponding horizontal pixels.

[0033] The vertical pixel coordinates and their corresponding contour feature projection values ​​are arranged sequentially to generate a structural reference sequence characterizing the physical structure of the insulator skirt. For the obtained spatial electric field intensity sequence, since the abrupt change between the air and ceramic media at the edge of the insulator skirt causes regular distortion of the spatial electric field, a differential gradient algorithm is used to process the spatial electric field intensity sequence to find the distortion extrema. The extremum determination formula is as follows: ; In the formula: This represents the extremum determination coefficient of the i-th sampling point in the spatial electric field intensity sequence; This represents the corrected spatial electric field intensity at the i-th sampling point in the spatial electric field intensity sequence; and These represent the corrected spatial electric field intensities of the preceding and following adjacent sampling points in the spatial electric field intensity sequence, respectively.

[0034] Filter out all that meet the requirements Sampling points with values ​​less than zero are treated as local peaks and troughs, and are sequentially concatenated according to their original spatial sampling coordinates to generate an electric field extremum sequence. To eliminate spatial nonlinear misalignment caused by differences in sensor sampling frequencies or relative displacement, the values ​​of the electric field extremum sequence and the structural reference sequence are first normalized to unify their dimensions. Then, a dynamic time warping algorithm is used to calculate the distance cost matrix between the normalized electric field extremum sequence and the structural reference sequence. The calculation formula is as follows: ; In the formula: This represents the cumulative distance cost between the m-th extreme point in the electric field extreme value sequence and the n-th reference point in the structural reference sequence; This represents the characteristic value of the m-th extreme point in the normalized electric field extreme value sequence; This represents the feature projection value of the nth reference point in the normalized structural reference sequence; This represents the absolute distance cost between the current normalized extreme point and the reference point's feature values; This represents the minimum cumulative distance cost of three adjacent historical state grids in the distance cost matrix.

[0035] After calculating the distance cost matrix, a reverse search is performed from the endpoint coordinates to the starting coordinates. Each time, the node with the minimum cumulative distance cost among adjacent grid cells is selected as the coordinate correspondence that meets the preset minimum value condition. All the coordinate correspondences traversed are then combined in reverse order to generate a dynamic alignment path. Based on this dynamic alignment path, a piecewise linear interpolation algorithm is used to resample the output spatial electric field intensity sequence. The calculation formula is as follows: ; In the formula: This represents the resampled electric field intensity at the j-th coordinate node in the aligned electric field sequence; This represents the axial pixel coordinates of the target insulator in the dynamic alignment path mapping. and These represent the image axial pixel coordinates corresponding to the two adjacent electric field anchor points of the current coordinate j in the dynamic alignment path; and These represent the corrected spatial electric field intensities corresponding to the two adjacent electric field anchor points. The resampling calculation outputs an aligned electric field sequence with the exact same length as the structural reference sequence.

[0036] It should be noted that by aligning the one-dimensional electric field signal with the two-dimensional visible light image on the physical structure of the insulator, the geometric coordinate misalignment error of the multi-source data was eliminated.

[0037] In this embodiment, step S3 above, using the visible light image as a reference, registers the infrared and ultraviolet images, the alignment electric field sequence, and the hydrophobic water droplet image to the umbrella skirt position to construct a registration space model, including: Extract corresponding feature points from visible light images, infrared thermal images, and ultraviolet discharge images, calculate the spatial transformation matrix between corresponding feature points, and use the spatial transformation matrix to superimpose the infrared thermal image and ultraviolet discharge image onto the pixel coordinate system of the visible light image to generate a multispectral fusion image. The axial center coordinates of the target skirt position in the multispectral fusion image are extracted, and the data nodes in the aligned electric field sequence are sequentially assigned to the axial center coordinates according to the position mapping relationship to generate an electric field distribution topology map. Extract the acquisition pose parameters of the hydrophobic water droplet image, anchor the hydrophobic water droplet image to the corresponding target umbrella skirt position in the multispectral fusion image based on the acquisition pose parameters, and perform coordinate system one on the multispectral fusion image, electric field distribution topology map and anchored hydrophobic water droplet image to output the registration space model.

[0038] Specifically, firstly, visible light images, infrared thermal images, ultraviolet discharge images, and hydrophobic water droplet images are acquired. Then, an edge detection algorithm is used to extract the skirt contour coordinates of the target insulator from the visible light images as the target skirt position.

[0039] Using the visible light image as a reference, a scale-invariant feature transformation algorithm is employed to extract corresponding feature points between the visible light image and the infrared thermal image, as well as between the visible light image and the ultraviolet discharge image. Based on the extracted corresponding feature points, a spatial transformation matrix is ​​calculated. This spatial transformation matrix is ​​then used to map and superimpose the pixel data of the infrared thermal image and the ultraviolet discharge image onto the pixel coordinate system of the visible light image, generating a multispectral fused image. The spatial transformation formula is as follows: ; In the formula: and This represents the horizontal and vertical pixel coordinates in the visible light image reference coordinate system; and These represent the horizontal and vertical pixel coordinates of the original image input into the infrared thermal image or ultraviolet discharge image to be registered, respectively. , , and This represents the parameters of the spatial rotation and scaling matrix calculated using the least squares method based on the corresponding feature points; and This represents the horizontal and vertical spatial translation offset parameters caused by differences in the physical installation positions of different spectral sensors.

[0040] For the generated multispectral fusion image, extract the left and right edge pixel coordinates corresponding to each vertical pixel coordinate at the target umbrella skirt position, and calculate the mean of the left and right edge pixel coordinates as the axial center coordinate under the current vertical pixel coordinate.

[0041] Obtain the output aligned electric field sequence, and assign values ​​to the corresponding axial center coordinates of the data nodes in the aligned electric field sequence according to the position mapping relationship to generate an electric field distribution topology map. The topology assignment formula is as follows: ; In the formula: This represents the electric field topology attribute value at the j-th vertical pixel coordinate node along the axial direction of the target insulator in the electric field distribution topology map. This represents the coordinate of the j-th vertical pixel along the physical axis of the insulator in the multispectral fusion image. This coordinate is completely consistent with the vertical pixel coordinate data when extracting the structural reference sequence in step S2. The vertical pixel coordinates in the multispectral fused image are: The horizontal center coordinates of the target umbrella skirt structure are calculated from the average pixel coordinates of the left and right edges. Represents the vertical pixel coordinates of the resampled output. The resampled electric field intensity at the j-th coordinate node in the corresponding aligned electric field sequence.

[0042] The acquisition pose parameters of the UAV-mounted gimbal were used to acquire images of hydrophobic water droplets. Based on these acquisition pose parameters, the acquired local high-resolution hydrophobic water droplet images were anchored to the corresponding target umbrella skirt position in the multispectral fusion image. The anchoring mapping formula is as follows: ; In the formula: and This represents the globally anchored horizontal and vertical pixel coordinates of a hydrophobic water droplet image mapped to a multispectral fused image reference. and Represents the local horizontal and vertical pixel coordinates within an image of hydrophobic water droplets; The image scaling factor is dynamically determined by the ratio of the focal length of the gimbal macro lens to the focal length of the visible light wide-angle lens. This indicates the relative rotational yaw angle of the gimbal when acquiring images of visible light and hydrophobic water droplets; and This represents the horizontal and longitudinal translation compensation amounts calculated based on the three-dimensional relative displacement between the gimbal's pointing center and the UAV.

[0043] The multispectral fused image, electric field distribution topology map, and anchored hydrophobic water droplet image are subjected to coordinate system one in global anchoring coordinates, and a registration space model containing multidimensional physical information of the insulator is output.

[0044] In this embodiment, step S4 above, which extracts surface, temperature, discharge, electric field features, and water droplet contact angle based on the registration space model to generate a multi-source feature vector set, includes: Analyze and register the spatial model to locate the target umbrella skirt position; Grayscale and texture parameters are extracted from the visible light image of the target umbrella skirt location as surface features, temperature distribution parameters are extracted from the infrared thermal image of the target umbrella skirt location as temperature features, and photon density parameters are extracted from the ultraviolet discharge image of the target umbrella skirt location as discharge features. The amplitude distortion rate is extracted as an electric field feature from the aligned electric field sequence at the target umbrella skirt position, and the tangent angle is calculated as the water droplet contact angle by edge fitting of the hydrophobic water droplet image at the target umbrella skirt position. The surface features, temperature features, discharge features, electric field features, and water droplet contact angle are numerically normalized and serialized according to preset feature dimensions, and then spliced ​​together to output a multi-source feature vector group.

[0045] Specifically, the output registration spatial model is first parsed to locate the target insulator's skirt position. Visible light image channel data, infrared thermal image channel data, and ultraviolet discharge image channel data, containing spatial location mapping relationships, are extracted from this target skirt position. From the visible light image channel data, the grayscale mean and texture contrast parameters of the target skirt position are extracted as surface features. From the infrared thermal image channel data, the difference between the local maximum temperature and the ambient reference temperature at the target skirt position is extracted as a temperature feature. From the ultraviolet discharge image channel data, the effective discharge photon count per unit area at the target skirt position is calculated as a discharge feature.

[0046] The electric field topology attribute values ​​of the electric field distribution topology map in the registration spatial model at the target umbrella skirt position are extracted, and the amplitude distortion rate is calculated as an electric field feature using the range statistics algorithm. The calculation formula is as follows: ; In the formula: This represents the extracted electric field characteristics, i.e., the amplitude distortion rate; This represents the maximum electric field topology attribute value within the target umbrella skirt location of the electric field distribution topology map generated in step S3; This represents the average electric field topology attribute value within the target umbrella skirt location of the electric field distribution topology map.

[0047] For hydrophobic water droplet images with globally anchored coordinates in the registration spatial model, the edge contour pixels of the water droplets on the dielectric surface of the insulator skirt are extracted. A polynomial curve is fitted using the globally anchored horizontal coordinates output in step S3 as the independent variable and the globally anchored vertical coordinates as the dependent variable to construct an edge function. The tangent angle of this edge function at the gas-liquid-solid three-phase interface is then calculated as the water droplet contact angle. The calculation formula is as follows: ; In the formula: This represents the calculated water droplet contact angle. This represents the horizontal coordinate of the edge anchoring point at the gas-liquid-solid three-phase interface in the image of hydrophobic water droplets. The marginal function obtained by fitting In the horizontal coordinate The value of the first derivative at point, where and These are the global anchor horizontal coordinates and global anchor vertical coordinates output in step S3, respectively.

[0048] Subsequently, the historical prior distribution ranges of surface features, temperature features, discharge features, electric field features, and water droplet contact angle under the power distribution line inspection environment were obtained. Based on these historical prior distribution ranges, a nonlinear stretching function was used to stretch the values ​​of the five types of features to generate a normalized feature subset. The calculation formula is as follows: ; In the formula: The index number represents the feature dimension, with values ​​of 1, 2, 3, 4, and 5, corresponding to surface features, temperature features, discharge features, electric field features, and water droplet contact angle, respectively. This represents the normalized value of the k-th feature dimension in the normalized feature subset; This represents the original calculated value of the k-th feature dimension extracted before the nonlinear stretching transformation; and Let represent the prior mean and prior standard deviation, respectively, for the k-th feature dimension within the historical prior distribution interval. This represents the preset nonlinear mapping smoothing coefficient; It represents the base of the natural logarithm.

[0049] The operating parameters of the target insulator in the distribution network are obtained, and the weight coefficients of each feature dimension in the normalized feature subset are dynamically calculated using the operating parameters. The calculation formula is as follows: ; In the formula: This represents the weight coefficient calculated for the k-th feature dimension; This indicates the current actual operating voltage of the target insulator; Indicates the rated reference voltage of the target insulator; This indicates the real-time relative humidity of the environment in which the insulator is located; and These represent the pre-calibrated sensitivity adjustment constants of the k-th feature dimension for voltage and humidity, respectively.

[0050] The calculated weight coefficients are used to weight and modulate the corresponding feature dimensions in the normalized feature subsets. The weighted and modulated surface features, temperature features, discharge features, electric field features, and water droplet contact angles are then sequentially concatenated according to a preset feature dimension order to output a multi-source feature vector group. The concatenation formula is as follows: ; In the formula: This represents the output multi-source feature vector set; to These represent, in order, the surface features, temperature features, discharge features, electric field features, and water droplet contact angle in the normalized feature subset; to These represent the weighting coefficients calculated based on the five corresponding feature dimensions according to the operating condition parameters.

[0051] It should be noted that by extracting and quantifying the physical state parameters of the target insulator in multiple dimensions such as surface morphology, thermodynamics, electromagnetics and hydrophobicity of materials, the dimensional differences of different sensor characteristics were eliminated and the influence of real-time operating conditions was integrated, providing standardized comprehensive data input for subsequent insulator defect diagnosis.

[0052] In this embodiment, step S5 above, which uses the temperature and electric field features in the multi-source feature vector set to correct the water droplet contact angle and generate a compensation feature vector set, includes: Analyze the multi-source feature vector group to extract the local temperature rise parameter corresponding to the temperature feature and the electric field distortion amplitude corresponding to the electric field feature; The local temperature rise parameter is input into the preset surface tension attenuation model to calculate the thermodynamic contact angle deformation, and the electric field distortion amplitude is input into the preset electrowetting effect equation to calculate the electrodynamic contact angle offset. The corrected water droplet contact angle is calculated by inversely superimposing and compensating the deformation of the water droplet contact angle, the thermodynamic contact angle, and the electrodynamic contact angle offset. The water droplet contact angle within the multi-source feature vector group is updated to the corrected water droplet contact angle, while keeping other feature dimensions within the multi-source feature vector group unchanged, and the compensated feature vector group is output.

[0053] Specifically, firstly, the original feature mapping data contained in the multi-source feature vector group output in step S4 is analyzed to extract the local temperature rise parameter corresponding to the temperature feature and the electric field distortion amplitude corresponding to the electric field feature. Specifically, the original extracted values ​​corresponding to the temperature feature in the multi-source feature vector group are extracted as the local temperature rise parameter, and the original amplitude distortion rate and average electric field topological attribute value corresponding to the electric field feature in the multi-source feature vector group are extracted. The electric field distortion amplitude is obtained through inverse calculation, using the following formula: ; ; In the formula: This indicates the extracted local temperature rise parameter; This represents the difference between the local maximum temperature calculated in step S4 and the ambient reference temperature, which is the original temperature characteristic calculation value. This represents the calculated amplitude of the electric field distortion; This represents the original electric field characteristics, i.e., amplitude distortion rate, calculated in step S4; This represents the average electric field topology attribute value of the electric field distribution topology map within the target umbrella skirt location in step S4.

[0054] Substituting the local temperature rise parameter into the partial derivative of the solid-liquid interfacial energy in the modified Young's equation, the surface tension attenuation coefficient as a function of temperature is calculated. The thermodynamic contact angle deformation of the water droplet is then derived using this coefficient. The calculation formula is as follows: ; ; In the formula: This represents the surface tension attenuation coefficient as it changes with temperature. This represents the partial derivative constant of the solid-liquid interfacial energy of the pre-calibrated target insulator skirt material with respect to temperature. This represents the local temperature rise parameter extracted by the above formula; This represents the thermodynamic contact angle deformation of the calculated output; This represents the liquid-gas surface tension constant at a preset standard ambient reference temperature; This represents the calculated value of the original water droplet contact angle extracted in step S4. It represents pi (π).

[0055] Historical surface feature samples and corresponding historical dielectric constant test samples of the target insulator under different standard salt density and ash density equivalent surface contamination were obtained. The gray-level mean sample and texture contrast sample from the historical surface feature samples were used as independent variables, and the perturbation difference of the corresponding historical dielectric constant test samples was used as the dependent variable. A multiple linear regression algorithm was used to fit the coefficients, constructing a pre-defined surface contamination-dielectric constant mapping model. The gray-level mean and texture contrast parameters contained in the surface features of the multi-source feature vector group were analyzed. The pre-defined surface contamination-dielectric constant mapping model was then called to calculate the dielectric constant perturbation value of the water droplet contact angle based on the surface features. The calculation formula is as follows: ; In the formula: This represents the dielectric constant perturbation caused by contamination buildup on the insulator surface; and These represent the original grayscale mean and the original texture contrast parameter obtained when extracting the surface features of the target insulator in step S4, respectively. and These represent the grayscale mapping constant and texture mapping constant, respectively, which were calibrated in advance based on the insulator surface contamination experiment.

[0056] Substituting the dielectric constant perturbation value and the electric field distortion amplitude into the Lippmann electrowetting equation, the electrostatic adsorption stress on the liquid surface is calculated. The electrodynamic contact angle offset of the water droplet is then derived using this electrostatic adsorption stress. The calculation formula is as follows: ; ; In the formula: This indicates the electrostatic adsorption stress on the liquid surface; This represents the basic dielectric constant of pure water under standard conditions. This represents the perturbation value of the dielectric constant calculated by the above formula; This represents the calculated amplitude of the electric field distortion. This represents the calculated electrodynamic contact angle offset. This represents the equivalent liquid layer thickness constant on the insulator surface, based on pre-calibrated microscopic observations of hydrophobic water droplets. This represents the preset surface tension constant of the liquid-gas mixture; This represents the calculated value of the original water droplet contact angle extracted in step S4. It represents pi (π).

[0057] The original water droplet contact angle, thermodynamic contact angle deformation, and electrodynamic contact angle offset are inversely superimposed and compensated to obtain the corrected water droplet contact angle. The calculation formula is as follows: ; In the formula: This indicates the corrected water droplet contact angle; This represents the calculated value of the original water droplet contact angle; This represents the calculated thermodynamic contact angle deformation. This represents the calculated electrodynamic contact angle offset.

[0058] Based on the nonlinear stretching transformation mechanism in step S4 and the historical prior distribution range, the corrected water droplet contact angle is recalculated using normalization to generate an updated normalized value for the water droplet contact angle. The calculation formula is as follows: ; In the formula: This represents the normalized value corresponding to the corrected water droplet contact angle; This represents the corrected water droplet contact angle calculated using the above formula; and These represent the prior mean and prior standard deviation of the water droplet contact angle feature dimension obtained in step S4, respectively. This represents the preset nonlinear mapping smoothing coefficient; It represents the base of the natural logarithm.

[0059] The original water droplet contact angle component in the multi-source feature vector group is updated to the weighted normalized component corresponding to the corrected water droplet contact angle, while keeping the other four feature dimensions in the multi-source feature vector group unchanged. The output is a compensated feature vector group. The feature vector update concatenation formula is as follows: ; In the formula: This represents the final output set of compensated feature vectors; to as well as to The weight coefficients and normalized values ​​of the corresponding first four feature dimensions obtained in step S4 are used directly. This represents the normalized value corresponding to the corrected water droplet contact angle calculated by the above formula.

[0060] It should be noted that by eliminating the interference errors caused by local heating and electric field distortion on the physical morphology of water droplets under operating conditions, the true hydrophobic characterization of the insulator was restored, providing reliable characteristic data for subsequent accurate comprehensive diagnosis.

[0061] In this embodiment, step S6 above, which involves calculating the distance to the compensation feature vector group and generating a feature space distance matrix, includes: Extract the coordinate positions of the compensation feature vector group in the corresponding feature dimension space as test sample points, and call multiple preset standard defect center clusters; The difference between the feature vectors of the test sample points and each standard defect center cluster is calculated, and the difference between the feature vectors is weighted by the pre-set covariance matrix of each feature dimension to obtain the Mahalanobis distance. Based on Mahalanobis distance, the distance metric of the test sample points relative to each standard defect center cluster is mapped to a feature space distance matrix.

[0062] Specifically, the compensation feature vector set, the final output of step S5, is extracted first. The serialized values ​​of the compensation feature vector set in the corresponding five-dimensional feature space are transposed to generate column vector coordinate positions for the test sample points. The formula for establishing the coordinate system of the test sample points is as follows: ; In the formula: This represents the test sample points as a column vector with five rows and one column. This represents the compensation feature vector group with one row and five columns output from step S5. It contains the weighted surface features, weighted temperature features, weighted discharge features, weighted electric field features, and weighted corrected water droplet contact angle of the insulator. This represents the matrix transpose operator.

[0063] Subsequently, multiple standard defect center clusters preset in the insulator maintenance database are invoked. These standard defect center clusters represent the multidimensional characteristic benchmark data distribution of insulators under different typical degradation states and normal states. The eigenvector difference between the test sample point and the cluster center mean vector of each standard defect center cluster is calculated. Then, using the preset covariance matrices of each feature dimension, this eigenvector difference is weighted non-independently to calculate the Mahalanobis distance between the test sample point and each standard defect center cluster. The calculation formula is as follows: ; In the formula: This represents the Mahalanobis distance of the test sample point relative to the m-th standard defect center cluster; The index variable represents the standard defect center cluster, and its value ranges from 1 to M, where M represents the total number of preset standard defect center cluster categories; This represents the column vector of test sample points calculated by the above formula; This represents the mean vector of the cluster centers of the preset m-th standard defect center cluster in the five-row, one-column feature dimension space; This represents the difference between the calculated eigenvectors of the test sample points and the m-th standard defect center cluster; This represents the matrix transpose operator. This represents a one-row, five-column transpose vector representing the eigenvector difference after transposition. This represents the five-by-five covariance matrix corresponding to the m-th standard defect center cluster. This matrix is ​​used to perform a non-independent weighted measurement of five physically related feature dimensions: surface features, temperature features, discharge features, electric field features, and corrected water droplet contact angle. The operator for inverting a matrix. It represents the inverse of a five-by-five covariance matrix.

[0064] Based on the calculated Mahalanobis distances, the distance metrics of the test sample points relative to each standard defect center cluster are combined in order of category number to generate a feature space distance matrix. The mapping formula is as follows: ; In the formula: This represents the feature space distance matrix with one row and M columns generated by the mapping; to These represent the Mahalanobis distances calculated relative to the first to the Mth standard defect center clusters, respectively. Index variable representing the standard defect center cluster; This represents the Mahalanobis distance of the test sample point relative to the m-th standard defect center cluster; This indicates the total number of categories in the preset standard defect center cluster.

[0065] It should be noted that this step, by combining the correlation between the multidimensional physical parameters of the insulator, objectively quantifies the similarity of the current target insulator to the distribution of various typical known defects in the five-dimensional feature space, providing an accurate data matching basis for the final output defect diagnosis results.

[0066] In this embodiment, the insulator defect classification result output in step S7 above, based on the feature space distance matrix classification mapping, includes: Parse the feature space distance matrix output in step S6, and extract the Mahalanobis distance elements of the test sample points relative to each standard defect center cluster contained in the feature space distance matrix. Traverse all Mahalanobis distance values ​​in the feature space distance matrix, use a minimization optimization algorithm to determine the minimum Mahalanobis distance, and extract the category index number of the standard defect center cluster corresponding to the minimum value. The calculation formula is as follows: ; In the formula: This represents the category index number of the standard defect center cluster corresponding to the calculated minimum Mahalanobis distance; This represents the mathematical operator for finding the minimum objective value. When multiple identical minimum Mahalanobis distances exist during the optimization process, the index number with the smallest value is extracted as the unique output primary key by default to ensure that the data structure of the subsequent hash mapping matches. The discrete index variable represents the standard defect center cluster, and its value is a set of positive integers from 1 to M; This indicates the total number of categories of the standard defect center clusters determined in step S6; This represents the feature space distance matrix that is the final output of step S6. The m-th Mahalanobis distance element in; It represents the set of discrete positive integers from 1 to M.

[0067] Extract the pre-defined defect status lookup table from the insulator maintenance database. This lookup table records a unique mapping relationship between category index numbers and actual operating status labels of insulators using a key-value pair data structure. Substitute the solved category index number as the input primary key into the defect status lookup table for matching and addressing, mapping and outputting the final defect classification result of the target insulator. The mapping formula is as follows: ; In the formula: This represents the defect classification result of the target insulator in the mapping output, and its corresponding text label is the insulator's normal operating status, surface area contamination and deterioration, aging of the shed material or internal insulation breakdown. This represents a computer hash mapping function composed of a pre-defined defect status lookup table; This represents the category index number of the standard defect center cluster corresponding to the minimum Mahalanobis distance calculated by the above formula.

[0068] It should be noted that this step transforms abstract multidimensional feature spatial distance values ​​into intuitive equipment status labels, outputting clear target insulator defect categories, and providing a direct basis for subsequent status assessment and maintenance strategy formulation for power distribution lines.

[0069] Example 3: The above is an illustrative scheme of a method for the coordinated detection of three-light-one-field and hydrophobic properties of distribution network insulators according to this embodiment. It should be noted that the technical solution of the coordinated detection system for three-light-one-field and hydrophobic properties of distribution network insulators and the technical solution of the above-described method for the coordinated detection of three-light-one-field and hydrophobic properties of distribution network insulators belong to the same concept. Details not described in detail in the technical solution of the coordinated detection system for three-light-one-field and hydrophobic properties of distribution network insulators in this embodiment can be found in the description of the technical solution of the above-described method for the coordinated detection of three-light-one-field and hydrophobic properties of distribution network insulators.

[0070] This embodiment also provides a collaborative detection system for the three-light field and hydrophobicity of distribution network insulators, including: The insulator global sensing module is used to acquire visible light, infrared, and ultraviolet images, axial spatial electric field sequences, and hydrophobic water droplet images of the target insulator. The electric field-skirt dynamic warping module is used to extract the visible light skirt outline to generate a structural reference sequence, align the electric field sequence with the structural reference sequence, and generate an aligned electric field sequence. The multi-source image fusion and registration module is used to register infrared, ultraviolet, aligned electric field sequences and hydrophobic water droplet images to the umbrella skirt position based on visible light images, and to build a registration space model. The composite feature vector generation module is used to extract surface, temperature, discharge, electric field features and water droplet contact angle based on the registration space model, and generate a multi-source feature vector set. The hydrophobic contact angle compensation module is used to correct the water droplet contact angle by utilizing the temperature and electric field characteristics in the multi-source feature vector set, and to generate a compensation feature vector set. The Mahalanobis distance evaluation module is used to calculate the distance to the compensated feature vector group and generate a feature space distance matrix. The defect status output module is used to output the insulator defect classification results based on the feature space distance matrix classification mapping.

[0071] This embodiment also provides a computer device applicable to a method for the coordinated detection of three-light field and hydrophobicity of distribution network insulators, including: The system includes a memory and a processor. The memory stores computer-executable instructions, and the processor executes these instructions to implement a method for the coordinated detection of three-light field and hydrophobicity of distribution network insulators, as proposed in the above embodiments.

[0072] This embodiment also provides a storage medium storing a computer program that, when executed by a processor, implements a method for the coordinated detection of three-light field and hydrophobicity of distribution network insulators as proposed in the above embodiments.

[0073] The storage medium proposed in this embodiment belongs to the same inventive concept as the method for coordinated detection of three-light field and hydrophobicity of distribution network insulators proposed in the above embodiments. Technical details not described in detail in this embodiment can be found in the above embodiments, and this embodiment has the same beneficial effects as the above embodiments.

[0074] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for the coordinated detection of three-light field and hydrophobicity of distribution network insulators, characterized in that, include: Acquire visible light, infrared, and ultraviolet images, axial spatial electric field sequences, and hydrophobic water droplet images of the target insulator; Extract the visible light umbrella skirt contour to generate a structural reference sequence, align the electric field sequence with the structural reference sequence to generate an aligned electric field sequence; Based on the visible light image, the infrared and ultraviolet images, the alignment electric field sequence, and the hydrophobic water droplet image are registered to the umbrella skirt position to construct a registration space model; Based on the registration spatial model, surface, temperature, discharge, electric field features and water droplet contact angle are extracted to generate a multi-source feature vector set; The contact angle of water droplets is corrected by using the temperature and electric field features in the multi-source feature vector set, and a compensation feature vector set is generated. Distance calculation is performed on the compensated feature vector group to generate a feature space distance matrix; Based on the feature space distance matrix classification mapping, the insulator defect classification results are output.

2. The method for synergistic detection of three-light field and hydrophobicity of distribution network insulators as described in claim 1, characterized in that, The step of extracting the visible light umbrella skirt contour to generate a structural reference sequence, aligning the electric field sequence with the structural reference sequence to generate an aligned electric field sequence includes: Extract the edge contour of the umbrella skirt from the visible light image, and generate a structural reference sequence by projection integration along the insulator axis; Local peaks and valleys are identified from the spatial electric field intensity sequence based on differential gradients and then merged to generate an electric field extremum sequence. After normalizing the structural reference sequence and the electric field extremum sequence, dynamic time warping is used to calculate the cumulative distance cost matrix, and a dynamic alignment path is generated by reverse search. The original electric field sequence is resampled by piecewise linear interpolation based on the dynamic alignment path, and the output is an aligned electric field sequence of the same length as the structural reference sequence.

3. The method for synergistic detection of three-light field and hydrophobicity of distribution network insulators as described in claim 2, characterized in that, The process of registering infrared and ultraviolet images, alignment electric field sequences, and hydrophobic water droplet images to the umbrella skirt position, based on visible light images, to construct a registration space model includes: Based on the visible light image, the corresponding feature points between the visible light image, the infrared thermal image, and the ultraviolet discharge image are extracted, the spatial transformation matrix is ​​calculated, and the infrared and ultraviolet images are mapped and superimposed onto the pixel coordinate system of the visible light image to generate a multispectral fusion image. Extract the axial center coordinates of the target umbrella skirt position in the multispectral fusion image, map the data nodes in the aligned electric field sequence to the axial center coordinates according to their positions, and generate an electric field distribution topology map; Acquire the acquisition pose parameters of the hydrophobic water droplet image, perform scaling, rotation and translation transformations on the hydrophobic image, and anchor it to the corresponding target umbrella skirt position in the multispectral fusion image; The multispectral fused image, the electric field distribution topology map, and the anchored hydrophobic water droplet image are combined into a coordinate system to output a registration space model.

4. The method for synergistic detection of three-light field and hydrophobicity of distribution network insulators as described in claim 3, characterized in that, The process involves extracting surface, temperature, discharge, electric field features, and water droplet contact angle based on a registration spatial model to generate a multi-source feature vector set, including: Gray-scale parameters and texture parameters are extracted from the visible light image channel at the target umbrella skirt location as surface features; The difference between the local maximum temperature and the ambient reference temperature is extracted from the infrared thermal image channel as a temperature feature; The number of effective discharge photons per unit area is extracted from the ultraviolet discharge image channel as the discharge feature. The electric field amplitude distortion rate within the target umbrella skirt location is extracted as an electric field feature from the electric field distribution topology map in the registration space model. The water droplet edge contour is extracted from the hydrophobic water droplet image after being anchored in the registration space model. The tangent angle of the water droplet edge contour at the gas-liquid-solid three-phase interface is calculated by fitting the edge function and used as the water droplet contact angle. Based on the historical prior distribution interval of each feature, the numerical values ​​of each feature are nonlinearly stretched and normalized to generate a normalized feature subset. The system acquires the operating parameters of the insulator, dynamically calculates the weight coefficients of each feature dimension, performs weighted modulation on the normalized feature subsets, and splices them in a preset order to output a multi-source feature vector group.

5. The method for synergistic detection of three-light field and hydrophobicity of distribution network insulators as described in claim 4, characterized in that, The step of using temperature and electric field features from a multi-source feature vector set to correct the water droplet contact angle and generate a compensation feature vector set includes: Analyze the multi-source feature vector set to extract local temperature rise parameters and electric field distortion amplitude; Thermodynamic contact angle deformation is calculated based on local temperature rise parameters, and electrodynamic contact angle offset is calculated based on electric field distortion amplitude and surface characteristics. The corrected water droplet contact angle is obtained by subtracting the thermodynamic deformation and electrodynamic offset from the original water droplet contact angle. The original contact angle in the multi-source feature vector set is updated to the corrected contact angle, and the compensated feature vector set is output.

6. The method for synergistic detection of three-light field and hydrophobicity of distribution network insulators as described in claim 5, characterized in that, The step of calculating the distance to the compensated feature vector group and generating a feature space distance matrix includes: The coordinate positions of the compensation feature vector group in the feature dimension space are extracted as test sample points, and multiple preset standard defect center clusters are called. Each standard defect center cluster corresponds to a typical deterioration state or normal state of an insulator. The eigenvector difference between the test sample point and the cluster center mean vector of each standard defect center cluster is calculated. The eigenvector difference is weighted by the pre-set covariance matrix of each feature dimension. The correlation effect and dimensional difference between feature dimensions are eliminated by the Mahalanobis distance calculation formula to obtain the Mahalanobis distance of the test sample point relative to each standard defect center cluster. The Mahalanobis distances are combined according to the category order of the corresponding standard defect center clusters to generate a feature space distance matrix.

7. The method for synergistic detection of three-light field and hydrophobicity of distribution network insulators as described in claim 6, characterized in that, The classification mapping based on the feature space distance matrix outputs the insulator defect classification results, including: Analyze the feature space distance matrix and extract the Mahalanobis distance elements of the test sample points relative to each standard defect center cluster; Traverse all Mahalanobis distance values, use a minimization optimization strategy to determine the minimum value, and extract the category index number of the standard defect center cluster corresponding to the minimum value. The system calls a preset defect status lookup table, uses the extracted category index number as the input primary key, performs matching and addressing in the defect status lookup table, and maps and outputs the final defect classification result of the target insulator. The defect classification result includes normal operation status, surface area contamination and deterioration, aging of the skirt material, or internal insulation breakdown.

8. A synergistic detection system for the three-light field and hydrophobicity of distribution network insulators, using the method described in any one of claims 1 to 7, characterized in that, include: The insulator global sensing module is used to acquire visible light, infrared, and ultraviolet images, axial spatial electric field sequences, and hydrophobic water droplet images of the target insulator. The electric field-skirt dynamic warping module is used to extract the visible light skirt outline to generate a structural reference sequence, and then align the electric field sequence with the structural reference sequence to generate an aligned electric field sequence. The multi-source image fusion and registration module is used to register infrared, ultraviolet, alignment electric field sequences and hydrophobic water droplet images to the umbrella skirt position based on visible light images, and to build a registration space model. The composite feature vector generation module is used to extract surface, temperature, discharge, electric field features and water droplet contact angle based on the registration space model to generate a multi-source feature vector set; The hydrophobic contact angle compensation module is used to correct the water droplet contact angle by utilizing the temperature and electric field characteristics in the multi-source feature vector set, and to generate a compensation feature vector set. The Mahalanobis distance evaluation module is used to calculate the distance to the compensated feature vector group and generate a feature space distance matrix. The defect status output module is used to output the insulator defect classification results based on the feature space distance matrix classification mapping.

9. A computer device, characterized in that, include: Memory and processor; The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions, which, when executed by the processor, implement the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, It stores computer-executable instructions that, when executed by a processor, implement the steps of the method according to any one of claims 1 to 7.

Images