High-voltage power line high-temperature fault detection system and method
The high-voltage power line high-temperature fault detection system utilizes UAV aerial infrared image processing and temperature detection to solve the problem of limited image processing accuracy in existing technologies, achieving efficient and accurate high-temperature fault detection. It is suitable for detecting overhead high-voltage power lines in complex environments.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- STATE GRID JIANGSU ELECTRIC POWER CO LTD RESEARCH INSTITUTE
- Filing Date
- 2023-04-18
- Publication Date
- 2026-06-23
AI Technical Summary
Existing drone inspection technology is limited in high-voltage power line fault detection due to image processing accuracy issues and susceptibility to background interference, making it difficult to achieve efficient and accurate high-temperature fault detection.
A high-voltage power line high-temperature fault detection system was designed, including a conductor preprocessing module, a drain wire preprocessing module, a maximum connected component acquisition module, a conductor position acquisition module, a drain wire position acquisition module, and a high-temperature fault judgment module. By processing infrared images taken by UAV and temperature detection, and using algorithms such as edge detection, gamma transform, gradient detection, and depth-first search, the system can accurately locate the positions of the conductor and drain wire and judge the high-temperature fault.
It enables high-temperature anomaly detection of overhead high-voltage conductors under complex backgrounds, with strong anti-interference ability, good robustness, and accurate acquisition of conductor temperature, thus improving detection accuracy and efficiency.
Smart Images
Figure CN116499591B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of image processing and pattern recognition technology, specifically to a high-voltage power line high-temperature fault detection system and method. Background Technology
[0002] Overhead transmission lines are typically characterized by complex structures, difficult installation, large scale, and the ability to be repaired. The complexity and difficulty of installation stem from the numerous components that make up the line, primarily including overhead conductors, towers, guy wires, hardware, and insulators. Any change in the condition of any of these components will affect the normal operation of the line. The large scale refers to the wide geographical area they traverse and the complex and variable climates along the routes, making it difficult to uniformly address geographical factors across different regions.
[0003] Overhead transmission lines, as an essential component of power transmission, are particularly susceptible to interference from internal and external factors, leading to abnormal operation or even malfunctions. In severe cases, this can result in power outages, seriously disrupting people's normal lives. Therefore, early prevention, monitoring, and systemic troubleshooting of overhead line faults play a crucial role in improving power transmission quality, reliability, and economy.
[0004] In the daily inspection of high-voltage transmission lines, traditional methods mainly rely on manual inspection. Researchers both domestically and internationally have shifted their focus from initial robotic inspection to drone inspection technology. Drone inspection not only overcomes the shortcomings of traditional manual line inspection in power systems but also improves inspection efficiency, reduces costs, and increases uptime. Its application in modern smart grids indirectly demonstrates the continuous improvement of technologies used in power line inspection. Domestic research on drone-based intelligent line inspection technology started relatively late. Currently, domestic drone line inspection technology mainly involves manipulating drones to take aerial photos near power lines and transmitting the images to a data center. In the data center, image processing algorithms and temperature data provided by the drone manufacturer are used to determine the conductor's heating status for fault detection. However, this method is limited by the detection accuracy of image processing and is easily affected by background interference. Summary of the Invention
[0005] The purpose of this invention is to provide a high-voltage power line high-temperature fault detection system and method. This invention determines the location of the overhead power line by using infrared images obtained from drone aerial photography and obtains the temperature of the overhead power line through a temperature detection SDK, thereby achieving accurate detection of high-voltage power line high-temperature faults.
[0006] To achieve this objective, the high-voltage power line high-temperature fault detection system designed in this invention includes a conductor preprocessing module, a drain wire preprocessing module, a maximum connectivity acquisition module, a conductor position acquisition module, a drain wire position acquisition module, and a high-temperature fault judgment module.
[0007] The conductor preprocessing module uses an edge detection algorithm to extract the conductor mask region from the image of the overhead conductor area, performs image morphological dilation on the conductor mask region, and then uses median filtering to denoise the image after the image morphological dilation operation to obtain the conductor preprocessed image.
[0008] The preprocessing module for the drainage line is used to adjust the grayscale value of the grayscale image of the overhead conductor area using gamma transform. It designs a gradient detection template based on the shape characteristics of the drainage line, extracts the boundary gradient of the drainage line in the grayscale image of the overhead conductor area after grayscale value adjustment using the gradient detection template, and generates a drainage line area mask based on the boundary gradient of the drainage line.
[0009] The maximum connected component acquisition module performs image erosion on the preprocessed image of the conductor to remove the connectivity between the insulator and other objects, and determines the position of the largest insulator in the image by the maximum connected component within the conductor mask area.
[0010] The conductor position acquisition module is used to start from the boundary of the insulator on the conductor side and end at the intersection of the conductor and the boundary of the conductor preprocessed image. It uses an 8-connected scanning algorithm to track adjacent edge pixels to form lines, thereby converting the conductor preprocessed image into a line graph and determining the conductor position based on the line graph.
[0011] The drainage line location acquisition module is used to perform a depth-first search for the drainage line location on the drainage line region mask, and retrieves the drainage line location using the drainage line slope as the termination condition.
[0012] The high-temperature fault diagnosis module is used to obtain the temperature of all points at the conductor position and the drain line position, and to diagnose temperature faults.
[0013] The beneficial effects of this invention are:
[0014] This invention is applicable to the detection of high-temperature anomalies in overhead high-voltage power lines under complex backgrounds. It accurately obtains the coordinates of the overhead power line through image enhancement, gradient edge extraction, and horizontal projection, and uses a temperature SDK provided by a drone manufacturer to obtain the corresponding temperature. Based on the high-temperature anomaly criteria for overhead power lines provided by the State Grid Corporation of China, it determines whether an anomaly exists in the overhead power line. This invention has strong anti-interference capabilities, good robustness, and can achieve high-temperature anomaly detection of overhead high-voltage power lines in complex environments. Attached Figure Description
[0015] Figure 1 This is a schematic diagram of the structure of the present invention;
[0016] Figure 2 This is the edge detection map of the Canny invention;
[0017] Figure 3 This is a pre-processed diagram of the drainage line of the present invention;
[0018] Figure 4 This is a diagram showing the insulator test results of the present invention;
[0019] Figure 5 This is a diagram showing the final test results of the wire of this invention;
[0020] Figure 6 This is a diagram showing the final test results of the drainage line of this invention; Detailed Implementation
[0021] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments:
[0022] like Figure 1 The high-voltage power line high-temperature fault detection system shown includes an RGB channel separation module, an ROI region acquisition module, a conductor preprocessing module, a drain wire preprocessing module, a maximum connected component acquisition module, a conductor position acquisition module, a drain wire position acquisition module, and a high-temperature fault judgment module.
[0023] The RGB channel separation module is used to separate the RGB channels of the infrared image of the overhead conductor and extract the G channel information, which is most sensitive to temperature information, as a grayscale image, so as to facilitate the unified processing of the conductor in subsequent inspection.
[0024] The ROI region acquisition module uses mean drift filtering and watershed algorithm to cluster pixels with gray value differences within a threshold of 50 in the grayscale image of the overhead conductor infrared image, and then segments the clustered region from the grayscale image to obtain the overhead conductor region image, which can remove the interference of background items such as towers and ground;
[0025] The overhead conductor preprocessing module uses the Canny edge detection algorithm to extract the conductor mask region from the image of the overhead conductor area. It then performs morphological dilation on the conductor mask region to enhance the conductor portion while weakening the background area. Median filtering is then used to denoise the morphologically dilated image to obtain the preprocessed conductor image. This process removes cloud interference and provides grayscale features for the conductor location acquisition module. The processing result is as follows: Figure 2 ;
[0026] The preprocessing module for the drainage line is used to adjust the grayscale values of the grayscale image of the overhead conductor region using gamma transform. After adjustment, bright areas in the image become brighter and dark areas become darker, mainly to enhance contrast. A gradient detection template is designed based on the shape characteristics of the drainage line. The gradient detection template is used to extract the gradient of the drainage line boundary in the grayscale image of the overhead conductor region after grayscale adjustment, and a drainage line region mask is generated based on the gradient of the drainage line boundary to provide grayscale features for the drainage line position acquisition module.
[0027] The maximum connected component acquisition module performs image erosion on the preprocessed image of the conductor to remove the connectivity between the insulator and other objects. It determines the position of the largest insulator in the image based on the maximum connected component within the conductor mask area. Based on the topological relationship between the insulator and the conductor position, it provides spatial features for the conductor position acquisition module.
[0028] The conductor position acquisition module is used to start from the boundary of the insulator on the conductor side and end at the intersection of the conductor and the boundary of the conductor preprocessed image. It uses the 8-connected scanning algorithm in image morphology to track adjacent edge pixels and form lines, thereby converting the conductor preprocessed image (edge map) into a line map and determining the conductor position based on the line map.
[0029] The drainage line position acquisition module is used to perform a depth-first search for the drainage line position on the drainage line region mask, and uses the drainage line slope as the termination condition (two pixels (6 pixels apart) are drawn from the starting point, one in front of the other, and the slope between them is calculated each time. The search ends when they tend to be horizontal).
[0030] The high-temperature fault diagnosis module is used to obtain the temperature of all points at the conductor position and the drain line position, and to diagnose temperature faults.
[0031] In the above technical solution, the temperature of all locations in the overhead conductor is extracted using DJI / Keyi's SDK (Software Development Kit), and the presence of a high-temperature fault in the overhead conductor is determined by the criteria for high-temperature anomalies in the State Grid standard.
[0032] In the above technical solution, the line graph is represented by a two-dimensional linked list, where each node in the outer linked list corresponds to a line in the image, and each node in the inner linked list corresponds to an edge pixel that makes up the line.
[0033] After recording the coordinates and number of wires in the line graph, extract the grayscale value of each pixel under the coordinates of each wire. If the grayscale value d>160, increment the count of the high-resolution pixels by 1. Finally, compare the count of the high-resolution pixels of each wire, and the wire with the largest count is the detected wire.
[0034] In the above technical solution, the specific method for adjusting the grayscale values of the overhead conductor area using gamma transformation in the preprocessing module is as follows:
[0035] The grayscale values of the overhead conductor area are adjusted using gamma transform, and the formula is as follows:
[0036]
[0037] In this algorithm, f(x,y) represents the input image pixel value, F(x,y) represents the output image pixel value, and γ is the gamma factor, which controls the image enhancement intensity. When γ < 1, it is used to expand the low grayscale part of the image; when γ > 1, it is used to expand the high grayscale part of the image. By using different γ values, the details of low or high grayscale parts can be enhanced. In this algorithm, the γ value is set to 0.5. After applying the gamma transform, the wire boundary is extracted. The gamma transform can improve the contrast of the boundary. The gamma transform is a non-linear transform that maps part of the grayscale region to a wider or narrower region to enhance bright areas (wires, insulators, etc.) and weaken dark areas (sky, background, etc.).
[0038] Inspired by traditional edge detection algorithms, this invention proposes an edge detection template based on the characteristics of drainage line images. Analysis of drainage line images reveals that drainage lines are often arc-shaped, gradually transitioning from vertical to horizontal, and their width is typically 2-3 pixels. The grayscale value is higher closer to the center of the drainage line. Based on the characteristic of high grayscale at the center of the drainage line, the gradient detection template used is as follows:
[0039]
[0040] The gradient detection template is convolved with the grayscale image of the overhead conductor region (ROI region grayscale image) after grayscale adjustment to extract the gradient of the drainage line boundary in the grayscale image of the ROI region, thereby improving the contrast between the drainage line and the background. Finally, median filtering is applied to the drainage line boundary gradient (edge image), and the drainage line region mask is obtained based on the filtered drainage line boundary gradient. The preprocessed drainage line image is shown below. Figure 3 As shown.
[0041] In the above technical solution, in practical application scenarios, the larger the insulator in the obtained conductor image, the closer the insulator is to the camera lens, and the more accurate the temperature field data obtained from the infrared image. Since the insulator and conductor have a strong spatial correlation in topology, the conductor connected to the insulator with the most accurate temperature can be determined based on the size of the insulator's outline. Considering that the largest area in the image in step 3 is the insulator, but it is connected to the conductor and the drain wire, and the image is already a binary image with a conductor width generally of 2-3 pixels, the specific method for image erosion of the conductor preprocessing image by the maximum connected component acquisition module is as follows: traverse each pixel in the conductor preprocessing image, count the number of pixels whose pixel value exceeds a threshold (grayscale value d>160) within a 3-pixel range of each pixel in the conductor preprocessing image, and when the number of pixels whose pixel value exceeds the threshold is less than a preset value (number of pixels <15), set the grayscale of that pixel to 0. At this time, the conductor and clouds in the background of the conductor preprocessing image are completely eroded away. Then, by filtering the outlines, the insulator with the largest connected component is detected. The processing result is as follows Figure 4 As shown.
[0042] In the above technical solution, the conductor position acquisition module starts from the boundary of the insulator on the conductor side and ends at the intersection of the conductor and the boundary of the preprocessed conductor image. Using an 8-connectivity scanning algorithm, it tracks adjacent edge pixels to form a line vector map. After obtaining the line vector map, the algorithm performs preliminary filtering based on the characteristics of the lines. Observing the edge image, it is noted that the lines are generally parallel and the background is a clean sky. Therefore, lines can be filtered according to the following three rules: 1. Delete lines with a pixel width greater than 5 to avoid interference from other objects, such as clouds in the sky, towers, and the ground. 2. Delete conductors among non-parallel lines. 3. Since a conductor may be detected multiple times, lines with repeated coordinates need to be deleted.
[0043] The specific method for ultimately determining the position of the conductor is as follows:
[0044] The 8-connected scan algorithm is expressed as follows: First, based on the scale factor γ (γ>1), all lines in the original wire preprocessing image with width W and height H are mapped to a line distribution thumbnail S with width M and height N;
[0045] M = W / γ, N = H / γ;
[0046] If the center position of the original line L is (x0, y0), and it maps to the cell at position (x, y) in the line distribution thumbnail S, the mapping rule is (x, y) = f(x0, y0) = (x0 / γ, y0 / γ). Each cell in the line distribution thumbnail S contains all the lines in the original image that are mapped to that cell, which can be represented as follows:
[0047] S(x,y)={L(x0,y0)|x=x0 / γ,y=y0 / γ}
[0048] Then, cells S(x,y) in the line distribution thumbnail S are treated as scanning elements. An 8-connectivity scanning algorithm is used to divide the line distribution thumbnail S into several connected regions, i.e., candidate wire regions. Each region contains several adjacent lines. To avoid interference from noisy lines near the wire regions, the horizontal distance is set to 5 pixels when generating the wire connected regions. After determining the number of wires, the number of pixels on these lines with grayscale values exceeding the grayscale threshold (d>200) is counted. The wire with the most pixels is the detected wire. The detection result is as follows: Figure 5 The left side shows the specific location of the separated wire, and the right side shows the location of the wire in the original diagram.
[0049] In the above technical solution, during the depth-first search retrieval of the drainage line position in the drainage line region mask by the drainage line position acquisition module, the depth-first search algorithm is improved using the pruning optimization principle. The process is as follows:
[0050] Based on the number of wires, find the starting point of the drainage line corresponding to each wire. Below the wires extracted by wire detection, use a rectangular sliding window (size 3*20) to traverse all points on the wires in turn. Where there is a drainage line, it will coincide with the slider, thus obtaining the starting point of the drainage line.
[0051] After obtaining the starting point, the depth-first traversal algorithm is used to traverse the gray points downwards, and constraints are used to ensure that the traversal always takes place on the flow line.
[0052] During the traversal of the drainage line, if the curvature of two points approaches 0, the retrieval of the drainage line ends.
[0053] Repeat the above steps and count the length of the drainage line each time. The longest length indicates the location of the detected drainage line.
[0054] If multiple drain wires need to be inspected, the relative position of each wire to the wire below it is calculated to determine which wire should be on the left and which on the right. Then, the drain wire closer to the insulator is slightly adjusted from its previous starting point, and a new drain wire search is performed to avoid overlap. Finally, the found drain wires are repositioned. This method determines which nodes should be visited and which should be discarded, thus avoiding unnecessary search processes and improving search efficiency. The final drain wire inspection results are as follows: Figure 6 The left side shows the specific location of the segmented drainage line, and the right side shows the location of the drainage line in the original image.
[0055] A method for detecting high-temperature faults in high-voltage power lines, comprising the following steps:
[0056] Step 1: Use drones to collect infrared images of overhead power lines, replacing the previous manual inspection method, improving inspection efficiency and reducing costs; perform RGB channel separation on the infrared images of overhead power lines, extract the G channel information as a grayscale image, which facilitates unified processing of subsequent power line inspections and enhances the robustness of the algorithm.
[0057] Step 2: Using mean-shift filtering and watershed algorithm, the pixels with gray value differences within the threshold in the grayscale image of the overhead conductor infrared image are clustered, and the clustered areas are segmented from the grayscale image to obtain the overhead conductor area image.
[0058] Step 3: Use the Canny edge detection algorithm to extract the overhead wire mask region from the image of the overhead wire area, and perform image morphological dilation on the overhead wire mask region to enhance the wire part and weaken the background area in the image. Then, use median filtering to denoise the image after the image morphological dilation operation to obtain the preprocessed image of the overhead wire.
[0059] Step 4: Use gamma transform to adjust the grayscale value of the grayscale image of the overhead conductor area. Design a gradient detection template according to the shape characteristics of the drainage line. Extract the boundary gradient of the drainage line in the grayscale image of the overhead conductor area after grayscale adjustment using the gradient detection template. Generate a drainage line area mask based on the boundary gradient of the drainage line to improve the contrast between the drainage line and the background.
[0060] Step 5: Perform image erosion on the preprocessed image of the conductor to remove the connectivity between the insulator and other objects, and determine the position of the largest insulator in the image by the largest connected region within the conductor mask area;
[0061] Step 6: Starting from the boundary of the insulator on the conductor side and ending at the intersection of the conductor and the boundary of the conductor preprocessed image, use the 8-connected scanning algorithm to track adjacent edge pixels to form lines, convert the conductor preprocessed image into a line graph, and determine the conductor position based on the line graph.
[0062] Step 7: Perform a depth-first search for the location of the drainage line in the drainage line region mask, and use the slope of the drainage line as the termination condition to retrieve the location of the drainage line.
[0063] Step 8: Obtain the temperature of all points at the conductor position and the drain line position, and perform temperature fault judgment. If the detected conductor temperature is within the judgment range, the conductor is judged to be normal temperature; otherwise, it is an abnormal temperature.
[0064] A computer-readable storage medium storing a computer program, characterized in that: when the computer program is executed by a processor, it implements the steps of the above-described method.
[0065] The contents not described in detail in this specification are existing technologies known to those skilled in the art.
Claims
1. A high-voltage power line high-temperature fault detection system, characterized in that: It includes a conductor preprocessing module, a drain wire preprocessing module, a maximum connected component acquisition module, a conductor position acquisition module, a drain wire position acquisition module, and a high-temperature fault judgment module; The conductor preprocessing module uses an edge detection algorithm to extract the conductor mask region from the image of the overhead conductor area, performs image morphological dilation on the conductor mask region, and then uses median filtering to denoise the image after the image morphological dilation operation to obtain the conductor preprocessed image. The preprocessing module for the drainage line is used to adjust the grayscale value of the grayscale image of the overhead conductor area using gamma transform. It designs a gradient detection template based on the shape characteristics of the drainage line, extracts the boundary gradient of the drainage line in the grayscale image of the overhead conductor area after grayscale value adjustment using the gradient detection template, and generates a drainage line area mask based on the boundary gradient of the drainage line. The maximum connected component acquisition module performs image erosion on the preprocessed image of the conductor to remove the connectivity between the insulator and other objects, and determines the position of the largest insulator in the image by the maximum connected component within the conductor mask area. The conductor position acquisition module is used to start from the boundary of the insulator on the conductor side and end at the intersection of the conductor and the boundary of the conductor preprocessed image. It uses an 8-connected scanning algorithm to track adjacent edge pixels to form lines, thereby converting the conductor preprocessed image into a line graph and determining the conductor position based on the line graph. The drainage line location acquisition module is used to perform a depth-first search for the drainage line location on the drainage line region mask, and retrieves the drainage line location with the drainage line slope approaching 0 as the termination condition. The high-temperature fault diagnosis module is used to obtain the temperature of all points at the locations of the conductor and the drain line, and to diagnose temperature faults.
2. The high-voltage power line high-temperature fault detection system according to claim 1, characterized in that: It also includes an RGB channel separation module, which is used to separate the RGB channels of the infrared image of the overhead power line and extract the G channel information as a grayscale image.
3. The high-voltage power line high-temperature fault detection system according to claim 1, characterized in that: It also includes a Region of Interest (ROI) acquisition module, which uses mean-shift filtering and watershed algorithm to cluster pixels in the grayscale image of the overhead conductor infrared image whose grayscale value difference is within a threshold, and then segments the clustered region from the grayscale image to obtain the overhead conductor region image.
4. The high-voltage power line high-temperature fault detection system according to claim 1, characterized in that: The specific method used by the preprocessing module for adjusting the grayscale values of the overhead conductor area using gamma transformation is as follows: The grayscale values of the overhead conductor area are adjusted using gamma transform, and the formula is as follows: in, For the input image pixel values, To output image pixel values, The gamma factor controls the image enhancement intensity. After applying the gamma transform, wire boundary extraction is performed.
5. The high-voltage power line high-temperature fault detection system according to claim 1 or 4, characterized in that: The gradient detection template is: The gradient detection template is convolved with the grayscale image of the overhead conductor region after grayscale adjustment to extract the gradient of the guide line boundary of the grayscale image of the overhead conductor region after grayscale adjustment. Finally, the gradient of the guide line boundary is filtered by median filtering, and the guide line region mask is obtained based on the filtered gradient of the guide line boundary.
6. The high-voltage power line high-temperature fault detection system according to claim 1, characterized in that: The specific method for the maximum connected component acquisition module to perform image erosion on the preprocessed image of the conductor is as follows: traverse each pixel in the preprocessed image of the conductor, count the number of pixels whose pixel value exceeds the threshold within a distance of 3 pixels from each pixel in the preprocessed image of the conductor, and when the number of pixels whose pixel value exceeds the threshold is less than a preset value, set the gray level of the point to 0. At this time, the conductor and the clouds in the background of the preprocessed image of the conductor are completely eroded away.
7. The high-voltage power line high-temperature fault detection system according to claim 1, characterized in that: The conductor position acquisition module starts at the boundary of the insulator on the conductor side and ends at the intersection of the conductor and the boundary of the preprocessed conductor image. It uses an 8-connected scanning algorithm to track adjacent edge pixels, forming lines, and ultimately determines the conductor position. The specific method is as follows: The 8-connected scan algorithm is expressed as follows: First, based on the scale factor γ, all lines in the original wire preprocessing image with width W and height H are mapped to a line distribution thumbnail S with width M and height N; M = W / γ, N = H / γ; If the center position of the original line L is ( , This is mapped to the cell at position (x,y) in the line distribution thumbnail S, with the mapping rule (x,y)=f ( , )=( / γ, / γ), each cell of the line distribution thumbnail S contains all the lines in the original image that are mapped to that cell, which can be represented as c Then, the cell S(x, y) in the line distribution thumbnail S is regarded as a scanning element. The line distribution thumbnail S is divided into several connected regions, namely candidate wire regions, using the 8-connected scanning algorithm. Each region contains several adjacent lines. When generating the wire connected regions, the horizontal distance is set to a set value. After determining the number of wires, the number of pixels on these lines whose gray values exceed the gray value threshold is counted. The wire with the most pixels is the detected wire.
8. The high-voltage power line high-temperature fault detection system according to claim 1, characterized in that: The drainage line location acquisition module improves the depth-first search algorithm by using pruning optimization principles during the depth-first search retrieval of the drainage line region mask. The process is as follows: Based on the number of wires, find the starting point of the drainage line corresponding to each wire. Below the wires extracted by wire detection, use a rectangular sliding window to traverse all points on the wires in turn. Where there is a drainage line, it will coincide with the slider, thus obtaining the starting point of the drainage line. After obtaining the starting point, the depth-first traversal algorithm is used to traverse the gray points downwards, and constraints are used to ensure that the traversal always takes place on the flow line. During the traversal of the drainage lines, if the slope between two points approaches 0, the retrieval of the drainage lines ends. Repeat the above steps and count the length of the drainage line each time. The longest length indicates the location of the detected drainage line. If multiple lead wires need to be tested, the relative position of each wire to the wire below will be counted to determine which wire should be on the left and which should be on the right. Then, the lead wire closer to the insulator side will be slightly adjusted from its previous starting point and a new lead wire will be searched to avoid multiple lead wires overlapping. Finally, the found lead wires will be repositioned.
9. A method for detecting high-temperature faults in high-voltage power lines, characterized in that, It includes the following steps: Step 1: Separate the RGB channels of the infrared image of the overhead power line and extract the G channel information as a grayscale image; Step 2: Using mean-shift filtering and watershed algorithm, the pixels with gray value differences within the threshold in the grayscale image of the overhead conductor infrared image are clustered, and the clustered areas are segmented from the grayscale image to obtain the overhead conductor area image. Step 3: Extract the conductor mask region from the image of the overhead conductor area using the edge detection algorithm, perform image morphological dilation on the conductor mask region, and then denoise the image after the image morphological dilation operation by median filtering to obtain the preprocessed image of the conductor. Step 4: Use gamma transform to adjust the grayscale value of the grayscale image of the overhead conductor area, design a gradient detection template according to the shape characteristics of the diversion line, extract the gradient of the diversion line boundary in the grayscale image of the overhead conductor area after grayscale value adjustment through the gradient detection template, and generate a diversion line area mask according to the gradient of the diversion line boundary. Step 5: Perform image erosion on the preprocessed image of the conductor to remove the connectivity between the insulator and other objects, and determine the position of the largest insulator in the image by the largest connected region within the conductor mask area; Step 6: Starting from the boundary of the insulator on the conductor side and ending at the intersection of the conductor and the boundary of the conductor preprocessed image, use the 8-connected scanning algorithm to track adjacent edge pixels to form lines, convert the conductor preprocessed image into a line graph, and determine the conductor position based on the line graph. Step 7: Perform a depth-first search for the location of the drainage line in the drainage line region mask, and use the slope of the drainage line as the termination condition to retrieve the location of the drainage line. Step 8: Obtain the temperature at all points on the conductor and drain line locations, and determine the temperature fault.
10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, it implements the steps of the method as described in claim 9.