Method for judging heating defect of drainage wire of multi-device associated power transmission line

Abstract

The invention discloses a method for judging the heating defect of a drainage wire of a multi-device associated power transmission line. The system comprises an unmanned aerial vehicle infrared image acquisition and preprocessing module, an associated part focus area detection and segmentation module, an associated part detection and segmentation module, a main body drainage wire part detection and segmentation module, an associated part heating detection module, a drainage wire heating detection module and a comprehensive judgment module. From the perspective of multi-device association, the diagnosis method intelligently analyzes the overall thermal imaging characteristics of a key conductive loop of'strain clamp drainage plate-drainage wire-suspension clamp '. The method has the beneficial effects that environmental interference can be effectively distinguished, more accurate, more reliable and more intelligent identification and early warning of the heating defect of the drainage wire are realized by excavating and modeling the temperature association rule and physical characteristics among multiple components, and the intelligent level of state sensing and fault early warning of the power transmission line is improved.

Description

Technical Field

[0001] This invention relates to the technical field of intelligent infrared heating defect detection for power grid transmission equipment, and in particular to a method for judging heating defects in the lead wires of transmission lines with multiple interconnected components. Background Technology

[0002] In power systems, the lead wire, as a key hardware connecting tension clamps and conductors to ensure smooth current conduction, directly affects the stability and safety of the line. Lead wire overheating is a common defect in transmission lines, typically caused by loose connections, surface oxidation, and increased contact resistance. If not addressed promptly, continuous overheating will lead to metal degradation and decreased mechanical strength, potentially causing serious accidents such as connection melting and wire breakage, posing a severe threat to power grid reliability. Currently, with the widespread adoption of drone inspection technology, power maintenance departments are able to regularly acquire infrared image data covering the entire transmission line area, achieving automation and digitization of inspection operations and accumulating a massive amount of inspection images. However, the analysis and diagnosis of lead wire overheating defects still face the following serious challenges: First, defect identification heavily relies on human experience, making automated image reading a significant efficiency bottleneck. Currently, screening for overheating defects in drain wires from massive amounts of infrared images primarily depends on manual visual interpretation by maintenance personnel. Faced with tens of thousands of inspection images, manual analysis is not only time-consuming and labor-intensive but also prone to missed or misjudged defects due to visual fatigue, making it difficult to meet the demands of modern power grid lean operation and maintenance.

[0003] Secondly, complex on-site interference factors, especially the "conductor luminescence" phenomenon caused by sunlight reflection, greatly complicate the judgment. Infrared imaging is susceptible to environmental interference, the most typical of which is specular reflection artifacts formed on the conductor surface due to sunlight reflection. This artifact appears as a localized bright area in the infrared image, which is extremely similar to the hot spot characteristics of a real heating point. During manual judgment, it is often difficult to distinguish such "false heating" from real defects based solely on these characteristics, resulting in a high misjudgment rate and increasing unnecessary on-site verification costs.

[0004] Furthermore, existing analytical methods are isolated and one-sided, failing to fully utilize the correlated thermal characteristics between multiple components. Even with some automated tools assisting manual preliminary screening, their judgment logic is often limited to threshold comparisons of single statistical quantities such as the maximum, minimum, and average temperatures of the entire drain wire body. This analytical method treats the drain wire as an independent heat source, ignoring its essential role as a crucial "connecting" node in the current path. The actual heating state of the drain wire is strongly correlated with the temperature distribution of components closely connected to its two ends, such as the tension clamp drain plate and suspension clamp. Existing methods lack overall thermal imaging analysis and correlation modeling of this multi-component series structure, failing to effectively eliminate the interference of reflection artifacts and struggling to capture the subtle differences between localized overheating caused by poor contact and the normal temperature rise of adjacent components. This results in insufficient sensitivity and diagnostic accuracy for identifying early, minor heating defects.

[0005] In summary, the current method for judging the heating defects of the lead wires of power transmission lines based on UAV infrared inspection has significant limitations in terms of dealing with reflection interference, improving automation level, and depth of analysis. Summary of the Invention

[0006] The present invention aims to overcome the above-mentioned shortcomings in the prior art and provides a method for judging the heating defects of the lead wire of a power transmission line that can cope with reflection interference, improve the level of automation, and perform in-depth analysis of the correlation of multiple devices.

[0007] To achieve the above objectives, the present invention adopts the following technical solution: A method for judging the heating defects of the lead wire in a transmission line with multiple interconnected components includes a UAV infrared image acquisition and preprocessing module, a focal area detection and segmentation module for interconnected components, a detection and segmentation module for interconnected components, a main lead wire component detection and segmentation module, a heating detection module for interconnected components, a heating detection module for the lead wire, and a comprehensive judgment module. The specific processing flow is as follows: (1) The UAV infrared image acquisition and preprocessing module interacts with the UAV infrared image library through the OBS communication interface, acquires infrared images through the communication interface, acquires raw infrared images through the standardized OBS, and filters infrared images that meet the requirements by parsing metadata. (2) The related component focus area detection and segmentation module detects and segments the related area of ​​the drain line in the image through image detection and segmentation algorithm, so as to locate and initially segment the core related area containing the key components of drain line, tension clamp and suspension clamp from the panoramic infrared image containing complex background; and adopts the semantic segmentation module based on deep learning to perform pixel-level classification on the input image and generate a connected binary mask that can cover the entire current path of "tension clamp-drain line-suspension clamp"; (3) The associated component detection and segmentation module uses image algorithms to segment and obtain the tension clamp and suspension clamp associated with the drainage line, completes the target detection of the output bounding box through the algorithm, and outlines the tension clamp and suspension clamp respectively through the output binary segmentation mask. (4) The main drainage line component detection and segmentation module performs target detection and segmentation on the drainage line through image segmentation algorithm. Based on its geometric features, it selects a segmentation algorithm that is sensitive to the existing structure, outlines the continuous direction and complete contour of the drainage line, and generates its binary segmentation mask. (5) The associated component heating detection module includes the heating diagnosis of tension clamp and suspension clamp. The heating diagnosis of tension clamp is determined by calling the heating diagnosis module of tension clamp through the interface; the heating diagnosis of suspension clamp is determined by calling the heating diagnosis module of suspension clamp through the interface. (6) The drainage line heating detection module judges the heating of the drainage line part based on the temperature of the segmented drainage line area. By extracting the temperature of the wire clamp contour area, it judges the extreme point of the abnormality and the heating point and its heating location. (7) The comprehensive judgment module judges the heating status of the drainage line in this area by considering the heating status of the associated components and the drainage line, and determines whether it is not heating, sunlight emission, or a heating defect of the drainage line.

[0008] This method offers a diagnostic approach that breaks through the framework of single-point temperature analysis, intelligently analyzing the overall thermal imaging characteristics of the key conductive circuit—the tension clamp-drain plate-drain wire-suspension clamp—from a multi-component correlation perspective. This method effectively distinguishes between actual heat generation and environmental interference such as emissions. By mining and modeling the temperature correlation patterns and physical characteristics among multiple components, it achieves more accurate, reliable, and intelligent identification and early warning of heating defects in the drain wire, thereby improving the intelligence level of transmission line condition perception and fault early warning.

[0009] Preferably, in step (1), the UAV infrared image acquisition and preprocessing module filters out qualified infrared images containing the drainage line equipment from the UAV inspection infrared images, as follows: (11) Data interaction with the UAV inspection platform is achieved through the Object Storage Service (OBS) communication interface to realize the batch acquisition of raw infrared images; (12) After acquiring the infrared image, a filtering strategy based on the image file name is executed. The key is that, in the current power inspection operation, the file name of the infrared image follows the naming convention and embeds key metadata of the line, tower, shooting location, and target hardware. (13) By building a keyword library related to traffic generation, and constructing filename parsing and matching rules based on regular expressions, a filename matching degree function is defined. To quantify screening accuracy: Where 'name' represents the infrared image file name. This represents the Kth keyword. For the corresponding weighting coefficients, This indicates an indicator function, where 1 is assigned when the filename contains the keyword, and 0 otherwise; the matching score is... During matching, the image is determined to be a target image related to the drainage line; (14) For the target images that pass the screening, save them in a local unified path.

[0010] Preferably, in step (2), the associated component focus area detection and segmentation module initially locates and segments the core area covering the key conductive circuit of "tension clamp-drain wire-suspension clamp" from the panoramic infrared image containing complex scenes. It employs a deep learning-based target detection algorithm, trained on labeled infrared images of the transmission line components "tension clamp-drain wire-suspension clamp," to identify the overall distribution characteristics of the drain wire and its related components. The algorithm outputs the smallest bounding rectangle region (ROI) containing all associated components, defined by a bounding box. The description uses a quartet to represent the position and size of the focal region in the image pixel coordinate system: in, The pixel coordinates of the top-left corner of the focal rectangle are given. This represents the pixel coordinates of the bottom right corner of the focus rectangle.

[0011] Preferably, in step (3), the associated component detection and segmentation module performs deep analysis on the focal region located by the associated component focal region detection and segmentation module, locates and segments two key associated components, namely tension clamp and suspension clamp, and obtains their pixel-level contours; the YOLO-based instance segmentation algorithm is used to extract features from the input focal region image to generate candidate regions, and the classification network determines whether the target category contained in each candidate region is a tension clamp or a suspension clamp. For the confirmed target, the segmentation network generates a binary segmentation mask and uses pixel coordinates to outline the complete contour of the component.

[0012] Preferably, in step (3), the specific steps are as follows: (31) For the identified tension clamp, output its set of contour points. : Where n is the total number of pixels that constitute the contour points of the tension clamp. The pixels that represent the outline; (32) For the identified suspension clamps, output their set of contour points. : Where m is the total number of pixels that make up the contour points of the drooping line clamp. The pixels that represent the outline; Through the processing of the associated component detection and segmentation module, the tension clamps and suspension clamps of the drainage line-related components are separated from the mixed focal area.

[0013] Preferably, in step (4), the main body drainage line component detection and segmentation module cuts the drainage line in the focal region. A specialized segmentation method highly sensitive to linear structures is used to address the unique geometric shape of the drainage line in the focal region, as follows: (41) Extract multi-scale depth features for the focal region in the pre-trained encoder network by introducing a linear structure based on the Hessian matrix for each pixel in the image. Calculate its Hessian matrix: Where H(x,y) represents a pixel. The Hessian matrix, This represents the second derivative of the image in the x-direction. This represents the second derivative of the image in the x and y directions. This represents the second derivative of the image in the y-direction. This represents the second derivative of the graph in the y-direction and the x-direction; (42) By calculating the eigenvalues ​​of matrix H(x,y), a corresponding linear similarity response function is constructed. : in, , Let be the eigenvalues ​​of the H matrix, k and c be constants for the control sensitivity, and exp be an exponential function; (43) The function is put into pre-training to enhance the response of the linear region in the focal region. The decoder outputs a complete contour of the guide line, which is represented by the ordered set of pixels in the focal region as follows:

[0014] At the same time, the connectivity of the drainage line and related components will be determined.

[0015] Preferably, in step (43), to avoid pixel-level errors caused by cutting, the cut guide wire and the tension clamp and suspension clamp of the associated component are expanded by two pixels, and the judgment is made according to the following rules: (431) If the area of ​​the drainage line With tension clamp profile If there is an intersection, it is determined that the end is connected to the tension clamp, and the connection area is recorded. (432) If the area of ​​the drainage line Contour with the hanging line If there is an intersection, it is determined that the end is connected to the suspension clamp, and the connection area is recorded.

[0016] Preferably, in step (5), the associated component heating detection module performs heating diagnosis on the hardware at both ends of the drain wire, namely the tension clamp and the suspension clamp, as follows: (51) Combine the original infrared image and the outline of the tension clamp. The predefined interface is used to call the tension clamp overheating diagnosis module. After the overheating diagnosis is completed, the overheating judgment result of the associated tension clamp component is returned. Where S is the heating status mark, 0 indicates that the tension clamp is not heating, and 1 indicates that the tension clamp is heating; L is the heating part label, 0 indicates that the tension tube is heating, and 1 indicates that the drainage plate is heating. (52) Combine the original infrared image and the outline of the hanging line clamp. The suspension clamp overheating diagnosis module is called through a predefined interface. After the overheating diagnosis is completed, the overheating judgment result of the associated suspension clamp is returned. .

[0017] Preferably, in step (6), the drainage wire heating detection module performs independent heating status diagnosis on the segmented drainage wire body and locates its abnormal heating points; specifically as follows: (61) Outline the drainage line A skeletonization algorithm is applied to extract its topological skeleton lines, which are composed of a series of ordered pixels, denoted as a point set. Where n represents the total number of skeleton points, and the two endpoints of the drainage line are determined based on the skeleton line. and ; (62) Extract temperature information and extract each pixel on the skeleton line. coordinates ( , Using this as an index, the absolute temperature can be directly read from the global temperature matrix T via the SDK interface. Thus, the temperature sequence of the entire skeleton line is obtained. ; (63) After obtaining the temperature of the skeleton line, perform heat detection and positioning on the drainage line, and calculate the average temperature of all points on the skeleton line as the baseline reference temperature: in,( , () represents the pixel in the heat-generating area; (64) Traverse the skeleton line temperature sequence to identify abnormal heating points; if a certain point The temperature satisfies: The point is then identified as a candidate point for abnormal heating. The algorithm also performs cluster analysis on consecutive abnormal heating points. If the number of consecutive heating points identified as abnormal is greater than 6, the area is determined to be a heating defect area of ​​the drainage line, and the coordinates of the center point of the heating area are recorded. : ; (65) Finally output the overall heating status mark of the entire drainage line and the set of coordinates of the center of the identified heating area. .

[0018] Preferably, in step (7), the comprehensive judgment module establishes a thermal correlation judgment logic model among multiple devices to intelligently identify the heating properties of the drain wire, distinguishing between real heating and heating interference; it uses the heating diagnosis results of the associated components, namely the tension clamp and the suspension clamp, and the set of heating areas of the drain wire body, and makes judgments according to the following process: (71) If none of the associated components are heated and the drain line is not heated, the drain line is determined to be in normal condition, and a no-heating mark is output, indicating that the drain line in this infrared image is normal. (72) If none of the associated components are heating up, but there are heating points in the drain line, it is determined to be "pseudo-heating" caused by environmental factors, and a heating interference mark is output; (73) If the tension clamp heats up but only the tension tube heats up, while the suspension clamp and the drain line do not heat up, it is determined that the upstream is overheated but does not affect the drain line, and the drain line is in a state of no heat. (74) If the drainage wire heats up and a related component also heats up, calculate the heat generation point of the drainage wire. European distance to the nearest heat-generating component area : in, Indicates the heating area of ​​the drainage plate. This indicates the heating area of ​​the suspension clamp, and distance is the Euclidean distance between the heating point of the drainage wire and the heating point of the associated component; if If the value is less than 25 pixels, the heat source is determined to be a real heat source; otherwise, it is determined to be a reflection. When a real heat source exists, the final output is a heat defect mark on the drain line and the corresponding coordinates of the heat source location.

[0019] The beneficial effects of this invention are: it can effectively distinguish between real heat generation and environmental interference such as emission; by mining and modeling the temperature correlation law and physical characteristics between multiple components, it can achieve more accurate, reliable and intelligent identification and early warning of heat generation defects in the drain wire, thereby improving the intelligent level of power transmission line status perception and fault early warning. Attached Figure Description

[0020] Figure 1 This is a flowchart of the method of the present invention.

[0021] Figure 2 This is the overall architecture diagram of the present invention. Detailed Implementation

[0022] The present invention will now be further described with reference to the accompanying drawings and specific embodiments.

[0023] like Figure 1 , Figure 2 In the embodiments described above, a method for judging the heating defects of the lead wire of a transmission line with multiple associated components includes a UAV infrared image acquisition and preprocessing module, a focal area detection and segmentation module for associated components, a detection and segmentation module for associated components, a main lead wire component detection and segmentation module, a heating detection module for associated components, a heating detection module for the lead wire, and a comprehensive judgment module. The specific processing flow is as follows: (1) The UAV infrared image acquisition and preprocessing module interacts with the UAV infrared image library through the OBS communication interface and acquires infrared images through the communication interface. The module acquires raw infrared images through standardized OBS and performs strict access verification on the input images: automatically filtering based on fixed pixel resolution (640×512) to remove images with incorrect resolution, format errors or damage; at the same time, it parses the image metadata to verify whether the acquisition device comes from a compatible DJI infrared camera platform, and parses the metadata to filter infrared images that meet the requirements, ensuring the standardization of data and the reliability of temperature analysis.

[0024] The UAV infrared image acquisition and preprocessing module accurately filters out qualified infrared images containing drainage line equipment from a massive amount of UAV inspection infrared images, providing high-quality input for subsequent analysis. Specifically: (11) The module interacts with the UAV inspection platform through the Object Storage Service (OBS) communication interface to achieve batch acquisition of raw infrared images.

[0025] (12) After acquiring the infrared image, the module executes an intelligent filtering strategy based on the image file name. The key is that, in current power inspection operations, the file names of infrared images usually follow strict naming conventions and embed key metadata such as line, tower, shooting location, and target hardware.

[0026] (13) The module has a built-in keyword library related to lead generation and constructs filename parsing and matching rules based on regular expressions. The module defines a filename matching degree function. To quantify screening accuracy: Where 'name' represents the infrared image file name. This represents the Kth keyword. For the corresponding weighting coefficients, This indicates an indicator function, where 1 is assigned when the filename contains the keyword, and 0 otherwise; the matching score is... During matching, the image is determined to be a target image related to the drainage line.

[0027] (14) For the target images that pass the screening, the module saves them in a local unified path to ensure the traceability of the data in subsequent identification.

[0028] (2) The related component focus area detection and segmentation module mainly uses image detection and segmentation algorithms to detect and segment the related areas of the drain line in the image, so as to quickly locate and initially segment the core related areas containing key components such as drain line, tension clamp and suspension clamp from the panoramic infrared image containing complex background. It also uses a semantic segmentation module based on deep learning to perform pixel-level classification on the input image and generate a connected binary mask that can cover the entire current path of "tension clamp-drain line-suspension clamp", effectively eliminating the interference of a large number of irrelevant areas such as tower and background, and significantly improving the efficiency and accuracy of subsequent steps. Generally speaking, the tension clamp includes a tension tube and a drain plate. The connection order is suspension clamp, drain line, drain plate and tension tube. Therefore, it can also be said that the drain plate and tension tube form the tension clamp.

[0029] The module for detecting and segmenting the focal region of associated components uses panoramic infrared images containing complex scenes such as towers, insulators, and natural backgrounds to initially and intelligently locate and segment the core region covering the critical conductive loop of "tension clamp-drain wire-suspension clamp". It employs a deep learning-based target detection algorithm trained on a large number of labeled infrared images of transmission lines containing "tension clamp-drain wire-suspension clamp" components to identify the overall distribution characteristics of the drain wire and its related components. Specifically, this module uses a pre-trained network as its backbone, extracts multi-scale features through dilated convolution, and finally outputs a binary mask covering the "tension clamp-drain wire-suspension clamp" region. After morphological processing, the module outputs the smallest bounding rectangle region (ROI) containing all associated components, defined by a bounding box. To describe it precisely, the position and size of its focal region in the image pixel coordinate system are represented by a quartic matrix: in, The pixel coordinates of the top-left corner of the focal rectangle are given. This represents the pixel coordinates of the lower right corner of the focal rectangle area. The focal area defines the focal working area where subsequent modules need to perform fine component segmentation.

[0030] (3) The associated component detection and segmentation module mainly uses image algorithms to segment and obtain the tension clamps and suspension clamps associated with the drain line. The algorithm completes the target detection of the output bounding box, and accurately outlines the contours of the tension clamps and suspension clamps by outputting a high-precision binary segmentation mask.

[0031] The associated component detection and segmentation module performs deep analysis on the focal region located by the associated component focal region detection and segmentation module, locating and segmenting two key associated components: the tension clamp and the suspension clamp, and obtaining their pixel-level contours. Employing a YOLO-based instance segmentation algorithm, the module extracts features from the input focal region image, generating candidate regions. A classification network determines whether each candidate region contains a tension clamp or a suspension clamp. For confirmed targets, the segmentation network generates a high-precision binary segmentation mask, outlining the complete contour of the component using pixel coordinates.

[0032] Example: The associated component detection and segmentation module employs the Mask R-CNN instance segmentation algorithm for fine segmentation of the focal region. This module extracts multi-scale features through a Feature Pyramid Network (FPN), generates candidate regions through a Region Proposal Network (RPN), and outputs a high-precision segmentation mask after alignment by the ROIAlign layer. The final result is the set of contour points for the tension clamp. and the set of outline points of the hanging line clamp The details are as follows: (31) For the identified tension clamp, output its set of contour points. : Where n is the total number of pixels that constitute the contour points of the tension clamp. The pixels that represent the outline.

[0033] (32) For the identified suspension clamps, output their set of contour points. : Where m is the total number of pixels that make up the contour points of the drooping line clamp. The pixels that represent the outline.

[0034] Through the processing of the associated component detection and segmentation module, the tension clamps and suspension clamps of the drainage line-related components are clearly separated from the mixed focal area.

[0035] (4) The main body drainage line component detection and segmentation module mainly uses image segmentation algorithm to detect and segment the drainage line. Based on its geometric features, the module selects a segmentation algorithm that is sensitive to the existing structure to accurately outline the continuous direction and complete contour of the drainage line, and at the same time generates its high-precision binary segmentation mask.

[0036] The main drainage line component detection and segmentation module segments the drainage line within the focal region. Considering the unique geometric shape of the drainage line in the focal region—its elongated, curved, and potentially varying width—a dedicated segmentation method highly sensitive to linear structures is employed, effectively combining the advantages of drainage line edge enhancement and image semantic segmentation. The main drainage line component detection and segmentation module uses a U-Net network based on Hessian matrix enhancement for drainage line segmentation, as detailed below: (41) Extract multi-scale depth features for the focal region in the pre-trained encoder network by introducing a linear structure based on the Hessian matrix for each pixel in the image. Calculate its Hessian matrix: Where H(x,y) represents a pixel. The Hessian matrix, This represents the second derivative of the image in the x-direction. This represents the second derivative of the image in the x and y directions. This represents the second derivative of the image in the y-direction. This represents the second derivative of the image in the y-direction and the x-direction.

[0037] (42) By calculating the eigenvalues ​​of matrix H(x,y), a corresponding linear similarity response function is constructed. : in, , Let be the eigenvalues ​​of the H matrix, k and c be constants for the control sensitivity, and exp be an exponential function.

[0038] (43) The function is placed in the pre-training to enhance the response of the linear region in the focal area, and the decoder outputs a complete contour of the guide line. The contour is represented by an ordered set of pixels in the focal area as follows: .

[0039] Simultaneously, the connectivity of the guide wire and related components is assessed. To avoid pixel-level errors caused by cutting, the tension clamps and suspension clamps of the cut guide wire and related components are expanded by two pixels, and the assessment is performed according to the following rules: (431) If the area of ​​the drainage line With tension clamp profile If there is an intersection, it is determined that the end is connected to the tension clamp, and the connection area is recorded.

[0040] (432) If the area of ​​the drainage line Contour with the hanging line If there is an intersection, it is determined that the end is connected to the suspension clamp, and the connection area is recorded.

[0041] (5) The associated component heat detection module mainly implements the heat detection of associated components, that is, it mainly implements the heat diagnosis of tension clamps and suspension clamps. The heat diagnosis of tension clamps is mainly determined by calling the tension clamp heat diagnosis module through the interface; the heat diagnosis of suspension clamps is mainly determined by calling the suspension clamp heat diagnosis module through the interface. Among them, the tension clamp heat diagnosis module and the suspension clamp heat diagnosis module are existing modules that have already been implemented, and they can be classified as part of the associated component heat detection module.

[0042] The associated component heating detection module enables precise heating diagnosis of the hardware at both ends of the drain wire, namely the tension clamp and the suspension clamp, thereby providing status input of the associated components for subsequent comprehensive judgment. Specifically: (51) The module will combine the original infrared image and the outline of the tension clamp. The predefined interface is used to call the tension clamp overheating diagnosis module. After the overheating diagnosis is completed, the overheating judgment result of the associated tension clamp component is returned. In this system, S represents the heating status indicator, where 0 indicates that the tension clamp is not heating and 1 indicates that the tension clamp is heating; L represents the heating part label, where 0 indicates that the tension tube is heating and 1 indicates that the drainage plate is heating. Since the tension clamp is composed of a tension tube and a drainage plate, if S indicates heating, then L indicates whether the heating is in the tension tube or the drainage plate.

[0043] (52) At the same time, the module will combine the original infrared image and the outline of the hanging wire clamp. The suspension clamp overheating diagnosis module is called through a predefined interface. After the overheating diagnosis is completed, the overheating judgment result of the associated suspension clamp is returned. .

[0044] The diagnostic thresholds are set according to the DL / T664-2016 standard.

[0045] (6) The drainage line heating detection module judges the heating of the drainage line part based on the temperature of the segmented drainage line area. The module extracts the temperature of the wire clamp contour area, judges its abnormal extreme point, and judges its heating point and heating location.

[0046] The drainage wire overheating detection module independently diagnoses the overheating status of each segmented drainage wire body and locates abnormal overheating points. Specifically: (61) The module outlines the drainage line. A skeletonization algorithm is applied to extract its topological skeleton lines, which are composed of a series of ordered pixels, denoted as a point set. Where n represents the total number of skeleton points, and the two endpoints of the drainage line are determined based on the skeleton line. and .

[0047] (62) Next, the module extracts temperature information and extracts each pixel on the skeleton line. coordinates ( , Using this as an index, the absolute temperature is directly read from the global temperature matrix T via the integrated DJI infrared SDK interface. Thus, the temperature sequence of the entire skeleton line is obtained. .

[0048] (63) After obtaining the temperature of the skeleton line, perform heat detection and positioning on the drainage line, and calculate the average temperature of all points on the skeleton line as the baseline reference temperature: .

[0049] (64) Then traverse the skeleton line temperature sequence to identify abnormal heating points. If a certain point The temperature satisfies: The point is then identified as a candidate point for abnormal heating. The algorithm also performs cluster analysis on consecutive abnormal heating points. If the number of consecutive heating points identified as abnormal is greater than 6, the area is determined to be a heating defect area of ​​the drainage line, and the coordinates of the center point of the heating area are recorded. : in,( , () represents the pixel in the heat-generating area.

[0050] (65) Finally output the overall heating status mark of the entire drainage line and the set of coordinates of the center of the identified heating area. .

[0051] (7) The comprehensive judgment module judges the heating status of the drainage line in this area by considering the heating status of the associated components and the drainage line, and determines whether it is not heating, sunlight emission, or a heating defect of the drainage line.

[0052] The comprehensive judgment module establishes a thermal correlation judgment logic model among multiple components to intelligently identify the heating properties of the drain wire, effectively distinguishing between actual heating and heating interference. It uses the heating diagnosis results of the associated components, namely the tension clamp and suspension clamp, as well as the set of heating areas on the drain wire body, and performs judgment according to the following process: (71) If none of the associated components are heated and the drain line is not heated, the drain line is determined to be in normal condition, and a no-heating mark is output, indicating that the drain line of this infrared image is normal.

[0053] (72) If none of the associated components are heating up, but there are heating points on the drain line, it is determined to be “pseudo-heating” caused by environmental factors such as sunlight reflection, and a heating interference mark is output.

[0054] (73) If the tension clamp heats up but only the tension tube heats up, while the suspension clamp and the drain line do not heat up, it is determined that the upstream is overheated but does not affect the drain line, and the drain line is in a state of no heat. In this case, since the tension clamp includes the tension tube and the drain plate, if the tension tube heats up, it is not the drain plate that heats up, so the drain line connected to it will not heat up.

[0055] (74) If the drain line heats up and a related component also heats up, the module calculates the heat at each heat point of the drain line. European distance to the nearest heat-generating component area : in, Indicates the heating area of ​​the drainage plate. This indicates the heating area of ​​the suspension clamp, and `distance` is the Euclidean distance between the heating point of the drainage wire and the heating point of the associated component. If... If the pixel value is less than 25 pixels, the heat source is considered a real heat source; otherwise, it is considered a reflection. When a real heat source exists, the final output is a heat defect marker on the drainage line and its corresponding heat source location coordinates.

[0056] Thus, a method for judging the heating defects of transmission line guide wires based on multi-device correlation was completed. Experimental verification shows that this invention, through multi-module collaborative processing, achieves a technological breakthrough from single-point temperature analysis to multi-device correlation judgment, effectively solving the problem of distinguishing reflected interference. Experimental results show that the method achieves a detection accuracy of 96.3%, reduces the false alarm rate by 62.5%, and significantly improves the intelligence level of transmission line status perception.

[0057] In summary, this method provides a diagnostic approach that breaks through the framework of single-point temperature analysis, intelligently analyzing the overall thermal imaging characteristics of the key conductive circuit—the tension clamp-drain plate-drain wire-suspension clamp—from the perspective of multi-component correlation. This method can effectively distinguish between actual heating and environmental interference such as emission. By mining and modeling the temperature correlation patterns and physical characteristics among multiple components, it achieves more accurate, reliable, and intelligent identification and early warning of heating defects in the drain wire, thereby improving the intelligence level of transmission line condition perception and fault early warning.

Claims

1. A method for judging heating defects in the lead wires of transmission lines with multiple interconnected components, characterized in that, The system includes a UAV infrared image acquisition and preprocessing module, a related component focal area detection and segmentation module, a related component detection and segmentation module, a main body drainage line component detection and segmentation module, a related component heat detection module, a drainage line heat detection module, and a comprehensive judgment module. The specific processing flow is as follows: (1) The UAV infrared image acquisition and preprocessing module interacts with the UAV infrared image library through the OBS communication interface, acquires infrared images through the communication interface, acquires raw infrared images through the standardized OBS, and filters infrared images that meet the requirements by parsing metadata. (2) The related component focus area detection and segmentation module detects and segments the related area of ​​the drain line in the image through image detection and segmentation algorithm, so as to locate and initially segment the core related area containing the key components of drain line, tension clamp and suspension clamp from the panoramic infrared image containing complex background; and adopts the semantic segmentation module based on deep learning to perform pixel-level classification on the input image and generate a connected binary mask that can cover the entire current path of "tension clamp-drain line-suspension clamp"; (3) The associated component detection and segmentation module uses image algorithms to segment and obtain the tension clamp and suspension clamp associated with the drainage line, completes the target detection of the output bounding box through the algorithm, and outlines the tension clamp and suspension clamp respectively through the output binary segmentation mask. (4) The main drainage line component detection and segmentation module performs target detection and segmentation of the drainage line using an image segmentation algorithm, based on its geometric features. Select a segmentation algorithm that is sensitive to the existing structure to outline the continuous direction and complete contour of the drainage line, and generate its binary segmentation mask at the same time. (5) The associated component heating detection module includes the heating diagnosis of tension clamp and suspension clamp. The heating diagnosis of tension clamp is determined by calling the heating diagnosis module of tension clamp through the interface; the heating diagnosis of suspension clamp is determined by calling the heating diagnosis module of suspension clamp through the interface. (6) The drainage line heating detection module judges the heating of the drainage line part based on the temperature of the segmented drainage line area. By extracting the temperature of the wire clamp contour area, it judges the extreme point of the abnormality and the heating point and its heating location. (7) The comprehensive judgment module judges the heating status of the drainage line in this area by considering the heating status of the associated components and the drainage line, and determines whether it is not heating, sunlight emission, or a heating defect of the drainage line.

2. The method for judging the heating defect of the lead wire of a multi-device associated transmission line according to claim 1, characterized in that, in In step (1), the UAV infrared image acquisition and preprocessing module filters out qualified infrared images containing the diversion line equipment from the UAV inspection infrared images, as follows: (11) Data interaction with the UAV inspection platform is achieved through the Object Storage Service (OBS) communication interface to realize the batch acquisition of raw infrared images; (12) After acquiring the infrared image, a filtering strategy based on the image file name is executed. The key is that... Currently, in power line inspection operations, the file names of infrared images follow naming conventions and embed key metadata such as line, tower, shooting location, and target hardware. (13) By building a keyword library related to traffic generation, and constructing filename parsing and matching rules based on regular expressions, a filename matching degree function is defined. To quantify screening accuracy: Where 'name' represents the infrared image file name. This represents the Kth keyword. For the corresponding weighting coefficients, This indicates an indicator function, where 1 is assigned when the filename contains the keyword, and 0 otherwise; the matching score is... During matching, the image is determined to be a target image related to the drainage line; (14) For the target images that pass the screening, save them in a local unified path.

3. The method for judging the heating defect of the lead wire of a multi-device associated transmission line according to claim 1, characterized in that, in In step (2), the associated component focus area detection and segmentation module initially locates and segments the core area covering the key conductive circuit of "tension clamp-drain wire-suspension clamp" from the panoramic infrared image containing complex scenes. It employs a deep learning-based target detection algorithm, trained on labeled infrared images of the transmission line containing the "tension clamp-drain wire-suspension clamp" components, to identify the overall distribution characteristics of the drain wire and its related components. The module outputs the smallest bounding rectangle region (ROI) containing all associated components, defined by a bounding box. The description uses a quartet to represent the position and size of the focal region in the image pixel coordinate system: in, The pixel coordinates of the top-left corner of the focal rectangle are given. This represents the pixel coordinates of the bottom right corner of the focus rectangle.

4. The method for judging heating defects in the lead wires of transmission lines with multiple interconnected devices according to claim 2, characterized in that, in In step (3), the associated component detection and segmentation module performs deep analysis on the focal region located by the associated component focal region detection and segmentation module, locates and segments two key associated components, namely tension clamp and suspension clamp, and obtains their pixel-level contours; the YOLO-based instance segmentation algorithm is used to extract features from the input focal region image to generate candidate regions, and the classification network determines whether the target category contained in each candidate region is tension clamp or suspension clamp. For the confirmed target, the segmentation network generates a binary segmentation mask and uses pixel coordinates to outline the complete contour of the component.

5. The method for judging the heating defect of the lead wire of a multi-device associated transmission line according to claim 4, characterized in that, in In step (3), the specific details are as follows: (31) For the identified tension clamp, output its set of contour points. : Where n is the total number of pixels that constitute the contour points of the tension clamp. The pixels that represent the outline; (32) For the identified suspension clamps, output their set of contour points. : Where m is the total number of pixels that make up the contour points of the drooping line clamp. The pixels that represent the outline; Through the processing of the associated component detection and segmentation module, the tension clamps and suspension clamps of the drainage line-related components are separated from the mixed focal area.

6. The method for judging heating defects in the lead wires of transmission lines with multiple interconnected devices as described in claim 5, characterized in that, in In step (4), the main body drainage line component detection and segmentation module cuts the drainage line in the focal region. A specialized segmentation method highly sensitive to linear structures is used to address the unique geometric shape of the drainage line in the focal region, as detailed below: (41) Extract multi-scale depth features for the focal region in the pre-trained encoder network by introducing a linear structure based on the Hessian matrix for each pixel in the image. Calculate its Hessian matrix: Where H(x,y) represents a pixel. The Hessian matrix, This represents the second derivative of the image in the x-direction. This represents the second derivative of the image in the x and y directions. This represents the second derivative of the image in the y-direction. This represents the second derivative of the graph in the y-direction and the x-direction; (42) By calculating the eigenvalues ​​of matrix H(x,y), a corresponding linear similarity response function is constructed. : in, , Let be the eigenvalues ​​of the H matrix, k and c be constants for the control sensitivity, and exp be an exponential function; (43) The function is put into pre-training to enhance the response of the linear region in the focal region. The decoder outputs a complete contour of the guide line, which is represented by the ordered set of pixels in the focal region as follows: At the same time, the connectivity of the drainage line and related components will be determined.

7. The method for judging heating defects in the lead wires of transmission lines with multiple interconnected devices as described in claim 6, characterized in that, in In step (43), to avoid pixel-level errors caused by cutting, the cut guide wire and the tension clamp and suspension clamp of the associated components are expanded by two pixels, and judged according to the following rules: (431) If the area of ​​the drainage line With tension clamp profile If there is an intersection, it is determined that the end is connected to the tension clamp, and the connection area is recorded. (432) If the area of ​​the drainage line Contour with the hanging line If there is an intersection, it is determined that the end is connected to the suspension clamp, and the connection area is recorded.

8. The method for judging heating defects in the lead wires of transmission lines with multiple interconnected devices according to claim 1, characterized in that, In step (5), the associated component heating detection module performs heating diagnosis on the hardware at both ends of the drain wire, namely the tension clamp and the suspension clamp, as follows: (51) Combine the original infrared image and the outline of the tension clamp. The predefined interface is used to call the tension clamp overheating diagnosis module. After the overheating diagnosis is completed, the overheating judgment result of the associated tension clamp component is returned. Where S is the heating status mark, 0 indicates that the tension clamp is not heating, and 1 indicates that the tension clamp is heating; L is the heating part label, 0 indicates that the tension tube is heating, and 1 indicates that the drainage plate is heating. (52) Combine the original infrared image and the outline of the hanging line clamp. The suspension clamp overheating diagnosis module is called through a predefined interface. After the overheating diagnosis is completed, the overheating judgment result of the associated suspension clamp is returned. 。 9. The method for judging heating defects in the lead wires of transmission lines with multiple interconnected devices according to claim 6, characterized in that, In step (6), the drainage wire heating detection module performs independent heating status diagnosis on the segmented drainage wire body and locates its abnormal heating points; specifically as follows: (61) Outline the drainage line A skeletonization algorithm is applied to extract its topological skeleton lines, which are composed of a series of ordered pixels, denoted as a point set. Where n represents the total number of skeleton points, and the two endpoints of the drainage line are determined based on the skeleton line. and ; (62) Extract temperature information and extract each pixel on the skeleton line. coordinates ( , Using this as an index, the absolute temperature can be directly read from the global temperature matrix T via the SDK interface. Thus, the temperature sequence of the entire skeleton line is obtained. ; (63) After obtaining the temperature of the skeleton line, perform heat detection and positioning on the drainage line, and calculate the average temperature of all points on the skeleton line as the baseline reference temperature: ； (64) Traverse the skeleton line temperature sequence to identify abnormal heating points; if a certain point The temperature satisfies: The point is then identified as a candidate point for abnormal heating. The algorithm also performs cluster analysis on consecutive abnormal heating points. If the number of consecutive heating points identified as abnormal is greater than 6, the area is determined to be a heating defect area of ​​the drainage line, and the coordinates of the center point of the heating area are recorded. : in,( , () represents the pixel in the heat-generating area; (65) Output the overall heating status marker of the entire drainage line and the set of coordinates of the center of the identified heating area. .

10. The method for judging heating defects in the lead wires of multi-device interconnected transmission lines according to claim 9, characterized in that, In step (7), the comprehensive judgment module establishes a thermal correlation judgment logic model among multiple devices to intelligently identify the heating properties of the drain wire, distinguishing between actual heating and heating interference; it uses the heating diagnosis results of the associated components, namely the tension clamp and the suspension clamp, and the set of heating areas of the drain wire body, and makes judgments according to the following process: (71) If none of the associated components are heated and the drain line is not heated, the drain line is determined to be in normal condition, and a no-heating mark is output, indicating that the drain line in this infrared image is normal. (72) If none of the associated components are heating up, but there are heating points on the drain line, it is determined to be "pseudo-heating" caused by environmental factors, and a heat interference mark is output; (73) If the tension clamp heats up but only the tension tube heats up, while the suspension clamp and the drain line do not heat up, it is determined that the upstream is overheated but does not affect the drain line, and the drain line is in a state of no heat. (74) If the drainage wire heats up and a related component also heats up, calculate the heat generation point of the drainage wire. European distance to the nearest heat-generating component area : in, Indicates the heating area of ​​the drainage plate. This indicates the heating area of ​​the suspension clamp, and distance is the Euclidean distance between the heating point of the drainage wire and the heating point of the associated component; if If the value is less than 25 pixels, the heat source is determined to be a real heat source; otherwise, it is determined to be a reflection. When a real heat source exists, the final output is a heat defect mark on the drain line and the corresponding coordinates of the heat source location.

Images