Method for automatically correcting and updating route of electric power inspection unmanned aerial vehicle

Abstract

The invention relates to a method for automatically correcting and updating the route of an electric power inspection unmanned aerial vehicle, which comprises the following steps of: S1, collecting and integrating geographic information system (GIS) data of an inspection area, and constructing a topological graph; s2, obtaining detailed point cloud data of the ground and the power grid facilities, and constructing a 3D model according to the obtained point cloud data; s3, based on the topological graph and the 3D model, extracting a key position from the 3D model by using an algorithm, determining an initial waypoint, extracting waypoints from the key position of the power grid equipment, connecting all the extracted waypoints by using a path planning algorithm, and generating an initial inspection route; s4, correcting the initial inspection route in combination with real-time weather, equipment priority and historical data; and S5, combining the combination degree RTK GPS and an environment adaptive VIO algorithm to realize sub-meter positioning precision, and after the key position centralization is successful, automatically updating the waypoint and immediately uploading the waypoint to a control center to realize automatic updating of the route. The accuracy and efficiency of unmanned aerial vehicle inspection are effectively improved.

Description

Technical Field

[0001] The present invention relates to the field of electric power inspection, and in particular to a method for automatically correcting and updating a route of an electric power inspection unmanned aerial vehicle. Background Art

[0002] In recent years, drone technology for power inspection has become a promising innovative tool in the power industry. With the complexity of power grid systems and the wide geographical distribution, traditional manual operations have gradually exposed problems such as low efficiency and safety hazards. The introduction of drone technology aims to inspect and monitor overhead lines, substations and related equipment from an aerial perspective, providing power companies with an efficient, accurate and safe solution.

[0003] The current power drone inspection system mainly relies on pre-planned routes. However, due to the dynamic changes in environmental and on-site conditions, such as weather changes, sudden obstacles or temporary changes in power equipment, a simple preset route may not be sufficient to meet all actual needs. Therefore, the ability to automatically correct and update unmanned routes has become an important direction to improve the flexibility and reliability of drone inspection systems. Summary of the invention

[0004] In order to solve the above problems, the purpose of the present invention is to provide a method for automatically correcting and updating the route of a power inspection drone, so as to effectively improve the accuracy and efficiency of drone inspections.

[0005] To achieve the above object, the present invention adopts the following technical solutions:

[0006] A method for automatically correcting and updating the route of a power inspection drone, characterized in that it includes the following steps:

[0007] Step S1: Collect and integrate the Geographic Information System (GIS) data of the inspection area and construct a topological map;

[0008] Step S2: Obtaining detailed point cloud data of the ground and power grid facilities, and constructing a 3D model based on the obtained point cloud data;

[0009] Step S3: Based on the topology map and the 3D model, an algorithm is used to extract key locations from the 3D model, initial waypoints are determined, waypoints are extracted from key locations of power grid equipment, and a path planning algorithm is used to connect all extracted waypoints to generate an initial inspection route;

[0010] Step S4: Using AI algorithms, combined with real-time weather, equipment priority, and historical data, correct the initial inspection route;

[0011] Step S5: The UAV performs inspection based on the corrected inspection route, and combines RTK GPS with the environment-adaptive VIO algorithm to achieve sub-meter positioning accuracy. After successful centralization at the key position, the waypoints are automatically updated and immediately uploaded to the control center to achieve automatic route update.

[0012] Furthermore, step S1 is specifically as follows:

[0013] Obtain Geographic Information System (GIS) data of the inspection area, including topography, infrastructure location, power towers, line connection points and their spatial coordinates;

[0014] Convert all data formats to common GIS formats and project them to the same coordinate system;

[0015] In GIS tools, create multi-level data views, including basic terrain, line networks, and power facility distribution;

[0016] Establish the topological relationship between towers, line connection points and terrain changes, construct a topological map, and use graph algorithms to check the connectivity of the topological map to ensure that there are paths between all power grid nodes.

[0017] Furthermore, the topological relationship between the towers, line connection points and terrain changes is established, a topological map is constructed, and the connectivity of the topological map is checked using a graph algorithm to ensure that there are paths between all grid nodes, as follows:

[0018] The towers and line connection points are defined as nodes in the graph, and adjacent nodes are connected into edges according to the power lines;

[0019] Attach attributes to each node, including location coordinates, height, and tower type; attach attributes to each edge, including length, cable type, and voltage level;

[0020] Use DEM data to identify terrain changes, including valleys and ridges, and calculate the slope and aspect of each location to assess the impact on towers and paths;

[0021] In the topological graph, terrain changes are introduced, and each node not only saves its position, but also the height and slope of the ground;

[0022] Construct a topological graph G = (V, E), where V is the node, including the tower and the connection point, and E is the edge, that is, the connecting line. Both nodes and edges carry terrain-related information.

[0023] Based on DFS, the entire graph is explored to identify whether each node can be accessed through a direct or indirect path.

[0024] Furthermore, step S2 is specifically as follows:

[0025] Use LiDAR technology to collect fine point cloud data of power grids and terrain, and transform the points collected by LiDAR from the sensor coordinate system to the global coordinate system:

[0026]

[0027] Among them, (x g ,y g ,z g ) is the global coordinate, (x t ,y t ,z t ) is the sensor coordinate, R is the rotation matrix, and t is the translation vector;

[0028] Use Voxel Grid simplification algorithm to reduce data volume while retaining structural integrity;

[0029] Apply terrain analysis algorithms to separate ground points from the data;

[0030] Identify and separate power grid infrastructure, including power lines and towers, and group and label them using clustering methods;

[0031] The point cloud data is converted into a three-dimensional surface model using the surface reconstruction algorithm Delaunay triangulation, and the reconstruction is optimized in combination with normal vector estimation to obtain the final 3D model.

[0032] Furthermore, the facility data is converted into a three-dimensional surface model using the surface reconstruction algorithm Delaunay triangulation, as follows:

[0033] Construct the initial hypertetrahedron, which contains all the input points. For each new point, find the tetrahedron area where it is located and re-split it;

[0034] Determine the visible polyhedra adjacent to the new point and remove them to create a new polyhedron;

[0035] After inserting the new point, recheck and correct the affected area to ensure that the empty sphere property is satisfied, that is, there are no other points in the circumscribed sphere of any triangle. For the tetrahedron T = △ABCD, calculate the circumscribed sphere of the tetrahedron and determine whether correction is needed:

[0036]

[0037] Among them, (x a ,y a ,z a ),(x b ,y b ,z b ),(x c ,y c ,z c ),(xd ,y d ,z d ) are the three-dimensional coordinates of the four vertices A, B, C, and D, which form an existing tetrahedron; (x e ,y e ,z e ) is the three-dimensional coordinate of the newly inserted point E. It is necessary to determine whether this point is inside the circumscribed sphere of the tetrahedron formed by points A, B, C, and D. If the value of the determinant is > 0, then the new point E(x e ,y e ,z e ) is inside the circumscribed sphere. If ≤ 0, then point E is on or outside the sphere;

[0038] Based on the Owyer-Watson algorithm, the subdivision is gradually optimized by searching for tetrahedrons and correcting them point by point; and the three-dimensional surface generation is optimized by calculating the normal vector to better represent the actual terrain and the surface of the object.

[0039] Furthermore, based on the Owyer-Watson algorithm, the tetrahedron is searched and corrected point by point, the subdivision is gradually optimized, and the three-dimensional surface generation is optimized by calculating the normal vector to better represent the actual terrain and the surface of the object, as follows:

[0040] Select a new point P from the point cloud and insert it into the current mesh structure;

[0041] Find all tetrahedrons that contain the circumsphere of point P. These tetrahedrons will be invalidated by the introduction of new points and need to be removed from the structure.

[0042] Create new tetrahedrons on the boundaries of the holes left by the previous deletion step;

[0043] For each edge (e) on the boundary of the hole, generate a new tetrahedron from P to the endpoints of (e).

[0044] For the newly created tetrahedrons, check the Delaunay conditions. If there are non-Delaunay tetrahedrons in a tetrahedron, correct them by flipping the edges.

[0045] By calculating the normal vector, the orientation of each face of the 3D mesh is obtained;

[0046] For the triangular face △ABC, its normal vector N is calculated by the cross product of the two edges:

[0047]

[0048] Among them, (BA)×(CA) is a vector perpendicular to the plane of △ABC△ABC, and its direction follows the right-hand rule;

[0049] Taking the vertex as the center, take the average of the normal vectors of all surrounding triangles as the smooth normal vector of this vertex;

[0050]

[0051] Among them, N i is the normal vector of the i-th adjacent triangle around the vertex;

[0052] Noise is reduced by iteratively moving vertices to the average position of neighboring vertices;

[0053]

[0054] Where neighbors(V) is the set of all vertices directly connected to vertex V; λ is the smoothing factor; V is the current vertex; V j is the neighboring vertex of the current vertex;

[0055] Merges two adjacent vertices into one to reduce the polygon count while maintaining the overall shape.

[0056] Furthermore, step S3 is specifically as follows:

[0057] The tower base and tower top are identified by analyzing the tower geometry and point density in the 3D model. The cable connection points are identified from the model based on the connection line features. The endpoints are determined by gradually tracing the linear structure in the 3D model.

[0058] Ensure that waypoints include all identified key locations so that all required checkpoints are covered. The extracted waypoints should ensure that high-risk points throughout the facility are covered.

[0059] Additional waypoints were added at points of spatial inhomogeneity to ensure the integrity of the inspection path;

[0060] The path planning algorithm Dijkstra is used to plan the optimal inspection route and generate the initial inspection route.

[0061] Further, step S4 is as follows:

[0062] Obtain weather information in real time through API, including wind speed, rainfall, and temperature, and determine the maintenance priority of each device based on the current operating status and failure probability data of the device;

[0063] Combine historical fault records with previous inspection results to identify high-risk equipment and areas;

[0064] Let state S be the current inspection location / point and its current environmental factors;

[0065] Action A is to select the next inspection destination;

[0066] The reward function R takes into account inspection efficiency, safety, and equipment priority, and designs positive and negative incentives:

[0067] R=α×Prioritize(E)+β×WeatherFactor(W)-γ×Cost(C);

[0068] Among them, Prioritize (E) is the positive benefit brought by the equipment priority; WeatherFactor (W) is the importance adjusted according to weather conditions; Cost (C) is the comprehensive cost of inspection time, distance and risk;

[0069] The policy gradient method is used to adjust and optimize the path selection to minimize the cost, maximize the path effectiveness, and correct the initial inspection route.

[0070] Furthermore, the policy gradient method is used to adjust and optimize the path selection to minimize the cost, maximize the path effectiveness, and correct the initial inspection route, as follows:

[0071] Initialize the parameters θ to generate a random strategy;

[0072] In each round, according to the current strategy π θ (a|s) generates a complete path;

[0073] At the end of the path, calculate the cumulative reward R t , represents the total reward starting from time t:

[0074]

[0075] Where γ is the discount factor, r(s k ,a k ) is to perform action a k The reward you get after that;

[0076] Update the policy parameters θ using the gradient ascent method:

[0077] θ←θ+α▽ θ J(θ);

[0078] Where α is the learning rate; θ J(θ) is the policy gradient:

[0079]

[0080] Estimated θ Use Monte Carlo methods to sample and estimate policy gradients from simulated paths;

[0081] Repeat path generation, evaluation, and strategy update until convergence or the maximum number of iterations is reached, and correct the initial inspection route.

[0082] Furthermore, step S5 is specifically as follows:

[0083] Based on differential GPS, it provides centimeter-level absolute position accuracy through real-time communication between ground base stations and drones; combined with visual sensors and IMU, it achieves high-frequency, environment-adaptive position estimation;

[0084] While the UAV is flying along the planned path, path tracking is performed through RTK and VIO joint positioning;

[0085] Perform centering operations at key locations on the planned path to ensure that the capture or detection device is in the center of the field of view; and use feature detection to analyze the video frame to ensure that the key target is in the center of the image;

[0086] When the drone is successfully centered at the key position, the state vector is updated in real time based on VIO combined with vision and IMU data. According to the VIO real-time update result and the deviation of the centralized position (if it is not completely centralized), it is updated to calculate the new waypoint; the VIO real-time update state vector, including position p, speed v, and attitude q, is updated in real time through the following equations:

[0087]

[0088] v t+1 =v t +a t Δt;

[0089]

[0090] Among them, a t is the acceleration measurement, Δθ t is the integral of the angular velocity, Represents quaternion multiplication;

[0091] The newly generated waypoints are immediately uploaded to the ground control center through the wireless communication module.

[0092] The present invention has the following beneficial effects:

[0093] 1. The present invention significantly improves the accuracy and efficiency of drone inspections. This synergy is reflected in the fact that drones combine adaptive adjustment with intelligent optimization to ensure the dynamic optimization of inspection paths, thus achieving efficient and accurate inspections of power grid infrastructure;

[0094] 2. The present invention converts point cloud data into a three-dimensional surface model in real time, ensures the generation of high-quality meshes by using Delaunay triangulation, and improves the resolution and authenticity of the three-dimensional model by combining the Bowyer-Watson algorithm and normal vector optimization technology;

[0095] 3. The present invention uses reinforcement learning to correct and optimize the route, dynamically corrects the initial inspection route, ensures its reliability and efficiency in complex environments, and combines real-time data. Reinforcement learning and multi-objective optimization can dynamically adapt to changes, improve the accuracy and safety of inspection plan execution, and enhance the level and effectiveness of the overall inspection task. BRIEF DESCRIPTION OF THE DRAWINGS

[0096] Figure 1 The figure is a flow chart of the method of the present invention. DETAILED DESCRIPTION

[0097] The present invention is further described in detail below with reference to the accompanying drawings and specific embodiments:

[0098] refer to Figure 1 In this embodiment, a method for automatically correcting and updating the route of a power inspection drone is provided, comprising the following steps:

[0099] Step S1: Collect and integrate the Geographic Information System (GIS) data of the inspection area and construct a topological map;

[0100] Step S2: Obtaining detailed point cloud data of the ground and power grid facilities, and constructing a 3D model based on the obtained point cloud data;

[0101] Step S3: Based on the topology map and the 3D model, an algorithm is used to extract key locations (such as tower bases, vertices, and cable connection points) from the 3D model, and initial waypoints are determined. Waypoints are extracted from key locations of power grid equipment (such as power towers, line connection points, etc.), and a path planning algorithm (such as Dijkstra or A*) is used to connect all the extracted waypoints to generate an initial inspection route;

[0102] Step S4: Using AI algorithms, combined with real-time weather, equipment priority, and historical data, correct the initial inspection route;

[0103] Step S5: The UAV performs inspection based on the corrected inspection route, and combines RTK GPS with the environment-adaptive VIO algorithm to achieve sub-meter positioning accuracy. After successful centralization at the key position, the waypoints are automatically updated and immediately uploaded to the control center to achieve automatic route update.

[0104] In this implementation, step S1 is specifically as follows:

[0105] Obtain Geographic Information System (GIS) data of the inspection area, including topography, infrastructure location, power towers, line connection points and their spatial coordinates;

[0106] Convert all data formats to a common GIS format such as Shapefile, KML or GeoJSON for easier processing, and project them to the same coordinate system (usually using WGS 84) to ensure consistency

[0107] In GIS tools, create multi-level data views, including basic terrain, line networks, and power facility distribution;

[0108] Establish the topological relationship between towers, line connection points and terrain changes, construct a topological map, and use graph algorithms to check the connectivity of the topological map to ensure that there are paths between all power grid nodes.

[0109] In this implementation, the topological relationship between the towers, line connection points and terrain changes is established, a topological map is constructed, and the connectivity of the topological map is checked using a graph algorithm to ensure that there are paths between all grid nodes, as follows:

[0110] The towers and line connection points are defined as nodes in the graph, and adjacent nodes are connected into edges according to the power lines;

[0111] Attach attributes to each node, including location coordinates, height, and tower type; attach attributes to each edge, including length, cable type, and voltage level;

[0112] Use DEM data to identify terrain changes, including valleys and ridges, and calculate the slope and aspect of each location to assess the impact on towers and paths;

[0113] In the topological graph, terrain changes are introduced, and each node not only saves its position, but also the height and slope of the ground;

[0114] Construct a topological graph G = (V, E), where V is the node, including the tower and the connection point, and E is the edge, that is, the connecting line. Both nodes and edges carry terrain-related information.

[0115] Based on DFS, the entire graph is explored to identify whether each node can be accessed through a direct or indirect path.

[0116] In this implementation, step S2 is specifically as follows:

[0117] Use LiDAR technology to collect fine point cloud data of power grids and terrain, and transform the points collected by LiDAR from the sensor coordinate system to the global coordinate system:

[0118]

[0119] Among them, (x g ,y g ,z g ) is the global coordinate, (x t ,y t,z t ) is the sensor coordinate, R is the rotation matrix, and t is the translation vector;

[0120] Use Voxel Grid simplification algorithm to reduce data volume while retaining structural integrity;

[0121] Apply terrain analysis algorithms (such as RANSAC or height-based filtering methods) to separate ground points from the data;

[0122] Identify and separate power grid infrastructure, including power lines and towers, and group and label them using clustering methods;

[0123] The point cloud data is converted into a three-dimensional surface model using the surface reconstruction algorithm Delaunay triangulation, and the reconstruction is optimized in combination with normal vector estimation to obtain the final 3D model.

[0124] In this implementation, the surface reconstruction algorithm Delaunay triangulation is used to convert the facility data into a three-dimensional surface model as follows:

[0125] Construct the initial hypertetrahedron, which contains all the input points. For each new point, find the tetrahedron area where it is located and re-split it;

[0126] Determine the visible polyhedra adjacent to the new point and remove them to create a new polyhedron;

[0127] After inserting the new point, recheck and correct the affected area to ensure that the empty sphere property is satisfied, that is, there are no other points in the circumscribed sphere of any triangle. For the tetrahedron T = △ABCD, calculate the circumscribed sphere of the tetrahedron and determine whether correction is needed:

[0128]

[0129] Among them, (x a ,y a ,z a ),(x b ,y b ,z b ),(x c ,y c ,z c ),(x d ,y d ,z d ) are the three-dimensional coordinates of the four vertices A, B, C, and D, which form an existing tetrahedron; (x e ,y e ,z e) is the three-dimensional coordinate of the newly inserted point E. It is necessary to determine whether this point is inside the circumscribed sphere of the tetrahedron formed by points A, B, C, and D. If the value of the determinant is > 0, then the new point E(x e ,y e ,z e ) is inside the circumscribed sphere. If ≤ 0, then point E is on or outside the sphere;

[0130] Based on the Owyer-Watson algorithm, the subdivision is gradually optimized by searching for tetrahedrons and correcting them point by point; and the three-dimensional surface generation is optimized by calculating the normal vector to better represent the actual terrain and the surface of the object.

[0131] In this implementation, based on the Owyer-Watson algorithm, the tetrahedron is searched and corrected point by point, the subdivision is gradually optimized, and the three-dimensional surface generation is optimized by calculating the normal vector to better represent the actual terrain and the surface of the object, as follows:

[0132] Select a new point P from the point cloud and insert it into the current mesh structure;

[0133] Find all tetrahedrons that contain the circumsphere of point P. These tetrahedrons will be invalidated by the introduction of new points and need to be removed from the structure.

[0134] Create new tetrahedrons on the boundaries of the holes left by the previous deletion step;

[0135] For each edge (e) on the boundary of the hole, generate a new tetrahedron from P to the endpoints of (e).

[0136] For the newly created tetrahedrons, check the Delaunay condition (empty sphericity). If there are non-Delaunay tetrahedrons in a tetrahedron, correct them by flipping the edges.

[0137] By calculating the normal vector, the orientation of each face of the 3D mesh is obtained;

[0138] For the triangular face △ABC, its normal vector N is calculated by the cross product of the two edges:

[0139]

[0140] Among them, (BA)×(CA) is a vector perpendicular to the plane of △ABC△ABC, and its direction follows the right-hand rule;

[0141] Taking the vertex as the center, take the average of the normal vectors of all surrounding triangles as the smooth normal vector of this vertex;

[0142]

[0143] Among them, N iis the normal vector of the i-th adjacent triangle around the vertex;

[0144] Noise is reduced by iteratively moving vertices to the average position of neighboring vertices;

[0145]

[0146] Where neighbors(V) is the set of all vertices directly connected to vertex V; λ is the smoothing factor; V is the current vertex; V j is the neighboring vertex of the current vertex;

[0147] Merges two adjacent vertices into one to reduce the polygon count while maintaining the overall shape.

[0148] In this implementation, step S3 is specifically as follows:

[0149] The tower base and tower top are identified by analyzing the tower geometry and point density in the 3D model. The cable connection points are identified from the model based on the connection line features. The endpoints are determined by gradually tracing the linear structure in the 3D model.

[0150] The key point positions are identified by feature point detection algorithms such as SIFT and SURF. For 3D points, accurate extraction is required based on normal vector analysis or point cloud slicing;

[0151] Ensure that waypoints include all identified key locations so that all required checkpoints are covered. The extracted waypoints should ensure that high-risk points throughout the facility are covered.

[0152] Additional waypoints were added at points of spatial inhomogeneity to ensure the integrity of the inspection path;

[0153] The path planning algorithm Dijkstra is used to plan the optimal inspection route and generate the initial inspection route.

[0154] In this implementation, S4 is as follows:

[0155] Obtain weather information in real time through API, including wind speed, rainfall, and temperature, and determine the maintenance priority of each device based on the current operating status and failure probability data of the device;

[0156] Combine historical fault records with previous inspection results to identify high-risk equipment and areas;

[0157] Let state S be the current inspection location / point and its current environmental factors (such as weather, priority, etc.);

[0158] Action A is to select the next inspection destination;

[0159] The reward function R takes into account inspection efficiency, safety, and equipment priority, and designs positive and negative incentives:

[0160] R=α×Prioritize(E)+β×WeatherFactor(W)-γ×Cost(C);

[0161] Among them, Prioritize (E) is the positive benefit brought by the equipment priority; WeatherFactor (W) is the importance adjusted according to weather conditions; Cost (C) is the comprehensive cost of inspection time, distance and risk;

[0162] The policy gradient method is used to adjust and optimize the path selection to minimize the cost, maximize the path effectiveness, and correct the initial inspection route.

[0163] In this implementation, the policy gradient method is used to adjust and optimize the path selection to minimize the cost, maximize the path effectiveness, and correct the initial inspection route, as follows:

[0164] Initialize the parameters θ to generate a random strategy;

[0165] In each round, according to the current strategy π θ (a|s) generates a complete path;

[0166] At the end of the path, calculate the cumulative reward R t , represents the total reward starting from time t:

[0167]

[0168] Where γ is the discount factor, r(s k ,a k ) is to perform action a k The reward you get after that;

[0169] Update the policy parameters θ using the gradient ascent method:

[0170] θ←θ+α▽ θ J(θ);

[0171] Where α is the learning rate; θ J(θ) is the policy gradient:

[0172]

[0173] Estimated θ Use Monte Carlo methods to sample and estimate policy gradients from simulated paths;

[0174] Repeat path generation, evaluation, and strategy update until convergence or the maximum number of iterations is reached, and correct the initial inspection route.

[0175] In this implementation, step S5 is specifically as follows:

[0176] Based on differential GPS, it provides centimeter-level absolute position accuracy through real-time communication between ground base stations and drones; combined with visual sensors and IMU, it achieves high-frequency, environment-adaptive position estimation;

[0177] While the UAV is flying along the planned path, path tracking is performed through RTK and VIO joint positioning;

[0178] Perform centering operations at key locations on the planned path to ensure that the capture or detection device is in the center of the field of view; and use feature detection to analyze the video frame to ensure that the key target is in the center of the image;

[0179] When the drone is successfully centered at the key position, the state vector is updated in real time based on VIO combined with vision and IMU data. According to the VIO real-time update result and the deviation of the centralized position (if it is not completely centralized), it is updated to calculate the new waypoint; the VIO real-time update state vector, including position p, speed v, and attitude q, is updated in real time through the following equations:

[0180]

[0181] v t+1 =v t +a t Δt;

[0182]

[0183] Among them, a t is the acceleration measurement, Δθ t is the integral of the angular velocity, Represents quaternion multiplication;

[0184] Upload the newly generated waypoints to the ground control center immediately through the wireless communication module

[0185] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0186] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0187] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0188] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0189] The above is only a preferred embodiment of the present invention, and does not limit the present invention in other forms. Any technician familiar with the profession may use the above disclosed technical content to change or modify it into an equivalent embodiment with equivalent changes. However, any simple modification, equivalent change and modification made to the above embodiment according to the technical essence of the present invention without departing from the technical solution of the present invention still belongs to the protection scope of the technical solution of the present invention.

Claims

1. A method for automatically correcting and updating the route of a power inspection drone, characterized in that: The following steps are involved: Step S1: Collect and integrate the Geographic Information System (GIS) data of the inspection area and construct a topological map; Step S2: Obtaining detailed point cloud data of the ground and power grid facilities, and constructing a 3D model based on the obtained point cloud data; Step S3: Based on the topology map and the 3D model, an algorithm is used to extract key locations from the 3D model, initial waypoints are determined, waypoints are extracted from key locations of power grid equipment, and a path planning algorithm is used to connect all extracted waypoints to generate an initial inspection route; Step S4: Using AI algorithms, combined with real-time weather, equipment priority, and historical data, correct the initial inspection route; Step S5: The UAV performs inspection based on the corrected inspection route, and combines RTK GPS with the environment-adaptive VIO algorithm to achieve sub-meter positioning accuracy. After successful centralization at the key position, the waypoints are automatically updated and immediately uploaded to the control center to achieve automatic route update.

2. The method for automatically correcting and updating the route of a power inspection drone according to claim 1 is characterized in that: The step S1 is specifically as follows: Obtain Geographic Information System (GIS) data of the inspection area, including topography, infrastructure location, power towers, line connection points and their spatial coordinates; Convert all data formats to common GIS formats and project them to the same coordinate system; In GIS tools, create multi-level data views, including basic terrain, line networks, and power facility distribution; Establish the topological relationship between towers, line connection points and terrain changes, construct a topological map, and use graph algorithms to check the connectivity of the topological map to ensure that there are paths between all power grid nodes.

3. The method for automatically correcting and updating the route of a power inspection drone according to claim 2 is characterized in that: The topological relationship between the towers, line connection points and terrain changes is established, a topological map is constructed, and the connectivity of the topological map is checked using a graph algorithm to ensure that there are paths between all power grid nodes, as follows: The towers and line connection points are defined as nodes in the graph, and adjacent nodes are connected into edges according to the power lines; Attach attributes to each node, including location coordinates, height, and tower type; attach attributes to each edge, including length, cable type, and voltage level; Use DEM data to identify terrain changes, including valleys and ridges, and calculate the slope and aspect of each location to assess the impact on towers and paths; In the topological graph, terrain changes are introduced, and each node not only saves its position, but also the height and slope of the ground; Construct a topological graph G = (V, E), where V is the node, including the tower and the connection point, and E is the edge, that is, the connecting line. Both nodes and edges carry terrain-related information. Based on DFS, the entire graph is explored to identify whether each node can be accessed through a direct or indirect path.

4. The method for automatically correcting and updating the route of a power inspection drone according to claim 1 is characterized in that: The step S2 is specifically as follows: Use LiDAR technology to collect fine point cloud data of power grids and terrain, and transform the points collected by LiDAR from the sensor coordinate system to the global coordinate system: Among them, (x g ,y g ,z g ) is the global coordinate, (x t ,y t ,z t ) is the sensor coordinate, R is the rotation matrix, and t is the translation vector; Use VoxelGrid simplification algorithm to reduce data volume while preserving structural integrity; Apply terrain analysis algorithms to separate ground points from the data; Identify and separate power grid infrastructure, including power lines and towers, and group and label them using clustering methods; The point cloud data is converted into a three-dimensional surface model using the surface reconstruction algorithm Delaunay triangulation, and the reconstruction is optimized in combination with normal vector estimation to obtain the final 3D model.

5. The method for automatically correcting and updating the route of a power inspection drone according to claim 4 is characterized in that: The facility data is converted into a three-dimensional surface model using the surface reconstruction algorithm Delaunay triangulation, as follows: Construct the initial hypertetrahedron, which contains all the input points. For each new point, find the tetrahedron area where it is located and re-split it; Determine the visible polyhedra adjacent to the new point and remove them to create a new polyhedron; After inserting the new point, recheck and correct the affected area to ensure that the empty sphere property is satisfied, that is, there are no other points in the circumscribed sphere of any triangle. For the tetrahedron T = △ABCD, calculate the circumscribed sphere of the tetrahedron and determine whether correction is needed: Among them, (x a ,y a ,z a ),(x b ,y b ,z b ),(x c ,y c ,z c ),(x d ,y d ,z d ) are the three-dimensional coordinates of the four vertices A, B, C, and D, which form an existing tetrahedron; (x e ,y e ,z e ) is the three-dimensional coordinate of the newly inserted point E. It is necessary to determine whether this point is inside the circumscribed sphere of the tetrahedron formed by points A, B, C, and D. If the value of the determinant is > 0, then the new point E(x e ,y e ,z e ) is inside the circumscribed sphere. If ≤ 0, then point E is on or outside the sphere; Based on the Owyer-Watson algorithm, the subdivision is gradually optimized by searching for tetrahedrons and correcting them point by point; And optimize the 3D surface generation by calculating normal vectors to better represent the surface of actual terrain and objects.

6. The method for automatically correcting and updating the route of a power inspection drone according to claim 5 is characterized in that: The method based on the Owyer-Watson algorithm optimizes the subdivision step by step by searching for tetrahedrons and correcting them point by point, and optimizes the generation of three-dimensional surfaces by calculating normal vectors, so as to better represent the actual terrain and the surface of objects, as follows: Select a new point P from the point cloud and insert it into the current mesh structure; Find all tetrahedrons that contain the circumsphere of point P. These tetrahedrons will be invalidated by the introduction of new points and need to be removed from the structure. Create new tetrahedrons on the boundaries of the holes left by the previous deletion step; For each edge (e) on the boundary of the hole, generate a new tetrahedron from P to the endpoints of (e). For the newly created tetrahedrons, check the Delaunay conditions. If there are non-Delaunay tetrahedrons in a tetrahedron, correct them by flipping the edges. By calculating the normal vector, the orientation of each face of the 3D mesh is obtained; For the triangular face △ABC, its normal vector N is calculated by the cross product of the two edges: Among them, (BA)×(CA) is a vector perpendicular to the plane of △ABC△ABC, and its direction follows the right-hand rule; Taking the vertex as the center, take the average of the normal vectors of all surrounding triangles as the smooth normal vector of this vertex; Among them, N i is the normal vector of the i-th adjacent triangle around the vertex; Noise is reduced by iteratively moving vertices to the average position of neighboring vertices; Where neighbors(V) is the set of all vertices directly connected to vertex V; λ is the smoothing factor; V is the current vertex; V j is the neighboring vertex of the current vertex; Merges two adjacent vertices into one to reduce the polygon count while maintaining the overall shape.

7. The method for automatically correcting and updating the route of a power inspection drone according to claim 1 is characterized in that: The step S3 is specifically as follows: The tower base and tower top are identified by analyzing the tower geometry and point density in the 3D model. The cable connection points are identified from the model based on the connection line features. The endpoints are determined by gradually tracing the linear structure in the 3D model. Ensure that waypoints include all identified key locations so that all required checkpoints are covered. The extracted waypoints should ensure that high-risk points throughout the facility are covered. Additional waypoints were added at points of spatial inhomogeneity to ensure the integrity of the inspection path; The path planning algorithm Dijkstra is used to plan the optimal inspection route and generate the initial inspection route.

8. The method for automatically correcting and updating the route of a power inspection drone according to claim 1 is characterized in that: The S4 is specifically as follows: Obtain weather information in real time through API, including wind speed, rainfall, and temperature, and determine the maintenance priority of each device based on the current operating status and failure probability data of the device; Combine historical fault records with previous inspection results to identify high-risk equipment and areas; Let state S be the current inspection location / point and its current environmental factors; Action A is to select the next inspection destination; The reward function R takes into account inspection efficiency, safety, and equipment priority, and designs positive and negative incentives: R=α×Prioritize(E)+β×WeatherFactor(W)-γ×Cost(C); Among them, Prioritize (E) is the positive benefit brought by the equipment priority; WeatherFactor (W) is the importance adjusted according to weather conditions; Cost (C) is the comprehensive cost of inspection time, distance and risk; Use the policy gradient method to adjust and optimize path selection to minimize cost, maximize path effectiveness, and correct the initial inspection route.

9. The method for automatically correcting and updating the route of a power inspection drone according to claim 8 is characterized in that: The policy gradient method is used to adjust and optimize the path selection to minimize the cost, maximize the path effectiveness, and correct the initial inspection route, as follows: Initialize the parameters θ to generate a random strategy; In each round, according to the current strategy π θ (a|s) generates a complete path; At the end of the path, calculate the cumulative reward R t , represents the total reward starting from time t: Where γ is the discount factor, r(s k ,a k ) is to perform action a k The reward you get after that; Update the policy parameters θ using the gradient ascent method: α Where, is the learning rate; is the policy gradient: estimate Use Monte Carlo methods to sample and estimate policy gradients from simulated paths; Repeat path generation, evaluation, and strategy update until convergence or the maximum number of iterations is reached, and correct the initial inspection route.

10. The method for automatically correcting and updating the route of a power inspection drone according to claim 1, characterized in that: The step S5 is specifically as follows: Based on differential GPS, it provides centimeter-level absolute position accuracy through real-time communication between ground base stations and drones; combined with visual sensors and IMU, it achieves high-frequency, environment-adaptive position estimation; While the UAV is flying along the planned path, path tracking is performed through RTK and VIO joint positioning; Perform centering operations at key locations along the planned path to ensure that the shooting or detection equipment is in the center of the field of view; And use feature detection to analyze the video frames to ensure that the key target is in the center of the image; When the drone is successfully centered at the key position, the state vector is updated in real time based on VIO combined with vision and IMU data. According to the VIO real-time update result and the deviation of the centralized position (if it is not completely centralized), it is updated to calculate the new waypoint; the VIO real-time update state vector, including position p, speed v, and attitude q, is updated in real time through the following equations: v t+1 =v t +a t Δt; Among them, a t is the acceleration measurement, Δθ t is the integral of the angular velocity, Represents quaternion multiplication; The newly generated waypoints are immediately uploaded to the ground control center through the wireless communication module.

Images