A method and system for trajectory planning of multi-UAV collaborative inspection of distribution network lines

Abstract

The present invention discloses a method and system for trajectory planning of multi-UAV collaborative inspection of distribution network lines, which relates to the technical field of UAV trajectory planning, including: collecting point cloud data based on UAVs, and performing data denoising sampling; performing adaptive clustering according to the reflection intensity of the sampled data, introducing position coding to describe point coordinate information and constructing point cloud vectors, and labeling the point cloud; and obtaining a power line model based on the power line label point cloud fitting quadratic curve equation. The method of the present invention combines the adaptive clustering method with the mean shift algorithm to make the interval division of the point cloud reflection intensity more adaptive, ensure that the initialization of the cluster center is based on the data distribution characteristics, and combine RANSAC fitting, power line point cloud classification and intersection area detection to make the modeling of the power line point cloud more accurate, while ensuring that the trajectories of different power lines are clearly separable, and reducing the misjudgment of the line intersection area.

Description

Technical Field

[0001] The present invention relates to the technical field of unmanned aerial vehicle (UAV) track planning, and in particular to a track planning method and system for cooperative inspection of distribution network lines by multiple unmanned aerial vehicles. Background Art

[0002] With the rapid development of smart grids and the continuous expansion of distribution network lines, traditional manual inspection methods can no longer meet the needs of modern power grids for efficient, accurate and safe inspections. UAV-based distribution network inspection technology has gradually matured and is widely used in high-voltage transmission line inspections, fault detection and power facility monitoring.

[0003] However, the existing drone inspection methods still mainly rely on the single-machine inspection mode, and their track planning usually adopts simple preset routes or flight paths based on experience, failing to fully consider the complex structure of power lines, environmental obstacles, drone endurance, and global optimization of task scheduling. The existing drone inspection track planning still has problems, and the inspection task area division method is relatively simple, failing to fully consider the power line topology, tower distribution, and the influence of environmental obstacles such as vegetation and buildings, resulting in low track planning accuracy, affecting the reliability of inspection tasks. Summary of the invention

[0004] In view of the above existing problems, the present invention is proposed.

[0005] Therefore, the present invention provides a trajectory planning method for multi-UAV collaborative inspection of distribution network lines to solve the problem that the existing UAV inspection method still mainly relies on the single-machine inspection mode. Its trajectory planning usually adopts a simple preset route or an experience-based flight path, and fails to fully consider the complex structure of the power line, environmental obstacles, UAV endurance and the global optimization of task scheduling. The existing UAV inspection trajectory planning also has the problem that the inspection task area division method is relatively simple, and fails to fully consider the influence of the power line topology, tower distribution and environmental obstacles such as vegetation and buildings, resulting in low trajectory planning accuracy, which affects the reliability of the inspection task.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0007] In a first aspect, the present invention provides a method for planning a track for collaborative inspection of distribution network lines by multiple UAVs, comprising:

[0008] Collect point cloud data based on drones and perform data denoising sampling;

[0009] Adaptive clustering is performed based on the reflection intensity of the sampled data, position coding is introduced to describe the point coordinate information and construct point cloud vectors, and the point cloud is labeled and classified;

[0010] The power line model is obtained by fitting the quadratic curve equation based on the power line label point cloud, and the intersection area is identified and marked according to the intersection distance of the power line point cloud;

[0011] Clustering is performed based on the point cloud of towers, vegetation, and building labels. The task area is planned based on the tower clustering, and the task allocation weight of the drone is calculated to perform the drone task allocation.

[0012] Perform preliminary path planning through the mission area, determine the Euclidean distance matrix between all inspection points, use genetic algorithm to obtain the optimal inspection sequence of different inspection points, and determine the optimal control instructions based on minimizing track deviation and control input as the objective function;

[0013] The control instruction inspection path and inspection point sequence data are securely transmitted and stored.

[0014] As a preferred solution of the trajectory planning method for the coordinated inspection of distribution network lines by multiple UAVs of the present invention, wherein: the point cloud data is collected based on the UAV, and data denoising sampling is performed, and the flight height of the UAV is set based on the maximum inspection height and the minimum safety height of the multiple UAVs in the inspection task area;

[0015] The drone is equipped with a laser radar to collect point cloud data, and the scanning range radius is determined based on the viewing angle of the laser radar and the flight altitude of the drone;

[0016] Obtain point cloud data of the scanning range, including point cloud coordinates and reflection intensity;

[0017] The point cloud data is downsampled using a three-dimensional voxel grid to obtain the sampling point coordinates based on the voxel grid, and then denoised by statistical filtering.

[0018] As a preferred solution of the track planning method for the multi-UAV collaborative inspection of the distribution network line of the present invention, wherein: the adaptive clustering is performed according to the reflection intensity of the sampled data, the position coding is introduced to describe the point coordinate information and construct the point cloud vector, the point cloud is labeled and classified, and the center value of the intensity interval is determined based on the reflection intensity of the denoised point cloud coordinates by the adaptive clustering method, which is expressed as:

[0019] ;

[0020] in represents the central value of the j-th interval intensity, The total number of bins representing the range of reflection intensity, and Respectively represent the maximum and minimum values ​​of the reflection intensity;

[0021] Calculate the number of points in each intensity interval for histogram counting, expressed as:

[0022] ;

[0023] ;

[0024] in The reflection intensity of the interval is The number of point clouds, Represents the total number of point cloud data, and represents the i-th reflection intensity value, represents the center value of the jth reflection intensity interval, represents the indicator function, It represents the ratio of the difference between the maximum and minimum values ​​of the reflection intensity to the total number of intervals B;

[0025] based on Discretize the reflection intensity to form a histogram, and calculate the proportion of the intensity range in the entire point cloud according to the number of point clouds with different reflection intensities based on the total number of point clouds;

[0026] Based on the sum of the proportion mean and standard deviation as the proportion threshold, the number of interval reflection intensity point clouds greater than the proportion threshold is As the window center for initializing mean shift clustering, the mean shift algorithm is used for adaptive clustering, in which the weighted mean in the window is calculated by the density and function, expressed as:

[0027] ;

[0028] in represents the updated kth window center, n represents the total number of reflection intensities, represents the density kernel function, represents the kth window center of mean shift clustering, and W represents the window width;

[0029] Iterate and update the window center. When the difference of the updated window center is no longer obvious during the continuous iteration, stop the iteration and classify the point cloud data into the closest cluster center through the attribution relationship.

[0030] The Transformer model is used to introduce position coding to describe the point coordinate information, which is encoded using trigonometric functions and expressed as:

[0031] ;

[0032] in Represents the position encoding of the i-th point cloud coordinates, , and are the three-dimensional point cloud coordinate information of the i-th point cloud, represents the frequency parameter;

[0033] Construct a point cloud vector, including point cloud coordinates, reflection intensity, cluster label and position encoding, use the self-attention mechanism based on the Transformer model to determine the query, key and value, and calculate the attention relationship between point clouds, expressed as:

[0034] ;

[0035] ;

[0036] in represents the attention weight between the i-th point cloud and the j-th point cloud, and Represents the query and key of the i-th point cloud and the j-th point cloud respectively. The T in represents the transpose calculation, represents the dimension of the Key matrix, represents the Transformer output of the i-th point cloud, Represents the value of the j-th point cloud;

[0037] Use multi-layer perceptron MLP to After the final classification, different classification labels are obtained, including label classification of poles, power lines, vegetation, ground and buildings.

[0038] As a preferred solution of the trajectory planning method for the multi-UAV collaborative inspection of distribution network lines of the present invention, wherein: the power line model is obtained by fitting a quadratic curve equation based on the power line label point cloud, the intersection area is identified and marked according to the intersection distance of the power line point cloud, and RANSAC is used to model the point cloud data based on the power line label, the power line point cloud is projected onto the XOY plane, and the two-dimensional trajectory of the power line is extracted;

[0039] Use RANSAC to fit the line, randomly sample two points in the power line point cloud, and calculate the assumed line equation. Calculate the vertical error from all points of the power line label point cloud data to the assumed line equation. Determine the line threshold based on historical data, and count the number of fitted point clouds that conform to the line. Repeat the sampling, and take the assumed line equation with the largest number of fitted point clouds as the fitting result, and obtain the optimal line variance model.

[0040] Based on the straight line variance model, the dichotomy method is used to optimize the modeling of the power line, calculate and determine the point farthest from the fitting line, and set the dividing point for the power line, which is expressed as:

[0041] ;

[0042] in Indicates the x-coordinate of the boundary point cloud, Indicates the x-coordinate of the point cloud with the maximum error, Indicates the minimum x value of the current fitting area. represents the adjustment factor;

[0043] The fitting curves of the two parts are calculated based on the dividing point, which are expressed as:

[0044] ;

[0045] ;

[0046] in and Respectively represent the power line data points on both sides of the dividing point, and is the coordinate of the i-th data point;

[0047] The quadratic curves are fitted using the least squares method, and the quadratic curve equations on both sides of the power line are solved. The fitting error is calculated using the mean square error (MSE). The error threshold is determined based on historical data. If the fitting error is greater than or equal to the error threshold, the binary method is continued for redivision and fitting. If the fitting error is less than the error threshold, the calculation is terminated and the final power line model is output, which is expressed as:

[0048] ;

[0049] in represents the final power line function, , as well as are the quadratic function coefficients of the curve on one side, , as well as represents the quadratic function coefficient of the curve on the other side, and x represents the horizontal coordinate of the power line point cloud data;

[0050] Determine different power line equations based on the final power line model, calculate the vertical distance from the data of each power line point cloud to the power line equation, determine the distance threshold based on historical data, and if the calculated vertical distance is less than or equal to the distance threshold, determine that the point cloud belongs to the power line fitted by the power line equation;

[0051] After determining the power line to which the power line point cloud belongs, the Euclidean distance of point cloud pairs of different power line lines is calculated, and the shortest crossing distance is calculated. The crossing threshold is determined based on historical data. If the shortest crossing distance is less than or equal to the crossing threshold, it is judged that the power line lines to which the point cloud pair belongs have regional intersections, and the point clouds in the intersection area are marked.

[0052] As a preferred solution of the trajectory planning method for the multi-UAV collaborative inspection of distribution network lines described in the present invention, wherein: the point cloud based on the pole tower, vegetation and building labels is clustered and distinguished, the task area is planned based on the tower clustering, the task allocation weight of the UAV is calculated, and the UAV task is allocated, based on the point cloud of the pole tower, vegetation and building labels, different pole towers, vegetation and buildings are marked respectively by DBSCAN clustering, and the covariance matrix is ​​calculated using PCA, the center points of different pole towers, vegetation and buildings are calculated respectively, and the heights of different pole towers, vegetation and buildings are determined;

[0053] Based on the minimum safe distance of drones, set a no-fly zone for the point cloud in the intersection area, and set a safe flight altitude directly over towers, vegetation and buildings;

[0054] The maximum distance between different towers is calculated based on the center point of the tower, and K-means++ is used to cluster the task areas. The distance from each tower to the center of the nearest task area is calculated, which is expressed as:

[0055] ;

[0056] in represents the central tower position of the kth mission area, represents the tower location data set, p represents the center point of the tower, and c represents the center tower location of the candidate task area;

[0057] Repeatedly update the mission area center, expressed as:

[0058] ;

[0059] in represents all towers in the task area k;

[0060] When the center of the task area no longer changes significantly, the clustering process ends, and after clustering the task areas, the tower set of each task area is determined;

[0061] The task allocation weight is calculated based on the task area and the UAV, which is expressed as:

[0062] ;

[0063] in represents the allocation weight of UAV j to perform mission area k, represents the shortest path distance from the starting point of UAV j to the tower i, represents the cruising speed of UAV j, represents the total inspection time of all towers in task area k;

[0064] Based on the allocation weight, the Hungarian algorithm is used to optimize the allocation, maximize the allocation weight of the UAV, and complete the allocation of the UAV in the mission area.

[0065] As a preferred solution of the track planning method for the multi-UAV collaborative inspection of the distribution network line of the present invention, wherein: the preliminary path planning is performed through the mission area, the Euclidean distance matrix between all inspection points is determined, the optimal inspection order of different inspection points is obtained by using a genetic algorithm, and the optimal control instruction is determined based on minimizing the track deviation and control input as the objective function, and the RRT is used based on the UAVs in different mission areas. The fast random tree is used for preliminary path planning. At the same time, based on all the inspection points in the preliminary path planning, the Euclidean distance matrix between all the inspection points is determined. The traveling salesman problem TSP is used to set the minimization of the total inspection path as the goal to optimize the UAV inspection path.

[0066] The genetic algorithm GA-TSP is used for approximate solution, in which multiple random path arrangements are generated, the total inspection distance of each path is calculated, and the crossover mutation operation is performed through partial matching crossover PMX to obtain the optimal inspection sequence and corresponding inspection length of different inspection points;

[0067] Based on the inspection trajectory of the UAV determined by the optimal inspection sequence, the MPC model predictive control is used to smooth the trajectory and control the speed of the UAV. The objective function is based on minimizing the track deviation and control input, which is expressed as:

[0068] ;

[0069] Where J represents the target deviation value, represents the total number of time steps predicted by the MPC model, The position of the drone at time g, represents the position of the drone’s inspection point at the gth moment, represents the speed of the drone at the gth moment;

[0070] Based on the objective function of minJ, the control instructions of the optimal control input, including linear velocity and angular velocity, are determined through the quadratic programming QP solver;

[0071] The calculated control instructions are input and sent to the UAV to adjust the flight trajectory.

[0072] As a preferred solution of the trajectory planning method for the multi-UAV collaborative inspection of the distribution network line of the present invention, wherein: the secure transmission and storage of the control instruction inspection path and the inspection point sequence data refers to the end-to-end encryption of the control instruction data using AES-256, and the inspection path and the inspection point sequence data are encrypted using AES-128 and written into a read-only database;

[0073] MQTT is used to securely transmit encrypted data, which is then transmitted to the cloud via wireless transmission technology for log data storage.

[0074] In a second aspect, the present invention provides a system for a method for trajectory planning of a multi-UAV collaborative inspection of a distribution network line, comprising:

[0075] Data processing module, for collecting and denoising point cloud data;

[0076] Point cloud feature extraction module, which classifies the collected point cloud data and extracts spatial location information;

[0077] Cross-recognition module, which models power lines and identifies cross-recognition areas;

[0078] The UAV mission area planning module divides the UAV mission area and assigns UAVs to perform inspection tasks;

[0079] Path planning module, generating the best inspection path and optimizing the flight trajectory of the drone;

[0080] The control command execution module calculates the optimal control command and sends it to the UAV for task execution;

[0081] The data security transmission module is used to securely transmit and store UAV inspection data, control instructions and mission paths.

[0082] In a third aspect, the present invention provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program is executed by the processor, it implements any step of the trajectory planning method for collaborative inspection of distribution network lines by multiple UAVs as described in the first aspect of the present invention.

[0083] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the trajectory planning method for collaborative inspection of distribution network lines by multiple UAVs as described in the first aspect of the present invention.

[0084] The beneficial effects of the present invention are as follows: by combining the adaptive clustering method with the mean shift algorithm, the interval division of the point cloud reflection intensity is made more adaptive, ensuring that the initialization of the cluster center is based on the data distribution characteristics; by combining RANSAC fitting, power line point cloud classification and intersection area detection, the modeling of the power line point cloud is more accurate, while ensuring that the trajectories of different power lines are clearly separable, reducing the misjudgment of line intersection areas; the binary optimization modeling is combined with error control, so that the power line model can reduce the consumption of computing resources while ensuring accuracy; through the combination of task weight calculation and Hungarian algorithm optimization task allocation, the entire UAV inspection task has accurate task area division, efficient task scheduling and safe flight planning; by using the genetic algorithm GA-TSP for approximate solution, a plurality of path arrangements are randomly generated and the total inspection distance of each path is calculated, so that the optimization of the inspection sequence can converge quickly, ensuring that the UAV inspection task can both improve the coverage rate and reduce energy consumption. BRIEF DESCRIPTION OF THE DRAWINGS

[0085] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0086] Figure 1 This is a flow chart of the method for planning the track of multiple UAVs coordinating inspection of distribution network lines in Example 1;

[0087] Figure 2 This is a structural diagram of the trajectory planning system for collaborative inspection of distribution network lines by multiple UAVs in Example 2. DETAILED DESCRIPTION

[0088] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the accompanying drawings.

[0089] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0090] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0091] Example 1, reference Figure 1 and Figure 2 , which is the first embodiment of the present invention, and provides a method for planning a track for a multi-UAV collaborative inspection of a distribution network line, comprising the following steps:

[0092] S1, collect point cloud data based on drones and perform data denoising sampling;

[0093] Preferably, point cloud data is collected based on drones, and data denoising sampling is performed. The flight height of the drones is set based on the maximum inspection height and the minimum safety height of multiple drones in the inspection task area, which is expressed as:

[0094] ;

[0095] in Indicates the flight altitude of the drone. Indicates the minimum safe altitude. Indicates the maximum inspection height;

[0096] The UAV is equipped with a laser radar to collect point cloud data. The scanning range radius is determined based on the viewing angle of the laser radar and the flight altitude of the UAV, which is expressed as:

[0097] ;

[0098] in Indicates the scanning range radius, represents the laser radar viewing angle;

[0099] Obtain point cloud data of the scanning range, including point cloud coordinates and reflection intensity;

[0100] The point cloud data is downsampled using a three-dimensional voxel grid to obtain the sampling point coordinates based on the voxel grid, and then denoised by statistical filtering.

[0101] By setting the minimum safety altitude and maximum inspection altitude, the drone can be prevented from flying too low, reducing the risk of hitting buildings, vegetation, power lines, etc., ensuring the safe operation of the drone, ensuring that the drone does not fly too high and exceed the effective scanning range of the lidar, and ensuring the quality of data collection;

[0102] By calculating the scanning range radius, the flight spacing of the drone can be reasonably adjusted to maximize the scanning range of each flight, thereby reducing the number of routes and improving inspection efficiency. By obtaining point cloud coordinates and reflection intensity through lidar, high-precision three-dimensional modeling of power lines, poles, and crossing areas can be achieved to ensure the integrity of inspection data. The reflection intensity can be used to identify structures such as conductors, insulators, and crossing points, thereby improving the drone's ability to automatically identify power line defects.

[0103] S2, adaptive clustering is performed according to the reflection intensity of the sampled data, position coding is introduced to describe the point coordinate information and construct point cloud vectors, and label classification is performed on the point cloud;

[0104] Preferably, adaptive clustering is performed according to the reflection intensity of the sampled data, position coding is introduced to describe the point coordinate information and construct a point cloud vector, the point cloud is labeled and classified, and the center value of the intensity interval is determined based on the reflection intensity of the denoised point cloud coordinates by an adaptive clustering method, which is expressed as:

[0105] ;

[0106] in represents the central value of the j-th interval intensity, The total number of intervals representing the range of reflection intensity, determined based on historical data, and Respectively represent the maximum and minimum values ​​of the reflection intensity;

[0107] Calculate the number of points in each intensity interval for histogram counting, expressed as:

[0108] ;

[0109] ;

[0110] in The reflection intensity of the interval is The number of point clouds, Represents the total number of point cloud data, represents the i-th reflection intensity value, represents the center value of the jth reflection intensity interval, represents the indicator function, It represents the ratio of the difference between the maximum and minimum values ​​of the reflection intensity to the total number of intervals B;

[0111] based on Discretize the reflection intensity to form a histogram, and calculate the proportion of the intensity range in the entire point cloud according to the number of point clouds with different reflection intensities based on the total number of point clouds;

[0112] Based on the sum of the proportion mean and standard deviation as the proportion threshold, the number of interval reflection intensity point clouds greater than the proportion threshold is As the window center for initializing mean shift clustering, the mean shift algorithm is used for adaptive clustering, in which the weighted mean in the window is calculated by the density and function, expressed as:

[0113] ;

[0114] in represents the updated kth window center, n represents the total number of reflection intensities, Indicates that the density kernel function uses a Gaussian kernel, , x represents the Gaussian kernel calculation term, represents the kth window center of mean shift clustering, W represents the window width, which is determined based on historical values;

[0115] Iterate and update the window center. When the difference of the updated window center is no longer obvious during the continuous iteration, stop the iteration and classify the point cloud data into the closest cluster center through the attribution relationship.

[0116] The Transformer model is used to introduce position coding to describe the point coordinate information, which is encoded using trigonometric functions and expressed as:

[0117] ;

[0118] in Represents the position encoding of the i-th point cloud coordinates, , and are the three-dimensional point cloud coordinate information of the i-th point cloud, Represents the frequency parameter, which is usually set to , where d is the feature dimension of the Transformer model, usually set to 64, 128, or 256, and k represents the index of the encoding channel at the current position;

[0119] Construct a point cloud vector, including point cloud coordinates, reflection intensity, cluster labels, and position encoding, and use the self-attention mechanism based on the Transformer model to determine the query, key, and value. , , , where Q, K, and V represent query, key, and value vectors respectively. , as well as Represent the weights of query, key, and value matrices respectively, Represents the point cloud vector and calculates the attention relationship between the point clouds, expressed as:

[0120] ;

[0121] ;

[0122] in represents the attention weight between the i-th point cloud and the j-th point cloud, and Represents the query and key of the i-th point cloud and the j-th point cloud respectively. The T in represents the transpose calculation, represents the dimension of the Key matrix, represents the Transformer output of the i-th point cloud, Represents the value of the j-th point cloud;

[0123] Use multi-layer perceptron MLP to After the final classification, different classification labels are obtained, including label classification of poles, power lines, vegetation, ground and buildings.

[0124] By using the reflection intensity of the denoised point cloud coordinates based on the adaptive clustering method to determine the center value of the intensity interval, the discreteness of the reflection intensity value can be reduced, the classification accuracy of the point cloud data can be improved, the number of points in each intensity interval is calculated and the histogram is counted, making the distribution of the point cloud data more interpretable, and providing the initial window center for the subsequent mean shift clustering, improving the stability of the clustering algorithm, and using the mean shift Mean-Shift algorithm for adaptive clustering. The density kernel function is used to calculate the weighted mean in the window, so that the clustering process can adapt to the changes in the point cloud density in different areas, thereby improving the clustering accuracy and avoiding the problem that the traditional clustering method needs to preset the number of categories. The Transformer model processes point cloud information and uses trigonometric functions for position encoding, which can better describe point cloud coordinate information in high-dimensional space and avoid numerical instability caused by directly using coordinate values. The self-attention mechanism calculates the attention relationship between point clouds, which can mine the correlation between point clouds, improve the accuracy of point cloud classification, and enhance the adaptability to point cloud data in complex environments. Finally, the multi-layer perceptron MLP is used to classify the output of the Transformer, making the classification of poles, power lines, vegetation, ground and buildings more accurate, improving the generalization ability of the classification model, and reducing the classification error rate caused by noisy point clouds.

[0125] The adaptive clustering method is combined with the mean shift algorithm to make the interval division of the point cloud reflection intensity more adaptive, ensuring that the initialization of the cluster center is based on the data distribution characteristics rather than fixed rules, thereby improving the accuracy and stability of clustering. The mean shift clustering algorithm calculates the weighted mean through the density kernel function, which can ensure that the point cloud cluster center is continuously optimized during the iteration process, so that the final classification result is more in line with the actual distribution of the data. On this basis, the Transformer model introduces position encoding, and gives spatial information to the point cloud data through trigonometric functions, so that the network can learn the relative position information between point cloud points, thereby improving the robustness of point cloud classification. In addition, the Transformer model is combined with point cloud data modeling, so that high-dimensional features can be further classified through the multi-layer perceptron MLP, ensuring that the final classification result is more stable, and avoiding the failure of traditional Euclidean distance-based classification methods when point cloud data is sparse or complex.

[0126] S3, fitting a quadratic curve equation based on the power line label point cloud to obtain a power line model, and identifying and marking the intersection area according to the intersection distance of the power line point cloud;

[0127] Preferably, a power line model is obtained by fitting a quadratic curve equation based on the power line label point cloud, the intersection area is identified and marked according to the intersection distance of the power line point cloud, RANSAC is used to model the point cloud data based on the power line label, the power line point cloud is projected onto the XOY plane, and the two-dimensional trajectory of the power line is extracted;

[0128] Use RANSAC for straight line fitting, randomly sample two points in the power line point cloud, and calculate the hypothetical straight line equation, which can be: , and calculate the vertical error from all points of the power line label point cloud data to the assumed straight line equation, determine the straight line threshold based on historical data, and count the number of fitted point clouds that conform to the straight line, repeat sampling, and take the assumed straight line equation with the largest number of fitted point clouds as the fitting result, and obtain the optimal straight line variance model;

[0129] Based on the straight line variance model, the dichotomy method is used to optimize the modeling of the power line, calculate and determine the point farthest from the fitting line, and set the dividing point for the power line, which is expressed as:

[0130] ;

[0131] in Indicates the x-coordinate of the boundary point cloud, Indicates the x-coordinate of the point cloud with the maximum error, Indicates the minimum x value of the current fitting area. represents the adjustment factor, which is determined based on historical data;

[0132] The fitting curves of the two parts are calculated based on the dividing point, which are expressed as:

[0133] ;

[0134] ;

[0135] in and Respectively represent the power line data points on both sides of the dividing point, and is the coordinate of the i-th data point;

[0136] The quadratic curves are fitted using the least squares method, and the quadratic curve equations on both sides of the power line are solved. The fitting error is calculated using the mean square error (MSE). The error threshold is determined based on historical data. If the fitting error is greater than or equal to the error threshold, the binary method is continued for redivision and fitting. If the fitting error is less than the error threshold, the calculation is terminated and the final power line model is output, which is expressed as:

[0137] ;

[0138] in represents the final power line function, , as well as are the quadratic function coefficients of the curve on one side, determined by the least squares method, , as well as represents the quadratic function coefficient of the curve on the other side, which is determined by the least squares method, and x represents the horizontal coordinate of the power line point cloud data;

[0139] Determine different power line equations based on the final power line model, calculate the vertical distance from the data of each power line point cloud to the power line equation, determine the distance threshold based on historical data, and if the calculated vertical distance is less than or equal to the distance threshold, determine that the point cloud belongs to the power line fitted by the power line equation;

[0140] After determining the power line to which the power line point cloud belongs, the Euclidean distance of point cloud pairs of different power line lines is calculated, and the shortest crossing distance is calculated. The crossing threshold is determined based on historical data. If the shortest crossing distance is less than or equal to the crossing threshold, it is judged that the power line lines to which the point cloud pair belongs have regional intersections, and the point clouds in the intersection area are marked.

[0141] In point cloud data, power lines are usually in the shape of catenary curves, but in local areas, they can be approximated as straight line segments. Using RANSAC for straight line fitting can identify and remove abnormal points, including noise and obstacles, and obtain the local straight line segment model of the power line. Power lines can usually be approximated as a straight line in a small range. The fitted straight line can be used for subsequent curve modeling. The local straight line model provided by RANSAC can be used as the initial input of the dichotomy optimization for subsequent curvature adjustment.

[0142] Through the RANSAC straight line fitting method, the optimal power line trajectory can be found in the noisy point cloud data. The straight line variance model and error calculation are used to screen the optimal fitting results to improve the accuracy and robustness of power line modeling. The power line is optimized and modeled based on the dichotomy method, which can further refine the local structure of the power line, so that the curve fitting can be more in line with the actual power line shape and reduce the overall fitting error. The combination of RANSAC fitting, power line point cloud classification and intersection area detection makes the modeling of power line point cloud more accurate, while ensuring that the trajectories of different power lines are clearly separable and reducing the misjudgment of line intersection areas. The dichotomy optimization modeling is combined with error control, so that the power line model can reduce the consumption of computing resources and improve the modeling efficiency while ensuring accuracy. Finally, through the synergy of vertical distance calculation, power line attribution determination and intersection area marking, a complete power line topology structure can be constructed to provide accurate data support for subsequent drone inspections and fault detection.

[0143] S4, clustering and distinguishing point clouds based on poles, vegetation and building labels, planning task areas based on pole clustering, calculating task allocation weights for drones, and performing drone task allocation;

[0144] Preferably, clustering and differentiation are performed based on the point cloud of pole tower, vegetation and building labels, task area is planned based on the tower clustering, task allocation weight of the UAV is calculated, and task allocation of the UAV is performed, and different pole towers, vegetation and buildings are marked respectively by DBSCAN clustering based on the point cloud of the tower, vegetation and building labels, and the covariance matrix is ​​calculated using PCA, the center points of different pole towers, vegetation and buildings are calculated respectively, and the heights of different pole towers, vegetation and buildings are determined;

[0145] Based on the minimum safe distance of drones, set a no-fly zone for the point cloud in the intersection area, and set a safe flight altitude directly over towers, vegetation and buildings;

[0146] The maximum distance between different towers is calculated based on the center point of the tower, and K-means++ is used to cluster the task areas. The distance from each tower to the center of the nearest task area is calculated, which is expressed as:

[0147] ;

[0148] in represents the central tower position of the kth mission area, represents the tower location data set, p represents the center point of the tower, and c represents the center tower location of the candidate task area;

[0149] Repeatedly update the mission area center, expressed as:

[0150] ;

[0151] in represents all towers in the task area k;

[0152] When the center of the task area no longer changes significantly, the clustering process ends, and after clustering the task areas, the tower set of each task area is determined;

[0153] The task allocation weight is calculated based on the task area and the UAV, which is expressed as:

[0154] ;

[0155] in represents the allocation weight of UAV j to perform mission area k, represents the shortest path distance from the starting point of UAV j to the tower i, represents the cruising speed of UAV j, represents the total inspection time of all towers in task area k;

[0156] Based on the allocation weight, the Hungarian algorithm is used to optimize the allocation, maximize the allocation weight of the UAV, and complete the allocation of the UAV in the mission area.

[0157] By using DBSCAN clustering to classify the point cloud data of poles, vegetation and buildings, and calculating the covariance matrix through PCA to determine their respective center points and heights, point cloud information of different categories can be accurately attributed, and the structural features of poles, vegetation and buildings can be accurately extracted, providing an accurate geographic information basis for subsequent UAV mission planning. Based on the center point information of poles, vegetation and buildings;

[0158] Based on the calculation of the maximum distance between different towers at the center point of the tower, the task area can be reasonably divided, and the task area can be optimized through K-means++ clustering, so that the drone can work efficiently in the assigned task area, reduce repeated inspections, and improve the overall task efficiency. In the task allocation stage, by calculating the allocation weights of the task area and the drone, the adaptability of different drones to perform different task areas can be quantified to ensure the rationality of task allocation. The Hungarian algorithm is used for task optimization allocation, which can maximize the allocation weight of the drone, ensure the load balance of the drone task, reduce the overall task time, and improve the execution efficiency of the inspection task. Through the combination of DBSCAN clustering, PCA calculation, K-means++ task area division, task weight calculation and Hungarian algorithm optimization task allocation, the entire drone inspection task has accurate task area division, efficient task scheduling and safe flight planning. Based on the point cloud classification of towers, vegetation and buildings, combined with the flight altitude and safety requirements of drones, the risk of inspection tasks can be effectively reduced and the reliability of drone task execution can be improved. At the same time, the synergy of task area clustering and task optimization allocation can reduce resource waste during the execution of drone inspection tasks and improve the overall inspection coverage and inspection efficiency.

[0159] S5, perform preliminary path planning through the mission area, determine the Euclidean distance matrix between all inspection points, use genetic algorithm to obtain the optimal inspection order of different inspection points, and determine the optimal control command based on minimizing track deviation and control input as the objective function;

[0160] Preferably, preliminary path planning is performed through the mission area to determine the Euclidean distance matrix between all inspection points, and the optimal inspection order of different inspection points is obtained by using genetic algorithm. Based on minimizing the track deviation and control input as the objective function, the optimal control command is determined. Based on the drones in different mission areas, RRT is used The fast random tree is used for preliminary path planning. At the same time, based on all the inspection points in the preliminary path planning, the Euclidean distance matrix between all the inspection points is determined. The traveling salesman problem TSP is used to set the minimization of the total inspection path as the goal to optimize the UAV inspection path.

[0161] The genetic algorithm GA-TSP is used for approximate solution, in which multiple random path arrangements are generated, the total inspection distance of each path is calculated, and the crossover mutation operation is performed through partial matching crossover PMX to obtain the optimal inspection sequence and corresponding inspection length of different inspection points;

[0162] Based on the inspection trajectory of the UAV determined by the optimal inspection sequence, the MPC model predictive control is used to smooth the trajectory and control the speed of the UAV. The objective function is based on minimizing the track deviation and control input, which is expressed as:

[0163] ;

[0164] Where J represents the target deviation value, represents the total number of time steps predicted by the MPC model, The position of the drone at time g, represents the position of the drone’s inspection point at the gth moment, represents the speed of the drone at the gth moment;

[0165] Based on the objective function of minJ, the control instructions of the optimal control input, including linear velocity and angular velocity, are determined through the quadratic programming QP solver;

[0166] The calculated control instructions are input and sent to the UAV to adjust the flight trajectory.

[0167] By adopting RRT Perform preliminary path planning to enable the drone to generate a collision-free and feasible inspection path, and calculate the Euclidean distance matrix between all inspection points based on the path to ensure that the distance relationship between the inspection points is clear, thus providing a basis for subsequent path optimization. By setting the TSP traveling salesman problem to minimize the total inspection path as the goal, the total flight distance of the drone can be reduced, the inspection efficiency can be improved, and unnecessary repeated flights can be avoided;

[0168] The genetic algorithm GA-TSP is used for approximate solution, multiple path arrangements are randomly generated and the total inspection distance of each path is calculated, so that the optimization of the inspection order can converge quickly. The crossover mutation operation is performed through partial matching crossover PMX, which can retain the better individuals and introduce appropriate randomness to improve the diversity of the final solution, thereby effectively avoiding local optimality and ensuring the global optimal solution of the UAV inspection task path. The optimal inspection order finally ensures that the UAV performs inspections along the shortest path, thereby reducing energy consumption and improving the execution efficiency of the inspection task.

[0169] Based on the MPC model predictive control, the trajectory of the UAV is smoothed and the speed is controlled, so that the UAV can fly more stably along the optimized inspection path, with the goal of minimizing the track deviation and control input, ensuring that the UAV is as close to the optimal path as possible during the inspection process, avoiding violent turns and speed fluctuations, and improving flight stability. The optimal control instructions, including linear velocity and angular velocity, are determined by the quadratic programming QP solver, making the motion control of the UAV more precise, thereby improving the reliability and safety of the inspection task. Finally, the calculated control instructions are sent to the UAV to adjust the flight trajectory in real time to ensure that the inspection task can still be completed efficiently in a complex environment. The combined effect of path planning, TSP shortest path optimization, genetic algorithm for solving inspection sequence and MPC trajectory smoothing control enables the UAV inspection mission to have efficient path planning, high-precision task scheduling and stable flight control capabilities. By reducing the total inspection distance, optimizing the inspection sequence, controlling track deviation, etc., it ensures that the UAV inspection mission can not only improve coverage but also reduce energy consumption, while ensuring the safety and accuracy of the inspection mission.

[0170] S6, securely transmit and store the control instruction inspection path and inspection point sequence data;

[0171] Preferably, the control instruction inspection path and inspection point sequence data are securely transmitted and stored, which means that the control instruction data is end-to-end encrypted using AES-256, and the inspection path and inspection point sequence data are encrypted using AES-128 and written into a read-only database;

[0172] MQTT is used to securely transmit encrypted data, which is then transmitted to the cloud via wireless transmission technology for log data storage.

[0173] By using AES-256 end-to-end encryption to protect the drone's control command data, it can ensure that the control commands will not be tampered with or stolen during the transmission and storage process, thereby improving the security of the drone's flight commands. MQTT (Message Queue Telemetry Transport Protocol) is used for secure transmission, so that encrypted data can communicate efficiently and stably in a low-bandwidth, high-latency environment, so that log data can be stored in real time, and remote access and task backtracking are supported, thereby improving the manageability and long-term storage capacity of drone inspection data, ensuring that inspection task data is auditable and recoverable, and supporting subsequent task analysis.

[0174] This embodiment also provides a system for a method for planning a track for collaborative inspection of distribution network lines by multiple UAVs, including:

[0175] Data processing module, for collecting and denoising point cloud data;

[0176] Point cloud feature extraction module, which classifies the collected point cloud data and extracts spatial location information;

[0177] Cross-recognition module, which models power lines and identifies cross-recognition areas;

[0178] The UAV mission area planning module divides the UAV mission area and assigns UAVs to perform inspection tasks;

[0179] Path planning module, generating the best inspection path and optimizing the flight trajectory of the drone;

[0180] The control command execution module calculates the optimal control command and sends it to the UAV for task execution;

[0181] The data security transmission module is used to securely transmit and store UAV inspection data, control instructions and mission paths.

[0182] This embodiment also provides a computer device, which is suitable for the case of a trajectory planning method for collaborative inspection of distribution network lines by multiple UAVs, including: a memory and a processor; the memory is used to store computer executable instructions, and the processor is used to execute computer executable instructions to implement the trajectory planning method for collaborative inspection of distribution network lines by multiple UAVs as proposed in the above embodiment.

[0183] The computer device may be a terminal, and the computer device includes a processor, a memory, a communication interface, a display screen and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device may be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device may be a touch layer covered on the display screen, or a key, trackball or touchpad provided on the housing of the computer device, or an external keyboard, touchpad or mouse, etc.

[0184] This embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, the method for realizing the trajectory planning of multi-UAV collaborative inspection of distribution network lines proposed in the above embodiment is implemented; the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (Static Random Access Memory, referred to as SRAM), electrically erasable programmable read-only memory (Electrically Erasable Programmable Read-Only Memory, referred to as EEPROM), erasable programmable read-only memory (Erasable Programmable Read Only Memory, referred to as EPROM), programmable read-only memory (Programmable Red-Only Memory, referred to as PROM), read-only memory (Read-Only Memory, referred to as ROM), magnetic storage, flash memory, disk or optical disk.

[0185] In summary, the present invention combines the adaptive clustering method with the mean shift algorithm to make the interval division of the point cloud reflection intensity more adaptive, ensure that the initialization of the cluster center is based on the data distribution characteristics, and combine RANSAC fitting, power line point cloud classification and intersection area detection to make the modeling of the power line point cloud more accurate, while ensuring that the trajectories of different power lines are clearly separable, reducing the misjudgment of line intersection areas, and combining binary optimization modeling with error control, so that the power line model can reduce the consumption of computing resources while ensuring accuracy. Through the combination of task weight calculation and Hungarian algorithm optimization task allocation, the entire UAV inspection task has accurate task area division, efficient task scheduling and safe flight planning. By using the genetic algorithm GA-TSP for approximate solution, multiple path arrangements are randomly generated and the total inspection distance of each path is calculated, so that the optimization of the inspection order can converge quickly, ensuring that the UAV inspection task can both improve the coverage rate and reduce energy consumption.

[0186] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A method for trajectory planning of multiple UAVs for collaborative inspection of distribution network lines, characterized in that: include: Collect point cloud data based on drones and perform data denoising sampling; Adaptive clustering is performed based on the reflection intensity of the sampled data, position coding is introduced to describe the point cloud coordinate information and construct point cloud vectors, and the point cloud is labeled and classified; The power line model is obtained by fitting the quadratic curve equation based on the power line label point cloud, and the intersection area is identified and marked according to the intersection distance of the power line point cloud; Clustering is performed based on the point cloud of towers, vegetation, and building labels. The task area is planned based on the tower clustering, and the task allocation weight of the drone is calculated to perform the drone task allocation. Perform preliminary path planning through the mission area, determine the Euclidean distance matrix between all inspection points, use genetic algorithm to obtain the optimal inspection sequence of different inspection points, and determine the optimal control instructions based on minimizing track deviation and control input as the objective function; Safely transmit and store control instruction inspection paths and inspection point sequence data; The preliminary path planning is carried out through the mission area to determine the Euclidean distance matrix between all inspection points, and the optimal inspection order of different inspection points is obtained by using genetic algorithm. The optimal control instruction is determined based on minimizing the track deviation and control input as the objective function. Based on the drones in different mission areas, RRT is used The fast random tree is used for preliminary path planning. At the same time, based on all the inspection points in the preliminary path planning, the Euclidean distance matrix between all the inspection points is determined. The traveling salesman problem TSP is used to set the minimization of the total inspection path as the goal to optimize the UAV inspection path. The genetic algorithm GA-TSP is used for approximate solution, in which multiple random path arrangements are generated, the total inspection distance of each path is calculated, and the crossover mutation operation is performed through partial matching crossover PMX to obtain the optimal inspection sequence and corresponding inspection length of different inspection points; Based on the inspection trajectory of the UAV determined by the optimal inspection sequence, the MPC model predictive control is used to smooth the trajectory and control the speed of the UAV. The objective function is based on minimizing the track deviation and control input, which is expressed as: ； Where J represents the target deviation value, represents the total number of time steps predicted by the MPC model, The position of the drone at time g, represents the position of the drone’s inspection point at the gth moment, represents the speed of the drone at the gth moment; Based on the objective function of minJ, the control instructions of the optimal control input, including linear velocity and angular velocity, are determined through the quadratic programming QP solver; The calculated control instructions are input and sent to the UAV to adjust the flight trajectory.

2. The method for planning the track of multiple UAVs coordinating inspection of distribution network lines as claimed in claim 1, characterized in that: The point cloud data is collected based on the drone, and data denoising sampling is performed, and the flight height of the drone is set based on the maximum inspection height and the minimum safety height of multiple drones in the inspection task area; The drone is equipped with a laser radar to collect point cloud data, and the scanning range radius is determined based on the viewing angle of the laser radar and the flight altitude of the drone; Obtain point cloud data of the scanning range, including point cloud coordinates and reflection intensity; The point cloud data is downsampled using a three-dimensional voxel grid to obtain the sampling point cloud coordinates based on the voxel grid, and then denoised by statistical filtering.

3. The method for planning the track of multi-UAV collaborative inspection of distribution network lines as claimed in claim 2, characterized in that: The adaptive clustering is performed according to the reflection intensity of the sampled data, the position coding is introduced to describe the point cloud coordinate information and the point cloud vector is constructed, the point cloud is labeled and classified, and the center value of the reflection intensity interval is determined based on the reflection intensity of the denoised point cloud coordinates through the adaptive clustering method, which is expressed as: ； in represents the center value of the jth reflection intensity interval, The total number of bins representing the range of reflection intensity, and Respectively represent the maximum and minimum values ​​of the reflection intensity; Calculate the number of points in each reflection intensity interval for histogram counting, expressed as: ； ； in The reflection intensity range is The number of point clouds, Represents the total number of point cloud data, represents the i-th reflection intensity value, represents the center value of the jth reflection intensity interval, represents the indicator function, It represents the ratio of the difference between the maximum and minimum values ​​of the reflection intensity to the total number of intervals B; based on Discretize the reflection intensity to form a histogram, and calculate the proportion of the reflection intensity range in the entire point cloud according to the total number of point clouds with different reflection intensities; Based on the sum of the proportion mean and the standard deviation as the proportion threshold, the number of interval reflection intensity point clouds greater than the proportion threshold is As the window center for initializing mean shift clustering, the mean shift algorithm is used for adaptive clustering, in which the weighted mean in the window is calculated by the density and function, expressed as: ； in represents the updated kth window center, n represents the total number of reflection intensities, represents the density kernel function, represents the kth window center of mean shift clustering, and W represents the window width; Iterate and update the window center. When the difference of the updated window center is no longer obvious during the continuous iteration, stop the iteration and classify the point cloud data into the closest cluster center through the attribution relationship. The Transformer model is used to introduce position encoding to describe the point cloud coordinate information, and trigonometric functions are used for encoding, which is expressed as: ； in Represents the position encoding of the i-th point cloud coordinates, , and are the three-dimensional point cloud coordinate information of the i-th point cloud, represents the frequency parameter; Construct a point cloud vector, including point cloud coordinates, reflection intensity, cluster label and position encoding, use the self-attention mechanism based on the Transformer model to determine the query, key and value, and calculate the attention relationship between point clouds, expressed as: ； ； in represents the attention weight between the i-th point cloud and the j-th point cloud, and Represents the query and key of the i-th point cloud and the j-th point cloud respectively. The T in represents the transpose calculation, represents the dimension of the Key matrix, represents the Transformer output of the i-th point cloud, Represents the value of the j-th point cloud; Use multi-layer perceptron MLP to After the final classification, different classification labels are obtained, including label classification of poles, power lines, vegetation, ground and buildings.

4. The method for planning the track of multiple UAVs coordinating inspection of distribution network lines as claimed in claim 3, characterized in that: The power line model is obtained by fitting a quadratic curve equation based on the power line label point cloud, identifying and marking the intersection area according to the intersection distance of the power line point cloud, modeling is performed using RANSAC based on the point cloud data of the power line label, the power line point cloud is projected onto the XOY plane, and the two-dimensional trajectory of the power line is extracted; Use RANSAC to fit the line, randomly sample two points in the power line point cloud, and calculate the assumed line equation. Calculate the vertical error from all points of the power line label point cloud data to the assumed line equation. Determine the line threshold based on historical data, and count the number of fitted point clouds that conform to the line. Repeat the sampling, and take the assumed line equation with the largest number of fitted point clouds as the fitting result, and obtain the optimal line variance model. Based on the straight line variance model, the dichotomy method is used to optimize the modeling of the power line, calculate and determine the point farthest from the fitting line, and set the dividing point for the power line, which is expressed as: ； in Indicates the x-coordinate of the boundary point cloud, Indicates the x-coordinate of the point cloud with the maximum error, Indicates the minimum x value of the current fitting area. represents the adjustment factor; The fitting curves of the two parts are calculated based on the dividing point, which are expressed as: ； ； in and Respectively represent the power line data points on both sides of the dividing point, and is the coordinate of the i-th data point cloud; The quadratic curves are fitted using the least squares method, and the quadratic curve equations on both sides of the power line are solved. The fitting error is calculated using the mean square error (MSE). The error threshold is determined based on historical data. If the fitting error is greater than or equal to the error threshold, the binary method is continued for redivision and fitting. If the fitting error is less than the error threshold, the calculation is terminated and the final power line model is output, which is expressed as: ； in represents the final power line function, , as well as are the quadratic function coefficients of the curve on one side, , as well as represents the quadratic function coefficient of the curve on the other side, and x represents the horizontal coordinate of the power line point cloud data; Determine different power line equations based on the final power line model, calculate the vertical distance from the data of each power line point cloud to the power line equation, determine the distance threshold based on historical data, and if the calculated vertical distance is less than or equal to the distance threshold, determine that the point cloud belongs to the power line fitted by the power line equation; After determining the power line to which the power line point cloud belongs, the Euclidean distance of point cloud pairs of different power line lines is calculated, and the shortest crossing distance is calculated. The crossing threshold is determined based on historical data. If the shortest crossing distance is less than or equal to the crossing threshold, it is judged that the power line lines to which the point cloud pair belongs have regional intersections, and the point clouds in the intersection area are marked.

5. The method for planning the track of multiple UAVs coordinating inspection of distribution network lines as claimed in claim 4, characterized in that: The point cloud based on the tower, vegetation and building labels is clustered and distinguished, the task area is planned based on the tower clustering, the task allocation weight of the drone is calculated, and the drone task allocation is performed. Based on the point cloud of the tower, vegetation and building labels, different towers, vegetation and buildings are marked respectively by DBSCAN clustering, and the covariance matrix is ​​calculated by PCA, the center points of different towers, vegetation and buildings are calculated respectively, and the heights of different towers, vegetation and buildings are determined; Based on the minimum safe distance of drones, set a no-fly zone for the point cloud in the intersection area, and set a safe flight altitude directly over towers, vegetation and buildings; The maximum distance between different towers is calculated based on the center point of the tower, and K-means++ is used to cluster the task areas. The distance from each tower to the center of the nearest task area is calculated, which is expressed as: ； in represents the central tower position of the kth mission area, represents the tower location data set, p represents the center point of the tower, and c represents the center tower location of the candidate task area; Repeatedly update the mission area center, expressed as: ； in represents all towers in the task area k; When the center of the task area no longer changes significantly, the clustering process ends, and after clustering the task areas, the tower set of each task area is determined; The task allocation weight is calculated based on the task area and the UAV, which is expressed as: ； in represents the allocation weight of UAV j to perform mission area k, represents the shortest path distance from the starting point of UAV j to the tower i, represents the cruising speed of UAV j, represents the total inspection time of all towers in task area k; Based on the allocation weight, the Hungarian algorithm is used to optimize the allocation, maximize the allocation weight of the UAV, and complete the allocation of the UAV in the mission area.

6. The method for planning the track of multiple UAVs coordinating inspection of distribution network lines as claimed in claim 1, characterized in that: The secure transmission and storage of the control instruction inspection path and inspection point sequence data refers to the end-to-end encryption of the control instruction data using AES-256, and the inspection path and inspection point sequence data are encrypted using AES-128 and written into a read-only database; MQTT is used to securely transmit encrypted data, which is then transmitted to the cloud via wireless transmission technology for log data storage.

7. A system for a method for planning a track for collaborative inspection of distribution network lines by multiple unmanned aerial vehicles, based on the method for planning a track for collaborative inspection of distribution network lines by multiple unmanned aerial vehicles as claimed in any one of claims 1 to 6, characterized in that: include, Data processing module, for collecting and denoising point cloud data; Point cloud feature extraction module, which classifies the collected point cloud data and extracts spatial location information; Cross-recognition module, which models power lines and identifies cross-recognition areas; The UAV mission area planning module divides the UAV mission area and assigns UAVs to perform inspection tasks; Path planning module, generating the best inspection path and optimizing the flight trajectory of the drone; The control command execution module calculates the optimal control command and sends it to the UAV for task execution; The data security transmission module is used to securely transmit and store UAV inspection data, control instructions and mission paths.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the trajectory planning method for collaborative inspection of distribution network lines by multiple UAVs according to any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, the steps of the trajectory planning method for collaborative inspection of distribution network lines by multiple UAVs according to any one of claims 1 to 6 are implemented.

Images