An AI-based electrical circuit image analysis and comparison method and system

Abstract

The present application provides an AI-based electric line image analysis and comparison method and system, including: identifying and locating the damaged area of ​​the cable through an image processing method, obtaining the contour information of the damaged area through an edge detection algorithm, and judging whether the damaged shape is circular or star-shaped according to the contour shape characteristics; segmenting the damaged area through an image segmentation algorithm, and obtaining the size of the damaged area by calculating the pixel area of ​​the segmented area; if it is determined that the internal wire insulation layer is damaged, the damaged area is enlarged and clarified through an image enhancement algorithm, the number and position of the damaged wires are analyzed, and whether the copper core is intact is judged; after completing the insulation wrapping of a single wire, the entire damaged area is wrapped with multiple layers of PVC tape through automated equipment, and the number of winding layers is determined according to the importance of the cable and environmental conditions, and the number of winding layers is at least two layers.

Description

Technical Field

[0001] The present invention relates to the field of information technology, and in particular to an AI-based electric circuit image analysis and comparison method and system. Background Art

[0002] In the image of the power line, the outer sheath damage and the internal wire insulation damage are two common types of defects. When the exposed defect area is equal, the difference in the shape of these two defects will lead to different safety hazards and repair difficulties. A single circular damage to the outer sheath is usually caused by external impact or wear, and the damage range is relatively concentrated. In this case, the integrity of the internal wire insulation layer is the key to judging the severity of the defect. If the internal wire insulation layer is not damaged, the wire core is not exposed, and the cable is used in a dry environment, the repair difficulty is relatively low, and it can be repaired by partial wrapping. If the outer sheath and the internal wire insulation layer are damaged in a star-shaped point shape, it may be caused by multiple strands of small foreign objects piercing. In this case, although the defect area is the same, the damage points are many and scattered, which is more likely to cause insulation breakdown. At the same time, the damage to the internal wire insulation layer means that the core wire is exposed, and there is a risk of leakage and short circuit. Even if the wire core is intact, the repair is difficult, and each wire needs to be treated separately before the cable is repaired as a whole. In addition, the use environment will also affect the severity of the defect. A humid environment will increase the risk of insulation breakdown, while frequent bending and vibration may cause defects to expand. Therefore, when the defect area is the same, star-shaped damage is more difficult to repair than circular damage, and the safety hazard is greater. Summary of the invention

[0003] The present invention provides an AI-based electrical circuit image analysis and comparison method, which mainly includes:

[0004] The damaged area of ​​the cable is identified and located by image processing methods, and the contour information of the damaged area is obtained by edge detection algorithm. The damaged shape is determined to be circular or star-shaped according to the contour shape characteristics.

[0005] The damaged area is segmented by using the image segmentation algorithm, and the size of the damaged area is obtained by calculating the pixel area of ​​the segmented area;

[0006] According to the damage shape and area information, combined with the preset judgment conditions, the severity of the cable damage is determined, including if the damage area is smaller than the preset damage area threshold and the shape is regular, it is determined that the outer sheath is damaged; if the damage area is larger than the preset damage area threshold and the shape is irregular, it is determined that the internal wire insulation layer is damaged;

[0007] If the outer sheath is determined to be damaged, the integrity of the internal wire insulation layer is detected through an image analysis algorithm. If the internal insulation layer is intact, the humidity information of the cable environment is obtained. If the humidity is lower than the preset humidity threshold, local repair is performed;

[0008] If it is determined that the insulation layer of the internal wire is damaged, the damaged area is enlarged and clarified through the image enhancement algorithm, the number and location of the damaged wires are analyzed, and whether the copper core is intact is determined;

[0009] After completing the insulation wrapping of a single wire, use automated equipment to wrap the entire damaged area with multiple layers of PVC tape. The number of layers to be wrapped is determined based on the importance of the cable and environmental conditions. The number of layers to be wrapped is at least two.

[0010] For each damaged area detected, its area and length dimension features are extracted, combined with the model and material information of the wire, and by querying the pre-established safety threshold database, it is determined whether the size of the damage exceeds the safety range, and the severity of the damage is preliminarily assessed;

[0011] The shape, size and depth information of the damage are comprehensively compared and matched with the pre-established defect pattern library. The specific type of damage, including wear, scratches and fractures, is identified through similarity comparison. The impact of the damage on the safety performance of the wire is quantitatively evaluated based on the risk level corresponding to the damage type.

[0012] The present invention provides an AI-based electrical circuit image analysis and comparison system, which mainly includes:

[0013] Image preprocessing module, used to reduce noise and enhance the cable image to improve image quality;

[0014] Edge detection and shape recognition module, used to identify the outline and shape of cable damage and determine the type of damage;

[0015] Image segmentation and area calculation module, used to segment the damaged area and calculate its area to assess the degree of damage;

[0016] Damage type and risk assessment module, used to assess safety risks based on damage characteristics and database matching;

[0017] Internal structure inspection module, used to detect whether the internal insulation layer is damaged to ensure the integrity of the wire;

[0018] The repair and protective measures module is used to implement specific repair work and protective measures.

[0019] The technical solution provided by the embodiment of the present invention may have the following beneficial effects:

[0020] The present invention discloses an AI-based electric line image analysis and comparison method, which comprehensively considers factors such as the damage shape, distribution location, internal wire insulation integrity and use environment to formulate a reasonable repair plan. The contour information of the cable damage area is obtained by the edge detection algorithm, and the contour shape feature is used to determine whether the damage is circular or star-shaped. This shape judgment is very critical for subsequent damage classification. According to the shape and area of ​​the damage, the degree of damage to the cable is determined in combination with preset judgment conditions. If the damage area is less than the preset threshold and the shape is regular, it is usually determined that the outer sheath is damaged; if the damage area is large and the shape is irregular, it may be that the internal wire insulation layer is damaged. For the case of damage to the outer sheath, the integrity of the internal wire insulation layer is further detected by the image analysis algorithm. If the internal insulation layer is not damaged, it is also necessary to obtain environmental humidity information. If the humidity is lower than the threshold, local repair can be performed. When it is determined that the internal wire insulation layer is damaged, the damaged area is enlarged and clarified by the image enhancement algorithm, the number and position of the damaged wires are analyzed, and whether the copper core is intact is determined. Next, use automated equipment to wrap the damaged area with multiple layers of PVC tape, at least two layers, to ensure the reliability and durability of the repair. In addition, the area and length dimensional features of each damaged area detected are extracted, combined with the model and material information of the wire, and the pre-established safety threshold database is used to determine whether the damage size exceeds the safety range, and the severity of the damage is preliminarily assessed. Finally, by matching and comparing the similarity with the defect pattern library, the specific type of damage such as wear, scratches or breaks is identified, and its impact on the safety performance of the wire is quantitatively evaluated according to the risk level of the damage type.

[0021] In general, the present invention identifies, evaluates and repairs cable damage, greatly improves the efficiency and reliability of cable maintenance, and ensures the stable operation of the power system. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 The present invention is a flowchart of an AI-based electrical circuit image analysis and comparison method.

[0023] Figure 2 It is a schematic diagram of an AI-based electrical circuit image analysis and comparison method and system of the present invention.

[0024] Figure 3 This is another schematic diagram of an AI-based electrical circuit image analysis and comparison method and system of the present invention. DETAILED DESCRIPTION

[0025] The following will describe the technical solutions in the embodiments of the present invention in detail in conjunction with the accompanying drawings in the embodiments of the present invention. The described embodiments are only part of the embodiments of the present invention.

[0026] like Figure 1-3In this embodiment, an AI-based electrical circuit image analysis and comparison method and system may specifically include:

[0027] Step S101, identifying and locating the damaged area of ​​the cable through an image processing method, using an edge detection algorithm to obtain contour information of the damaged area, and judging whether the damaged shape is circular or star-shaped according to the contour shape features.

[0028] The damaged area of ​​the cable is identified and located by image processing methods, and the edge detection algorithm is used to obtain the contour information of the damaged area. According to the contour shape characteristics, it is determined whether the damaged shape is circular or star-shaped. If the damaged shape is circular, the Hough circle transform algorithm is used to locate the circular damaged area to obtain the center coordinates and radius parameter information; if the damaged shape is star-shaped, the skeleton of the star-shaped area is extracted through morphological operations to obtain the coordinate information of the skeleton pixel points. According to the center coordinates and radius information of the circular damaged area, the position and size of the damaged area are determined; according to the coordinate information of the skeleton pixel points of the star-shaped area, the distribution range and density of the damaged area are determined. If the radius of the circular damaged area is greater than the preset threshold, the area is marked as severely damaged; if the skeleton pixel density of the star-shaped area is greater than the preset threshold, the area is marked as multiple damages. For circular areas marked as severely damaged, the regional growing algorithm is used to expand the damaged area to obtain a more accurate damage boundary; for star-shaped areas marked as multiple damages, the skeleton pixels are grouped by the clustering analysis algorithm to determine the location of each damage point. The enlarged circular damaged area and the clustered star-shaped damaged area are merged to obtain a complete cable damaged area; the degree of cable damage is determined based on the proportion of the merged damaged area to the total area of ​​the cable. According to the degree of cable damage, combined with the cable material and service life attribute information, the decision tree algorithm is used to give suggestions for cable repair or replacement, and the location information of the damaged area is fed back to the maintenance personnel to guide them to perform targeted repair or replacement operations.

[0029] For example, in image processing technology, commonly used edge detection algorithms include the Canny algorithm and the Sobel algorithm. Taking the Canny algorithm as an example, it smoothes the image through Gaussian filtering, calculates the image gradient amplitude and direction, performs non-maximum suppression on the gradient amplitude, and then uses the double threshold method to detect and connect edges. Assuming that the cable image size is 500×500 pixels, the standard deviation of the Gaussian filter is 1.5, the low threshold is 50, and the high threshold is 100, edge detection can be achieved by calling the Canny(50,100,1.5) function. For the positioning of circular damaged areas, the Hough circle transform can be used. The algorithm maps each pixel point in the edge image to a conic surface in the parameter space, and then votes in the parameter space to find the center and radius of the circle. Assuming that the radius of the circular damaged area ranges from 20 to 50 pixels, the circular positioning can be achieved by calling the HoughCircles(50,100,20,50) function, where 50 and 100 are the low threshold and high threshold of the Canny algorithm, respectively. For the extraction of star-shaped damaged areas, the skeleton extraction algorithm in morphological operations can be used. This algorithm obtains the skeleton of the image by corroding and dilating the binary image until the image no longer changes. A 3×3 cross-shaped structural element can be used for skeleton extraction, and the number of iterations is 5. When judging the degree of damage, some quantitative indicators can be set. When the radius of the circular damaged area is greater than 30 pixels, it is considered to be severely damaged; when the skeleton pixel density of the star-shaped area is greater than 0.05, it is considered to be multiple damages. The density here refers to the ratio of the number of skeleton pixels to the number of pixels in the entire area. For severely damaged circular areas, the regional growth algorithm can be used to expand the damage range. Starting from the seed point, the algorithm continuously incorporates adjacent pixels into the area until it can no longer be expanded. The center of the circle can be selected as the seed point, and the growth standard is that the difference between the grayscale value and the seed point is less than 20. For star-shaped areas with multiple damages, the k-means clustering algorithm can be used to group the damaged points. The algorithm continuously iterates to assign pixels to the nearest cluster center and updates the location of the cluster center until the cluster center no longer changes. The number of clusters k can be set to 3, and the number of iterations is 10. When giving repair or replacement recommendations, the decision tree algorithm can be used. The algorithm generates a decision tree by recursively partitioning the data set. When the cable damage exceeds 30% and the service life exceeds 10 years, it is recommended to replace the cable; otherwise, it is recommended to repair. In the process of building the decision tree, information gain or Gini index can be used as a partitioning indicator, and the recursive depth is set to 3. When the location information of the damaged area is fed back to the maintenance personnel, it can be in the form of a rectangular bounding box or a polygonal bounding box. For a circular damaged area, its circumscribed rectangle can be used as the bounding box; for a star-shaped point area, its convex hull polygon can be used as the bounding box.The coordinate information of the bounding box can be transmitted in JSON format, such as {"type":"Rectangle","x":100,"y":200,"width":50,"height":80} to represent a rectangular bounding box.

[0030] Step S102, segmenting the damaged area using an image segmentation algorithm, and obtaining the size of the damaged area by calculating the pixel area of ​​the segmented area.

[0031] The damaged area is segmented by using an image segmentation algorithm based on region growth. If the segmentation result is over-segmented or under-segmented, the segmentation result is optimized by using morphological operations according to the shape characteristics of the damaged area. Small areas and holes are eliminated by erosion and expansion operations to obtain an accurate segmented area. According to the boundary coordinate information of the segmented area, the contour tracing algorithm is used to extract the contour of the segmented area, and a contour vector composed of a sequence of pixel point coordinates is obtained, and the perimeter and area geometric features of the contour are calculated. If the contour of the segmented area is discontinuous or has breakpoints, the contour is repaired by using a contour repair algorithm, and the coordinates of the missing contour points are estimated by interpolation or fitting methods to ensure the integrity and continuity of the contour. According to the repaired contour vector, the polygon approximation algorithm is used to simplify the contour. By recursively merging adjacent contour points, the polygon obtained is used to represent the approximate shape of the segmented area. The size of the segmented area, that is, the number of pixels in the damaged area, is obtained by calculating the pixel area of ​​the polygon. At the same time, according to the spatial resolution information of the image, the pixel area is converted into an area value under the actual physical size to judge the severity of the damaged area. If the damaged area exceeds the preset threshold, the segmented area is marked as severely damaged, and its location and size information is recorded; otherwise, the segmented area is marked as slightly damaged, and the next damaged area is processed until all damaged areas are identified and quantified, and a damage distribution map and area statistics report are obtained.

[0032] Exemplarily, the damaged area is segmented by using an image segmentation algorithm based on region growth. By setting seed points and growth criteria, adjacent pixels are included in the segmented area until it can no longer be expanded. When accurately segmenting the damaged area, a region growing algorithm can be used. First, a pixel point located inside the damaged area is selected as a seed point, and then a growth criterion is set, for example, the difference between the grayscale value of the adjacent pixel point and the grayscale value of the seed point is less than 20. Then, starting from the seed point, adjacent pixels that meet the growth criterion are continuously added to the segmented area until it can no longer be expanded. If the segmentation result is over-segmented, that is, the segmented area contains a large number of background pixels, morphological operations can be used to correct it. For example, the segmentation result is corroded, using a 3x3 square structure element, iterated 3 times, and small areas and holes in the segmented area can be removed. Then, the result after corrosion is expanded, and the boundary shape of the segmented area can be restored using the same structure element and number of iterations. When extracting the contour of the segmented area, the Moore neighborhood tracking algorithm can be used. The algorithm starts from the boundary pixel point of the segmented area and searches for the next boundary pixel point in a clockwise or counterclockwise direction until it returns to the starting point. During the search, the chain code of the 8-neighborhood is used to represent the direction of the contour, for example, 0 represents right, 1 represents upper right, 2 represents upward, and so on. Finally, a contour vector consisting of a chain code sequence is obtained, for example, [0,0,1,2,2,3,4,4,5,6,6,7]. According to the contour vector, the perimeter and area of ​​the contour can be calculated. The calculation formula of the perimeter is P = n + sqrt (2) * m, where n is the number of times 0, 2, 4, 6 appear in the chain code, and m is the number of times 1, 3, 5, 7 appear in the chain code. If the perimeter of an isosceles right triangle with a 45-degree angle is calculated, assuming that the length of the right angle side is n, the length of the hypotenuse is naturally n * sqrt (2). The calculation formula of the area is A = SB / 2, where S is the number of pixels in the segmented area and B is the perimeter of the contour. If the contour is discontinuous or has breakpoints, the Bezier curve fitting method can be used to repair it. First, segment the contour vector and find the breakpoints. Then, in each segment, select the starting point, end point and control point, and use the cubic Bezier curve formula to generate interpolation points to connect the broken contours. The cubic Bezier curve formula is P(t) = (1-t)^3P0+3t*(1-t)^2P1+3t^2*(1-t)P2+t^3P3, where P0, P1, P2, P3 are the starting point, the first control point, the second control point and the end point respectively, and t is a parameter with a value range of [0,1]. By adjusting the position of the control point, the shape of the curve can be controlled to make it fit the original contour as much as possible. For the repaired contour, the Douglas-Peucker algorithm can be used for polygonal approximation.The algorithm recursively deletes points on the contour to obtain an approximate contour composed of fewer points. The specific steps can be to first connect the starting point and the end point of the contour into a straight line, and calculate the distance from the remaining points to the straight line; then find the point with the largest distance. If the distance is greater than a given threshold, such as 2 pixels, then use the point as a new segmentation point and recursively perform the same operation on the two sub-segments; then if the distance from all points to the straight line is less than the threshold, use the starting point and the end point as the vertices of the polygon and end the recursion. Finally, a polygon composed of fewer vertices is obtained, which approximately represents the shape of the original contour. According to the vertex coordinates of the polygon, its area can be calculated as the area of ​​the damaged area. The formula for calculating the area of ​​the polygon is A=0.5*, where (x1,y1), (x2,y2),..., (xn,yn) are the vertex coordinates of the polygon, arranged in a clockwise or counterclockwise direction. Assuming that the spatial resolution of the image is 0.1 mm / pixel, the physical size of the damaged area is A*0.1^2 square millimeters. If the area is larger than 100 square millimeters, it is marked as severe damage, otherwise it is marked as light damage. After segmenting, quantifying and classifying all damaged areas, a damage distribution map can be obtained, and the number, total area and proportion of various types of damage can be counted to generate a damage analysis report.

[0033] Step S103, based on the damage shape and area information, combined with the preset judgment conditions, determine the severity of the cable damage, including if the damage area is smaller than the preset damage area threshold and the shape is regular, it is determined that the outer sheath is damaged; if the damage area is larger than the preset damage area threshold and the shape is irregular, it is determined that the internal wire insulation layer is damaged.

[0034] The shape characteristics of the damaged area are analyzed using the Hu invariant moment method. By comparing the Hu invariant moment of the damaged area with the preset threshold, it is determined whether its shape is regular. At the same time, it is determined whether the area of ​​the damaged area is greater than the preset threshold. If the damaged area is less than the preset damage area threshold and the shape is regular, it is determined that the outer sheath is damaged. If the damaged area is greater than the preset damage area threshold and the shape is irregular, it is determined that the internal wire insulation layer is damaged. A rule-based reasoning method is used to make a preliminary assessment of the severity of the damage. The shape characteristics and area information of the damaged area are input into the rule base, and the preliminary judgment result of the damage type is obtained by forward reasoning or backward reasoning. Combined with the hazard level and environmental factors, the severity score of each damaged area is calculated, and it is divided into corresponding severity levels according to the severity score. Maintenance personnel quickly formulate maintenance plans to reduce the risk of accidents.

[0035] Exemplarily, when obtaining the shape features of the damaged area, the Hu invariant moment method can be used. This method calculates the second-order central moment of the image to obtain a set of eigenvalues ​​that are independent of image rotation, translation, and scaling. For a 256x256 binary image, its Hu invariant moment can be expressed as a 7-dimensional vector [h1, h2, h3, h4, h5, h6, h7], where h1 reflects the ductility of the image, h2 reflects the symmetry of the image, and h3-h7 reflects other high-order features of the image. By comparing the Hu invariant moment of the damaged area with a preset threshold, it can be determined whether its shape is regular. When h1>0.3 and h2<0.1, it can be considered that the damaged shape is relatively regular. When calculating the damaged area, the pixel counting method can be used. For a binary image composed of 0-1 pixels, the damaged area is the number of pixels with a pixel value of 1. Assuming that the spatial resolution of the image is 0.1 mm / pixel, the physical size of the damaged area is A = N*0.1^2 mm2, where N is the number of pixels. When A is less than 100 mm2, it can be judged as a small area of ​​damage. When making a preliminary assessment of the severity of damage, a rule-based reasoning method can be used. Based on expert experience, a rule base containing three variables: shape regularity, area size, and damage type is constructed. A rule can be expressed as IF the shape is regular AND the area is small THEN the damage type is outer sheath damage. Then, the shape features and area information of the damaged area are input into the rule base, and the preliminary judgment result of the damage type is obtained by forward reasoning or backward reasoning. If the confidence of the reasoning result is low, it is necessary to further extract the texture features of the damaged area. Commonly used texture features include gray level co-occurrence matrix (GLCM) and local binary pattern (LBP). Taking GLCM as an example, the pixel association probability matrix of the image in different directions (0°, 45°, 90°, 135°) and distances (1, 2, 3...) can be calculated to obtain a set of eigenvalues ​​reflecting texture roughness, contrast, correlation and other attributes. Then, using classifiers such as K nearest neighbor (KNN) or support vector machine (SVM), the texture features are matched with samples of known damage types to obtain the confidence of the damage type. For damage types with a confidence exceeding the threshold, such as 0.8, it can be directly used as the final judgment result. For damage types with lower confidence, it is necessary to combine other factors for comprehensive analysis. According to the material of the cable, such as copper core, aluminum core and use, such as high voltage, low voltage, the danger level of different damage types is queried from the expert knowledge base to obtain a 5-level score, 1-5 points, the higher the score, the greater the danger. Then, according to the environment of the damaged area, such as temperature, humidity, vibration, etc., the weighting coefficients of different damage types are queried from the environmental factor library (between 0-1, the larger the value, the greater the impact of the environment on this type of damage).Finally, the weighted coefficients of the hazard level and environmental factors are multiplied to obtain a comprehensive score, and the score is divided into different severity levels according to the score size, such as 1-2 points for mild damage, 2-3 points for moderate damage, and 3-5 points for severe damage. When generating a cable damage diagnosis report, the severity scores of each damaged area can be summarized by weighted average. Let the area of ​​the i-th damaged area be Ai, and the severity score be Si, then the comprehensive damage severity score of the entire cable is S = (A1S1 + A2S2 + ... + An*Sn) / (A1 + A2 + ... + An), where n is the number of damaged areas. According to the value of the comprehensive score, corresponding repair or replacement suggestions can be given. For example, when S < 2, it is recommended to observe regularly and repair locally; when 2 < = S < 3, it is recommended to remove and rewrap the damaged part as soon as possible; when S > = 3, it is recommended to cut and replace the severely damaged part, and check and preventive maintenance of the entire cable. These suggestions can help operation and maintenance personnel quickly develop maintenance plans and reduce the risk of accidents.

[0036] Step S104, if it is determined that the outer sheath is damaged, the integrity of the internal wire insulation layer is detected through an image analysis algorithm. If the internal insulation layer is intact, the humidity information of the cable environment is obtained. If the humidity is lower than the preset humidity threshold, local repair is performed.

[0037] Determine whether the damaged area of ​​the cable is the outer sheath damage. If the shape and size of the damaged area meet the characteristics of the outer sheath damage, it is preliminarily determined to be the outer sheath damage. Otherwise, determine whether it is other types of damage. For the damaged area of ​​the outer sheath, use X-ray imaging or ultrasonic scanning detection method to obtain image data of the internal wire insulation layer; according to the material properties and defect characteristics of the internal wire insulation layer, select the corresponding image segmentation algorithm. The image segmentation algorithm includes threshold segmentation and edge detection, extract the area of ​​the insulation layer, and calculate the integrity index of the insulation layer, including connectivity and circularity geometric parameters. Compare the integrity index with the preset threshold to determine whether the internal insulation layer is intact. If the index meets the threshold condition, the insulation layer is considered to be intact, otherwise the insulation layer is considered to be damaged or defective. If the internal insulation layer is intact, the humidity value of the cable environment is obtained through the built-in temperature and humidity sensor of the cable or the external environment monitoring equipment, and it is compared with the humidity threshold. If the humidity value is lower than the threshold, the cable is judged to be in a dry environment and local repairs are performed using methods including electrical tape. The repaired cable is subjected to insulation resistance and voltage withstand tests to ensure that its insulation performance meets the standard requirements, and the damaged parts are inspected and maintained regularly. If the tape is found to be aging or falling off, it is replaced to ensure the safe operation of the cable.

[0038] For example, when judging whether the damaged area of ​​the cable is the damaged outer sheath, an image segmentation algorithm based on region growth can be used. First, a pixel point inside the damaged area is selected as the seed point, and then a growth criterion is set, such as the difference between the grayscale value of the adjacent pixel point and the grayscale value of the seed point is less than 20. Then, starting from the seed point, the adjacent pixel points that meet the growth criterion are continuously added to the segmented area until they can no longer be expanded. Finally, the geometric features of the segmented area, such as area, perimeter, circularity, etc., are calculated. Assuming that the typical area of ​​the damaged outer sheath is 50-100 square millimeters and the circularity is greater than 0.8, when the area and circularity of the segmented area meet these conditions, it can be preliminarily determined that the outer sheath is damaged. For areas suspected of damaged outer sheaths, X-ray imaging technology can be used to obtain images of the internal wire insulation layer. The principle of X-ray imaging is to use the attenuation characteristics of X-rays when penetrating the object being measured, and reconstruct the image of the internal structure of the object according to the received X-ray intensity distribution. In order to improve the image quality, a histogram equalization algorithm can be used to enhance the image to make its grayscale distribution more uniform. At the same time, the median filter algorithm is used to remove the salt and pepper noise in the image to reduce the impact of noise on subsequent analysis. When extracting the internal wire insulation layer area, the Otsu threshold segmentation algorithm can be used. This algorithm automatically determines an optimal threshold by maximizing the inter-class variance and divides the image into two parts: foreground (insulation layer) and background. Assume that the grayscale value range of the image is [0, L-1], the number of pixels is N, and the number of pixels with grayscale value i is ni, then the average grayscale value μ of the image is μ=∑(ini) / N. Assuming that the segmentation threshold is t, the average grayscale values ​​of the foreground and background are μ0=∑(ini) / (∑ni), i=0,1,...,t; μ1=∑(ini) / (∑ni), i=t+1,t+2,...,L-1. The inter-class variance is g = P0(μ0-μ)^2+P1*(μ1-μ)^2, where P0 = ∑ni / N,i = 0,1,...,t; P1 = ∑ni / N,i = t+1,t+2,...,L-1. The optimal threshold t* maximizes the inter-class variance g. P0 and P1 are the proportions of foreground and background pixels to the total pixels, μ0 and μ1 are the average grayscale values ​​of the foreground and background, and μ is the average grayscale value of the entire image. The larger the inter-class variance g, the more obvious the distinction between the foreground and background. Based on the segmentation results, the integrity indicators of the insulating layer area, such as connectivity and circularity, can be calculated. Connectivity reflects the degree of connection between the pixels inside the region, which can be measured by the ratio of the number of pixels inside the region to the area of ​​the region. Circularity reflects the closeness of the shape of the region to the circle, which can be measured by the ratio of the area of ​​the region to the area of ​​its circumscribed circle. Assuming that the connectivity of the complete insulation layer is greater than 0.9 and the circularity is greater than 0.85, when the segmented area meets these indicators, the internal insulation layer can be considered to be complete. If the internal insulation layer is intact, it is necessary to obtain the humidity information of the cable environment.Humidity information can be collected in real time through the temperature and humidity sensor built into the cable, or it can be queried from the database of the environmental monitoring system. If the humidity value is lower than the threshold, it is judged that the cable is in a dry environment, and a method including electrical tape is used for local repair. The specific method is to wrap the tape around the damaged part of the outer sheath and heat it with a hot air gun to make it tightly bonded. Assuming that the cable has a built-in SHT31 temperature and humidity sensor, the sensor data is read through the I2C bus, and the humidity value obtained is 45%. According to the instructions for use of electrical tape, when the ambient humidity is lower than 50%, it can be used to repair the insulating sheath. Therefore, it can be judged that the cable is in a dry environment and is suitable for repair with electrical tape. When repairing, first wipe the damaged part with a clean cloth or cotton paper, then cut the electrical tape of appropriate length, wrap it around the damaged part, and press it with your hands to make it tightly bonded. When winding, it should be noted that each circle of tape should have a certain overlap, and the general overlap width is 1 / 3 to 1 / 2 of the tape width. After the winding is completed, use a hot air gun to heat the tape so that it fully melts and penetrates into the gap of the sheath to form a stronger bond. The heating temperature is controlled at 150-200 degrees Celsius for 30-60 seconds. After the repair is completed, the cable should be tested for insulation performance to verify the repair effect. Common test methods include insulation resistance test and withstand voltage test. The insulation resistance test is to measure the resistance value between the cable conductor and the outer sheath with an ohmmeter under a specified voltage to determine whether the insulation performance meets the standard. The withstand voltage test is to apply voltage for a period of time under conditions higher than the rated voltage to observe whether the insulation layer breaks down. The test voltage and time are selected according to the rated voltage level and material properties of the cable, generally 2-3 times the rated voltage, and the duration is 1-5 minutes. If the test results are qualified, it means that the insulation performance of the repaired cable meets the requirements and can be put back into use.

[0039] Step S105, if it is determined that the internal wire insulation layer is damaged, the damaged area is enlarged and clarified through an image enhancement algorithm, the number and position of the damaged wires are analyzed, and it is determined whether the copper core is intact.

[0040] The image data of the internal wire insulation layer is obtained. If the image resolution is low or the contrast is poor, the image interpolation algorithm and histogram equalization algorithm are used to preprocess the image to improve the image quality. According to the density and attenuation coefficient differences of the internal wire insulation layer materials, the region of interest is extracted by the threshold segmentation algorithm to obtain the binary image of the suspected insulation layer damage area, and the small area noise area is removed by morphological filtering operation. The geometric feature analysis of the suspected damaged area is performed to calculate the area, aspect ratio, and circularity parameters of the area, and match them with the preset insulation layer damage feature model. If the matching degree exceeds the threshold, the area is determined to be an insulation layer damage area. The image pyramid algorithm is used to decompose the damaged area at multiple scales, select the corresponding scale layer to enlarge the damaged area, and use the Gaussian filter to smooth the enlarged image to remove the noise and artifacts generated during the enlargement process. Perform edge detection and contour extraction on the magnified and smoothed damaged area image to obtain the boundary of the damaged area. According to the shape characteristics of the boundary, determine whether the damage type is a crack, break or hole; segment the wires in the damaged area using the watershed algorithm or the regional growing algorithm, mark the position and number of each wire, and calculate the number of damaged wires, and save the segmentation results in the form of a mask image. Perform grayscale statistics and texture analysis on each segmented wire, calculate its average grayscale value, standard deviation, and gradient direction histogram features, and determine the integrity of the wire. If the feature value is abnormal, it is considered that the wire is broken or damaged, and the condition of its copper core needs to be checked.

[0041] For example, after obtaining the image of the internal wire insulation layer, if the image resolution is low, the bicubic interpolation algorithm can be used to enlarge the image. The algorithm obtains the enlarged image by performing cubic interpolation on the image in the horizontal and vertical directions. Assuming that the pixel coordinates of the original image are (x, y) and the magnification is n, the pixel coordinates of the enlarged image are (nx, ny). The interpolation formula is f(x, y) = ∑∑a_ijx^iy^j, where i, j = 0, 1, 2, 3, and a_ij is the interpolation coefficient, which can be obtained by solving the grayscale values ​​of 16 adjacent pixels in the original image. If the original image resolution is 500x500, after being enlarged by 2 times, the resolution is increased to 1000x1000, and the image details are clearer. For images with low contrast, the histogram equalization algorithm can be used for enhancement. The algorithm stretches the grayscale distribution range of the image to make the histogram of the image flat, thereby improving the contrast of the image. Suppose the grayscale value range of the image is [0, L-1], the number of pixels is N, and the number of pixels with grayscale value i is n_i. Then the probability of the occurrence of a pixel with grayscale value i is p_i = n_i / N, and the cumulative distribution function is cdf_i = ∑p_j, where j = 0, 1, ..., i. p_j is the probability of the occurrence of a pixel with grayscale value j. The grayscale value after equalization is h_i = round((L-1)*cdf_i). Where L is the total number of grayscale levels, for example, for an 8-bit grayscale image, L = 256. The round function is used to round the result to the nearest integer, because the grayscale level must be an integer. For an image with a grayscale range of [0, 255], if the distribution of the equalized histogram at each grayscale level is more uniform, the contrast of the image is enhanced. When extracting suspected insulation layer damage areas, the maximum entropy threshold segmentation algorithm can be used. The algorithm maximizes the sum of the entropy values ​​of the foreground and background, and adaptively determines an optimal threshold to binarize the image. Assuming the probability distribution of the foreground and background are p0 and p1, respectively, and the number of pixels are n0 and n1, respectively, the total entropy H = ∑p0_ilog(p0_i) + ∑p1_jlog(p1_j), where p0_i = n0_i / n0, p1_j = n1_j / n1, n0_i and n1_j are the number of pixels with gray values ​​i and j in the foreground and background, respectively. The optimal threshold maximizes the total entropy H. If the insulating layer material is polyvinyl chloride, its density is 1.4g / cm^3, and the density of other parts is less than 1g / cm^3, then a suitable threshold can be selected to segment the insulating layer area. After segmenting the suspected damaged area, the small area of ​​noise area can be removed by morphological operations. Commonly used morphological operations include corrosion and dilation. The erosion operation uses a structural element to convolve with the image and takes the minimum value in the area covered by the structural element as the output pixel value, which can remove the noise area smaller than the structural element.The dilation operation takes the maximum value in the area covered by the structural element as the output pixel value, which can fill the small holes in the target area. Using a 3x3 cross-shaped structural element to erode the image can remove areas with an area of ​​less than 9 pixels, and then performing the dilation operation can fill holes with an area of ​​less than 9 pixels. When judging whether the suspected damaged area is a real insulation layer damage, the geometric feature parameters such as the circularity and rectangularity of the area can be calculated and compared with the preset threshold. The circularity C represents the similarity between the area and its circumscribed circle. The calculation formula is C = 4πS / L^2, where S is the area of ​​the area and L is the perimeter of the area. The rectangularity R represents the degree of fit between the area and its minimum circumscribed rectangle. The calculation formula is: R = S / (w*h), where S represents the area of ​​the area, and w and h are the width and height of the circumscribed rectangle, respectively. For a complete circular insulation layer, its circularity should be close to 1, while the circularity of the damaged area will be significantly smaller. If the circularity is less than 0.8 and the rectangularity is greater than 1.5, the area can be determined as an insulation layer damage. For the damaged areas, the image pyramid algorithm can be used for multi-scale analysis. The image pyramid consists of a series of images of different resolutions, and each layer is the downsampling result of the previous layer. The generation process of the Gaussian pyramid is G_i=Downsample(G_{i-1}), where G_i is the i-th layer pyramid image, and Downsample is a downsampling operation, usually using mean filtering or Gaussian filtering. The Laplacian pyramid can be obtained by subtracting two adjacent layers of the Gaussian pyramid to obtain L_i=G_i-Upsample(G_{i+1}), where Upsample is an upsampling operation, usually using nearest neighbor interpolation or bilinear interpolation. By enlarging a layer of the Laplacian pyramid and fusing it with the original image, an enlarged local area image can be obtained. If the third layer of the Laplacian pyramid is selected, enlarged by 2 times, and then denoised by Gaussian filtering, an enlarged and smooth damaged area image can be obtained. When extracting the boundary of the enlarged damaged area, the Canny algorithm can be used for edge detection first, and then the contour tracking algorithm can be used to extract the boundary. The Canny algorithm uses Gaussian filtering, gradient calculation, non-maximum suppression, double threshold detection and other steps to obtain a binary edge image. The contour tracing algorithm selects a starting point from the edge image, and then moves along the edge pixel points until it returns to the starting point to obtain a closed contour curve. According to the shape characteristics of the contour, such as area, perimeter, convexity, etc., the type of damage can be determined. If the contour is a thin strip with sharp ends, it may be a crack; if the contour is an irregular polygon with jagged edges, it may be a hole. When segmenting the wires in the damaged area, the watershed algorithm can be used. The algorithm regards the image as a terrain surface, where pixels with high grayscale values ​​correspond to peaks and pixels with low grayscale values ​​correspond to valleys. Then, starting from the local minimum, it floods upward until the water in different valleys converges to form a watershed.The watershed line divides the image into multiple areas, each corresponding to a wire. First, the gradient of the image is calculated, and the gradient value is used as the terrain height. Then, the pixel with the smallest gradient value is flooded to finally obtain the segmentation result. In order to avoid over-segmentation, the gradient image can be morphologically reconstructed to remove tiny local minima. Each segmented wire can be represented by a mask image. The area with a pixel value of 1 corresponds to the wire, and the area with a pixel value of 0 corresponds to the background. The grayscale and texture features of each segmented wire can be extracted to determine its integrity. The average grayscale value of the wire reflects the density of its material. Generally, the grayscale value of a complete wire is higher, while the grayscale value of a damaged or oxidized wire is lower. The grayscale standard deviation of the wire reflects the uniformity of its surface. The surface of a complete wire is smooth, the grayscale distribution is uniform, and the standard deviation is small. The surface of a damaged wire is rough, the grayscale distribution is uneven, and the standard deviation is large. The gradient direction histogram of the wire reflects the directionality of its surface texture. The surface texture of a complete wire is consistent, and the gradient direction distribution is concentrated, while the surface texture of a damaged wire is messy and the gradient direction distribution is dispersed. If the average grayscale value of a wire is 20% lower than that of other wires, the standard deviation is 50% larger, and the peak value of the gradient direction histogram is less than 0.5, the wire may be damaged and further inspection is needed to see if the copper core inside is damaged.

[0042] Step S106, after completing the insulation wrapping of a single wire, use automated equipment to wrap the entire damaged area with multiple layers of PVC tape. The number of layers to be wrapped is determined based on the importance of the cable and environmental conditions, and the number of layers to be wrapped is at least two.

[0043] Identify the location and number of each damaged wire, use template matching or feature extraction methods to determine the diameter and length geometric parameters of the wire, and obtain the three-dimensional spatial coordinate information of the wire. According to the spatial coordinates and geometric parameters of the wire, control the movement trajectory and speed of the automation equipment so that the insulation wrapping material can be put on each wire, and the insulation layer material is tightly fitted to the surface of the wire by heating or UV curing. Perform eddy current or ultrasonic testing on the single wire after insulation wrapping to evaluate the integrity and uniformity of the insulation layer. If the test results show that there are bubbles, impurities or uneven thickness defects in the insulation layer, control the automation equipment to rework and redo it until it is qualified. Bundle the single wires that have completed insulation wrapping together according to the original arrangement, use the image stitching algorithm to reconstruct the three-dimensional structural model of the entire damaged area, and compare it with the design drawings of the cable to verify the specifications and models of the wire bundle. Select the tape model and specifications that match the cable model from the PVC tape material database. According to the importance of the cable and environmental conditions, the environmental adjustment includes temperature, humidity, and vibration, calculate the optimal number of layers and winding angles of the tape winding, and generate a winding process parameter table. Place the wire bundle on the work surface of the automated winding equipment, and use visual guidance and servo control technology to control the tape reel to spiral along the surface of the wire bundle, and adjust the tightness and angle of the tape in real time to ensure that there are no wrinkles, overlaps, or gaps between the tape and the surface of the wire bundle during the winding process. After the winding is completed, spray a polymer waterproof coating on the surface of the tape, and apply sealant to the tape joints to improve the waterproof and insulation properties of the cable. The electrical parameters of the cable are collected through online monitoring equipment, including insulation resistance and leakage current, to determine whether the repaired cable meets the requirements for safe operation.

[0044] For example, after identifying the location and number of the damaged wire, the Hough transform algorithm can be used to extract the central axis of the wire. The algorithm finds the most likely straight line parameters, that is, the direction and position of the wire, by voting for the edge points in the image. For a wire with a diameter of 5 mm and a length of 100 mm, its central axis can be represented by the starting point coordinates (x0, y0) and the end point coordinates (x1, y1). These two coordinate points can be obtained through Hough transform to determine the position of the wire in space. According to the starting and end point coordinates of the wire, the motion trajectory of the automation equipment can be controlled to move along the axis direction of the wire. At the same time, it is also necessary to control the feeding speed and tension of the insulating wrapping material to ensure the uniformity of the insulating layer. For a wire with a diameter of 5 mm, a heat shrink tube with an inner diameter of 4.5 mm and a wall thickness of 0.5 mm can be selected as the insulating wrapping material. The feeding speed v of the heat shrink tube can be calculated based on the length L of the wire and the wrapping time t, that is, v=L / t. The tension F can be calculated based on the elastic modulus E, cross-sectional area A and elongation ε of the heat shrink tube, that is, F=EAε. By controlling the feeding speed and tension, the heat shrink tube can be uniformly shrunk on the surface of the wire to form a dense insulation layer. When performing eddy current detection on the wire after insulation wrapping, pulse eddy current technology can be used. This technology generates eddy currents inside the insulation layer by sending a short high-frequency pulse current to the surface of the wire, and detects the time constant of eddy current decay. If there are defects in the insulation layer, such as bubbles or conductive foreign matter, it will affect the propagation path and decay speed of the eddy current. For a wire, a square wave pulse with a frequency of 1MHz and a pulse width of 10μs is applied to its surface, and the eddy current signal is measured at a distance of 2mm from the surface of the wire. If the insulation layer is intact, the peak voltage of the eddy current signal is 2V and the decay time constant is 50μs; if there are pores of φ0.5mm in the insulation layer, the peak voltage of the eddy current signal is reduced to 1.5V, and the decay time constant is shortened to 30μs. By setting the qualified threshold of the eddy current signal, the quality of the insulation layer can be automatically judged, and rework or qualified release can be controlled. When reconstructing the three-dimensional structural model of the damaged area of ​​the cable, computed tomography (CT) technology can be used. CT technology obtains a series of two-dimensional cross-sectional images by performing X-ray scanning on the cross section of the cable, and then uses an image reconstruction algorithm to stitch these images into a three-dimensional model. For a cable with a length of 1m, it can be scanned every 10mm, and a total of 100 cross-sectional images can be obtained. The resolution of each image is 500×500 pixels and the grayscale is 256 levels. These images can be reconstructed into a triangular patch mesh model through the marchingcubes algorithm, and then the quality of the model is optimized through mesh simplification and smoothing algorithms, and finally a realistic three-dimensional model is obtained. By comparing the model with the CAD design drawings of the cable, the differences and defects in the internal structure of the cable can be found, providing a basis for subsequent repairs.When choosing the model and specification of PVC tape, it is necessary to consider the cable's rated voltage, current, temperature and other parameters. For a 10kV high-voltage cable, the conductor cross-sectional area is 120mm. 2, the maximum operating temperature is 90℃, then you can choose PVC tape with model PV100, thickness of 0.2mm and maximum operating temperature of 105℃. According to the number and cross-sectional area of ​​the cable core wires, the length required for winding the tape can be calculated. For a 10-core cable, the diameter of each core wire is 10mm, and the damaged length is 100mm. If the tape width is 50mm and the overlap rate is 50%, the length required for winding 1 layer of tape is L=nπd / (w50%), where n is the number of core wires, d is the core wire diameter, and w is the tape width. Substituting the data, we can get L=10π10 / (50*50%)=12.56mm. If 3 layers of tape need to be wound, the total length is 37.68mm. When winding the tape, the tension and angle of the tape need to be controlled. Too much tension will cause the tape to deform or tear, and too little tension will cause the tape to loosen or fall off. Too much angle will cause the tape to overlap, and too little angle will cause the tape to gap. In general, the tension of the tape should be controlled at 1-2N, and the angle should be controlled at 45-60°. For a roll of tape with an inner diameter of 76mm and an outer diameter of 150mm, its initial tension is 1.5N, the winding tension is 2N, the winding angle is 55°, and the winding speed is 10m / min. During the winding process, the tension and angle of the tape are measured in real time by the tension sensor and the angle sensor, and the speed and tension of the motor are adjusted by the PID control algorithm to keep the tape in the best state. The PID control algorithm can be expressed as u(t)=Kpe(t)+Ki∫e(t)dt+Kd*de(t) / dt, where u(t) is the control quantity, e(t) is the deviation, Kp, Ki, and Kd are the proportional, integral, and differential coefficients, respectively. By setting appropriate PID parameters, precise control of the tension and angle of the tape can be achieved. When monitoring the insulation performance of the repaired cable, online partial discharge detection technology can be used. Partial discharge refers to the local discharge phenomenon inside or on the surface of the insulating medium, which will generate a high-frequency pulse signal. By detecting and analyzing the amplitude, phase, frequency and other parameters of the partial discharge signal, the aging degree and defect location of the insulating medium can be evaluated. For a 10kV high-voltage cable, an ultra-high frequency sensor is coupled on its surface, with a frequency range of 300MHz-1500MHz and a sensitivity of 1pC. When there are defects such as air gaps or impurities inside the cable, a local high electric field will be generated under the action of the power frequency AC voltage, causing electron avalanche and corona discharge. These discharge pulse signals are collected and amplified by the sensor, and then through spectrum analysis and pattern recognition, the three-dimensional spatial distribution map and defect type of partial discharge can be obtained. If a partial discharge signal with an amplitude of 100pC, a phase of 45°, and a frequency of 500MHz is detected at a certain position, there may be a 0.1mm wide air gap defect. According to the degree of danger of the partial discharge signal, it can be judged whether the cable needs to be replaced or continued to be used. Through automated equipment and online monitoring technology, intelligent repair and quality control of cable damage can be achieved.This not only improves the reliability and safety of the cable, but also extends the service life of the cable and reduces the cost of maintenance and replacement.

[0045] Step S107, for each damaged area detected, extract its area and length size features, combine the model and material information of the wire, and query the pre-established safety threshold database to determine whether the size of the damage exceeds the safety range, and preliminarily evaluate the severity of the damage.

[0046] The image segmentation algorithm is used to identify and locate the damaged area of ​​the wire. By setting the grayscale threshold or color threshold, the damaged area is distinguished from the background to obtain the binary mask image of the damaged area. The binary mask image of the damaged area is subjected to morphological processing, which includes corrosion, expansion, opening operation, and closing operation to eliminate noise and interference in the image and obtain an accurate and smooth damaged area contour. According to the pixel coordinates of the damaged area contour, its geometric features are calculated, including area, perimeter, length and width of the minimum circumscribed rectangle, to obtain quantitative indicators reflecting the damage size. The model and material information of the wire are extracted from the mark on the surface of the wire. The model and material information of the wire include conductor cross-sectional area, insulation material type, and rated voltage parameters. The extracted wire model and material information are used as keywords to query in the pre-established safety threshold database to obtain the corresponding damage size safety range; the geometric feature value of the damaged area is compared with the queried safety threshold to determine whether the damage size exceeds the safety range. If the damage size does not exceed the safety range, it is classified as general damage; if the damage size exceeds the safety range, it is classified as severe damage. Based on the number and proportion of general damage and severe damage, combined with the environment and usage factors of the wires, a preliminary assessment of the severity of the damage is given. The preliminary assessment results include mild, moderate, and severe levels of damage, providing a reference for subsequent maintenance and replacement decisions.

[0047] For example, when automatically identifying the damaged area of ​​the wire, the Otsu threshold segmentation algorithm can be used. The algorithm adaptively determines a global threshold by maximizing the inter-class variance between the foreground and the background, and divides the image into two parts: the damaged area and the non-damaged area. The calculation formula of the inter-class variance is: σ_b^2=ω_0*(μ_0-μ_T)^2+ω_1*(μ_1-μ_T)^2, where ω_0 and ω_1 are the proportion of pixels in the foreground and background respectively, μ_0 and μ_1 are the average grayscale values ​​of the foreground and background respectively, and μ_T is the average grayscale value of the entire image. By traversing all possible thresholds, the threshold with the largest inter-class variance is found as the optimal segmentation threshold. For an 8-bit grayscale image, the grayscale range is 0-255, and there are 256 candidate thresholds. Set the points with pixel grayscale values ​​less than the threshold to 0, and the points with pixel grayscale values ​​greater than the threshold to 255, and obtain a binary damaged area mask image. When performing morphological processing on the damaged area mask image obtained by segmentation, the structural element can be used for erosion and dilation operations. The erosion operation can remove the burrs and noise on the edge of the damaged area to make the contour smoother. The dilation operation can fill the small holes inside the damaged area to make the area more complete. Selecting a 3x3 cross-shaped structural element and performing 3 erosions and 2 dilations on the damaged area can effectively smooth the contour and eliminate noise. The morphological operation can be expressed in mathematical form as Erosion: AB = {z|(B)_z A}; Dilation Where A is the original image, B is the structural element, z is the coordinate of the center position of B, and (B)_z represents the result after B is translated by z units. When calculating the geometric features of the damaged area, the moment of the image can be used. The moment is the weighted sum of the powers of the pixel coordinates in the image, which can reflect the shape, size, direction and other characteristics of the image. The 0th order moment m00 represents the area of ​​the damaged area, the 1st order moments m10 and m01 represent the centroid coordinates (x, y) of the damaged area, and the 2nd order central moments μ20, μ02 and μ11 represent the length, width and rotation angle of the damaged area. These moments can be calculated by the following formula m_pq=∑_x∑_yx^py^qI(x,y), x=m10 / m00, y=m01 / m00, μ_pq=∑_x∑_y*(xx)^p*(yy)^q*I(x,y), where I(x,y) is the pixel value of the point (x, y) on the mask image of the damaged area, and p+q is the order of the moment. By calculating these moments, the geometric features of the damaged area, such as area, length, and width, can be obtained for subsequent size judgment. When identifying the model and material identification on the surface of the wire, a convolutional neural network (CNN) can be used. CNN is a deep learning model that can automatically extract local features of the image through the combination of convolutional layers and pooling layers, and classify them through fully connected layers. For a 640x480 resolution wire surface image, a CNN model containing 3 convolutional layers, 3 pooling layers, and 2 fully connected layers can be designed. The first convolutional layer contains 16 5x5 convolutions, the second convolutional layer contains 32 3x3 convolutions, and the third convolutional layer contains 64 3x3 convolutions. The window size of the pooling layer is 2x2, the step size is 2, and the number of neurons in the fully connected layer is 128 and 10 respectively. The input image is forward propagated through the CNN model to obtain a 10-dimensional feature vector, which is then converted into a probability distribution through the Softmax function, and the category with the largest probability is taken as the recognition result. The training process of the CNN model optimizes the weight parameters through the back propagation algorithm to minimize the cross entropy loss between the predicted results and the true labels. When querying the security threshold database, the Yaohui index and binary search algorithm can be used to improve the query efficiency. First, the wire model and material information in the database are Yaohui mapped and converted into a fixed-length Yaohui code as the key value of the data. Then, the Yaohui codes are arranged in order from small to large to establish an ordered index table. When it is necessary to query the security threshold corresponding to a certain wire model and material, first calculate its Yaohui code, then use the binary search algorithm to locate the position of the Yaohui code in the index table, and finally return the security threshold data corresponding to the position. The time complexity of the binary search algorithm is O(logN), where N is the number of records in the database. Assuming that the Yaohui code of the wire model and material is 64 bits and there are 10,000 records in the database, the time complexity of a query is log_2(10000)≈13 comparisons, which is much lower than the 10,000 comparisons of linear search.When comparing the damage size with the safety threshold, a proportional coefficient can be set to enlarge or reduce the safety threshold by a certain multiple to adapt to different usage scenarios. For wires in important occasions, the safety threshold can be multiplied by 0.8 to increase the safety margin; for wires in minor occasions, the safety threshold can be multiplied by 1.2 to reduce the damage sensitivity. If the area of ​​the damaged area exceeds 1.2 times the area of ​​the safety threshold, or the length of the damaged area exceeds 1.2 times the length of the safety threshold, or the width of the damaged area exceeds 1.2 times the width of the safety threshold, it is judged as severe damage; otherwise, it is judged as general damage. When giving a preliminary assessment of the severity of the damage, a weighted average method can be used to comprehensively consider multiple influencing factors. For example, the number of severe damage is recorded as x1, the number of general damage is recorded as x2, the humidity of the wire environment is recorded as x3, the temperature of the wire environment is recorded as x4, and the use time of the wire is recorded as x5. The damage severity y can be expressed as y=w1x1+w2x2+w3x3+w4x4+w5*x5, where w1 to w5 are the weight coefficients of severe damage, general damage, humidity of the wire environment, wire environment, and wire use time, reflecting their influence on the severity of damage, which can be determined by expert experience or data statistics. When the value of y is less than or equal to 20, it is assessed as a slight degree of damage; when the value of y is greater than 20 and less than or equal to 50, it is assessed as a moderate degree of damage; when the value of y is greater than 50, it is assessed as a severe degree of damage. The evaluation result can be expressed in percentage form, that is, y / 100, and a damage severity index between 0-100 is obtained. In this way, automatic detection, quantitative analysis and severity evaluation of wire damage can be achieved.

[0048] Step S108: The shape, size and depth information of the damage are integrated and matched with the pre-established defect pattern library. Through similarity comparison, the specific type of damage is identified, including wear, scratches and fractures. According to the risk level corresponding to the damage type, the impact of the damage on the safety performance of the wire is quantitatively evaluated.

[0049] The three-dimensional reconstruction technology is adopted, which includes structured light scanning or laser scanning, to obtain the three-dimensional point cloud data of the damaged area of ​​the wire. Through point cloud registration and surface fitting algorithm, a three-dimensional model of the damaged area is constructed to extract the shape, size and depth geometric information of the damage. According to the material properties and processing technology of the wire, the three-dimensional model is physically simulated and analyzed, the stress distribution, deformation field and strain field of the damaged area are calculated, and the mechanical indicators reflecting the severity of the damage are obtained. The mechanical indicators include stress concentration coefficient and plastic strain. Among them, the three-dimensional models and mechanical indicator ranges of various typical damage types are stored in the defect pattern library. The damage types include wear, scratches, fractures, and corrosion. Each damage type corresponds to a risk level, indicating its impact on the safety performance of the wire. The damage model to be identified is matched with the model in the defect pattern library, and the three-dimensional shape descriptor is used. The three-dimensional shape descriptor is a rotationally invariant surface distribution histogram or spherical harmonic function to measure the similarity between the two models and obtain a similarity coefficient between 0 and 1. Select several models with the highest similarity to the damage model to be identified, and perform cluster analysis on the mechanical indicators of the models. The cluster analysis includes K-means clustering or density clustering. According to the clustering results, the type of damage model is determined and its risk level is judged. If the similarity between the damage model and the three-dimensional models of multiple damage types reaches the preset value, the risk levels of multiple types are fused, and a weighted average or maximum value method is used to obtain a comprehensive risk assessment result, which is quantified as a risk index between 0 and 5. The identified damage type and risk index are compared with the preset threshold. If the risk index exceeds the warning value, the high-risk damage area is marked on the three-dimensional model, a damage detection report is generated, and an early warning signal is triggered to notify the maintenance personnel to handle it.

[0050] For example, when obtaining the three-dimensional point cloud data of the damaged area of ​​the wire, the structured light scanning technology can be used. This technology projects a series of light beams of a specific pattern onto the surface of the object, and calculates the three-dimensional coordinates of the surface according to the deformation of the pattern. A sinusoidal grating stripe is projected onto the surface of the wire, the period of the stripe is 0.5mm, and the phase difference between adjacent stripes is π / 2. The stripe image is collected by a camera, the image is Fourier transformed, the phase information of the stripe is extracted, and then the three-dimensional coordinates (x, y, z) of each pixel are calculated according to the principle of triangulation to obtain the point cloud data of the wire surface. Point cloud data is usually stored in PLY or OBJ format, containing a large amount of three-dimensional point coordinates and normal vector information. In order to construct a three-dimensional model of the damaged area, the point cloud data needs to be preprocessed and the surface reconstructed. The preprocessing steps include removing noise points, such as isolated points and outliers; homogenizing the point cloud distribution, such as downsampling or upsampling the point cloud; and smoothing the point cloud surface, such as Laplace smoothing or moving least squares smoothing of the point cloud. Common algorithms for surface reconstruction include Poisson surface reconstruction, Malkinger triangulation, Bessel surface fitting, etc. Taking Poisson reconstruction as an example, the algorithm solves the Poisson equation, constructs an implicit function on the point cloud, extracts its isosurface, and obtains a closed triangular mesh surface. The Poisson equation can be expressed as The Laplacian operator is usually expressed as Where u is an implicit function and V is the normal vector field of the point cloud. Solving the Poisson equation can be transformed into solving the sparse linear equation system Ax=b, where A is the Laplace operator matrix and b is the normal vector divergence. Iterative optimization algorithms such as the conjugate gradient method can be used to solve the linear equation system. When performing physical simulation analysis on a three-dimensional model, the finite element method can be used. The finite element method divides a continuous object into a finite number of discrete units, such as tetrahedrons or hexahedrons, establishes a local coordinate system and interpolation function on each unit, discretizes physical field quantities such as displacement, stress, and strain into a finite number of node degrees of freedom, and then lists the equilibrium equations, constitutive equations, and coordination equations of the unit according to the laws of physics. They are assembled into a set of algebraic equations for the entire object, and the response of the object under load can be obtained by solving the set of equations. Taking the stress analysis of the damaged area as an example, assuming that the three-dimensional model of the damaged area is composed of 10,000 tetrahedral units, each unit has 4 nodes, and each node has 3 displacement degrees of freedom, the number of degrees of freedom of the entire model is 120,000. The relationship between unit stress and node displacement can be expressed by the stiffness matrix Ke: Ke = ∫BTDBdV, where B is the strain-displacement matrix and D is the elastic matrix. The equilibrium equation of the unit is Kede = Fe, where de is the unit node displacement vector and Fe is the unit node force vector. The equilibrium equations of all units are assembled to obtain the equilibrium equation of the entire model Kd = F, where K is the total stiffness matrix, d is the total displacement vector, and F is the total force vector. Solve the equilibrium equation to obtain the node displacement, and then calculate the unit stress based on the stress-strain relationship. The stress concentration factor of the damaged area can be defined as stress concentration factor = maximum stress in the damaged area / far-field stress, where the far-field stress is the average stress outside the damaged area. In the defect pattern library, each damage type has a three-dimensional model template and a set of mechanical index ranges. For wear damage, the three-dimensional model features are smooth surface, no sharp edges, stress concentration coefficient is usually less than 2, and plastic strain is less than 1%; for scratch damage, the three-dimensional model features are elongated grooves on the surface, stress concentration coefficient is usually greater than 3, and plastic strain is less than 5%; for fracture damage, the three-dimensional model features are discontinuous cracks on the surface, stress concentration coefficient is usually greater than 10, and plastic strain is greater than 10%. These damage templates and index ranges can be obtained by mathematical statistics through a large number of damage cases and simulation experiments. When matching the damage model, a three-dimensional shape descriptor based on the surface distribution histogram can be used. This descriptor counts the normal vector distribution of the model surface as a histogram, and each bin of the histogram represents an interval of the normal vector direction. The direction space of the normal vector is divided into 8 intervals, each interval corresponds to a bin of the histogram, and the area proportion of the normal vector on the model surface falling in each interval is counted to obtain an 8-dimensional feature vector. In order to ensure the rotation invariance of the descriptor, the model coordinate system can be aligned with its main direction, and the main direction can be obtained by principal component analysis (PCA).The similarity between two models can be measured by Euclidean distance, that is, similarity = 1 / (1+√(∑(hi-hi')2)), where hi and hi' are the characteristic vectors of the surface distribution histogram of the two models. When clustering the mechanical indicators of the damage model, a density clustering algorithm, such as DBSCAN, can be used. This algorithm calculates the distance between data points, finds the area with a density greater than a certain threshold, divides it into a cluster, and marks the points that do not belong to any cluster as noise points. The DBSCAN algorithm has two important parameters ε and MinPts, ε represents the maximum distance between two data points, and MinPts represents the minimum number of points in a cluster. For mechanical indicator clustering, the stress concentration factor and plastic strain can be normalized to the range of 0-1, and then they are combined into a two-dimensional feature space. Setting ε = 0.1 and MinPts = 3, the damage model can be clustered into several clusters such as wear, scratches, and fractures. The center point of each cluster represents the typical mechanical indicator value of this type of damage. When integrating multiple types of risk assessment results, a fuzzy comprehensive evaluation method can be used. This method constructs a fuzzy relationship matrix between the evaluation factors and the comment level, quantifies the qualitative indicators into membership, and then uses the weighted average operator to synthesize the membership of multiple indicators to obtain a comprehensive evaluation result. Among them, the damage type and mechanical index are used as evaluation factors, and the risk level is divided into 5 levels: very low, low, medium, high, and very high. Assume that the weight of the damage type is 0.6 and the weight of the mechanical index is 0.4. If the similarity of a damage model with the wear type is 0.2, the similarity with the fracture type is 0.8, its stress concentration coefficient is 5, and the plastic strain is 3%, then the fuzzy evaluation vector A1=(0,0,0.2,0.8,0) of the damage type on the risk level and the fuzzy evaluation vector A2=(0,0,0.6,0.4,0) of the mechanical index on the risk level can be obtained. A1 and A2 are weighted averaged to obtain the comprehensive evaluation vector A = 0.6A1 + 0.4A2 = (0, 0, 0.36, 0.64, 0), and the level with the largest membership is taken as the final evaluation result, that is, the risk level of the damage model is "high". When determining the warning threshold, different thresholds can be set according to the safety level and risk tolerance of the wire. For UHV transmission lines, the safety level requirement is the highest, and the risk index threshold can be set to 1.5; for ordinary civil wires, the safety level requirement is relatively low, and the risk index threshold can be set to 3. When the risk index of the damage model exceeds the threshold, the high-risk area is automatically highlighted in red on the three-dimensional model, and a detection report containing information such as damage type, location, size, and risk level is output. At the same time, a warning signal is sent to the upper-level monitoring platform to notify the operation and maintenance personnel to deal with it in time to avoid the expansion of damage or accidents.

[0051] It should be noted that the above examples are only some specific embodiments of the present invention. Obviously, the present invention is not limited to the above embodiments, and there are many variations. All variations that can be directly derived or associated with the content disclosed by a person skilled in the art should be considered as the protection scope of the present invention.

Claims

1. An AI-based electrical circuit image analysis and comparison method, characterized in that: The method comprises: The damaged area of ​​the cable is identified and located by image processing methods, and the contour information of the damaged area is obtained by edge detection algorithm. The damaged shape is determined to be circular or star-shaped according to the contour shape characteristics. The damaged area is segmented using an image segmentation algorithm, and the size of the damaged area is obtained by calculating the pixel area of ​​the segmented area; According to the damage shape and area information, combined with the preset judgment conditions, the severity of the cable damage is determined, including if the damage area is smaller than the preset damage area threshold and the shape is regular, it is determined that the outer sheath is damaged; if the damage area is larger than the preset damage area threshold and the shape is irregular, it is determined that the internal wire insulation layer is damaged; If the outer sheath is determined to be damaged, the integrity of the internal wire insulation layer is detected through an image analysis algorithm. If the internal insulation layer is intact, the humidity information of the cable environment is obtained. If the humidity is lower than the preset humidity threshold, local repair is performed; If it is determined that the insulation layer of the internal wire is damaged, the damaged area is enlarged and clarified through the image enhancement algorithm, the number and location of the damaged wires are analyzed, and whether the copper core is intact is determined; After completing the insulation wrapping of a single wire, use automated equipment to wrap the entire damaged area with multiple layers of PVC tape. The number of layers to be wrapped is determined based on the importance of the cable and environmental conditions. The number of layers to be wrapped is at least two. For each damaged area detected, its area and length dimension features are extracted, combined with the model and material information of the wire, and by querying the pre-established safety threshold database, it is determined whether the size of the damage exceeds the safety range, and the severity of the damage is preliminarily assessed; The shape, size and depth information of the damage are comprehensively compared and matched with the pre-established defect pattern library. The specific type of damage, including wear, scratches and fractures, is identified through similarity comparison. The impact of the damage on the safety performance of the wire is quantitatively evaluated based on the risk level corresponding to the damage type.

2. The method according to claim 1, wherein: The method of identifying and locating the damaged area of ​​the cable by an image processing method, obtaining the contour information of the damaged area by an edge detection algorithm, and determining whether the damaged shape is circular or star-shaped according to the contour shape features include: If the damaged shape is circular, the Hough circle transform algorithm is used to locate the circular damaged area to obtain the coordinates of the circle center and radius parameter information; If the damaged shape is a star-shaped point, the skeleton of the star-shaped point area is extracted through morphological operations to obtain the coordinate information of the skeleton pixel points; Determine the position and size of the damaged area according to the center coordinates and radius information of the circular damaged area; According to the coordinate information of the skeleton pixels in the star-shaped dot area, the distribution range and density of the damaged area are determined; If the radius of the circular damaged area is greater than a preset threshold, the area is marked as severely damaged; If the density of skeleton pixels in the star-shaped dotted area is greater than a preset threshold, the area is marked as having multiple damages; For the circular area marked as severely damaged, the region growing algorithm is used to expand the damaged area to obtain a more accurate damage boundary; For star-shaped areas marked with multiple damages, the skeleton pixels are grouped using a cluster analysis algorithm to determine the location of each damage point; The enlarged circular damaged area and the clustered star-shaped damaged area are merged to obtain the complete cable damaged area; Determine the degree of cable damage based on the proportion of the combined damaged area to the total cable area; According to the degree of damage of the cable, combined with the cable material and service life attribute information, a decision tree algorithm is used to give suggestions for cable repair or replacement, and the location information of the damaged area is fed back to the maintenance personnel to guide them to carry out targeted repair or replacement operations.

3. The method according to claim 1, wherein: The damaged area is segmented by an image segmentation algorithm, and the size of the damaged area is obtained by calculating the pixel area of ​​the segmented area, including: The damaged area is segmented using the image segmentation algorithm based on region growing; If the segmentation result is over-segmented or under-segmented, the segmentation result is optimized by morphological operations according to the shape characteristics of the damaged area, and small areas and holes are eliminated by corrosion and expansion operations to obtain accurate segmentation areas; According to the boundary coordinate information of the segmented area, the contour tracing algorithm is used to extract the contour of the segmented area, obtain the contour vector composed of the pixel point coordinate sequence, and calculate the perimeter and area geometric features of the contour; If the contour of the segmented area is discontinuous or has breakpoints, the contour is repaired using a contour repair algorithm, and the coordinates of the missing contour points are estimated by interpolation or fitting methods to ensure the integrity and continuity of the contour; According to the restored contour vector, the polygon approximation algorithm is used to simplify the contour. By recursively merging adjacent contour points, the obtained polygon is used to represent the approximate shape of the segmented area. By calculating the pixel area of ​​the polygon, the size of the segmented area, i.e., the number of pixels in the damaged area, is obtained; At the same time, according to the spatial resolution information of the image, the pixel area is converted into the area value under the actual physical size to judge the severity of the damaged area; If the damaged area exceeds the preset threshold, the segmented area is marked as severely damaged, and its location and size information are recorded; Otherwise, the segmented area is marked as slightly damaged, and the next damaged area is processed until all damaged areas are identified and quantified, and a damage distribution map and area statistics report are obtained.

4. The method according to claim 1, wherein: The severity of the cable damage is determined based on the damage shape and area information and the preset judgment conditions, including if the damage area is less than the preset damage area threshold and the shape is regular, it is determined that the outer sheath is damaged; if the damage area is greater than the preset damage area threshold and the shape is irregular, it is determined that the internal wire insulation layer is damaged, including: The shape characteristics of the damaged area are analyzed using the Hu invariant moment method. By comparing the Hu invariant moment of the damaged area with the preset threshold, it is determined whether its shape is regular. At the same time, it is determined whether the area of ​​the damaged area is greater than a preset threshold; If the damaged area is smaller than the preset damage area threshold and the shape is regular, it is determined that the outer sheath is damaged; if the damaged area is larger than the preset damage area threshold and the shape is irregular, it is determined that the inner wire insulation layer is damaged; A rule-based reasoning method is used to make a preliminary assessment of the severity of the damage. The shape characteristics and area information of the damaged area are input into the rule base, and the preliminary judgment result of the damage type is obtained through forward reasoning or backward reasoning. Combined with the hazard level and environmental factors, the severity score of each damaged area is calculated, and the damage area is divided into corresponding severity levels according to the severity score; Maintenance personnel can quickly develop maintenance plans to reduce the risk of accidents.

5. The method according to claim 1, wherein: If the outer sheath is determined to be damaged, the integrity of the internal wire insulation layer is detected by an image analysis algorithm. If the internal insulation layer is intact, the humidity information of the cable environment is obtained. If the humidity is lower than a preset humidity threshold, local repair is performed, including: Determine whether the damaged area of ​​the cable is the outer sheath damage. If the shape and size of the damaged area meet the characteristics of the outer sheath damage, it is preliminarily determined to be the outer sheath damage. Otherwise, determine whether it is other types of damage. For areas where the outer sheath is damaged, X-ray imaging or ultrasonic scanning detection methods are used to obtain image data of the inner wire insulation layer; According to the material properties and defect characteristics of the internal wire insulation layer, the corresponding image segmentation algorithm is selected. The image segmentation algorithm includes threshold segmentation and edge detection, extracts the area of ​​the insulation layer, and calculates the integrity index of the insulation layer, including connectivity and circularity geometric parameters; Compare the integrity index with the preset threshold to determine whether the internal insulation layer is intact. If the index meets the threshold condition, the insulation layer is considered intact; otherwise, the insulation layer is considered damaged or defective. If the internal insulation layer is intact, the humidity value of the cable environment is obtained through the built-in temperature and humidity sensor of the cable or the external environment monitoring device, and compared with the humidity threshold; If the humidity value is below the threshold, the cable is judged to be in a dry environment and local repairs are performed using methods including electrical tape; Carry out insulation resistance and withstand voltage tests on the repaired cable to ensure that its insulation performance meets the standard requirements, and regularly inspect and maintain the damaged parts. If the tape is found to be aging or falling off, replace it to ensure the safe operation of the cable.

6. The method according to claim 1, wherein: If it is determined that the insulation layer of the internal wire is damaged, the damaged area is enlarged and clarified through the image enhancement algorithm, the number and location of the damaged wires are analyzed, and whether the copper core is intact is determined, including: Obtain image data of the internal wire insulation layer. If the image resolution is low or the contrast is poor, use image interpolation algorithm and histogram equalization algorithm to pre-process the image to improve image quality; According to the difference in density and attenuation coefficient of the internal wire insulation material, the region of interest is extracted through the threshold segmentation algorithm to obtain a binary image of the suspected insulation damage area, and the small area of ​​noise is removed through morphological filtering operation; Perform geometric feature analysis on the suspected damaged area, calculate the area, aspect ratio and circularity parameters of the area, and match it with the preset insulation layer damage feature model. If the matching degree exceeds the threshold, the area is determined to be an insulation layer damage area; The image pyramid algorithm is used to decompose the damaged area at multiple scales, and the corresponding scale layer is selected to enlarge the damaged area. The enlarged image is smoothed with a Gaussian filter to remove the noise and artifacts generated during the enlargement process. Perform edge detection and contour extraction on the magnified and smoothed damaged area image to obtain the boundary of the damaged area, and determine whether the damage type is a crack, fracture or hole based on the shape characteristics of the boundary; The wires in the damaged area are segmented by the watershed algorithm or the region growing algorithm, the position and number of each wire are marked, the number of damaged wires is calculated, and the segmentation result is saved in the form of a mask image; Grayscale statistics and texture analysis are performed on each segmented wire to calculate its average grayscale value, standard deviation and gradient direction histogram features to determine the integrity of the wire. If the feature value is abnormal, it is considered that the wire is broken or damaged and the condition of its copper core needs to be checked.

7. The method according to claim 1, wherein: After the insulation wrapping of a single wire is completed, the entire damaged area is wrapped with multiple layers of PVC tape by automated equipment. The number of layers of wrapping is determined according to the importance of the cable and environmental conditions. The number of layers of wrapping is at least two, including: Identify the location and number of each damaged wire, use template matching or feature extraction methods to determine the wire's diameter and length geometric parameters, and obtain the wire's three-dimensional spatial coordinate information; According to the spatial coordinates and geometric parameters of the wires, the movement trajectory and speed of the automation equipment are controlled so that the insulation wrapping material can be put on each wire, and the insulation layer material is closely attached to the surface of the wire through heating or UV curing; Perform eddy current or ultrasonic testing on the single wire after insulation wrapping to evaluate the integrity and uniformity of the insulation layer. If the test results show that there are bubbles, impurities or uneven thickness defects in the insulation layer, the automated equipment will be controlled to rework until it is qualified; The single wires that have been insulated are bundled together according to the original arrangement, and the three-dimensional structural model of the entire damaged area is reconstructed using an image stitching algorithm. This is then compared with the cable design drawings to verify the specifications and models of the wire bundles. Select the tape model and specification that matches the cable model from the PVC tape material database, calculate the optimal number of layers and winding angle of the tape winding according to the importance of the cable and environmental conditions, including temperature, humidity and vibration, and generate a winding process parameter table; Place the wire bundle on the work surface of the automated winding equipment, control the tape reel to spiral along the surface of the wire bundle through visual guidance and servo control technology, and adjust the tightness and angle of the tape in real time to ensure that there are no wrinkles, overlaps, or gaps between the tape and the surface of the wire bundle during the winding process; After winding, spray a polymer waterproof coating on the surface of the tape and apply sealant on the tape joints to improve the waterproof and insulation performance of the cable; The electrical parameters of the cable, including insulation resistance and leakage current, are collected through online monitoring equipment to determine whether the repaired cable meets the requirements for safe operation.

8. The method according to claim 1, wherein: For each damaged area detected, the area and length characteristics are extracted, and combined with the model and material information of the wire, a pre-established safety threshold database is queried to determine whether the size of the damage exceeds the safety range, and the severity of the damage is preliminarily assessed, including: The image segmentation algorithm is used to identify and locate the damaged area of ​​the wire. By setting the grayscale threshold or color threshold, the damaged area is distinguished from the background to obtain a binary mask image of the damaged area. Perform morphological processing on the binary damaged area mask image, including corrosion, expansion, opening and closing operations, to eliminate noise and interference in the image and obtain an accurate and smooth damaged area contour; According to the pixel coordinates of the damaged area contour, its geometric features are calculated. Including area, perimeter, length and width of the smallest circumscribed rectangle, quantitative indicators reflecting the damage size are obtained; Extract the model and material information of the wire from the mark on the surface of the wire, the model and material information of the wire includes the conductor cross-sectional area, the type of insulation material and the rated voltage parameters; The extracted wire model and material information are used as keywords to query the pre-established safety threshold database to obtain the corresponding damage size safety range; Compare the geometric characteristic value of the damaged area with the queried safety threshold to determine whether the damage size exceeds the safety range; If the injury size does not exceed the safe range, it is classified as a general injury; If the injury size exceeds the safe range, it is classified as severe; Based on the number and proportion of general damage and severe damage, combined with the environment and usage factors of the wires, a preliminary assessment of the severity of the damage is given. The preliminary assessment results include mild, moderate and severe levels of damage, providing a reference for subsequent maintenance and replacement decisions.

9. The method according to claim 1, wherein: The shape, size and depth information of the comprehensive damage are matched with the pre-established defect pattern library; the specific type of damage is identified through similarity comparison, including wear, scratches, and fractures; according to the risk level corresponding to the damage type, the impact of the damage on the safety performance of the wire is quantitatively evaluated, including: Adopt 3D reconstruction technology, including structured light scanning or laser scanning, to obtain 3D point cloud data of the damaged area of ​​the wire. Through point cloud registration and surface fitting algorithm, construct a 3D model of the damaged area and extract the shape, size and depth geometric information of the damage. According to the material properties and processing technology of the wire, physical simulation analysis is performed on the three-dimensional model to calculate the stress distribution, deformation field and strain field in the damaged area, and obtain mechanical indicators reflecting the severity of the damage, including stress concentration coefficient and plastic strain; The defect pattern library stores the three-dimensional models and mechanical index ranges of various typical damage types, including wear, scratches, fractures and corrosion. Each damage type corresponds to a risk level, indicating its impact on the safety performance of the wire. The damage model to be identified is matched with the model in the defect pattern library, and a three-dimensional shape descriptor is used. The three-dimensional shape descriptor is a rotationally invariant surface distribution histogram or spherical harmonic function to measure the similarity between the two models and obtain a similarity coefficient between 0 and 1; Select several models with the highest similarity to the damage model to be identified, and perform cluster analysis on the mechanical indicators of the models. The cluster analysis includes K-means clustering or density clustering. According to the clustering results, the type of damage model is determined, and its risk level is judged. If the similarity between the damage model and the 3D models of multiple damage types reaches a preset value, the risk levels of multiple types are fused and a weighted average or maximum value method is used to obtain a comprehensive risk assessment result, which is quantified as a risk index between 0 and 5; The identified damage type and risk index are compared with the preset threshold. If the risk index exceeds the warning value, the high-risk damage area is marked on the 3D model, a damage detection report is generated, and an early warning signal is triggered to notify maintenance personnel to handle it.

10. An AI-based electrical circuit image analysis and comparison system, characterized in that: Utilizing the electrical circuit image analysis and comparison method according to any one of claims 1 to 9, the system comprises: Image preprocessing module, used to reduce noise and enhance the cable image to improve image quality; Edge detection and shape recognition module, used to identify the outline and shape of cable damage and determine the type of damage; Image segmentation and area calculation module, used to segment the damaged area and calculate its area to assess the degree of damage; Damage type and risk assessment module, used to assess safety risks based on damage characteristics and database matching; Internal structure inspection module, used to detect whether the internal insulation layer is damaged to ensure the integrity of the wire; The repair and protective measures module is used to implement specific repair work and protective measures.

Images