Power transmission line length auditing method and system based on unmanned aerial vehicle image positioning

Abstract

The invention discloses a power transmission line length auditing method and system based on unmanned aerial vehicle image positioning, and relates to the field of power transmission and distribution line auditing and acceptance. The method comprises the steps that an audit task background management system issues a task to an unmanned aerial vehicle remote controller end, and a pilot carries an unmanned aerial vehicle to work according to a specified line; the unmanned aerial vehicle flies along the line to collect tower pictures and call a tower type and insulator chain identification model to identify the tower type and the number of insulator chains; a tower head position control camera looks down to call a tower position positioning model to record tower position information; the background system receives and gathers various kinds of information collected by the unmanned aerial vehicle; the background system processes the data, calculates the length of wires among the multi-stage towers and generates a topological graph; summarizing the audit data to generate an audit report. The method and the system are helpful to guarantee auditing quality and compliance of investment auditing, and meet complex requirements of modern power transmission line auditing.

Description

Technical Field

[0001] The present invention relates to the field of power transmission and distribution line audit and acceptance, and in particular to a power transmission line length audit method and system based on drone image positioning. Background Art

[0002] With the development of new energy technology, grid technology and digital technology, the construction of new power systems continues to increase, and the increasing construction of overhead transmission line projects has brought many problems to the audit of transmission line length. In traditional power grid construction projects, the audit of transmission line length relies on on-site audit work by auditors. For some transmission line projects, the environmental conditions are complex, with high altitudes, steep terrain, poor traffic conditions, and difficulty in reaching manpower. The influence of multiple factors reduces the audit efficiency. In addition, there are obvious deficiencies in the cost investment and reliability of manual on-site audits. This makes it difficult to carry out audit work based on traditional technologies on time and in quality, resulting in a gap in project investment audits, which brings great challenges to the audit work of power grid construction project investment.

[0003] With the popularization of drone technology and its extensive application in transmission and distribution construction and inspection scenarios, drones are portable, flexible and have excellent on-site accessibility. They can be equipped with visible light cameras, infrared cameras, and RTK locators to remotely perform transmission and distribution line inspections and data collection tasks, greatly freeing up people's hands and feet to replace manual on-site execution of related tasks. Combined with the current on-site audit scenarios of transmission line construction projects, drones are used to assist manual remote line audit work, infinitely extending the hands and eyes of auditors, combining artificial intelligence target recognition algorithms with RTK positioning technology to greatly improve audit work efficiency, and automatically generating audit results with the background system is an excellent solution to improve audit efficiency.

[0004] At present, China Electric Power Research Institute has proposed a method and system for determining the length of a transmission line based on power image data (CN201910944991.1). This invention uses a human-machine collaborative method to quickly extract transmission tower equipment from power image data after fully automatic processing of satellite remote sensing or fixed camera remote sensing data, and calculates the length of the wire on the remote sensing image based on its longitude and latitude. This method uses the large positioning gap between high and low remote sensing accuracy and poor recognition accuracy, resulting in large deviations in wire length calculation.

[0005] The Electric Power Research Institute of State Grid Jiangxi Electric Power Co., Ltd. has proposed an intelligent acceptance method for distribution lines by drones (CN202110373707.7). The drone is equipped with an RFID scanning device to scan the RFID tags of the equipment to be accepted on the pole tower to identify the physical ID, and bind the physical ID to the pole tower coordinates, thereby locating the pole tower coordinates and finally forming a pole tower topology map for display. This method requires manual precision positioning of the RFID tags of the materials for identification, and the operating efficiency has not been rapidly improved.

[0006] Guangzhou Grid Huizhou Power Supply Company proposed an automated measurement method for construction parameters of transmission line projects based on laser point cloud (CN202211092372.2). It uses a drone equipped with a laser radar to obtain point cloud data of transmission line towers. The point cloud data is processed to obtain a tower point cloud model for evaluating the tower position and tower inclination angle, and for calibrating the tower position and angle. This method needs to overcome severe data noise through laser point cloud processing, and the point cloud processing requires a large amount of background computing power, resulting in poor real-time results. Summary of the invention

[0007] The purpose of the present invention is to provide a transmission line length audit method and system based on drone image positioning, which collects and identifies the tower type information of the transmission tower through the drone equipped with a high-definition camera combined with an artificial intelligence target recognition algorithm, and can accurately record the location information of the transmission tower. Through the target detection algorithm and the wire length calculation formula, the tower type of the transmission tower, the location information of the tower, and the length information of the wire are output to the background system. The system generates a transmission line tower line topology map based on the collected data and map data and compares it with the actual construction situation to form an audit task report. In this way, the investment audit project of the transmission line project is carried out efficiently, the quality of the front-line audit is guaranteed, and the authenticity, legality and compliance of the investment audit are ensured.

[0008] To achieve the above object, the present invention provides the following technical solutions: A transmission line length audit method based on drone image positioning, characterized in that it includes the following steps: S1: The drone obtains the audit task, and the drone pilot carries the drone to the designated line to perform the line audit operation according to the audit task; the audit task is an instruction to accurately obtain and analyze information related to the length of the transmission line. The drone pilot carries the drone to the designated line to perform the line audit operation according to the audit task. The designated line is determined according to the actual range of transmission lines that need to be audited. The pilot needs to take the drone to the corresponding location to start work to ensure that the audit operation covers the target line. The drone audit task in step S1 is issued by the audit task background management system and obtained by the drone remote control end. The audit task format is xx line, xx-xx tower.

[0009] S2: The drone pilot controls the drone to fly along the line, adjusts the camera angle of view in the transmission line channel to collect pictures of the transmission towers and calls the corresponding tower type and insulator string recognition model to automatically identify the type of transmission towers and the number of tower-mounted insulator strings; the drone flies along the transmission line and obtains clear pictures of the towers by adjusting the camera angle of view. The called recognition model is specially designed to identify the type of towers and the number of insulator strings. The collected pictures are analyzed using image recognition technology to obtain information on the type of towers and the number of insulator strings. This can avoid errors that may occur in manual recognition and improve recognition efficiency and accuracy. The step S2 specifically includes the following steps: S21: Perform tower target recognition. The drone uses the target detection algorithm to adaptively adjust the shooting position on the channel side. The drone flies to the transmission line channel and hovers. The drone camera lens is adjusted and the tower recognition model is called. A frame is captured from the video stream to perform tower target recognition, and the position coordinates of the tower and the tower head in the frame are output. S22: Adjust the shooting position, obtain the pole tower position coordinates and calculate the pixel-level offset of the target frame from the center of the video frame, obtain the adjustment amount of the shooting position parameter of the drone from the pole tower according to the relationship between the image coordinate system, the camera coordinate system and the world coordinate system, and fly the drone to the pole tower shooting position; S23: Perform tower type identification, adjust the tower full-view target frame to the center of the camera screen, capture the full-view image of the tower, and call the tower identification model algorithm to identify the tower model; S24: Identify the number of tower-mounted insulator strings. Calculate the adjustment amount of the tower head shooting position parameters according to the tower head position coordinates. Fly the drone to the tower head shooting position, adjust the tower head to the center of the drone camera screen, take a tower head image and call the insulator string recognition algorithm to identify the number of insulators.

[0010] S3: The drone pilot controls the drone to fly along the line, and adjusts the camera to look down at a right angle at the tower head to call the pole tower location positioning model, and automatically records the pole tower location information; the called pole tower location positioning model can accurately record the position of the pole tower in the geographic space, providing basic data for subsequent calculations and analysis. Step S3 specifically includes the following steps: S31: When the drone arrives at the hovering point on the top of the transmission tower, adjust the drone camera lens to look downward and call the tower top recognition model, take a frame from the video stream to identify the tower top target, and output the position coordinates of the tower top in the frame; S32: Adjust the position recording point, obtain the tower top position coordinates and calculate the pixel-level offset of the target frame from the center of the video frame, obtain the adjustment parameters of the drone's distance to the tower position recording point according to the relationship between the image coordinate system, the camera coordinate system and the world coordinate system, and fly the drone to the tower shooting position to record the tower position; S33: After arriving at the pole tower shooting position and the pole tower position recording point, the drone triggers the point recording function to record the latitude and longitude of the tower top and the relative height information.

[0011] S4: After the drone audit task is completed, the background system receives the image information, tower type information and location information, and tower-mounted insulator string quantity information collected and recorded by the drone; the background system is the information processing center of the entire audit process. It receives various data collected by the drone during the operation. These data will be used as input for subsequent steps to provide a basis for generating audit reports and calculating line lengths. Step S4 specifically refers to the output information of tower identification, insulator string identification, and tower location records during the execution of the audit task, which are uniformly summarized in the data summary unit. The tower models and insulator string quantities of different towers are bound to the transmission towers based on the tower longitude and latitude, forming a bill of materials for each level of towers.

[0012] S5: The background system automatically processes the data information, binds the tower location information, the number of tower insulators, and the tower type to the tower, and calculates the length of the conductors between the multi-level towers to generate a transmission line topology map; the background system integrates the received data, associates different types of information, and then calculates the length of the conductors between the multi-level towers based on this information and related algorithms. The generated transmission line topology map can intuitively display the structure of the transmission line and the connection relationship between the towers. Step S5 specifically includes the following steps: Calculate the spacing L between adjacent towers using the following formula: , Wherein S1 is the longitude of the current transmission tower, S2 is the longitude of the next transmission tower, W1 is the latitude of the current transmission tower, and W2 is the latitude of the next transmission tower; Calculate the relative height difference H0 of adjacent towers using the following formula: , Where H1 is the relative height of the current level transmission tower, and H2 is the relative height of the next level transmission tower; Calculate the height difference between adjacent towers , the formula is: , Where H0 is the relative height difference H0 between adjacent towers, and L is the spacing between adjacent towers; Calculate the approximate length of a single wire , the formula is: , Where L is the distance between adjacent towers. is the height difference between adjacent towers.

[0013] S6: The backend system will summarize the audit data and generate an audit report on the tower information and line length. All audit data will be summarized and organized to form a detailed report. The report content includes various information about the tower and the calculated line length, which provides a reference for the maintenance and management of the transmission line. Step S6 is specifically when the audit task is completed. The drone will send the captured images, tower type, number of insulator strings at each level of the tower, and conductor length audit data to the backend management system. The backend system will form a tower topology map and a bill of materials corresponding to the tower according to each level of the tower to form a line audit report.

[0014] The method identification model includes tower type identification, insulator string identification, and tower top identification; specifically includes the following steps: Measures the correlation between features and targets, that is, whether the target is a pole, insulator string, or tower top. For discrete random variables X, representing image features and discrete random variables Y, representing target categories, their mutual information The calculation formula is: , Where X is the set of quantized texture features of each region in the image, Y is 1 or -1, and its values ​​represent pole tower, insulator string, tower top or non-pole tower, insulator, and tower top, respectively, which are used to distinguish whether the target in the image is a pole tower, insulator, or tower top, x is the value of the random variable X, y is the value of the random variable Y, p(x,y) is the joint probability of X=x and Y=y, that is, the probability that the image feature value is x and the target category is y, p(x) is the marginal probability of X=x, that is, the probability that the image feature value is x, without considering the target category, and p(y) is the marginal probability of Y=y, that is, the probability that the target category is y, without considering the image features; Support vector machine is used as the recognition model for towers, insulators and tower tops. For the linearly separable case, the goal is to find a hyperplane to separate data of different categories. The training data set is set as , where x i is the image feature vector, Represents the category, where 1 represents a tower, insulator, or tower top, and -1 represents a non-tower, insulator, or tower top. The equation of the hyperplane can be expressed as: , Where w is the normal vector of the hyperplane and b is the bias term; The parameters of the hyperplane are determined by solving the following optimization problem: , where x irepresents the image feature vector of the i-th sample in the training data set. This vector contains various image features used to identify poles, insulators, and tower tops, including a quantized numerical combination of texture features and shape features. i It represents the category label of the i-th sample in the training data set, with a value of 1 or -1, which is used to distinguish whether it is a tower, insulator, tower top or non-tower, insulator, tower top. w is the normal vector of the hyperplane, b is the bias term of the hyperplane, and n represents the number of samples in the training data set. When determining the position of the tower and the top of the tower, since the tower and the top of the tower have many straight line structures, the Hough transform is used to detect the straight line equation in the image. , Hough transform converts it into parameter space; In Hough space, the straight line is determined by counting the intersection points in the parameter space. For the discretized Hough space, each point (m d , c d )'s vote count V(m d , c d ) is calculated as follows: , in is the set of original image points, is the Dirac function m d and c d are the discretized slope and intercept values ​​in Hough space, V(m d , c d ) means that in Hough space, the slope and intercept corresponding to the discretization are (m d , c d )’s votes; By counting the pixel points in the image that satisfy the straight line equation, the positions with high votes are obtained. The positions with high votes indicate the existence of corresponding straight lines, thereby determining the positions of the tower and the tower top.

[0015] A transmission line length audit method and system based on drone image positioning, comprising a server and a processor, wherein a system program is stored in the server, and wherein the processor implements the steps of the method described in any one of claims 1 to 9 when executing the computer program.

[0016] The tower location information, tower insulator string quantity information, tower type and tower are bound to form a complete tower information database.

[0017] According to the location and type information in the tower information database, the conductor length between multiple towers is calculated. The realization of this step depends on accurate measurement algorithms and mathematical models to ensure the accuracy of the calculation of conductor length.

[0018] Generate a transmission line topology map, which intuitively displays the direction of the transmission line, tower location and type, and other information, providing convenience for subsequent operation, maintenance and management.

[0019] The backend system aggregates the audit data and generates an audit report on tower information and line length.

[0020] The report content usually includes the type, quantity, location information of towers, conductor length and the overall condition of the transmission line.

[0021] The report can be used in the operation and maintenance management, asset management and cost analysis of transmission lines, providing strong data support for relevant decision-making.

[0022] In this method, the audit task background management system first issues the task of specifying the line and tower range to the drone remote controller, and the pilot carries the drone to the specified line for operation. During the operation, the drone uses the target detection algorithm to adaptively adjust the shooting position on the channel side, adjusts the camera lens after hovering, calls the tower recognition model to cut frames to identify the position coordinates of the tower and the tower head, adjusts the shooting position accordingly, and then adjusts the tower full-view target frame to the center of the picture to shoot an image to identify the tower model, and adjusts the position according to the tower head coordinates to shoot an image to identify the number of insulator strings. For the tower position record, the drone flies to the tower top hovering point and adjusts the camera downward, calls the tower top recognition model to take frames to identify the tower top coordinates, adjusts the position accordingly and triggers the dot recording function to record the longitude and latitude and relative height of the tower top. The relevant information in the audit process is summarized into the data summary unit, and the towers are distinguished by the longitude and latitude of the towers and the information is bound to form a bill of materials, and then the spacing, relative height difference, height difference and approximate length of a single conductor between adjacent towers are calculated by a specific formula. After the audit is completed, the drone transmits the data to the background management system, and the background generates a tower topology map and a bill of materials and forms an audit report.

[0023] Compared with the prior art, the present invention has the following beneficial effects: In terms of data collection, it is more efficient and accurate. The use of drones to fly along the transmission lines to collect data avoids the complex terrain and difficult access problems that may be faced by manual collection, greatly improving the efficiency of data acquisition. And through target detection algorithms and professional recognition models, it can accurately identify the type of tower, the number of insulator strings and the location of the tower. Compared with traditional manual recognition or other technologies, it reduces human errors and greatly improves the recognition accuracy.

[0024] The adaptive adjustment function of the shooting position, by deploying the target detection algorithm on the drone, the drone can automatically adjust the shooting position according to the position of the target object tower, tower head, and tower top, ensuring that each shot can obtain the best viewing angle and high-quality images, which is helpful for subsequent identification and analysis work and optimizes the entire audit process.

[0025] It has more advantages in data processing and audit report generation. The background system can automatically process the collected data, quickly and accurately calculate the length of the conductors between multi-level towers, generate a transmission line topology map, and summarize all audit data information to generate detailed tower information and line length audit reports. This not only saves a lot of manpower and time costs, but also generates comprehensive and accurate reports, providing a reliable basis for the maintenance, management and planning of transmission lines. Compared with existing technologies, it can better meet the complex needs of modern transmission line audits. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 This is a flow chart of a transmission line length audit method and system based on drone image positioning of the present invention; Figure 2 It is a diagram of a transmission line length audit method based on drone image positioning and a process of adaptively adjusting the shooting points of the system according to the present invention; Figure 3 The present invention provides a transmission line length audit method based on drone image positioning and a system tower position information adaptive adjustment process; Figure 4 A schematic diagram of a transmission line length audit method based on drone image positioning and a parameter diagram for calculating the length of conductors of adjacent towers in the system according to the present invention; Figure 5 A schematic diagram of a specific example of a transmission line length audit method and system based on drone image positioning according to the present invention; DETAILED DESCRIPTION

[0027] The technical solutions in the embodiments of the present invention will be fully described below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0028] like Figure 1-5 As shown, a transmission line length audit method based on drone image positioning includes the following steps: S1: The drone obtains the audit task, and the drone pilot carries the drone to the designated route to perform the route audit operation according to the audit task; S2: The drone pilot controls the drone to fly along the line, adjusts the camera angle in the transmission line channel to collect transmission tower images and calls the corresponding tower type and insulator string recognition model to automatically identify the type of transmission tower and the number of tower-mounted insulator strings; S3: The drone pilot controls the drone to fly along the line, adjusts the camera's right-angle downward view at the tower head position, calls the tower position positioning model, and automatically records the tower position information; S4: The drone audit task is completed, and the background system receives the image information collected and recorded by the drone, the tower type information and location information, and the number of tower-mounted insulators; S5: The backend system automatically processes the data information, binds the tower location information, the number of tower-mounted insulators, the tower type and the tower, and calculates the conductor length between multi-level towers to generate a transmission line topology diagram; S6: The backend system aggregates the audit data information and generates a tower information and line length audit report.

[0029] The drone audit task in step S1 is issued by the audit task background management system and obtained by the drone remote control end. The audit task format is xx line, xx-xx tower.

[0030] Through the comprehensive application of drone technology, image recognition technology, positioning technology and data processing technology, efficient and accurate auditing of transmission lines is achieved. The generated transmission line topology map and audit report provide strong technical support and data support for the operation, maintenance and management of the power system. At the same time, this method also improves the efficiency and accuracy of auditing work and reduces the cost and risk of manual auditing.

[0031] Step S2 specifically includes the following steps: S21: Perform tower target recognition. The drone uses the target detection algorithm to adaptively adjust the shooting position on the channel side. The drone flies to the transmission line channel and hovers. The drone camera lens is adjusted and the tower recognition model is called. A frame is captured from the video stream to perform tower target recognition, and the position coordinates of the tower and the tower head in the frame are output. S22: Adjust the shooting position, obtain the pole tower position coordinates and calculate the pixel-level offset of the target frame from the center of the video frame, obtain the adjustment amount of the shooting position parameter of the drone from the pole tower according to the relationship between the image coordinate system, the camera coordinate system and the world coordinate system, and fly the drone to the pole tower shooting position; S23: Perform tower type identification, adjust the tower full-view target frame to the center of the camera screen, capture the full-view image of the tower, and call the tower identification model algorithm to identify the tower model; S24: Identify the number of tower-mounted insulator strings. Calculate the adjustment amount of the tower head shooting position parameters according to the tower head position coordinates. Fly the drone to the tower head shooting position, adjust the tower head to the center of the drone camera screen, take a tower head image and call the insulator string recognition algorithm to identify the number of insulators.

[0032] The entire workflow is based on the comprehensive application of drone technology, image recognition technology, and coordinate transformation technology. The drone adaptively adjusts the flight position and camera lens to ensure that the image of the tower and tower head can be clearly captured. Then, the image is identified and processed using a pre-trained target recognition model to obtain the type of tower and the number of insulator strings. This method not only improves the efficiency and accuracy of transmission line audits, but also provides strong technical support for the operation, maintenance, and management of power systems.

[0033] The step S3 specifically includes the following steps: S31: When the drone arrives at the hovering point on the top of the transmission tower, adjust the drone camera lens to look downward and call the tower top recognition model, take a frame from the video stream to identify the tower top target, and output the position coordinates of the tower top in the frame; S32: Adjust the position recording point, obtain the tower top position coordinates and calculate the pixel-level offset of the target frame from the center of the video frame, obtain the adjustment parameters of the drone's distance to the tower position recording point according to the relationship between the image coordinate system, the camera coordinate system and the world coordinate system, and fly the drone to the tower shooting position to record the tower position; S33: After arriving at the pole tower shooting position and the pole tower position recording point, the drone triggers the point recording function to record the latitude and longitude of the tower top and the relative height information.

[0034] The step S4 is specifically to output the tower identification, insulator string identification, and tower position record during the audit task execution, which is uniformly summarized in the data summary unit, and the tower models and insulator string quantities of different towers are bound to the transmission towers based on the tower longitude and latitude to form a bill of materials for each level of towers.

[0035] The drone adaptively adjusts its flight position and camera lens to ensure that it can clearly capture images of the tower and the top of the tower. Then, it uses a pre-trained target recognition model to identify and process the image to obtain the type of tower, the number of insulator strings, and the precise location of the tower. Finally, this information is summarized and a bill of materials is generated, providing comprehensive data support for the operation, maintenance, and management of the power system.

[0036] The step S5 specifically includes the following steps: Calculate the spacing L between adjacent towers using the following formula: , Wherein S1 is the longitude of the current transmission tower, S2 is the longitude of the next transmission tower, W1 is the latitude of the current transmission tower, and W2 is the latitude of the next transmission tower; Calculate the relative height difference H0 of adjacent towers using the following formula: , Where H1 is the relative height of the current level transmission tower, and H2 is the relative height of the next level transmission tower; Calculate the height difference between adjacent towers , the formula is: , Where H0 is the relative height difference H0 between adjacent towers, and L is the spacing between adjacent towers; Calculate the approximate length of a single wire , the formula is: , Where L is the distance between adjacent towers. is the height difference between adjacent towers.

[0037] The step S6 specifically means that when the audit task is completed, the drone will send the captured images, tower type, number of insulator strings at each level of the tower, and conductor length audit data to the background management system. The background system will form a tower topology map and a bill of materials corresponding to the tower according to each level of the tower to form a line audit report.

[0038] The identification model includes tower type identification, insulator string identification, and tower top identification; specifically, it includes the following steps: Measures the correlation between features and targets, that is, whether the target is a pole, insulator, or tower top. For discrete random variables X, representing image features and discrete random variables Y, representing target categories, their mutual information The calculation formula is: , Where X is the set of quantized texture features of each region in the image, Y is 1 or -1, and its values ​​represent pole tower, insulator, tower top or non-pole tower, insulator, tower top, which are used to distinguish whether the target in the image is a pole tower, insulator, or tower top, x is the value of the random variable X, y is the value of the random variable Y, p(x, y) is the joint probability of X=x and Y=y, that is, the probability that the image feature value is x and the target category is y, p(x) is the marginal probability of X=x, that is, the probability that the image feature value is x, without considering the target category, and p(y) is the marginal probability of Y=y, that is, the probability that the target category is y, without considering the image features; Support vector machine is used as the recognition model for towers, insulators and tower tops. For the linearly separable case, the goal is to find a hyperplane to separate data of different categories. The training data set is set as , where x i is the image feature vector, Represents the category, where 1 represents a tower, insulator, or tower top, and -1 represents a non-tower, insulator, or tower top. The equation of the hyperplane can be expressed as: , Where w is the normal vector of the hyperplane and b is the bias term; The parameters of the hyperplane are determined by solving the following optimization problem: , where x irepresents the image feature vector of the i-th sample in the training data set. This vector contains various image features used to identify poles, insulators, and tower tops, including a quantized numerical combination of texture features and shape features. i It represents the category label of the i-th sample in the training data set, with a value of 1 or -1, which is used to distinguish whether it is a tower, insulator, tower top or non-tower, insulator, tower top. w is the normal vector of the hyperplane, b is the bias term of the hyperplane, and n represents the number of samples in the training data set. When determining the position of the tower and the top of the tower, since the tower and the top of the tower have many straight line structures, the Hough transform is used to detect the straight line equation in the image. , Hough transform converts it into parameter space; In Hough space, the straight line is determined by counting the intersection points in the parameter space. For the discretized Hough space, each point (m d， c d )'s vote count V(m d， c d )The calculation formula is as follows , in is the set of original image points, is the Dirac function. m d and c d are the discretized slope and intercept values ​​in Hough space, V(m d， c d ) means that in Hough space, the slope and intercept corresponding to the discretization are (m d， c d ), which is obtained by counting the number of pixels in the image that satisfy the straight line equation. A higher number of votes indicates that there is a higher probability that a corresponding straight line exists in the image. By traversing all pixel points and judging whether they satisfy the straight line equation, we vote for the points in the Hough space. Through voting statistics and straight line detection, we can obtain the straight line structure information of the tower and assist in determining the position coordinates of the tower and tower head.

[0039] The tower type and insulator string target recognition model in the method uses the open source yolov8s algorithm model, and the model construction includes the following steps: Sample library construction: Based on the drone inspection images of local power transmission lines, a data sample library is constructed; Image preprocessing: cropping, scaling, image enhancement and normalization; Model training: Use open source LabelImg software to identify the target and mark it as a rectangular annotation box. Convert the annotation information format into XML for storage. Divide the sample library into training set, validation set, and test set according to the ratio of 6:3:1. Convert the .xml format labels into .txt format and save them with the corresponding images into the training set, validation set, and test set.

[0040] Model evaluation calculates the ratio of the number of samples correctly predicted by the model to the total number of samples, thereby measuring the overall prediction accuracy of the model.

[0041] The specific implementation method is: S1: The drone obtains the audit task. The drone pilot carries the drone to the designated line to perform the line audit operation according to the audit task. The drone audit task is issued by the audit task background management system and obtained by the drone remote control. The audit task format is xx line, xx-xx tower.

[0042] S2: The drone pilot controls the drone to fly along the line, adjusts the camera angle in the transmission line channel to collect transmission tower images and calls the corresponding tower type and insulator string recognition model to automatically identify the type of transmission tower and the number of tower-mounted insulator strings; see Figure 2 , the step S2 specifically comprises the following steps: S21: Perform tower target recognition. The drone uses the target detection algorithm to adaptively adjust the shooting position on the channel side. The drone flies to the transmission line channel and hovers. The drone camera lens is adjusted and the tower recognition model is called. A frame is captured from the video stream to perform tower target recognition, and the position coordinates of the tower and the tower head in the frame are output. S22: Adjust the shooting position, obtain the pole tower position coordinates and calculate the pixel-level offset of the target frame from the center of the video frame, obtain the adjustment amount of the shooting position parameter of the drone from the pole tower according to the relationship between the image coordinate system, the camera coordinate system and the world coordinate system, and fly the drone to the pole tower shooting position; S23: Perform tower type identification, adjust the tower full-view target frame to the center of the camera screen, capture the full-view image of the tower, and call the tower identification model algorithm to identify the tower model; S24: Identify the number of tower-mounted insulator strings. Calculate the adjustment amount of the tower head shooting position parameters according to the tower head position coordinates. Fly the drone to the tower head shooting position, adjust the tower head to the center of the drone camera screen, take a tower head image and call the insulator string recognition algorithm to identify the number of insulators.

[0043] S3: The drone pilot controls the drone to fly along the line, adjusts the camera's right-angle downward view at the tower head position, calls the tower position positioning model, and automatically records the tower position information; see Figure 3 , the step S3 specifically comprises the following steps: S31: When the drone arrives at the hovering point on the top of the transmission tower, adjust the drone camera lens to look downward and call the tower top recognition model, take a frame from the video stream to identify the tower top target, and output the position coordinates of the tower top in the frame; S32: Adjust the position recording point, obtain the tower top position coordinates and calculate the pixel-level offset of the target frame from the center of the video frame, obtain the adjustment parameters of the drone's distance to the tower position recording point according to the relationship between the image coordinate system, the camera coordinate system and the world coordinate system, and fly the drone to the tower shooting position to record the tower position; S33: After arriving at the pole tower shooting position and the pole tower position recording point, the drone triggers the point recording function to record the latitude and longitude of the tower top and the relative height information.

[0044] S4: The drone audit task is completed, and the background system receives the image information collected and recorded by the drone, the tower type information and location information, and the number of tower-mounted insulators; The step S4 is specifically to collect the output information of pole tower identification, insulator string identification and pole tower location record in the process of audit task execution, and to collect them into the data collection unit. The pole tower models and the number of insulator strings of different pole towers are bound to the transmission pole towers based on the pole tower longitude and latitude to form a bill of materials for each level of pole towers. Figure 4 ; The step S5 specifically includes the following steps: Calculate the spacing L between adjacent towers using the following formula: , Wherein S1 is the longitude of the current transmission tower, S2 is the longitude of the next transmission tower, W1 is the latitude of the current transmission tower, and W2 is the latitude of the next transmission tower; Calculate the relative height difference H0 of adjacent towers using the following formula: , Where H1 is the relative height of the current level transmission tower, and H2 is the relative height of the next level transmission tower; Calculate the height difference between adjacent towers , the formula is: , Where H0 is the relative height difference H0 between adjacent towers, and L is the spacing between adjacent towers; Calculate the approximate length of a single wire , the formula is: , Where L is the distance between adjacent towers. is the height difference between adjacent towers.

[0045] S5: The backend system automatically processes the data information, binds the tower location information, the number of tower-mounted insulators, the tower type and the tower, and calculates the conductor length between multi-level towers to generate a transmission line topology diagram; S6: The backend system will summarize the audit data and generate an audit report on the tower information and line length. Specifically, when the audit task is completed, the drone will send the captured images, tower type, number of insulator strings at each level of tower, and conductor length audit data to the backend management system. The backend system will generate a tower topology map and a bill of materials corresponding to the tower according to each level of tower, and generate a line audit report.

[0046] When identifying the tower type, insulator string, and tower top, the following steps are included: Measures the correlation between features and targets, that is, whether the target is a pole, insulator, or tower top. For discrete random variables X, representing image features and discrete random variables Y, representing target categories, their mutual information The calculation formula is: , Where X is the set of quantized texture features of each region in the image, Y is 1 or -1, and its values ​​represent pole tower, insulator, tower top or non-pole tower, insulator, tower top, which are used to distinguish whether the target in the image is a pole tower, insulator, or tower top, x is the value of the random variable X, y is the value of the random variable Y, p(x, y) is the joint probability of X=x and Y=y, that is, the probability that the image feature value is x and the target category is y, p(x) is the marginal probability of X=x, that is, the probability that the image feature value is x, without considering the target category, and p(y) is the marginal probability of Y=y, that is, the probability that the target category is y, without considering the image features; Support vector machine is used as the recognition model for towers, insulators and tower tops. For the linearly separable case, the goal is to find a hyperplane to separate data of different categories. The training data set is set as , where x i is the image feature vector, Represents the category, where 1 represents a tower, insulator, or tower top, and -1 represents a non-tower, insulator, or tower top. The equation of the hyperplane can be expressed as: , Where w is the normal vector of the hyperplane and b is the bias term; The parameters of the hyperplane are determined by solving the following optimization problem: , where x irepresents the image feature vector of the i-th sample in the training data set. This vector contains various image features used to identify poles, insulators, and tower tops, including a quantized numerical combination of texture features and shape features. i It represents the category label of the i-th sample in the training data set, with a value of 1 or -1, which is used to distinguish whether it is a tower, insulator, tower top or non-tower, insulator, tower top. w is the normal vector of the hyperplane, b is the bias term of the hyperplane, and n represents the number of samples in the training data set. The tower type and insulator string target recognition model in the method uses the open source yolov8s algorithm model, and the model construction includes the following steps: S71: Sample library construction: Based on the drone inspection images of local power transmission lines, a data sample library is constructed; S72: image preprocessing, cropping and scaling the image, image enhancement and normalization; S73: Model training: Use the open source LabelImg software to identify the target and mark it as a rectangular annotation box. Convert the annotation information format into XML for storage. Divide the sample library into training set, validation set, and test set according to the ratio of 6:3:1. Convert the .xml format labels into .txt format and save them with the corresponding images into the training set, validation set, and test set.

[0047] S74: Model evaluation, calculate the ratio of the number of samples correctly predicted by the model to the total number of samples, so as to measure the overall prediction accuracy of the model.

[0048] When determining the position of the tower and the top of the tower, since the tower and the top of the tower have many straight line structures, the Hough transform is used to detect the straight line equation in the image. , Hough transform converts it into parameter space; In Hough space, the straight line is determined by counting the intersection points in the parameter space. For the discretized Hough space, each point (m d， c d )'s vote count V(m d， c d ) is calculated as follows: , in is the set of original image points, is the Dirac function. m d and c d are the discretized slope and intercept values ​​in Hough space, V(m d， c d ) means that in Hough space, the slope and intercept corresponding to the discretization are (m d， c d), which is obtained by counting the number of pixels in the image that satisfy the straight line equation. A higher number of votes indicates that there is a higher probability that a corresponding straight line exists in the image. By traversing all pixel points and judging whether they satisfy the straight line equation, we vote for the points in the Hough space. Through voting statistics and straight line detection, we can obtain the straight line structure information of the tower and assist in determining the position coordinates of the tower and tower head.

[0049] Specific embodiments, such as Figure 5 As shown, during the test of UAV inspection of power transmission lines in Shandong area, relevant information was successfully identified and the task summary was successfully exported.

Claims

1. A transmission line length audit method based on drone image positioning, characterized in that: The following steps are included: S1: The drone obtains the audit task, and the drone pilot carries the drone to the designated route to perform the route audit operation according to the audit task; S2: The drone pilot controls the drone to fly along the line, adjusts the camera angle in the transmission line channel to collect transmission tower images and calls the tower type and insulator string target recognition model to automatically identify the type of transmission tower and the number of tower-mounted insulator strings; S3: The drone pilot controls the drone to fly along the line, adjusts the camera's right-angle downward view to call the position positioning at the tower head position, and automatically records the tower position information; S4: The drone audit task is completed, and the background system receives the image information collected and recorded by the drone, the tower type information and location information, and the number of tower-mounted insulator strings; S5: The backend system automatically processes the data information, binds the tower location information, the number of tower-mounted insulator strings, the tower type and the tower, and calculates the conductor length between multi-level towers to generate a transmission line topology diagram; S6: The backend system aggregates the audit data information and generates an audit report on the tower information and line length.

2. According to claim 1, a transmission line length audit method based on drone image positioning is characterized in that: The drone audit task in step S1 is issued by the audit task background management system and obtained by the drone remote control end. The audit task format is xx line, xx-xx tower.

3. According to claim 1, a transmission line length audit method based on drone image positioning is characterized in that: The step S2 specifically includes the following steps: S21: Performing tower target recognition, the drone adjusts the shooting position on the channel side, flies to the transmission line channel and hovers, adjusts the drone camera lens and calls the target recognition model, cuts frames from the video stream to perform tower target recognition, and outputs the position coordinates of the tower and the tower head in the frame; S22: Adjust the shooting position, obtain the pole tower position coordinates and calculate the pixel-level offset of the target frame from the center of the video frame, obtain the adjustment amount of the shooting position parameter of the drone from the pole tower according to the relationship between the image coordinate system, the camera coordinate system and the world coordinate system, and fly the drone to the pole tower shooting position; S23: Performing pole tower recognition, adjusting the tower full-view target frame to the center of the camera screen, taking a full-view image of the pole, and calling the tower type and insulator string target recognition model to identify the pole tower model; S24: Calculate the adjustment amount of the tower head shooting position parameters according to the tower head position coordinates, fly the UAV to the tower head shooting position, adjust the tower head to the center of the UAV camera screen, shoot the tower head image and call the tower type and insulator string target recognition model to determine the number of insulators.

4. According to the method of transmission line length audit based on drone image positioning according to claim 1, it is characterized in that: The step S3 specifically includes the following steps: S31: When the UAV reaches the hovering point on the top of the transmission tower, the UAV camera lens is adjusted to look downward and the target recognition model is called to take a frame from the video stream to identify the target on the top of the tower, and the position coordinates of the tower top in the frame are output; S32: Adjust the position recording point, obtain the tower top position coordinates and calculate the pixel-level offset of the target frame from the center of the video frame, obtain the adjustment parameters of the drone's distance to the tower position recording point according to the relationship between the image coordinate system, the camera coordinate system and the world coordinate system, and fly the drone to the tower shooting position to record the tower position; S33: After arriving at the pole tower shooting position and the pole tower position recording point, the drone triggers the point recording function to record the latitude and longitude of the tower top and the relative height information.

5. According to claim 1, a transmission line length audit method based on drone image positioning is characterized in that: The step S4 is specifically to output the tower identification, insulator string identification, and tower position record during the audit task execution, which is uniformly summarized in the data summary unit, and the tower models and insulator string quantities of different towers are bound to the transmission towers based on the tower longitude and latitude to form a bill of materials for each level of towers.

6. The method for power transmission line length audit based on drone image positioning according to claim 1 is characterized in that: The identification model of the method includes identifying the tower, the insulator string, and the tower top respectively, and specifically includes the following steps: By measuring the correlation between the feature and the target, that is, whether the target is a pole, insulator string, or tower top, the discrete random variable X represents the image feature, and the discrete random variable Y represents the target category. The mutual information between them is The calculation formula is: ， Where X is the set of quantized texture features of each region in the image, Y is 1 or -1, and its values ​​represent pole tower, insulator string, tower top or non-pole tower, insulator string, tower top, which are used to distinguish whether the target in the image is a pole tower, insulator string, or tower top, x is the value of the random variable X, y is the value of the random variable Y, p(x,y) is the joint probability of X=x and Y=y, that is, the probability that the image feature value is x and the target category is y, p(x) is the marginal probability of X=x, that is, the probability that the image feature value is x, without considering the target category, and p(y) is the marginal probability of Y=y, that is, the probability that the target category is y, without considering the image features; Support vector machine is used as the recognition model of pole tower, insulator string and tower top. For the linearly separable case, the goal is to find a hyperplane to separate the data of different categories. The training data set is set as , where x i is the image feature vector, Represents the category, where 1 represents a tower, insulator string, or tower top, and -1 represents a non-tower, insulator string, or tower top. The equation of the hyperplane can be expressed as: ， Where w is the normal vector of the hyperplane and b is the bias term; The parameters of the hyperplane are determined by solving the following optimization problem: ， where x i represents the image feature vector of the i-th sample in the training data set. This vector contains various image features used to identify poles, insulator strings, and tower tops, including a quantized numerical combination of texture features and shape features. i It represents the category label of the i-th sample in the training data set, and takes the value of 1 or -1, which is used to distinguish whether it is a pole tower, insulator string, tower top or non-pole tower, insulator string, tower top. w is the normal vector of the hyperplane, b is the bias term of the hyperplane, and n represents the number of samples in the training data set.

7. The method for auditing the length of a power transmission line based on drone image positioning according to claim 1 is characterized in that: The tower type and insulator string target recognition model in the method uses the open source yolov8s algorithm model, and the model construction includes the following steps: S71: Sample library construction: Based on the drone inspection images of local power transmission lines, a data sample library is constructed; S72: image preprocessing, cropping and scaling the image, image enhancement and normalization; S73: Model training: Using the open source LabelImg software, the recognition target is marked as a rectangular annotation box, and the annotation information format is converted into XML for storage. The sample library is divided into training set, validation set, and test set according to the ratio of 6:3:1, and the .xml format labels are converted into .txt format and saved with the corresponding images into the training set, validation set, and test set. S74: Model evaluation, calculate the ratio of the number of samples correctly predicted by the model to the total number of samples, so as to measure the overall prediction accuracy of the model.

8. The method for power transmission line length audit based on drone image positioning according to claim 1 is characterized in that: When determining the position of the tower and the top of the tower, since the tower and the top of the tower have many straight line structures, the Hough transform is used to detect the straight line equation in the image. , Hough transform converts it into parameter space; In Hough space, the straight line is determined by counting the intersection points in the parameter space. For the discretized Hough space, each point (m d， c d )'s vote count V(m d， c d ) is calculated as follows: ， in is the set of original image points, is the Dirac function m d and c d are the discretized slope and intercept values ​​in Hough space, V(m d， c d ) means that in Hough space, the slope and intercept corresponding to the discretization are (m d， c d )’s votes; By counting the pixel points in the image that satisfy the straight line equation, the positions with high votes are obtained. The positions with high votes indicate the existence of corresponding straight lines, thereby determining the positions of the tower and the tower top.

9. The method for power transmission line length audit based on drone image positioning according to claim 1 is characterized in that: The step S5 specifically includes the following steps: Calculate the spacing L between adjacent towers using the following formula: ， Wherein S1 is the longitude of the current transmission tower, S2 is the longitude of the next transmission tower, W1 is the latitude of the current transmission tower, and W2 is the latitude of the next transmission tower; Calculate the relative height difference H0 of adjacent towers using the following formula: ， Where H1 is the relative height of the current level transmission tower, and H2 is the relative height of the next level transmission tower; Calculate the height difference between adjacent towers , the formula is: ， Where H0 is the relative height difference H0 between adjacent towers, and L is the spacing between adjacent towers; Calculate the approximate length of a single wire , the formula is: ， Where L is the distance between adjacent towers. is the height difference between adjacent towers.

10. A transmission line length audit system based on drone image positioning, comprising a server and a processor, wherein the server stores a system program, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 9 are implemented.

Images