Power transmission line sag identification and early warning method and device combined with zoom ranging

Abstract

The application discloses a power transmission line sag identification and early warning method and device combined with zoom ranging, relates to the technical field of power transmission line monitoring, and comprises the following steps: installing a binocular camera system in a target identification scene, adjusting variable focus lens parameters, calibrating the binocular camera system by using a checkerboard, and constructing a disparity-depth mapping relationship; collecting Q times of synchronous images by using the calibrated system, performing multi-point disparity analysis, obtaining disparity analysis results of a first suspension point, a second suspension point and a sag lowest point, and screening and determining target suspension point and sag lowest point analysis results; taking the results as indexes, searching the disparity-depth mapping relationship, and obtaining depth information; calculating target suspension point ranging results according to the depth information, calculating a sag value and judging whether the sag value exceeds a threshold value, and if not, triggering early warning, so as to solve the technical problem that a traditional ranging mode cannot feed back sag changes in real time, and thus cannot effectively early warn safety hazards caused by excessively large sag.

Description

Technical Field

[0001] This application relates to the field of transmission line monitoring technology, specifically to a method and device for identifying and warning of transmission line sag by combining zoom distance measurement. Background Technology

[0002] In the operation of power transmission lines, sag has always been a significant factor affecting the efficiency and safety of power transmission. Sag refers to the phenomenon of power line conductors drooping under their own weight. Especially in long-distance, high-voltage transmission lines, excessive sag reduces the safe distance between the conductor and the ground or other objects, increasing the risk of short circuits, electric shocks, and other safety accidents. Furthermore, excessive sag can also cause mechanical damage to the transmission line, further affecting the stability and efficiency of power transmission. Traditional sag detection methods mostly rely on fixed-location monitoring equipment, which has limitations such as insufficient accuracy, poor real-time performance, and difficulty in covering the entire transmission line. Under complex environmental conditions, such as weather changes and complex line layouts, existing sag monitoring technologies cannot meet the needs for real-time and accurate monitoring, failing to issue timely risk warnings and posing potential safety hazards to the power system. Summary of the Invention

[0003] This application provides a method and device for identifying and warning of sag in transmission lines by combining zoom ranging. It solves the technical problem that traditional ranging methods cannot provide real-time feedback on sag changes, resulting in an inability to effectively warn of safety hazards caused by excessive sag. It achieves the technical effect of accurately and in real-time monitoring of sag changes by combining a binocular camera system with zoom ranging, thereby improving the safety and efficiency of transmission lines.

[0004] This application provides a method for identifying and warning of sag in transmission lines using zoom ranging. The method includes: installing a binocular camera system in a target recognition scenario and adjusting the parameters of the zoom lens in the binocular camera system; calibrating the focused binocular camera system using a checkerboard pattern to construct a binocular disparity-depth mapping relationship, wherein the binocular camera system is configured with a zoom lens; using the calibrated binocular camera system to perform Q simultaneous image acquisitions on the target recognition scenario, obtaining Q sets of simultaneous images, where Q is an integer greater than or equal to 1; traversing the Q sets of simultaneous images to perform multi-point disparity analysis, obtaining Q binocular disparity analysis results for first suspension points, Q binocular disparity analysis results for second suspension points, and Q binocular disparity analysis results for the lowest sag points; and centrally filtering the Q binocular disparity analysis results for the first suspension points, Q binocular disparity analysis results for the second suspension points, and the Q binocular disparity analysis results for the lowest sag points to determine... The binocular parallax analysis results of the first suspension point, the second suspension point, and the lowest point of the sag are used as indexes to retrieve the binocular parallax-depth mapping relationship, thereby obtaining the depth information of the first suspension point, the second suspension point, and the lowest point of the sag. Based on these depth information, the distance measurement results of the first suspension point, the second suspension point, and the lowest point of the sag are obtained. The sag value of the target recognition scene is calculated based on these results, and it is determined whether the sag value meets a preset sag threshold. If not, a first warning message is obtained.

[0005] This application also provides a transmission line sag identification and early warning device combined with zoom ranging, comprising: a binocular camera calibration module: installing a binocular camera system in a target recognition scenario, adjusting the parameters of the zoom lens in the binocular camera system, calibrating the focused binocular camera system using a checkerboard pattern, and constructing a binocular disparity-depth mapping relationship, wherein the binocular camera system is configured with a zoom lens; a synchronous image acquisition module: using the calibrated binocular camera system to perform Q synchronous image acquisitions on the target recognition scenario, obtaining Q sets of synchronous images, where Q is an integer greater than or equal to 1; and a disparity analysis and filtering module: traversing the Q sets of synchronous images to perform multi-point disparity analysis, obtaining Q binocular disparity analysis results for first suspension points, Q binocular disparity analysis results for second suspension points, and Q binocular disparity analysis results for the lowest sag points, and performing centralized filtering on the Q binocular disparity analysis results for the first suspension points, Q binocular disparity analysis results for the second suspension points, and the Q binocular disparity analysis results for the lowest sag points. The system comprises the following modules: a depth retrieval module, a depth search module, and a distance calculation module. The distance calculation module calculates the distance measurement results for the first, second, and lowest points of the target's sag based on these results. A warning module calculates the sag value of the target recognition scene based on these results, determines whether the sag value meets a preset sag threshold, and if not, obtains a first warning message.

[0006] The proposed method and device for sag identification and early warning of transmission lines, combining zoom ranging, includes a binocular camera calibration module. This module involves: installing a binocular camera system in a target recognition scenario and adjusting the parameters of the zoom lens within the system; calibrating the focused binocular camera system using a checkerboard pattern to construct a binocular disparity-depth mapping relationship; wherein the binocular camera system is equipped with a zoom lens; using the calibrated binocular camera system to perform Q simultaneous image acquisitions on the target recognition scenario, obtaining Q sets of simultaneous images, where Q is an integer greater than or equal to 1; traversing the Q sets of simultaneous images to perform multi-point disparity analysis, obtaining Q binocular disparity analysis results for first suspension points, Q binocular disparity analysis results for second suspension points, and Q binocular disparity analysis results for the lowest sag points; and centrally processing these results. The process involves filtering and determining the binocular parallax analysis results for the first suspension point, the second suspension point, and the lowest point of the sag. Using these results as indexes, the binocular parallax-depth mapping relationship is retrieved to obtain the depth information for the first, second, and lowest points of the sag. Based on these depth information, the distance measurement results for the first, second, and lowest points of the sag are obtained. Finally, the sag value of the target recognition scene is calculated based on these results. It is then determined whether the sag value meets a preset sag threshold; if not, a first warning message is obtained. This invention solves the technical problem that traditional ranging methods cannot provide real-time feedback on sag changes, thus failing to effectively warn of safety hazards caused by excessive sag. It achieves the technical effect of accurately and in real-time monitoring of sag changes through a binocular camera system combined with zoom ranging, thereby improving the safety and efficiency of power transmission lines. Attached Figure Description

[0007] To more clearly illustrate the technical solutions of the embodiments of this disclosure, the accompanying drawings of the embodiments of this disclosure will be briefly described below. Flowcharts are used in this application to illustrate the operations performed by the apparatus according to the embodiments of this application. It should be understood that the preceding or following operations are not necessarily performed precisely in sequence. Instead, various steps can be processed in reverse order or simultaneously as needed. Furthermore, other operations can be added to these processes, or one or more steps can be removed from these processes.

[0008] Figure 1This is a schematic flowchart of a transmission line sag identification and early warning method combined with zoom ranging provided in an embodiment of this application.

[0009] Figure 2 This is a schematic diagram of the structure of a transmission line sag identification and early warning device combined with zoom ranging provided in an embodiment of this application.

[0010] Figure labeling: 11. Binocular camera calibration module; 12. Synchronous image acquisition module; 13. Parallax analysis and filtering module; 14. Depth retrieval module; 15. Distance calculation module; 16. Early warning module. Detailed Implementation

[0011] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, the following are specific embodiments of this application.

[0012] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description of this application will be provided in conjunction with the accompanying drawings. The described embodiments should not be considered as limitations on this application. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0013] In the following description, references to "some embodiments" describe a subset of all possible embodiments. However, it is understood that "some embodiments" can be the same or different subsets of all possible embodiments and can be combined with each other without conflict. The terms "first" and "second" are used merely to distinguish similar objects and do not represent a specific ordering of objects. The terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, apparatus, product, or server that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or modules not explicitly listed or inherent to these processes, methods, products, or apparatuses. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only.

[0014] This application provides a method for identifying and warning of sag in transmission lines by combining zoom ranging, such as... Figure 1 As shown, the method includes:

[0015] In a target recognition scenario, a binocular camera system is installed, and the parameters of the zoom lens in the binocular camera system are adjusted. The binocular camera system after focusing is calibrated using a checkerboard pattern to construct a binocular parallax-depth mapping relationship. The binocular camera system is equipped with a zoom lens.

[0016] In this embodiment, in a target recognition scenario (power transmission lines within a certain area), a binocular camera system is first installed at a designated location to ensure it can cover the area to be monitored. The binocular camera system consists of two cameras that capture the same scene from different perspectives to obtain depth information. After installation, the zoom lens in the binocular camera system is adjusted; that is, the focal length is adjusted according to the actual distance and shooting requirements to ensure a clear image is obtained within the target distance range, avoiding blurring of distant and near objects that could affect the ranging effect. After focusing, the camera is calibrated using a checkerboard pattern, which is a widely used... The calibration tool uses a known grid pattern where the size and spacing of each square are known. The checkerboard pattern typically consists of black and white squares, and the number and size of the squares can be adjusted as needed. The checkerboard pattern is placed within the field of view of the binocular camera system, ensuring it is within the system's focal range. To obtain more perspectives, the checkerboard pattern can be repositioned multiple times within the system's field of view to provide sufficient image data for the calibration process. Subsequently, multiple checkerboard images are simultaneously captured using the binocular camera system. Each pair of images corresponds to the two camera perspectives of the system, capturing the same checkerboard pattern in each pair. Each image should ideally encompass the entire checkerboard pattern. The image contains multiple corner points of a checkerboard pattern, ensuring that the captured image fully reflects the three-dimensional structure of the object. Methods such as Harris corner detection and Shi-Tomasi corner detection are then used to process the captured image to detect the checkerboard corners. Taking Harris corner detection as an example, the Sobel operator is used to calculate the gradient of the image (the change along the X and Y directions), and then the covariance matrix of the gradient (structure tensor) is calculated. The response value (R value) of each pixel is calculated using this matrix, and the strength of the corner is determined by the magnitude of the R value. When the R value is greater than a threshold, the point corresponding to that R value is identified as a corner. Corners are the points where black and white squares intersect in the checkerboard pattern. After obtaining all corner points, these points are used to perform camera calibration using a calibration algorithm (such as the Zhang Zhengyou calibration method). The relationship between the corner points in each image and their actual 3D coordinates is calculated using a camera projection model (i.e., the existing pinhole camera model). Then, the camera's intrinsic and extrinsic parameters are optimized using the least squares method. Intrinsic parameters include focal length, principal point position, and distortion coefficients, while extrinsic parameters include the camera's rotation matrix and translation vector, which describe the relationship between the camera coordinate system and the world coordinate system. After calibration, the calibrated focal length is used, combined with the baseline distance of the binocular camera system (i.e., the distance between the two cameras), to construct the disparity-depth mapping relationship of the binocular cameras. Where Z is the depth value, f is the focal length, B is the baseline distance, and d is the disparity value. Through this binocular disparity-depth mapping relationship, the depth information of different objects can be correctly calculated, ensuring that subsequent target recognition and measurement tasks can be completed accurately.

[0017] The calibrated binocular camera system is used to acquire Q synchronous images of the target recognition scene, resulting in Q sets of synchronous images, where Q is an integer greater than or equal to 1.

[0018] In one embodiment, after the binocular camera system is calibrated, it is used to acquire multiple images of the target recognition scene. During this process, the binocular camera system simultaneously captures two images of the target scene (left and right images), ensuring that these two images are acquired at the same time, i.e., synchronized images. This ensures that the spatial position and depth information of each scene element in the image are available at the same time. In this step, Q image acquisitions are performed, where Q represents the number of acquisitions, and the value of Q can be any integer greater than or equal to 1. Each acquisition yields a set of synchronized images (i.e., left and right images), forming Q sets of synchronized images. Through these synchronized images, the binocular camera system can capture multiple perspectives of the target scene, thereby providing more data for subsequent image processing and analysis. These image sets will serve as the basic data for subsequent depth calculation, disparity analysis, and target recognition.

[0019] Multi-point disparity analysis is performed on the Q sets of synchronized images to obtain Q binocular disparity analysis results for the first suspension point, Q binocular disparity analysis results for the second suspension point, and Q binocular disparity analysis results for the lowest point of the sag. The binocular disparity analysis results for the first suspension point, Q binocular disparity analysis results for the second suspension point, and the binocular disparity analysis results for the lowest point of the sag are then centrally filtered to determine the binocular disparity analysis results for the target's first suspension point, target's second suspension point, and target's lowest point of the sag.

[0020] In one embodiment, during the process of traversing Q sets of synchronized images and performing multi-point parallax analysis, the main objective is to analyze the parallax of multiple key points of the target scene from each pair of synchronized images. Specifically, for each pair of synchronized images (left and right images), parallax analysis is performed, focusing on three key points: the first suspension point, the second suspension point, and the lowest point of sag. The first suspension point refers to the suspension point in the transmission line that is closest to the power tower (or support). This point is usually the starting point of the line, the connection point between the upper end of the suspension wire and the tower or support. The second suspension point refers to a suspension point in the transmission line that is relatively far from the first suspension point, usually located at the next tower or support. The lowest sag point refers to the lowest point of the bend in the cable caused by gravity in the transmission line, usually located in the middle between two support towers or towers. By using steps such as nearest neighbor windows, cost matching, and difference calculation, the disparity of each key point in these synchronized image sets in the left and right images is calculated, thus obtaining Q binocular disparity analysis results for the first suspension points, Q binocular disparity analysis results for the second suspension points, and Q binocular disparity analysis results for the lowest sag points. The larger the disparity value, the closer the point is to the camera, and vice versa. After obtaining the Q binocular disparity analysis results for the first suspension points, Q binocular disparity analysis results for the second suspension points, and Q binocular disparity analysis results for the lowest sag points, in order to improve the accuracy of the analysis results... The system constructs an initial neighborhood using the mean of each analysis result as the starting point, and then builds an iterative neighborhood through iteration. In each iteration, the density of the iterative neighborhood is compared with that of the initial neighborhood. Based on the judgment result, subsequent steps are executed. For example, if the neighborhood density of the iterative neighborhood is greater than or equal to that of the initial neighborhood, a first suspension point binocular parallax analysis result is randomly extracted from the edge of the iterative neighborhood as an update iteration point, and centralized screening is performed again until the preset number of iterations is met. After completing the centralized screening of Q first suspension point binocular parallax analysis results, Q second suspension point binocular parallax analysis results, and Q sag lowest point binocular parallax analysis results, the target first suspension point binocular parallax analysis results, target second suspension point binocular parallax analysis results, and target sag lowest point binocular parallax analysis results are obtained. These final target parallax analysis results reflect the actual parallax information of each key point, and can more accurately reflect the depth information and positional changes of each key point in the target scene.

[0021] Furthermore, this application provides a method for performing multi-point disparity analysis by traversing the Q sets of synchronized images to obtain Q binocular disparity analysis results for first suspension points, Q binocular disparity analysis results for second suspension points, and Q binocular disparity analysis results for the lowest points of sag, including:

[0022] Traverse the Q sets of synchronized images and perform grayscale preprocessing to obtain Q sets of preprocessed synchronized images; extract Q left-side preprocessed images from the Q sets of preprocessed synchronized images and identify the first suspension point, the second suspension point, and the lowest point of the sag to obtain Q left-side first suspension point positions, Q left-side second suspension point positions, and Q left-side lowest sag point positions; based on the Q left-side first suspension point positions, Q left-side second suspension point positions, and Q left-side lowest sag point positions, perform nearest neighbor window identification in the Q left-side preprocessed images to obtain Q left-side first suspension point nearest neighbor windows, Q left-side second suspension point nearest neighbor windows, and Q left-side lowest sag point nearest neighbor windows; based on the Q left-side first suspension point nearest neighbor windows, Q left-side second suspension point nearest neighbor windows, and Q left-side lowest sag point nearest neighbor windows... For each point, a nearest neighbor window is used to perform cost matching with the Q right-side preprocessed images within the set of Q preprocessed synchronized images, resulting in Q right-side matching first hanging point nearest neighbor windows, Q right-side matching second hanging point nearest neighbor windows, and Q right-side matching sag lowest point nearest neighbor windows. For each of the Q left-side first hanging point nearest neighbor windows, Q left-side second hanging point nearest neighbor windows, and Q left-side sag lowest point nearest neighbor windows, a one-to-one pixel value difference calculation is performed between the corresponding Q right-side matching first hanging point nearest neighbor windows, Q right-side matching second hanging point nearest neighbor windows, and Q right-side matching sag lowest point nearest neighbor windows. The absolute value of the calculation result is used as the binocular disparity analysis result for the Q first hanging point points, the Q second hanging point points, and the Q sag lowest point points.

[0023] Preferably, in performing multi-point parallax analysis, each image in the Q synchronized image set is first preprocessed for grayscale. Grayscale conversion transforms a color image into a grayscale image to reduce computational complexity. A color image typically consists of three color channels: red (R), green (G), and blue (B). The color value of each pixel is determined by the intensity values ​​of these three colors (i.e., RGB values). During grayscale conversion, a weighted average is usually used to convert the RGB values ​​into a single brightness value. A commonly used weighting formula is... Where Y is the grayscale value, and R, B, and G are the values ​​of the red, green, and blue channels of each pixel in the original image, respectively. Through preprocessing grayscale conversion, Q sets of preprocessed synchronized images are obtained. Subsequently, Q left-side preprocessed images are extracted from these Q sets. Feature extraction algorithms (such as SIFT, SURF, or Harris corner detection) are used to extract salient features from the images, such as corners, edges, and textures. These features are then matched with feature templates (such as standard features of hanging points or the lowest point of sag, pre-set based on historical experience and expert decisions) to identify the first hanging point, the second hanging point, and the lowest point of sag. The positions of these key points are recorded, forming Q left-side first hanging point positions, Q left-side second hanging point positions, and Q left-side lowest point of sag positions. Afterwards, in each left-side image... A nearest neighbor window (e.g., a 5x5 area) is defined for the first suspension point, the second suspension point, and the lowest point of the sag. The size of this window is determined based on the required business accuracy to include sufficient surrounding pixel information for more accurate keypoint matching. For each keypoint (first suspension point, second suspension point, and lowest point of sag), the corresponding nearest neighbor window region is extracted from the left preprocessed image, forming Q left first suspension point nearest neighbor windows, Q left second suspension point nearest neighbor windows, and Q left lowest sag point nearest neighbor windows. The pixel values ​​within these regions are used for comparison with the right image. Then, each of the Q nearest neighbor windows in the left preprocessed image is subjected to cost matching with the corresponding regions in the Q right preprocessed images. Cost matching typically uses a metric to evaluate the similarity between the two windows; common metrics include SSD (Sum of Spanning Scale). The process of calculating the disparity (or squared difference) or NCC (Normalized Cross-Correlation) involves comparing the pixel value differences between the left and right window regions and selecting the best-matching right window region. This results in Q right-matching windows that are the nearest neighbors of the first dangling point, Q right-matching windows that are the nearest neighbors of the second dangling point, and Q right-matching windows that are the nearest neighbors of the lowest point of the sag. Further, for each pair of matched windows (left and right neighbor windows), the pixel value differences between them are calculated. This difference is typically calculated by comparing the pixel values ​​of the two windows pixel by pixel and calculating their absolute differences. The difference for each pair of pixels in the window is calculated, and the total difference for each window is taken as the disparity value for that point. The specific calculation formula is as follows: ;in, and These are the positions in the left and right windows, respectively. The pixel values ​​are given, and D is the disparity metric. Finally, by obtaining the disparity corresponding to each keypoint (first suspension point, second suspension point, and lowest sag point) through the above process, we can obtain Q binocular disparity analysis results for the first suspension points, Q binocular disparity analysis results for the second suspension points, and Q binocular disparity analysis results for the lowest sag points. In summary, by performing nearest-neighbor window recognition on the left image in the Q synchronized image set and cost matching with the right image, the binocular disparity analysis results for each keypoint can be effectively calculated. Using nearest-neighbor windows can enhance the accuracy of matching, reduce the impact of noise, and improve the reliability of disparity analysis results, laying the foundation for subsequent depth recognition and sag value calculation.

[0024] Furthermore, this application provides cost matching based on the Q left first suspension point nearest neighbor windows, Q left second suspension point nearest neighbor windows, and Q left sag lowest point nearest neighbor windows, with the Q right preprocessed images in the Q preprocessed synchronized image sets, to obtain Q right matching first suspension point nearest neighbor windows, Q right matching second suspension point nearest neighbor windows, and Q right matching sag lowest point nearest neighbor windows, including:

[0025] The matching cost analysis is performed between the Q left-side first suspension point nearest neighbor windows and the Q right-side first matching windows in the Q right-side preprocessed images to obtain Q right-side first matching costs. Based on the magnitude of the Q right-side first matching costs, the Q right-side first matching windows are iterated to obtain Q right-side second matching windows. The matching cost analysis is then performed between the Q left-side first suspension point nearest neighbor windows and the Q right-side second matching windows to obtain Q right-side second matching costs. Based on the magnitude of the Q right-side second matching costs, the Q right-side second matching windows are iterated until all Q right-side preprocessed images have been iterated. The matching window corresponding to the minimum matching cost during the iteration process is taken as the Q right-side matching first suspension point nearest neighbor windows. Based on the Q left-side second suspension point nearest neighbor windows and the Q left-side sag lowest point nearest neighbor windows, cost matching is performed with the Q right-side preprocessed images to obtain Q right-side matching second suspension point nearest neighbor windows and Q right-side matching sag lowest point nearest neighbor windows.

[0026] Optionally, for each pair of windows at corresponding positions in the left and right images, a matching cost analysis is first performed. Taking the first dangling point as an example, Q neighboring windows of the left first dangling point are matched with Q right first matching windows (randomly selected regions of the same size within the same vertical range as the neighboring windows of the left first dangling point) in the Q preprocessed right images. The matching cost is usually evaluated by calculating the similarity between windows (e.g., SSD or NCC). Taking SSD as an example, the calculation formula is: The goal of this step is to calculate the matching cost of each pair of windows, thus obtaining Q right-side first matching costs, reflecting the quality of each pair of window matches. Then, by analyzing the magnitude of the Q right-side first matching costs, the optimal matching window is selected. Window with a smaller cost indicates a better match, so the Q right-side first matching windows are moved with a smaller step size to obtain Q right-side second matching windows. Conversely, a larger step size is used to move the Q right-side first matching windows to find the best matching window. Afterwards, the Q right-side second matching windows, after iterations, are again subjected to cost matching analysis with the nearest neighbor windows of the left first dangling point, using the same calculation method, to obtain Q right-side second matching costs. Based on the magnitude of the Q right-side second matching costs, the selection of the Q right-side second matching windows is further optimized to obtain Q right-side third matching windows. The cost matching analysis is performed iteratively until the matching windows of the right-side image have undergone a preset number of iterations to ensure that the selected right-side first sag point nearest neighbor windows have the optimal matching degree. Through iterative calculation, Q right-side matching first sag point nearest neighbor windows, i.e., windows with the minimum matching cost, are finally obtained as the best matching result for each left-side first sag point. Then, the same method is used to match the second sag point nearest neighbor windows and the sag lowest point nearest neighbor windows. First, the matching cost of Q left-side second sag point nearest neighbor windows and Q left-side sag lowest point nearest neighbor windows with Q right-side preprocessed images is calculated. Then, through the iterative process, the Q right-side matching second sag point nearest neighbor windows and Q right-side matching sag lowest point nearest neighbor windows that best match the left-side windows are found, thus providing accurate matching data for subsequent disparity analysis, depth calculation, and sag value calculation.

[0027] Furthermore, this application provides a method for iterating through the Q right-side first matching windows based on the magnitude of the Q right-side first matching costs to obtain Q right-side second matching windows, including:

[0028] Determine whether the cost of the Q right-side first matching windows is less than or equal to a preset cost. If yes, move the Q right-side first matching windows by a first moving step size to obtain Q right-side second matching windows. If no, move the Q right-side first matching windows by a second moving step size to obtain Q right-side second matching windows, wherein the second moving step size is greater than the first moving step size.

[0029] Optionally, during the initial matching window iteration, the calculated costs of the Q right-side first matching windows are compared with a preset cost value (determined based on business requirements and expert decisions). If the cost of the right-side first matching window is less than or equal to the preset cost value, the matching quality is good, and further optimization can be performed using a first movement step. The first movement step is used to adjust the window position with a small movement range to refine the matching results and make the window position more accurate. The positions of the Q right-side first matching windows are adjusted according to this first movement step to obtain Q right-side second matching windows. If the cost of the right-side first matching window is greater than the preset cost value, the matching quality is not ideal and there may be a large matching error. In this case, a second movement step is needed to correct the matching results more significantly. This second movement step is used to move the window over a larger range to quickly find a better matching region. The window position is adjusted using this second movement step to obtain Q right-side second matching windows. By penalizing possible erroneous matches using the above method based on the first and second movement step, the matching accuracy can be improved, ensuring the accuracy of subsequent calculations.

[0030] Furthermore, this application provides and centrally filters the binocular parallax analysis results of the Q first suspension points, the Q second suspension points, and the Q lowest sag points to determine the binocular parallax analysis results of the target first suspension points, the target second suspension points, and the target lowest sag points, including:

[0031] Calculate the mean of the binocular disparity analysis results of the Q first suspension points to obtain the mean of the binocular disparity analysis results of the first suspension points; using the mean of the binocular disparity analysis results of the first suspension points as the starting point, construct a neighborhood in the binocular disparity analysis results of the Q first suspension points according to a preset neighborhood radius to obtain an initial neighborhood; randomly extract a binocular disparity analysis result of the first suspension point from the edge of the initial neighborhood as an iteration point, and construct a neighborhood in the binocular disparity analysis results of the Q first suspension points based on the preset neighborhood radius to obtain an iterative neighborhood; determine the neighborhood density of the iterative neighborhood. If the degree is greater than or equal to the initial neighborhood, then a first suspension point binocular parallax analysis result is randomly extracted from the edge of the iterative neighborhood as an update iteration point. Based on the update iteration point, centralized screening continues until the preset number of iterations is met. The update iteration point obtained in the last iteration is taken as the target first suspension point binocular parallax analysis result. The Q second suspension point binocular parallax analysis results and the Q sag lowest point binocular parallax analysis results are centrally screened to determine the target second suspension point binocular parallax analysis result and the target sag lowest point binocular parallax analysis result.

[0032] Preferably, when performing centralized screening of the binocular disparity analysis results of the Q first suspension points, the Q second suspension points, and the Q lowest sag points, taking the binocular disparity analysis results of the first suspension points as an example, firstly, the mean of the Q first suspension point binocular disparity analysis results is calculated. These disparity analysis results represent the disparity values ​​of the first suspension points in different sets of synchronized images. Calculating their mean yields a representative disparity value. Then, using the mean of the first suspension point binocular disparity analysis results as the starting point, a value matching the mean is found among the Q first suspension point binocular disparity analysis results according to a preset neighborhood radius (e.g., a specific disparity range). Similar disparity values ​​(i.e., the difference is within this radius) are selected, and the set of points corresponding to these disparity values ​​is used as the initial neighborhood. Then, a first dangling point binocular disparity analysis result is randomly selected from the edge of the initial neighborhood (i.e., at the maximum or minimum value in the neighborhood) as the initial iteration point. This point will serve as the basis for further iterative optimization. Based on the selected iteration point, a new iterative neighborhood is constructed again from the Q first dangling point binocular disparity analysis results using a preset neighborhood radius. This new neighborhood contains points with disparity values ​​similar to the iteration point. Finally, the neighborhood density of the initial neighborhood is calculated by ratioing the number of points in the initial neighborhood and the iterative neighborhood to the neighborhood area. The neighborhood density of the iterative neighborhood is considered; a higher density indicates greater consistency in disparity values ​​around the iteration point. This is achieved by comparing the neighborhood density of the initial neighborhood with that of the iterative neighborhood. If the density of the iterative neighborhood is greater than or equal to that of the initial neighborhood, a new binocular disparity analysis result is randomly extracted from the edge of the iterative neighborhood as the updated iteration point. Neighborhood construction and selection continue based on this updated iteration point. This process iteratively optimizes the selection of iteration points to ensure they are always within regions of higher neighborhood density, ultimately finding the most representative matching point. The process of matching disparity values ​​continues until the preset maximum number of iterations is reached, typically 10 or 20. After the iterations are completed, the disparity analysis result obtained from the last iteration is selected as the target first suspension point binocular disparity analysis result. For the second suspension point binocular disparity analysis result and the lowest sag point binocular disparity analysis result, a similar centralized screening process is adopted. By traversing Q analysis results, the most representative disparity value is selected as the target result. The screening of these two points (the second suspension point and the lowest sag point) can be optimized by calculating their mean and using neighborhood density to ensure the selection of the most accurate matching result.In summary, this process involves constructing neighborhoods, iteratively optimizing, and comparing densities based on the binocular disparity analysis results of the first suspension point, the second suspension point, and the lowest point of the sag. Through multiple rounds of iteration, the disparity value that best meets expectations is selected. This method ensures that, in the presence of multiple disparity solutions, the final selected disparity value has the highest accuracy and representativeness.

[0033] Furthermore, this application provides a method for determining whether the neighborhood density of the iterative neighborhood is greater than or equal to that of the initial neighborhood, including:

[0034] If not, a first random value and a second random value are obtained based on a random number generator. When the first random value is greater than or equal to the second random value, the starting point is taken as the binocular parallax analysis result of the target first suspension point. When the first random value is less than the second random value, a binocular parallax analysis result of the first suspension point is randomly extracted from the edge of the iterative neighborhood as an update iteration point. Based on the update iteration point, centralized screening continues until the preset number of iterations is met. The update iteration point obtained in the last iteration is taken as the binocular parallax analysis result of the target first suspension point.

[0035] Optionally, if in a certain iteration the neighborhood density of the iterative neighborhood is less than the neighborhood density of the starting neighborhood, the iterative optimization is not continued. Instead, the update method for the iteration point is determined based on a random number generator to ensure the randomness of the update. Two random values ​​are generated by the random number generator: a first random value and a second random value. This random number generator can use a standard random number algorithm (e.g., uniform distribution) to ensure that each generated random value is independent and has sufficient randomness. Subsequently, these two random values ​​are compared. If the first random value is greater than or equal to the second random value, the starting point is considered as the result of the binocular parallax analysis of the target first dangling point, i.e., no further iteration is needed, and the iteration process ends. Conversely, when the first random value is less than the second random value, the update is performed from the iterative neighborhood. A new binocular disparity analysis result for the first dangling point is randomly extracted from the edge (i.e., the extreme point within the neighborhood) as an update iteration point. This allows for continued searching for possible optimal matching results within the neighborhood through random selection. Based on the determined update iteration point, a centralized screening process continues. This means that new iteration points are used to further screen the Q binocular disparity analysis results for the first dangling point, further optimizing the matching results through neighborhood construction. This process can be repeated until the set maximum number of iterations is reached. When the preset number of iterations is met, iteration stops, and the update iteration point obtained from the last iteration is selected as the target binocular disparity analysis result for the first dangling point. The final target result is the optimal matching disparity value obtained through a series of iterative processes, random updates, and neighborhood screening. In summary, when the neighborhood density does not meet expectations, a random number generator is used to decide whether to use the current starting point as the target or to randomly extract a new iteration point from the neighborhood for further optimization. This approach avoids local optima when the neighborhood density is low, ensuring that a more suitable target binocular disparity analysis result for the first dangling point can be obtained through more iterations.

[0036] Based on the depth information of the first suspension point, the second suspension point, and the lowest point of the sag of the target, the distance measurement results of the first suspension point, the second suspension point, and the lowest point of the sag of the target are obtained.

[0037] In one embodiment, after obtaining the depth information of the first suspension point, the second suspension point, and the lowest point of the sag of the target, the angle between the Z-axis of the camera coordinate system of the binocular camera system and the ground is obtained, and expressed by the formula: Calculate the vertical distance corresponding to each key point, and use these vertical distances as the ranging results for the first suspension point, the second suspension point, and the lowest point of the sag, respectively. Here, d represents the vertical distance, and Z represents the depth information. The angle between the Z-axis of the camera coordinate system and the ground is used as the basis for subsequent sag calculations to ensure the accuracy of the early warning.

[0038] Based on the distance measurement results of the first suspension point of the target, the distance measurement results of the second suspension point of the target, and the distance measurement results of the lowest point of the target sag, the sag value of the target recognition scene is calculated, and it is determined whether the sag value meets the preset sag threshold. If not, the first warning information is obtained.

[0039] In one embodiment, after obtaining the distance measurement results of the first suspension point, the second suspension point, and the lowest point of the sag, the height difference between the first suspension point and the lowest point of the sag is calculated by subtracting the distance measurement results of the first suspension point and the lowest point of the sag. Similarly, the height difference between the second suspension point and the lowest point of the sag is calculated by subtracting the distance measurement results of the second suspension point and the lowest point of the sag. These two height differences are then added together, and the sum is compared with the horizontal distance between the suspension points to obtain the sag value in the target recognition scenario. After calculating the sag value, it is compared with a preset sag threshold, which is typically based on... According to engineering design standards, operating conditions, and safety requirements, if the calculated sag value is less than or equal to the preset threshold, it means that the cable sag is within an acceptable range, and the safety and stability of the power transmission line are guaranteed. If the calculated sag value is greater than the preset threshold, it means that the cable sag is excessive, which may affect the power transmission efficiency and even lead to damage to power equipment or safety accidents. At this time, a first warning message will be generated to remind relevant personnel to take necessary measures (such as adjusting cable tension, inspecting equipment, etc.) to prevent potential safety hazards. The first warning message may include the sag value, warning level (compared with the warning level based on the sag value), etc. The specific content is set according to the actual application scenario.

[0040] Furthermore, this application also includes:

[0041] The sag value is continuously monitored within a preset monitoring window to obtain a set of continuously monitored sag values; when the fluctuation variance of the continuously monitored sag value set is greater than or equal to a preset fluctuation variance, a second early warning message is obtained.

[0042] Preferably, when continuously monitoring the sag value of power lines, it is necessary to calculate the sag value of the cable within a preset monitoring window, and determine the fluctuation variance based on this data to ultimately decide whether to issue a second warning message. Specifically, firstly, within the preset monitoring window, the sag value of the cable is monitored periodically, for example, every 10 minutes or hour. Each measured sag value is stored, forming a continuously monitored sag value set, which contains all sag value data measured within a certain period. Subsequently, the variance of this continuously monitored sag value set is calculated. Variance is a statistical measure of the degree of fluctuation of a set of data. By comparing the calculated fluctuation variance with a preset fluctuation variance threshold, when the fluctuation variance of the calculated continuously monitored sag value set is greater than or equal to the preset fluctuation variance, it indicates abnormal sag value fluctuation, which may indicate cable slack or other abnormal conditions. At this time, a second warning message is generated to remind relevant personnel that the cable may have abnormal risks, such as excessive or insufficient cable tension. This process helps to monitor the safety of power transmission lines in real time, detect cable abnormalities in a timely manner, and prevent accidents.

[0043] In the above text, refer to Figure 1 A method for identifying and warning of sag in transmission lines incorporating zoom ranging according to an embodiment of the present invention is described in detail. Next, reference will be made to... Figure 2 This invention describes a transmission line sag identification and early warning device incorporating zoom ranging according to an embodiment of the present invention.

[0044] The transmission line sag identification and early warning device combining zoom ranging according to embodiments of the present invention solves the technical problem that traditional ranging methods cannot provide real-time feedback on sag changes, resulting in an inability to effectively warn of safety hazards caused by excessive sag. It achieves the technical effect of accurately and in real-time monitoring of sag changes through a binocular camera system combined with zoom ranging, thereby improving the safety and efficiency of transmission lines. The transmission line sag identification and early warning device combining zoom ranging includes: a binocular camera calibration module 11, a synchronous image acquisition module 12, a parallax analysis and filtering module 13, a depth retrieval module 14, a distance calculation module 15, and an early warning module 16.

[0045] Binocular camera calibration module 11: Installs a binocular camera system in a target recognition scenario and adjusts the parameters of the zoom lens in the binocular camera system. Calibrates the focused binocular camera system using a checkerboard pattern to construct a binocular disparity-depth mapping relationship. The binocular camera system is equipped with a zoom lens. Synchronous image acquisition module 12: Uses the calibrated binocular camera system to perform Q synchronous image acquisitions on the target recognition scenario, obtaining Q sets of synchronous images, where Q is an integer greater than or equal to 1. Disparity analysis and filtering module 13: Traverses the Q sets of synchronous images to perform multi-point disparity analysis, obtaining Q binocular disparity analysis results for the first suspension point, Q binocular disparity analysis results for the second suspension point, and Q binocular disparity analysis results for the lowest sag point. These results are then centrally filtered to determine the binocular disparity score of the target's first suspension point. The system analyzes the results of binocular parallax analysis of the target's second suspension point and the target's lowest sag point; the depth retrieval module 14 uses the binocular parallax analysis results of the target's first suspension point, the target's second suspension point, and the target's lowest sag point as indexes to retrieve the binocular parallax-depth mapping relationship and obtain the target's first suspension point depth information, target's second suspension point depth information, and target's lowest sag point depth information; the distance calculation module 15 obtains the target's first suspension point distance measurement result, target's second suspension point distance measurement result, and target's lowest sag point distance measurement result based on the target's first suspension point depth information, target's second suspension point depth information, and target's lowest sag point depth information; the warning module 16 calculates the sag value of the target recognition scene based on the target's first suspension point distance measurement result, target's second suspension point distance measurement result, and target's lowest sag point distance measurement result, and determines whether the sag value meets a preset sag threshold. If not, it obtains a first warning message.

[0046] Furthermore, the disparity analysis screening module 13 also includes:

[0047] Traverse the Q sets of synchronized images and perform grayscale preprocessing to obtain Q sets of preprocessed synchronized images; extract Q left-side preprocessed images from the Q sets of preprocessed synchronized images and identify the first suspension point, the second suspension point, and the lowest point of the sag to obtain Q left-side first suspension point positions, Q left-side second suspension point positions, and Q left-side lowest sag point positions; based on the Q left-side first suspension point positions, Q left-side second suspension point positions, and Q left-side lowest sag point positions, perform nearest neighbor window identification in the Q left-side preprocessed images to obtain Q left-side first suspension point nearest neighbor windows, Q left-side second suspension point nearest neighbor windows, and Q left-side lowest sag point nearest neighbor windows; based on the Q left-side first suspension point nearest neighbor windows, Q left-side second suspension point nearest neighbor windows, and Q left-side lowest sag point nearest neighbor windows... For each of the Q preprocessed images in the set of Q preprocessed synchronized images, a cost match is performed between the nearest neighbor window and the Q right-side preprocessed images to obtain Q right-side matching first hanging point nearest neighbor windows, Q right-side matching second hanging point nearest neighbor windows, and Q right-side matching sag lowest point nearest neighbor windows. Then, for each of the Q left-side first hanging point nearest neighbor windows, Q left-side second hanging point nearest neighbor windows, and Q left-side sag lowest point nearest neighbor windows, a one-to-one pixel value difference calculation is performed between the corresponding Q right-side matching first hanging point nearest neighbor windows, Q right-side matching second hanging point nearest neighbor windows, and Q right-side matching sag lowest point nearest neighbor windows. The absolute value of the calculation result is used as the binocular disparity analysis result for the Q first hanging point, the Q second hanging point, and the Q sag lowest point.

[0048] Furthermore, the disparity analysis screening module 13 also includes:

[0049] The matching cost analysis is performed between the Q left-side first suspension point nearest neighbor windows and the Q right-side first matching windows in the Q right-side preprocessed images to obtain Q right-side first matching costs. Based on the magnitude of the Q right-side first matching costs, the Q right-side first matching windows are iterated to obtain Q right-side second matching windows. The matching cost analysis is then performed between the Q left-side first suspension point nearest neighbor windows and the Q right-side second matching windows to obtain Q right-side second matching costs. Based on the magnitude of the Q right-side second matching costs, the Q right-side second matching windows are iterated until all Q right-side preprocessed images have been iterated. The matching window corresponding to the minimum matching cost during the iteration process is taken as the Q right-side matching first suspension point nearest neighbor windows. Based on the Q left-side second suspension point nearest neighbor windows and the Q left-side sag lowest point nearest neighbor windows, cost matching is performed with the Q right-side preprocessed images to obtain Q right-side matching second suspension point nearest neighbor windows and Q right-side matching sag lowest point nearest neighbor windows.

[0050] Furthermore, the disparity analysis screening module 13 also includes:

[0051] Determine whether the cost of the Q right-side first matching windows is less than or equal to a preset cost. If yes, move the Q right-side first matching windows by a first moving step size to obtain Q right-side second matching windows. If no, move the Q right-side first matching windows by a second moving step size to obtain Q right-side second matching windows, wherein the second moving step size is greater than the first moving step size.

[0052] Furthermore, the disparity analysis screening module 13 also includes:

[0053] Calculate the mean of the binocular disparity analysis results of the Q first suspension points to obtain the mean of the binocular disparity analysis results of the first suspension points; using the mean of the binocular disparity analysis results of the first suspension points as the starting point, construct a neighborhood in the binocular disparity analysis results of the Q first suspension points according to a preset neighborhood radius to obtain an initial neighborhood; randomly extract a binocular disparity analysis result of the first suspension point from the edge of the initial neighborhood as an iteration point, and construct a neighborhood in the binocular disparity analysis results of the Q first suspension points based on the preset neighborhood radius to obtain an iterative neighborhood; determine the neighborhood density of the iterative neighborhood. If the degree is greater than or equal to the initial neighborhood, then a first suspension point binocular parallax analysis result is randomly extracted from the edge of the iterative neighborhood as an update iteration point. Based on the update iteration point, centralized screening continues until the preset number of iterations is met. The update iteration point obtained in the last iteration is taken as the target first suspension point binocular parallax analysis result. The Q second suspension point binocular parallax analysis results and the Q sag lowest point binocular parallax analysis results are centrally screened to determine the target second suspension point binocular parallax analysis result and the target sag lowest point binocular parallax analysis result.

[0054] Furthermore, the disparity analysis screening module 13 also includes:

[0055] If not, a first random value and a second random value are obtained based on a random number generator. When the first random value is greater than or equal to the second random value, the starting point is taken as the binocular parallax analysis result of the target first suspension point. When the first random value is less than the second random value, a binocular parallax analysis result of the first suspension point is randomly extracted from the edge of the iterative neighborhood as an update iteration point. Based on the update iteration point, centralized screening continues until the preset number of iterations is met. The update iteration point obtained in the last iteration is taken as the binocular parallax analysis result of the target first suspension point.

[0056] Furthermore, the early warning module 16 also includes:

[0057] The sag value is continuously monitored within a preset monitoring window to obtain a set of continuously monitored sag values; when the fluctuation variance of the continuously monitored sag value set is greater than or equal to a preset fluctuation variance, a second early warning message is obtained.

[0058] The transmission line sag identification and early warning device combined with zoom ranging provided in this embodiment of the invention can execute the transmission line sag identification and early warning method combined with zoom ranging provided in any embodiment of the invention, and has the corresponding functional modules and beneficial effects of the method.

[0059] Although this application makes various references to certain modules in the apparatus according to the embodiments of this application, any number of different modules can be used and run on user terminals and / or servers. The various units and modules included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be achieved; in addition, the specific names of each functional unit are only for easy distinction between each other and are not intended to limit the scope of protection of this invention.

[0060] The specific embodiments described above do not constitute a limitation on the scope of protection of this application. Those skilled in the art should understand that various modifications, combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. Transmission line sag identification and early warning method combined with zoom ranging, characterized in that, The method includes: In a target recognition scenario, a binocular camera system is installed, and the parameters of the zoom lens in the binocular camera system are adjusted. The binocular camera system after focusing is calibrated using a checkerboard pattern to construct a binocular parallax-depth mapping relationship. The binocular camera system is equipped with a zoom lens. The calibrated binocular camera system is used to acquire Q synchronous images of the target recognition scene, resulting in Q sets of synchronous images, where Q is an integer greater than or equal to 1. Multi-point disparity analysis is performed on the Q sets of synchronized images to obtain Q binocular disparity analysis results for the first suspension point, Q binocular disparity analysis results for the second suspension point, and Q binocular disparity analysis results for the lowest point of the sag. The binocular disparity analysis results for the first suspension point, Q binocular disparity analysis results for the second suspension point, and Q binocular disparity analysis results for the lowest point of the sag are then centrally filtered to determine the binocular disparity analysis results for the target first suspension point, the target second suspension point, and the target lowest point of the sag. Using the binocular parallax analysis results of the first suspension point of the target, the binocular parallax analysis results of the second suspension point of the target, and the binocular parallax analysis results of the lowest point of the target sag as indexes, the binocular parallax-depth mapping relationship is retrieved to obtain the depth information of the first suspension point of the target, the depth information of the second suspension point of the target, and the depth information of the lowest point of the target sag. Based on the target first suspension point depth information, target second suspension point depth information and target sag lowest point depth information, obtain the target first suspension point distance measurement result, target second suspension point distance measurement result and target sag lowest point distance measurement result; Based on the distance measurement results of the first suspension point of the target, the distance measurement results of the second suspension point of the target, and the distance measurement results of the lowest point of the target sag, the sag value of the target recognition scene is calculated, and it is determined whether the sag value meets the preset sag threshold. If not, the first warning information is obtained.

2. The method of identifying and warning of power transmission line sag according to claim 1, wherein, Multi-point disparity analysis is performed by traversing the Q sets of synchronized images to obtain Q binocular disparity analysis results for the first suspension points, Q binocular disparity analysis results for the second suspension points, and Q binocular disparity analysis results for the lowest points of the sag, including: Traverse the Q sets of synchronized images and perform grayscale preprocessing to obtain Q sets of preprocessed synchronized images; Extract the Q left-side preprocessed images from the Q preprocessed synchronized image set and identify the first suspension point, the second suspension point, and the lowest point of the sag to obtain the Q left-side first suspension point positions, the Q left-side second suspension point positions, and the Q left-side lowest point of the sag. Based on the positions of Q left first suspension points, Q left second suspension points, and Q left lowest sag points, nearest neighbor windows are identified in the Q left preprocessed images to obtain Q left first suspension point nearest neighbor windows, Q left second suspension point nearest neighbor windows, and Q left lowest sag point nearest neighbor windows. Based on the Q left first suspension point nearest neighbor windows, Q left second suspension point nearest neighbor windows and Q left sag lowest point nearest neighbor windows, cost matching is performed with the Q right preprocessed images in the Q preprocessed synchronous image set to obtain Q right matching first suspension point nearest neighbor windows, Q right matching second suspension point nearest neighbor windows and Q right matching sag lowest point nearest neighbor windows; For the Q left first suspension point neighbor windows, Q left second suspension point neighbor windows, and Q left sag lowest point neighbor windows, and the corresponding Q right matching first suspension point neighbor windows, Q right matching second suspension point neighbor windows, and Q right matching sag lowest point neighbor windows, perform a one-to-one pixel value difference calculation. The absolute value of the calculation result is used as the binocular disparity analysis result of the Q first suspension points, the Q second suspension points, and the Q sag lowest point.

3. The method of claim 2, wherein the method further comprises: Based on the Q left first suspension point nearest neighbor windows, Q left second suspension point nearest neighbor windows, and Q left sag lowest point nearest neighbor windows, cost matching is performed with the Q right preprocessed images in the Q preprocessed synchronized image sets to obtain Q right matching first suspension point nearest neighbor windows, Q right matching second suspension point nearest neighbor windows, and Q right matching sag lowest point nearest neighbor windows, including: The matching cost analysis is performed between the Q left first suspension point nearest neighbor windows and the Q right first matching windows in the Q right preprocessed images to obtain the Q right first matching costs. Based on the magnitude of the Q right-side first matching costs, the Q right-side first matching windows are iterated to obtain Q right-side second matching windows; The matching cost analysis is performed between the Q left-side first suspension point nearest neighbor windows and the Q right-side second matching windows to obtain the Q right-side second matching costs. Based on the magnitude of the Q right-side second matching costs, the Q right-side second matching windows are iterated until all Q right-side preprocessed images have been iterated. The matching window corresponding to the minimum matching cost during the iteration process is taken as the Q right-side matching first dangling point nearest neighbor window. Based on the Q left second suspension point nearest neighbor windows and the Q left sag lowest point nearest neighbor windows, cost matching is performed with the Q right preprocessed images to obtain Q right matching second suspension point nearest neighbor windows and Q right matching sag lowest point nearest neighbor windows.

4. The method of identifying and warning of power transmission line sag according to claim 3, wherein, Based on the magnitude of the Q right-side first matching costs, the Q right-side first matching windows are iterated to obtain Q right-side second matching windows, including: Determine whether the cost of the Q right-side first matching windows is less than or equal to a preset cost. If so, move the Q right-side first matching windows according to the first moving step size to obtain Q right-side second matching windows. If not, then move the Q right-side first matching windows according to the second moving step size to obtain Q right-side second matching windows, wherein the second moving step size is greater than the first moving step size.

5. The method of identifying and warning of power line sag according to claim 1, wherein, The binocular parallax analysis results of the Q first suspension points, the Q second suspension points, and the Q lowest sag points are then centrally filtered to determine the binocular parallax analysis results of the target first suspension points, the target second suspension points, and the target lowest sag points, including: Calculate the mean of the binocular parallax analysis results of the Q first suspension points to obtain the mean of the binocular parallax analysis results of the first suspension points; Starting from the average value of the binocular parallax analysis results of the first suspension point, a neighborhood is constructed in the Q binocular parallax analysis results of the first suspension point according to a preset neighborhood radius to obtain the initial neighborhood. A binocular parallax analysis result of a first suspension point is randomly extracted from the edge of the initial neighborhood as an iteration point, and a neighborhood is constructed in the Q binocular parallax analysis results of the first suspension points based on a preset neighborhood radius to obtain the iterative neighborhood; Determine whether the neighborhood density of the iterative neighborhood is greater than or equal to that of the initial neighborhood. If so, randomly extract a first dangling point binocular parallax analysis result from the edge of the iterative neighborhood as an update iteration point. Based on the update iteration point, continue to perform centralized screening until the preset number of iterations is met. Use the update iteration point obtained in the last iteration as the target first dangling point binocular parallax analysis result. The binocular parallax analysis results of the Q second suspension points and the binocular parallax analysis results of the Q lowest sag points are centrally screened to determine the binocular parallax analysis results of the target second suspension points and the target lowest sag points.

6. The method of identifying and warning of power line sag according to claim 5, wherein, Determining whether the neighborhood density of the iterative neighborhood is greater than or equal to that of the initial neighborhood includes: If not, a first random value and a second random value are obtained based on a random number generator. When the first random value is greater than or equal to the second random value, the starting point is taken as the result of binocular parallax analysis of the first suspension point of the target. When the first random value is less than the second random value, a binocular parallax analysis result of the first suspension point is randomly extracted from the edge of the iterative neighborhood as an update iteration point. Based on the update iteration point, centralized screening continues until the preset number of iterations is met. The update iteration point obtained in the last iteration is taken as the binocular parallax analysis result of the target first suspension point.

7. The transmission line sag identification and early warning method combined with zoom ranging as described in claim 1, characterized in that, Also includes: Continuous monitoring of sag values ​​is performed within a preset monitoring window to obtain a set of continuously monitored sag values. When the variance of the continuously monitored sag value set is greater than or equal to the preset variance, a second early warning message is obtained.

8. A transmission line sag identification and early warning device combined with zoom ranging, characterized in that, The device is used to implement the transmission line sag identification and early warning method combined with zoom ranging as described in any one of claims 1-7, including: Binocular camera calibration module: Install a binocular camera system in a target recognition scenario, adjust the parameters of the zoom lens in the binocular camera system, calibrate the binocular camera system after focusing using a checkerboard pattern, and construct a binocular parallax-depth mapping relationship. The binocular camera system is equipped with a zoom lens. Synchronous image acquisition module: Using the calibrated binocular camera system, perform Q synchronous image acquisitions on the target recognition scene to obtain Q sets of synchronous images, where Q is an integer greater than or equal to 1; Parallax analysis filtering module: Iterates through the Q sets of synchronized images to perform multi-point parallax analysis, obtains Q first suspension point binocular parallax analysis results, Q second suspension point binocular parallax analysis results, and Q lowest point of sag binocular parallax analysis results, and performs centralized filtering on the Q first suspension point binocular parallax analysis results, Q second suspension point binocular parallax analysis results, and Q lowest point of sag binocular parallax analysis results to determine the target first suspension point binocular parallax analysis results, target second suspension point binocular parallax analysis results, and target lowest point of sag binocular parallax analysis results; Depth retrieval module: Using the binocular parallax analysis results of the first suspension point of the target, the binocular parallax analysis results of the second suspension point of the target, and the binocular parallax analysis results of the lowest point of the target sag as indexes, the binocular parallax-depth mapping relationship is retrieved to obtain the depth information of the first suspension point of the target, the depth information of the second suspension point of the target, and the depth information of the lowest point of the target sag; Distance calculation module: Based on the depth information of the first suspension point of the target, the depth information of the second suspension point of the target, and the depth information of the lowest point of the target sag, obtain the distance measurement results of the first suspension point of the target, the distance measurement results of the second suspension point of the target, and the distance measurement results of the lowest point of the target sag; Early warning module: Based on the distance measurement results of the first suspension point of the target, the distance measurement results of the second suspension point of the target, and the distance measurement results of the lowest point of the target sag, calculate the sag value of the target recognition scene, determine whether the sag value meets the preset sag threshold, and if not, obtain the first early warning information.

Images