Embedded method, system, equipment and storage medium for defect identification of insulator strings in transmission lines

Abstract

This invention discloses an embedded method, system, device, and storage medium for identifying defects in insulator strings of transmission lines, belonging to the field of defect identification technology. The method includes: acquiring an image of the insulator string; using color space changes and image processing to locate the insulator string region in the image to obtain the coordinates of the target insulator string; based on the coordinates of the target insulator string, extracting features from the insulator pieces to construct an insulator piece state vector; judging the insulator piece state vector to obtain a list of potentially defective insulators; calculating and detecting the surface of the insulator pieces in the list of potentially defective insulators to obtain the defect type and location on the surface of the insulator pieces; and generating structured transmission line insulator defect report data by associating defect category codes and descriptions based on the defect type and location. This improves the accuracy of defect identification, the precision of defect classification, and the practicality of the result report.

Description

Technical Field

[0001] This invention relates to the field of defect identification technology, specifically to a method, system, device, and storage medium for identifying defects in embedded power transmission line insulator strings. Background Technology

[0002] An embedded transmission line insulator string defect identification system is a specialized automated inspection device integrated into a specific hardware platform (such as a drone payload). Its main purpose is to automatically analyze image or video data of transmission line insulator strings collected by equipment such as drones, identify potential defects on the insulator discs, such as damage, cracks, dirt, flashover marks, and spontaneous explosions, in real time or near real time, and determine the type and location of the defects.

[0003] Current technologies lack mechanisms for actively optimizing image acquisition, leading to unstable image quality in complex and variable field environments. For example, motion blur caused by flight turbulence or overexposure and underexposure due to sudden changes in lighting directly affect the reliability of subsequent analysis and may obscure subtle defect features. Regarding target localization, when targets are obscured by tower structures or are similar in color to the background, localization errors can introduce non-target areas or miss target areas, affecting the completeness and accuracy of the analysis. Therefore, improvements are needed. Summary of the Invention

[0004] In view of the above-mentioned problems, the present invention is proposed.

[0005] Therefore, the technical problem solved by the present invention is to address the issue that the quality of images acquired in complex and ever-changing field environments is unstable due to the active optimization mechanism for image acquisition in existing technologies.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: an embedded method for identifying defects in transmission line insulator strings, comprising:

[0007] An image of an insulator string is acquired, and the insulator string region is located using color space transformation and image processing to obtain the coordinates of the target insulator string.

[0008] Based on the coordinates of the target insulator string, feature extraction is performed on the insulator segments to construct the insulator segment state vector;

[0009] The state vectors of the insulator pieces are judged to obtain a list of potentially defective insulators;

[0010] The insulator discs in the list of suspected defective insulators are calculated and their surfaces are inspected to obtain the defect types and locations on the insulator disc surfaces;

[0011] Based on the defect type and location, the defect category code and description are associated to generate structured transmission line insulator defect report data.

[0012] As a preferred embodiment of the embedded transmission line insulator string defect identification method of the present invention, the method for locating the insulator string region includes:

[0013] Read the channel pixel values ​​of the insulator string image, calculate the color elements corresponding to each pixel, and obtain the image in the color space;

[0014] Set a first threshold to filter the pixels of the image in the color space to obtain the target color pixel set;

[0015] Traverse the coordinate axes of the target color pixel set, determine the maximum and minimum coordinate information, define a rectangular area that encloses all target color pixels, record the vertex coordinates, width and height of the rectangular area, and obtain the preliminary candidate frame.

[0016] The gradient operator is applied to the image region within the initial candidate frame to calculate the gradient at each point. Non-maximum suppression and double thresholding are performed, and the edges are connected to form a continuous edge contour, which is confirmed as the boundary of the insulator string. The coordinates of the target insulator string are then established.

[0017] The beneficial effects of this preferred technical solution are that it can accurately locate the insulator string area through color screening and edge detection, effectively eliminate background interference, lay the foundation for subsequent defect identification, and significantly improve detection accuracy and efficiency.

[0018] As a preferred embodiment of the embedded transmission line insulator string defect identification method of the present invention, the construction of the insulator disc state vector includes:

[0019] Within the calibrated insulator string image area, the horizontal projection of the pixel intensity of each row is calculated. The gap between insulator pieces is located by detecting the periodic troughs of the projection curve. The sliding window matching is performed using the shape template of the pre-set hardware to locate the hardware. The independent image area and associated hardware area of ​​each insulator piece are segmented to obtain a single-piece image area set.

[0020] For each independent image region of an insulator piece in a single image region set, the pixel values ​​of the color region are converted into grayscale representation, a grayscale co-occurrence matrix is ​​constructed, texture feature parameters are calculated, and the average value and dispersion of the original region channel pixels are calculated respectively to obtain the visual parameters of the insulator piece.

[0021] Based on the independent image region and visual parameters of each insulator disc in the single image region set, the outer boundary point set of the insulator disc is obtained by using the edge detection operator, the outline of the shed is determined, the maximum span of the outline in the horizontal direction is calculated as the shed diameter, and the thickness of the central part along the insulator axis is measured. All parameters are integrated to construct the state vector of the insulator disc.

[0022] The beneficial effects of this preferred technical solution are that it comprehensively extracts the texture, color, and morphological features of insulator sheets, constructs a multi-dimensional state vector, and provides rich and accurate characterization information for defect detection.

[0023] As a preferred embodiment of the embedded transmission line insulator string defect identification method of the present invention, the determination of the insulator disc state vector includes:

[0024] A database for obtaining sample parameters of normal insulator discs;

[0025] The mean and standard deviation are calculated based on the second-order angular matrix and contrast texture features. Threshold intervals are defined, and fixed upper and lower limits of color standard deviation and umbrella skirt shape parameters are set as allowable deviations to establish a defect judgment threshold set.

[0026] Based on the defect judgment threshold set and the state vector of each insulator piece, the current values ​​of each parameter in the vector are read sequentially. The second moment of the angle and the contrast value are compared with the threshold range, the color standard deviation is compared with the upper limit threshold, and the skirt diameter and thickness values ​​are compared with the allowable deviation range. If any parameter value falls outside the corresponding normal range, the parameter is marked as abnormal, and a parameter deviation indication is obtained.

[0027] The marking result of each insulator is determined based on the parameter deviation indication of each insulator disc;

[0028] If at least one abnormal parameter is included, the insulator is identified as a suspected defective individual. The identifiers of all suspected defective insulators and their corresponding insulator state vectors are collected and compiled into a list to generate a list of suspected defective insulators.

[0029] The beneficial effects of this preferred technical solution are that by comparing multi-dimensional parameters with thresholds, suspected defective insulator pieces can be accurately screened, effectively reducing the false judgment rate and improving the accuracy and reliability of defect detection.

[0030] As a preferred embodiment of the embedded transmission line insulator string defect identification method of the present invention, the defect types and locations on the surface of the insulator discs include:

[0031] Gradient filters are applied to image blocks of suspected defective insulator pieces, and the pixel intensity change rate along the coordinate axis is calculated to synthesize gradient amplitude images, thus obtaining high-contrast surface images.

[0032] Based on the high-contrast surface image and the corresponding insulator state vector, the gradient pixels in the image are connected by chain code tracking or path search to identify linear structures whose length and curvature meet the crack conditions. Threshold segmentation is performed on the original color image region to extract abnormal dark patches. The geometric parameters and grayscale distribution of the patches are calculated to obtain the defect appearance feature set.

[0033] As a preferred embodiment of the embedded transmission line insulator string defect identification method of the present invention, it further includes:

[0034] Based on the defect appearance feature set, the similarity between the linear structure features, dark patch parameters and the quantitative features of various types of defects in the pre-stored defect sample library is calculated, the estimated defect category is assigned, and the degree of damage is evaluated based on the numerical range of the feature parameters to obtain the defect type and location.

[0035] As a preferred embodiment of the embedded transmission line insulator string defect identification method of the present invention, the generation of structured transmission line insulator defect report data includes:

[0036] Based on the defect type and location, a local image of the defective insulator piece is cropped from the original high-resolution image. At the same time, the installation sequence number of the defect within the insulator string, its three-dimensional spatial coordinates, the defect category identifier determined by the system, and the corresponding text description of the defect are retrieved to form a holographic file of the individual defect.

[0037] Based on the holographic archive of a single defect, a common data exchange format is selected, and the image slices, digitized serial numbers and coordinates, standardized defect codes and coded descriptive text in the archive are sequentially filled into a preset key-value pair template to form a record and obtain a standardized defect bar.

[0038] Based on the standardized defect bars, create a list or array structure, add each standardized defect entry as an element, sort the list according to the tower identification of the inspection task or the severity level of the defect, and generate structured transmission line insulator defect report data.

[0039] This invention provides an embedded insulator string defect identification system for power transmission lines.

[0040] To solve the above-mentioned technical problems, the present invention provides the following technical solution: an embedded transmission line insulator string defect identification system, comprising:

[0041] The positioning module is used to acquire images of insulator strings, perform insulator string region positioning on the insulator string images using color space changes and image processing, and obtain the coordinates of the target insulator string.

[0042] The feature extraction module is used to extract features from the insulator segments based on the coordinates of the target insulator string and construct the state vector of the insulator segments;

[0043] The potential defect detection module is used to judge the state vector of the insulator piece and obtain a list of insulators with suspected defects;

[0044] The defect category discrimination module is used to calculate and detect the surface of the insulator discs in the list of suspected defective insulators to obtain the defect type and location on the surface of the insulator discs.

[0045] The defect information output module is used to generate structured transmission line insulator defect report data by associating defect category codes and descriptions with the defect type and location.

[0046] The present invention provides a computer device, including a memory and a processor, wherein the memory stores a computer program, characterized in that the processor executes the computer program to implement the steps of the embedded transmission line insulator string defect identification method.

[0047] The present invention provides a computer-readable storage medium having a computer program stored thereon, characterized in that the computer program, when executed by a processor, implements the steps of the embedded transmission line insulator string defect identification method.

[0048] The beneficial effects of this invention are as follows: By actively adjusting the camera attitude and focal length, and combining the optimization of contrast and sharpness in the image sequence with specific image correction steps, the basic quality of the input image is ensured, laying the foundation for subsequent accurate analysis. Furthermore, color space transformation and threshold segmentation are used to initially screen target pixels. Combined with minimum bounding rectangle and in-frame gradient edge detection and shape analysis, the insulator string region is calibrated, effectively eliminating background interference. For the calibration region, individual insulator pieces and fittings are segmented by analyzing pixel projection and structural features. The gray-level co-occurrence matrix of each insulator piece is calculated to extract texture features. Combined with color statistics and morphological parameters such as skirt diameter and thickness, a comprehensive state vector is constructed, enriching the representation dimensions. This vector is then objectively compared with preset multi-dimensional thresholds, effectively filtering out potentially abnormal insulator pieces. Ultimately, by calculating the pixel directional derivative to enhance the surface details of suspected defect areas and detecting specific abnormal patterns such as continuous high-gradient line segments or irregular dark areas, accurate identification and localization of specific defect types such as cracks, flashovers, and spontaneous explosions were achieved. At the same time, combined with severity level assessment, structured data containing image slices, localization information, category codes, and descriptions were generated, improving the accuracy of defect identification, the precision of defect classification, and the practicality of the results report. Attached Figure Description

[0049] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying 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.

[0050] Figure 1 This is a flowchart illustrating the overall process of an embedded transmission line insulator string defect identification method according to an embodiment of the present invention. Detailed Implementation

[0051] 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.

[0052] Example 1, referring to Figure 1 This is one embodiment of the present invention, which provides an embedded method for identifying defects in transmission line insulator strings, including:

[0053] S100: Acquire an image of the insulator string, use color space transformation and image processing to locate the insulator string region in the image, and obtain the coordinates of the target insulator string;

[0054] S200: Based on the coordinates of the target insulator string, feature extraction is performed on the insulator pieces to construct the state vector of the insulator pieces;

[0055] S300: Judge the state vector of the insulator disc to obtain a list of suspected defective insulators;

[0056] S400: Calculate and inspect the surface of the insulator discs in the list of suspected defective insulators to obtain the type and location of defects on the surface of the insulator discs;

[0057] S500: Generates structured transmission line insulator defect report data by associating defect category codes and descriptions with defect type and location.

[0058] It should be noted that by actively optimizing image acquisition, accurately locating insulator string areas, comprehensively extracting features, and precisely judging defects, efficient and accurate identification of insulator defects in transmission lines has been achieved, significantly improving detection efficiency and reliability, reducing manual inspection costs, and ensuring the safe operation of the power grid.

[0059] In this embodiment of the invention, step S100 includes the following sub-steps A1-A4;

[0060] In A1: Read the channel pixel values ​​of the insulator string image, calculate the color elements corresponding to each pixel, and obtain the image in the color space;

[0061] In A2: Set a first threshold to filter the pixels of the image in the color space to obtain the target color pixel set;

[0062] In A3: Traverse the coordinate axes of the target color pixel set, determine the maximum and minimum coordinate information, define a rectangular area that surrounds all target color pixels, record the vertex coordinates, width and height of the rectangular area, and obtain the preliminary candidate frame;

[0063] In A4: Gradient operators are applied to the image regions within the initial candidate frames to calculate the gradients at each point, non-maximum suppression and double thresholding are performed, edges are connected to form continuous edge contours, which are then confirmed as the boundaries of the insulator string, and the coordinates of the target insulator string are established.

[0064] Specifically, acquiring images of insulator strings involves sending control commands to the camera gimbal's pitch and rotation motors to adjust the shooting angle based on the UAV's flight attitude and lens parameters. Simultaneously, the lens focusing motor is driven to adjust the relative position of the lens group, changing the focal length value, and obtaining the camera's ready state. Based on the camera's ready state, dynamic images of the transmission line insulator strings are continuously captured at an exposure time of 1 / 800 second and an ISO of 100. The image pixel data of each frame is stored in the onboard storage unit to generate the original image dataset. The standard deviation of the grayscale histogram is calculated for each frame as the contrast value, and the inter-frame pixel displacement is analyzed to assess motion blur. The frame with the largest standard deviation and the smallest displacement is selected, and geometric correction is applied using lens distortion model parameters to obtain a clear image of the insulator string.

[0065] Specifically, the intensity values ​​of the red, green, and blue channels of a clear insulator string image are read one by one. Typically, the value range of each channel is 0 to 255. For each pixel, the RGB (red, green, blue) value is converted to HSV (hue, saturation, lightness) value using a standard color space conversion algorithm. For example, the hue is calculated based on the maximum and minimum values ​​of red, green, and blue, and is usually mapped to a range of 0 to 360 degrees or normalized to a range of 0 to 1.

[0066] Saturation represents the purity of a color, and is calculated as follows:

[0067] ,

[0068] in, It is the minimum value among red, green, and blue. It is the maximum value among red, green, and blue;

[0069] Brightness is The normalized value is used to complete the conversion of the entire image and form an HSV image;

[0070] The first threshold is a preset hue range and minimum saturation threshold for the insulator string material. Pixels in the HSV image are filtered based on the first threshold. The hue range and minimum saturation threshold are obtained by statistical color analysis of a large number of known insulator sample images. For common brown or cyan ceramic insulators, the preset hue range is set, for example, a hue value of 15 to 35 for brown (e.g., hue range 0-360), or a hue value of 170 to 190 for cyan. The minimum saturation threshold is set empirically, for example, greater than 0.3 (saturation range 0-1), to exclude low-saturation background objects such as the sky or concrete towers. The brightness is usually set to a wide range, for example, 0.2 to 0.9, to adapt to changes in lighting and shadows. Pixels whose hue, saturation, and brightness all fall within the preset range are marked as belonging to the target color. All these marked pixels are aggregated to obtain the target color pixel set.

[0071] Iterate through every pixel in the target color pixel set and read the pixel's position in the original image. (Horizontal) and (Vertical) coordinate values: Before the traversal begins, initialize four variables: minimum... coordinate( ) is a value greater than the maximum width of the image (e.g., image width plus 1), maximum coordinate( The value is -1, the minimum. coordinate( () is a value greater than the maximum height of the image (e.g., image height plus 1), maximum coordinate( The value is -1, and during the traversal, for each pixel in the target color pixel set... ,Will Coordinates and current and If a comparison is made, Then update ,like Then update Similarly, Coordinates and current and If a comparison is made, Then update ,like Then update After all pixels in the target color pixel set have been traversed and compared, the obtained , , , These four values ​​constitute the boundary of the smallest bounding rectangle that completely encloses all target color pixels. The system then records the coordinates of the top-left vertex of the rectangular region. Then, the width and height of the rectangle are calculated to obtain the preliminary candidate frame. The width and height of the rectangle are expressed as:

[0072] ,

[0073] ,

[0074] in, for The maximum and minimum values ​​of the coordinates. for The maximum and minimum values ​​of the coordinates. For width, For height;

[0075] Extract the image sub-regions corresponding to the initial candidate bounding boxes, and apply gradient operators, such as the Sobel operator, to these sub-regions. This is done by convolving the image sub-regions with 3x3 Sobel horizontal and vertical convolution kernels, respectively, to calculate the gradient of each pixel. gradient components of direction and gradient components of direction Therefore, the gradient magnitude and gradient direction of a point are expressed as follows:

[0076] ,

[0077] ,

[0078] in, It is the azimuth function. For the gradient direction, This represents the gradient magnitude.

[0079] Next, non-maximum suppression (NMS) is performed. For each pixel, the gradient magnitude is checked along the gradient direction to see if it is greater than the gradient magnitudes of the two adjacent pixels. If not, the pixel's gradient magnitude is set to zero, thus refining the edges. Then, a dual-threshold method is used for edge connection, setting a high threshold and a low threshold. For example, the high threshold is set to the 90th percentile of the cumulative distribution histogram of all non-zero gradient magnitudes in the NMS-suppressed image, such as 120. The low threshold is set as a specific proportion of the high threshold, such as 0.4 × 120, or 48. Pixels with gradient magnitudes higher than the high threshold are considered strong edges and are retained directly. Pixels with gradient magnitudes lower than the low threshold are suppressed and not considered edges. Pixels with gradient magnitudes between the low and high thresholds are considered strong edges. Weak edge points are only retained as edge points when they are connected to strong edge points through an 8-neighborhood. All edge pixels meeting these criteria are traced and connected to form a continuous set of edge contours. Connectivity analysis is then performed on these contours, calculating the attributes of each connected component, such as area, perimeter, aspect ratio, and roundness. For example, the lower limit for the area of ​​an effective contour is set to 0.5% of the total image pixels (to eliminate excessively small noise), and the aspect ratio (major axis length / secondary axis length) is greater than 2.5 (insulator strings are typically slender structures). This aspect ratio threshold is obtained by statistically analyzing the geometric features of 100 typical insulator string sample images, taking the 10th percentile value of the distribution. The connected component that best meets these preset geometric constraints is selected as the boundary of the insulator string. Finally, the geometric center coordinates are calculated based on the confirmed boundary pixel set. The coordinates of the target insulator string can be established by using the coarse center of the insulator string as a reference point, or by determining the principal axis direction as the center line through principal component analysis (PCA) of the boundary points.

[0080] It should be noted that by actively adjusting the shooting parameters and optimizing the image processing flow, the acquisition quality and positioning accuracy of the insulator string images were significantly improved.

[0081] In this embodiment of the invention, step S200 includes the following sub-steps B1-B3;

[0082] In B1: Within the calibrated insulator string image area, the horizontal projection of the pixel intensity of each row is calculated. The gap between insulator pieces is located by detecting the periodic troughs of the projection curve. The sliding window matching is performed using the shape template of the preset hardware to locate the hardware. The independent image area and associated hardware area of ​​each insulator piece are segmented to obtain a single-piece image area set.

[0083] In B2: For the independent image region of each insulator piece in the single image region set, the pixel values ​​of the color region are converted into grayscale representation, a grayscale co-occurrence matrix is ​​constructed, texture feature parameters are calculated, and the average value and dispersion of the original region channel pixels are calculated respectively to obtain the visual parameters of the insulator piece.

[0084] In B3: Based on the independent image region of each insulator disc and the visual parameters of the insulator disc in the single image region set, the edge detection operator is used to obtain the outer boundary point set of the insulator disc, determine the outline of the shed, calculate the maximum span of the outline in the horizontal direction as the shed diameter, and measure the thickness of the central part along the insulator axis. All parameters are integrated to construct the state vector of the insulator disc.

[0085] Specifically, the marked insulator string image area is cropped from the complete aerial image. If the insulator string is tilted in the image, the principal axis direction is calculated and rotational correction is performed to make the insulator string vertical or horizontal. Within the corrected insulator string image area, it is converted into a grayscale image. Then, the sum of pixel intensity is calculated row by row to form a one-dimensional horizontal projection curve. The horizontal axis of the curve represents the row number, and the vertical axis represents the total grayscale value of the row pixels. Due to the material and structure, the grayscale value of the insulator sheet body is usually higher or lower than that of the inter-sheet gap area (depending on the background and insulator color), so it is represented on the projection curve. The current pattern consists of periodic peaks and troughs. The gaps between insulator discs are located by identifying these troughs. Specifically, the projection curve is smoothed, for example using Gaussian smoothing with a window size of 3 to 5 pixels to remove noise. Then, a peak-trough detection algorithm is used. A point is identified as a trough. If the projected value is less than the projected values ​​of the two nearest points on either side, and the projected value is below a preset trough depth threshold (set as 60% of the average of all local peaks on the projection curve), this percentage is determined by statistically analyzing the projection data of 50 different types of insulator strings to select the proportion that maximizes the gap detection accuracy. For example, if the average peak value... If the value is 20000, then the valley depth threshold is 12000. A continuous valley sequence that meets specific distance constraints (e.g., spacing between 30 and 80 pixels, estimated based on typical insulator disc height and shooting resolution) is used to initially segment the insulator discs. Simultaneously, the system loads various preset hardware shape templates, such as U-shaped hanging ring templates and ball-head hanging ring templates. These templates are small-sized (e.g., 30x40 pixels) grayscale images obtained by cropping and averaging a large number of standard hardware images. A sliding window is used at the top and bottom of the insulator string image area, and at locations where hardware connections exist according to the insulator string model. The method involves performing normalized cross-correlation (NCC) calculations on each fitting template and the corresponding image sub-region within the window. When the NCC value exceeds a preset matching similarity threshold, such as 0.7, it is considered that a fitting has been located. The threshold of 0.7 is obtained by performing receiver operating characteristic (ROC) curve analysis on a test set containing 100 known fittings and 100 non-fitting regions, selecting the NCC value corresponding to the point with the maximum Youden index. Based on the identified gap position and fitting position, the insulator string image is finally segmented into a series of independent insulator piece image regions and associated fitting image regions to obtain a single-piece image region set.

[0086] Based on the independent image region of each segmented insulator piece in the single-piece image region set, the pixel values ​​in the color image region are converted to grayscale representation using the formula: grayscale value = 0.299 × red channel value + 0.587 × green channel value + 0.114 × blue channel value. This yields the grayscale image of a single insulator piece. A grayscale co-occurrence matrix (GLCM) is then constructed on the grayscale image to compress the number of grayscale levels from 0-255 to a smaller range, such as 0-31 (32 grayscale levels). The compression method is linear mapping, i.e., the new grayscale levels... The statistical directions are selected as 0 degrees, 45 degrees, 90 degrees, and 135 degrees, with a fixed statistical distance of 1 pixel. The GLCM is calculated in each of these four directions, and the corresponding elements are summed and normalized to obtain an average GLCM. Based on this average GLCM, the second moment of the angle is calculated as the sum of the squares of all element values ​​in the matrix. For example, if the GLCM is a 2x2 matrix... Then the second moment of the angle is It also calculates contrast, which is the weighted sum of squares of the differences between different gray level pairs in GLCM, and entropy, which is the sum of the probability values ​​of each element in GLCM multiplied by the logarithm and then negative. Returning to the original color independent image region, it calculates the arithmetic mean and standard deviation of all pixel values ​​in each of the three channels: red (R), green (G), and blue (B). It combines the texture parameters (second moment of the angle, contrast, entropy) and color statistical parameters (average value of R channel, standard deviation of R channel, average value of G channel, standard deviation of G channel, average value of B channel, standard deviation of B channel) to obtain the visual parameters of the insulator sheet.

[0087] Based on the independent image regions of each insulator disc in the single-image region set and the visual parameters of the insulator disc, edge detection operators are applied to the independent image regions, such as the Canny edge detection algorithm. This includes smoothing the image using a Gaussian filter (e.g., kernel size 5x5, standard deviation 1.4), calculating gradient magnitude and direction, performing non-maximum suppression, and finally connecting the edges using a double thresholding method. The setting of the high and low thresholds is crucial for extracting accurate skirt contours. The high threshold can be set to the 85th percentile of the gradient magnitude histogram of the independent image region, for example, calculated to be 100. The low threshold is set to 0.4 times the high threshold, i.e., 40. This method obtains the set of boundary points of the outermost layer of the insulator disc, and the longest continuous contour is selected as the skirt contour line. Next, in order to calculate the skirt diameter, the principal axis direction of the insulator disc needs to be determined. This is obtained by performing principal component analysis (PCA) on the contour point set. The contour line is rotated so that it is perpendicular or horizontal to the principal axis direction, and then perpendicular to the... In the direction of the main axis (i.e., the radial direction of the skirt), calculate the maximum span of the contour line point set in that direction. For example, if the main axis is vertical, calculate the difference between the maximum and minimum abscissas of all points on the contour line. This difference is the skirt diameter. To measure the thickness of the central part along the insulator axis, first, in the direction parallel to the main axis, exclude the edge area of ​​the skirt (e.g., exclude 20% of the area at each end of the contour line in the direction of the main axis). In the remaining central area, measure multiple distances between the contour lines in the direction perpendicular to the main axis, and take the average or minimum value as the central part thickness. For example, in the central area of ​​the vertically aligned insulator disc, select 5 horizontal lines at equal intervals, calculate the distance between the left and right points of the contour on each horizontal line, and the average of these 5 distances is the central thickness. Finally, integrate the calculated skirt diameter, central part thickness, and all parameters such as the second moment of the angle, contrast, entropy value, average value and standard deviation of each color channel to form a multi-dimensional vector, and construct the insulator disc state vector.

[0088] It should be noted that by accurately segmenting images and extracting multi-dimensional features, a comprehensive state vector of the insulator sheet is constructed, providing rich and accurate characterization information for defect detection.

[0089] In this embodiment of the invention, step S300 includes the following sub-steps C1-C5;

[0090] In C1: Obtain the database of normal insulator sample parameters;

[0091] In C2: the mean and standard deviation are calculated based on the second-order angular matrix and contrast texture features, the threshold interval is defined, and the fixed upper and lower limits of the color standard deviation and the umbrella skirt shape parameters are set as allowable deviations to establish a defect judgment threshold set.

[0092] In C3: Based on the defect judgment threshold set and the state vector of each insulator piece, the current value of each parameter in the vector is read in sequence. The second moment of the angle and the contrast value are compared with the threshold range, the color standard deviation is compared with the upper limit threshold, and the skirt diameter and thickness values ​​are compared with the allowable deviation range. If any parameter value falls outside the corresponding normal range, the parameter is marked as abnormal, and the parameter deviation indication is obtained.

[0093] In C4: Based on the parameter deviation indication of each insulator disc, determine the marking result of each insulator disc;

[0094] In C5: If at least one abnormal parameter marker is included, the insulator is judged to be a suspected defective individual. Collect the identifiers of all suspected defective insulators and their corresponding insulator state vectors, compile them into a list, and generate a list of suspected defective insulators.

[0095] Specifically, the values ​​corresponding to all normal samples are extracted from the database of normal insulator sample parameters, and the arithmetic mean and standard deviation are calculated for each sample. A threshold interval is defined and set as the mean plus or minus... Multiple standard deviations, for example The threshold was determined by optimizing the balance between defect detection rate and false alarm rate on the validation set, ensuring coverage of approximately 98.8% of normal sample fluctuations. Taking the second moment of the angle as an example, if the mean of normal samples is 0.075 and the standard deviation is 0.01, then the threshold range is set as follows: ,Right now Regarding contrast, if the mean of normal samples is 30 and the standard deviation is 8, then the dynamic threshold range is... ,Right now For color standard deviations (e.g., the standard deviations of R, G, and B channels), since they often indicate dirt or uneven material, only an upper limit is usually set. This upper limit is taken as the 98th percentile of the corresponding color standard deviation value in the normal sample database. For example, if the 98th percentile of the standard deviation of the R channel in the normal sample is 18 (gray level 0-255), then the allowable upper limit of the standard deviation of the R channel is 18. For skirt morphology parameters such as skirt diameter and center thickness, fixed upper and lower limits are set as allowable deviation ranges based on the nominal dimensions and allowable tolerances in the design specifications corresponding to the insulator model, or in combination with the distribution of these parameters in the normal sample database (e.g., taking the 1% and 99th percentiles). For example, if the nominal skirt diameter of a certain model of insulator is 250 mm and the allowable tolerance is ±5 mm, then the allowable deviation range can be set as [245 mm, 255 mm]. The normal range or threshold of each key parameter in the state vector is obtained, and a defect judgment threshold set is established.

[0096] Based on the defect judgment threshold set and the insulator state vector of each insulator disc to be inspected, the system sequentially reads the current measured values ​​of each parameter within the state vector and compares them with the normal range of the corresponding parameter in the defect judgment threshold set. The normal range is the preset allowable value interval for each parameter in the defect judgment threshold set. For example, for the angular second moment parameter, if the measured value of the current insulator disc's angular second moment is 0.045, while the normal range of the angular second moment obtained from the defect judgment threshold set is [0.050, 0.100], since 0.045 is less than 0.050, the parameter value falls outside the normal range, and the angular second moment parameter of this insulator disc is marked as abnormal. Similarly, if the measured value of the contrast is 55, while the normal range of the contrast is [10, 50], since 55 is greater than 50, the contrast parameter is also marked as abnormal. For the color standard deviation, R-pass... Taking the standard deviation of the channel as an example, if the current value is 22, and the preset upper limit threshold of the standard deviation of the R channel is 18, since 22 is greater than 18, this parameter is also marked as abnormal. For the skirt diameter, if the measured value is 240 mm, and the allowable deviation range is [245 mm, 255 mm], since 240 mm is less than 245 mm, the skirt diameter is marked as abnormal. A similar comparison operation is performed on the thickness of the center part and all other parameters contained in the state vector. The current value of each parameter is checked against the normal range or unilateral threshold defined in the defect judgment threshold set. As long as the current value of any parameter exceeds the corresponding normal range boundary, the specific parameter is given an abnormal mark. After completing the item-by-item comparison of all parameters in the state vector of the insulator, the system integrates these marking information and generates a parameter deviation indication containing the normal or abnormal status of each parameter for the current insulator.

[0097] The markings on each insulator disc in the inspection images are checked. During the inspection, if at least one parameter of an insulator disc deviates from the indication and is marked as abnormal, then the insulator disc is initially identified as a suspected defective individual. For example, for insulator disc A, if the second moment of the angle is within the normal range, but the contrast is marked as abnormal, while the color standard deviation is normal, the skirt diameter is normal, and the center thickness is normal, then insulator disc A is classified as a suspected defective individual due to the abnormal contrast parameter. Similarly, for insulator disc B, if the skirt diameter exceeds the lower limit of the allowable deviation range, and the R channel color standard deviation also exceeds... If the upper limit is exceeded, even if all other parameters are within the normal range, insulator piece B is still judged as a suspected defective individual due to the presence of two abnormal parameter markers: the shed diameter and the standard deviation of the R channel color. For insulator piece C, if the parameter deviation indicator shows that all parameters are within their respective normal threshold ranges and there are no abnormal markers, then the insulator piece is not judged as a suspected defective individual. The unique identifiers and corresponding complete insulator piece state vectors of all insulator pieces judged as suspected defective individuals are collected and the information is recorded one by one, and compiled into a structured list. This list is the generated list of suspected defective insulators.

[0098] It should be noted that by scientifically setting thresholds and strictly comparing parameters, suspected defective insulators can be accurately screened, effectively improving the accuracy and reliability of detection.

[0099] In this embodiment of the invention, step S400 includes the following sub-steps D1-D3;

[0100] In D1: Gradient filters are applied to image blocks of suspected defective insulator pieces, the pixel intensity change rate in the coordinate axis direction is calculated respectively, and gradient amplitude images are synthesized to obtain high-contrast surface images;

[0101] In D2: Based on the high-contrast surface image and the corresponding insulator state vector, the gradient pixels in the image are connected by chain code tracking or path search to identify linear structures whose length and curvature meet the crack conditions. Threshold segmentation is performed on the original color image area to extract abnormal dark patches. The geometric parameters and grayscale distribution of the patches are calculated to obtain the defect appearance feature set.

[0102] In D3: Based on the defect appearance feature set, the similarity between the linear structure features, dark patch parameters and the quantitative features of each type of defect in the pre-stored defect sample library is calculated, the estimated defect category is assigned, and the degree of damage is evaluated based on the numerical range of the feature parameters to obtain the defect type and location.

[0103] Specifically, for each suspected defective insulator piece in the list of suspected defective insulators, an image patch is extracted from the original image pixel data. If the original image is color, it is converted to grayscale. A gradient filter is applied to the grayscale image patch; the Sobel operator can be used, by using a 3x3 Sobel horizontal convolution kernel, for example... and vertical convolution kernels, for example By performing two-dimensional convolution operations with each image patch, the pixel intensity change rate of each pixel in the image patch in the x-direction is calculated, denoted as . The pixel intensity change rate in the y direction is denoted as... The gradient magnitude is synthesized for each pixel, and the gradient magnitude is expressed as:

[0104] ,

[0105] in This refers to the gradient magnitude of a pixel, resulting in a new image where the pixel values ​​are all composed of the corresponding gradient magnitude. This significantly highlights areas in the original image where pixel intensity changes drastically, such as object edges, texture details, and surface irregularities like potential cracks or scratches, thus obtaining a high-contrast surface image.

[0106] For binarization of high-contrast surface images, the Otsu adaptive thresholding algorithm can be used. This algorithm automatically selects the optimal segmentation threshold by maximizing the inter-class variance. For example, if the calculated Otsu threshold is 70 (grayscale range 0-255), pixels with a gradient magnitude greater than 70 are marked as foreground, and the rest as background. Foreground pixels in the binarized image, a chain code tracking algorithm is used. Starting from a foreground point, the algorithm searches for the next foreground point along eight neighborhoods, encoding it with numbers representing directions (e.g., 0 for east, 1 for northeast, and so on up to 7 for southeast). The path is recorded until the starting point is reached or no foreground point can be connected, forming closed or open chains. These chains are then identified as linear structures. For each linear structure, the length (length of the chain code sequence) and average curvature (e.g., obtained by dividing the cumulative value of the chain code direction change by the length) are calculated. The length must exceed a minimum length threshold, for example, set to 15 pixels. This value is based on experience; linear structures with a length less than 15 pixels are usually noise or insignificant. The system identifies minor scratches, not typical cracks. A 15-pixel threshold is determined based on the analysis of 200 known minor crack sample images, selecting the minimum length that excludes over 90% of non-crack interference. Simultaneously, the average curvature must be less than a maximum curvature threshold, such as 0.2 radians / pixel, to ensure the straightness or gradual bending of the linear structure. This curvature threshold is set by analyzing the curvature distribution of 50 known real cracks, representing the maximum curvature value covering 95% of the real crack samples. Linear structures meeting these two conditions are initially identified as crack candidates. Simultaneously, the system returns to the original color image area of ​​the insulator sheet, converts it to the HSV color space, and performs threshold segmentation on the V (brightness) channel to extract abnormal dark patches. The segmentation threshold for dark patches is set to the 15th percentile value of the histogram of V channel pixel values ​​on the insulator sheet. For example, if this value is calculated to be 60, pixels with V channel values ​​less than 60 are considered dark area candidates. Connectivity analysis is performed on the formed dark areas to calculate the area, perimeter, and circularity of each connected dark patch. The average gray value, along with the average gray value, constitutes the defect appearance feature set;

[0107] Similarity calculations are performed between features extracted in real-time from the defect appearance feature set and quantitative features of various defect types in a pre-stored defect sample library. The pre-stored defect sample library is established by extracting features and performing statistical analysis on a large number (e.g., over 1000) images of defective insulators (such as spontaneous explosion, flashover, icing, contamination, and damage) whose types have been confirmed by experts. Each defect type in the library corresponds to one or more feature vector templates or statistical distribution models of feature parameters (e.g., mean and covariance matrix). Similarity calculations can use a trained classifier model, such as a Support Vector Machine (SVM) or a deep learning network (e.g., a Convolutional Neural Network, CNN). The model takes the defect appearance feature set as input and outputs the defect category. For example, an SVM classifier is trained as follows: using at least 80% of the data in the pre-stored defect sample library as the training set, extracting the defect appearance feature set of each sample to form a feature vector, assigning it a true defect category label, and selecting a radial basis function (RBF) as the kernel function. The optimal penalty coefficient C (e.g., C=10) and kernel parameter gamma (e.g., gamma=0.01) are determined through grid search and 10-fold cross-validation. A classification model is then trained, with the input being the apparent feature set of defects in the insulator under test. The output is the probability of each defect category or a direct category determination. When the probability of a certain category exceeds a preset confidence threshold (e.g., 0.75, which is adjusted on the remaining 20% ​​of the validation set to achieve the best balance between precision and recall), the category is assigned as the predicted defect category. After determining the defect category, the degree of damage is assessed based on the numerical range of key feature parameters related to the defect type. For example, if it is determined to be a crack, it is classified into micro cracks (e.g., less than 20 pixels in length), general cracks (20-50 pixels in length), or severe cracks (greater than 50 pixels in length) based on the crack length. Finally, by combining the predicted defect category and the assessed degree of damage with the pixel coordinates of the defect in the image, the confirmed defect type and location are obtained.

[0108] It should be noted that by enhancing image features and using intelligent classification algorithms, the type and location of insulator defects can be accurately identified, significantly improving detection accuracy and efficiency.

[0109] In this embodiment of the invention, step S500 includes the following sub-steps E1-E3;

[0110] In E1: Based on the defect type and location, a local image of the defective insulator piece is cropped from the original high-resolution image. At the same time, the installation sequence number of the defect within the insulator string, the three-dimensional spatial coordinates, the defect category identifier determined by the system, and the corresponding text description of the defect are retrieved to form a holographic file of the individual defect.

[0111] In E2: Based on the holographic archive of a single defect, a common data exchange format is selected, and the image slices, digitized serial numbers and coordinates, standardized defect codes and coded descriptive text in the archive are sequentially filled into a preset key-value pair template to form a record and obtain a standardized defect bar.

[0112] In E3: Based on standardized defect bars, create a list or array structure, add each standardized defect entry as an element, sort the list according to the tower identification of the inspection task or the severity level of the defect, and generate structured transmission line insulator defect report data.

[0113] Specifically, based on the positioning information in the insulator disc image, a local image slice containing the defect is cropped from the stored original high-resolution aerial image. During cropping, a certain margin is added to the defect range, such as extending by 30 pixels in each direction, to ensure the integrity of the defect context. At the same time, the installation sequence number of the insulator disc with the defect in the insulator string is retrieved from the image processing stage or metadata associated with the inspection task, and the three-dimensional spatial coordinates are extracted from the exchangeable image file format (EXIF) or synchronized flight logs recorded when the drone took the image. According to the determined defect type, the corresponding defect category identifier is found from the preset defect code-description mapping table. For example, "self-explosion" corresponds to the code "ZBA01". Each code in the table is associated with a standard text description of the defect situation. For example, "ZBA01" corresponds to "the insulator shed is completely broken and loses its insulation function". The image slice, installation sequence number, GPS three-dimensional spatial coordinates, defect category identifier, defect situation text description, original image file name, shooting timestamp, etc. are integrated to form an independent single defect holographic file for each confirmed defect.

[0114] Based on the individual defect holographic file of each defective insulator piece, a data exchange format is selected, such as JSON (text data exchange format). The data in the individual defect holographic file is organized according to a predefined key-value pair template, specifying the field name (key) and data type of each information element in the output data. For example, a preset JSON template structure is as follows: {"defect_id": "unique defect number", "tower_id": "tower identifier", "string_id": "serial number", "piece_index": installation serial number, "gps_longitude": longitude value, "gps_latitude": latitude value, "gps_altitude": altitude value, "capture_timestamp": "capture time", "defect_code": "defect category identifier", "defect_description": "defect description", "severity_ The `image_slice_base64` field converts the image slices in the individual defect holographic file into Base64 encoded strings and fills them into the `image_slice_base64` field. The digitized installation serial number and coordinates are filled into the corresponding numeric fields. The standardized defect category identifier is filled into the `defect_code` field, and the UTF-8 encoded text description of the defect is filled into the `defect_description` field. Other information such as tower identification, serial number, shooting time, and severity level code (e.g., "01" represents critical, "02" represents severe, and "03" represents moderate) is also filled in. Each completed entry of information for a defective insulator piece constitutes a structured JSON record, thus obtaining a standardized defect entry set containing multiple independent defect records.

[0115] Create a list or array structure in memory and add each standardized defect entry as an independent element to this list or array. After all standardized defect entries have been added, sort the list according to a preset or user-selected sorting rule. There are multiple sorting rules; for example, it can be sorted in ascending order based on the tower identification of the inspection task. Under the same tower identification, it can be sorted in descending order based on the severity level code of the defect (e.g., "01" critical is placed first, and "03" general is placed last). Finally, under the same severity level, it can be sorted in ascending order by insulator string number and piece number. (For example, the primary key is the tower identifier, the secondary key is the severity level, and the tertiary key is the serial number and piece number.) This facilitates maintenance personnel in planning maintenance work according to region and urgency. The severity level codes are sorted according to their degree of threat to the safe operation of the line. For example, "01-Critical" needs to be dealt with immediately, "02-Severe" needs to be dealt with as soon as possible, and "03-General" can be included in regular maintenance. Alternatively, the severity level code of the defect can be used as the primary sorting key for global descending sorting, placing all critical defects at the top, regardless of which tower they belong to. After sorting, the final structured transmission line insulator defect report data is obtained.

[0116] It should be noted that the generation of structured data enables efficient organization and output of defect information, facilitating maintenance personnel to quickly locate and handle defects, and improving the efficiency of power grid operation and maintenance.

[0117] The above is a schematic scheme of the embedded transmission line insulator string defect identification method of this embodiment. It should be noted that the technical solution of this embedded transmission line insulator string defect identification system and the technical solution of the embedded transmission line insulator string defect identification method described above belong to the same concept. For details not described in detail in the technical solution of the embedded transmission line insulator string defect identification system in this embodiment, please refer to the description of the technical solution of the embedded transmission line insulator string defect identification method described above.

[0118] The embedded transmission line insulator string defect identification system in this embodiment includes:

[0119] The positioning module is used to acquire images of insulator strings, perform insulator string region positioning on the insulator string images using color space changes and image processing, and obtain the coordinates of the target insulator string.

[0120] The feature extraction module is used to extract features from the insulator segments based on the coordinates of the target insulator string and construct the state vector of the insulator segments;

[0121] The potential defect detection module is used to judge the state vector of the insulator piece and obtain a list of insulators with suspected defects;

[0122] The defect category discrimination module is used to calculate and detect the surface of the insulator discs in the list of suspected defective insulators to obtain the defect type and location on the surface of the insulator discs.

[0123] The defect information output module is used to generate structured transmission line insulator defect report data by associating defect category codes and descriptions with the defect type and location.

[0124] This embodiment also provides a computer device applicable to the embedded transmission line insulator string defect identification method, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the embedded transmission line insulator string defect identification method proposed in the above embodiment.

[0125] This embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, it implements the embedded transmission line insulator string defect identification method proposed in the above embodiments.

[0126] The storage medium proposed in this embodiment and the embedded transmission line insulator string defect identification method proposed in the above embodiments belong to the same inventive concept. 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.

[0127] Based on the above description of the implementation methods, those skilled in the art can clearly understand that the present invention can be implemented using software and necessary general-purpose hardware, and of course, it can also be implemented using hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ROM), random access memory (RAM), flash memory, hard disk, or optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of the various embodiments of the present invention.

[0128] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not 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 identifying defects in embedded transmission line insulator strings, characterized in that, include: An image of an insulator string is acquired, and the insulator string region is located using color space transformation and image processing to obtain the coordinates of the target insulator string. Based on the coordinates of the target insulator string, feature extraction is performed on the insulator segments to construct the insulator segment state vector; The state vectors of the insulator pieces are judged to obtain a list of potentially defective insulators; The insulator discs in the list of suspected defective insulators are calculated and their surfaces are inspected to obtain the defect types and locations on the insulator disc surfaces; Based on the defect type and location, the defect category code and description are associated to generate structured transmission line insulator defect report data; Constructing the state vector of an insulator piece includes: Within the calibrated insulator string image area, the horizontal projection of the pixel intensity of each row is calculated. The gap between insulator pieces is located by detecting the periodic troughs of the projection curve. The sliding window matching is performed using the shape template of the pre-set hardware to locate the hardware. The independent image area and associated hardware area of ​​each insulator piece are segmented to obtain a single-piece image area set. For each independent image region of an insulator piece in a single image region set, the pixel values ​​of the color region are converted into grayscale representation, a grayscale co-occurrence matrix is ​​constructed, texture feature parameters are calculated, and the average value and dispersion of the original region channel pixels are calculated respectively to obtain the visual parameters of the insulator piece. Based on the independent image region and visual parameters of each insulator disc in the single image region set, the outer boundary point set of the insulator disc is obtained by using the edge detection operator, the outline of the shed is determined, the maximum span of the outline in the horizontal direction is calculated as the diameter of the shed, and the thickness of the central part along the insulator axis is measured. All parameters are integrated to construct the state vector of the insulator disc. Determining the state vector of the insulator disc includes: A database for obtaining sample parameters of normal insulator discs; The mean and standard deviation are calculated based on the second-order angular matrix and contrast texture features. Threshold intervals are defined, and fixed upper and lower limits of color standard deviation and umbrella skirt shape parameters are set as allowable deviations to establish a defect judgment threshold set. Based on the defect judgment threshold set and the state vector of each insulator piece, the current values ​​of each parameter in the vector are read sequentially. The second moment of the angle and the contrast value are compared with the threshold range, the color standard deviation is compared with the upper limit threshold, and the skirt diameter and thickness values ​​are compared with the allowable deviation range. If any parameter value falls outside the corresponding normal range, the parameter is marked as abnormal, and a parameter deviation indication is obtained. The marking result of each insulator is determined based on the parameter deviation indication of each insulator disc; If at least one abnormal parameter is included, the insulator piece is judged to be a suspected defective individual. Collect the identifiers of all suspected defective insulator pieces and their corresponding insulator piece state vectors, compile them into a list, and generate a list of suspected defective insulators. The types and locations of defects on the surface of the insulator discs were obtained, including: Gradient filters are applied to image blocks of suspected defective insulator pieces, and the pixel intensity change rate along the coordinate axis is calculated to synthesize gradient amplitude images, thus obtaining high-contrast surface images. Based on the high-contrast surface image and the corresponding insulator state vector, the gradient pixels in the image are connected by chain code tracking or path search to identify linear structures whose length and curvature meet the crack conditions. The original color image region is then thresholded to extract abnormal dark patches. The geometric parameters and grayscale distribution of the patches are calculated to obtain the defect appearance feature set. Also includes: Based on the defect appearance feature set, the similarity between the linear structure features, dark patch parameters and the quantitative features of each type of defect in the pre-stored defect sample library is calculated, the estimated defect category is assigned, and the degree of damage is evaluated based on the numerical range of the feature parameters to obtain the defect type and location. The data for generating a structured transmission line insulator defect report includes: Based on the defect type and location, a local image of the defective insulator piece is cropped from the original high-resolution image. At the same time, the installation sequence number of the defect within the insulator string, its three-dimensional spatial coordinates, the defect category identifier determined by the system, and the corresponding text description of the defect are retrieved to form a holographic file of the individual defect. Based on the holographic archive of a single defect, a common data exchange format is selected, and the image slices, digitized serial numbers and coordinates, standardized defect codes and coded descriptive text in the archive are sequentially filled into a preset key-value pair template to form a record and obtain a standardized defect bar. Based on the standardized defect bars, create a list or array structure, add each standardized defect entry as an element, sort the list according to the tower identification of the inspection task or the severity level of the defect, and generate structured transmission line insulator defect report data.

2. The embedded transmission line insulator string defect identification method as described in claim 1, characterized in that, Insulator string region positioning includes: Read the channel pixel values ​​of the insulator string image, calculate the color elements corresponding to each pixel, and obtain the image in the color space; Set a first threshold to filter the pixels of the image in the color space to obtain the target color pixel set; Traverse the coordinate axes of the target color pixel set, determine the maximum and minimum coordinate information, define a rectangular area that encloses all target color pixels, record the vertex coordinates, width and height of the rectangular area, and obtain the preliminary candidate frame. The gradient operator is applied to the image region within the initial candidate frame to calculate the gradient at each point. Non-maximum suppression and double thresholding are performed, and the edges are connected to form a continuous edge contour, which is confirmed as the boundary of the insulator string. The coordinates of the target insulator string are then established.

3. An embedded transmission line insulator string defect identification system, employing the embedded transmission line insulator string defect identification method as described in any one of claims 1 to 2, characterized in that, The positioning module is used to acquire images of insulator strings, perform insulator string region positioning on the insulator string images using color space changes and image processing, and obtain the coordinates of the target insulator string. The feature extraction module is used to extract features from the insulator segments based on the coordinates of the target insulator string and construct the state vector of the insulator segments; The potential defect detection module is used to judge the state vector of the insulator piece and obtain a list of insulators with suspected defects; The defect category discrimination module is used to calculate and detect the surface of the insulator discs in the list of suspected defective insulators to obtain the defect type and location on the surface of the insulator discs. The defect information output module is used to generate structured transmission line insulator defect report data by associating defect category codes and descriptions with the defect type and location.

4. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the embedded transmission line insulator string defect identification method according to any one of claims 1 to 2.

5. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the embedded transmission line insulator string defect identification method according to any one of claims 1 to 2.

Images