Substation unmanned aerial vehicle inspection path planning method, system, device, medium and product

Abstract

This invention relates to the field of power system inspection technology, and discloses a method, system, equipment, medium, and product for planning unmanned aerial vehicle (UAV) inspection paths for substations. The method includes: extracting the location information of each inspection target point based on a three-dimensional vector model of the target substation to construct a set of inspection target points; dividing the global space of the target substation into a multi-scale three-dimensional grid based on an octree structure, and marking the divided grid cells as obstacle grids to obtain a collision detection three-dimensional grid; selecting multiple grid cells not marked as obstacle grids as candidate waypoints from the collision detection three-dimensional grid; selecting the optimal observation waypoints of the inspection target points under each preset observation viewpoint based on the spatial geometric relationship between each candidate waypoint and the inspection target point; and planning the globally optimal inspection path from the set of optimal observation waypoints of each inspection target point, thereby improving the reliability of the inspection path planning.

Description

Technical Field

[0001] This invention relates to the field of power system inspection technology, and in particular to a method, system, equipment, medium and product for planning the inspection path of a substation using a drone. Background Technology

[0002] With the development of smart grids, drones have been widely used in the refined inspection of substations. Current intelligent inspections mainly rely on laser 3D point cloud views, where manual extraction of substation inspection points from the point cloud is used to generate drone flight paths. However, laser point clouds typically suffer from problems such as high noise levels and severe equipment obstruction. Flight path planning based on point clouds still requires significant manual intervention for waypoint selection and safety verification, and the level of automation and planning efficiency is far from meeting expectations.

[0003] Existing automated route planning methods typically generate waypoints using simple methods such as "normal vector offset based on equipment surface" or "grid sampling at fixed distances." However, given the intricate electrical connections, dense array of insulators and surge arresters within substations, this traditional forward planning method generates a massive number of redundant waypoints. Furthermore, it is highly prone to collisions at complex topologies and to causing safety accidents when traversing complex structures (such as busbar bridges and gantry cranes). Summary of the Invention

[0004] In view of this, in order to solve the above-mentioned technical problems, the present invention provides a method, system, equipment, medium and product for planning the inspection path of a substation by unmanned aerial vehicle.

[0005] The first aspect of this invention provides a method for planning the path of a substation unmanned aerial vehicle (UAV) inspection, comprising:

[0006] Obtain a three-dimensional vector model of the target substation, extract the location information of each inspection target point based on the three-dimensional vector model, and construct an inspection target point set;

[0007] The global space of the target substation is divided into a multi-scale three-dimensional mesh based on an octree structure, and obstacle meshes are marked on the divided mesh cells to obtain a collision detection three-dimensional mesh.

[0008] For each of the inspection target points in the set of inspection target points, based on the location information of the inspection target point and the preset camera parameters of the UAV, multiple grid cells that are not identified as obstacle grids are selected as candidate waypoints in the collision detection three-dimensional grid.

[0009] For each inspection target point, based on the spatial geometric relationship between each candidate waypoint and the inspection target point, the optimal observation waypoint for the inspection target point under each preset observation view is selected from all the candidate waypoints to obtain the optimal observation waypoint set;

[0010] Based on the heuristic cost algorithm and combined with the collision detection 3D mesh, a globally optimal inspection path connecting each of the optimal observation waypoints is planned from the optimal observation waypoint set of each of the inspection target points.

[0011] In one embodiment, the step of extracting the location information of each inspection target point based on the three-dimensional vector model and constructing an inspection target point set includes:

[0012] Obtain the geometric features and semantic tags of each device object in the three-dimensional vector model;

[0013] Based on the semantic tags, multiple equipment objects under the preset inspection equipment type are filtered, and based on the geometric features of the filtered equipment objects, the inspection target surface of the equipment objects is identified.

[0014] The position information of each inspection target point is determined based on the geometric center and normal vector direction of the inspection target surface;

[0015] Based on the location information of each inspection target point, the set of inspection target points is constructed.

[0016] In one embodiment, the step of dividing the global space of the target substation into a multi-scale three-dimensional mesh based on an octree structure, and marking the divided mesh cells with obstacle meshes to obtain a collision detection three-dimensional mesh, includes:

[0017] Based on the three-dimensional vector model of the target substation, determine the global three-dimensional spatial bounding box of the target substation;

[0018] The global 3D spatial bounding box is used as the root node of an octree, and the root node of the octree is recursively divided into octree child nodes to obtain a multi-scale 3D mesh.

[0019] Based on the three-dimensional vector model of the target substation, the geometric envelopes of all obstacles are extracted, and the geometric envelopes of the obstacles are subjected to safety distance expansion processing to generate obstacle expansion bodies;

[0020] The obstacle expansion is mapped onto the multi-scale three-dimensional mesh, and all mesh cells that intersect with the obstacle expansion are identified as obstacle meshes. The remaining mesh cells in the multi-scale three-dimensional mesh, excluding all obstacle meshes, are identified as free-flight meshes, thus obtaining the initial collision detection three-dimensional mesh.

[0021] A unique octree code is assigned to all grid cells in the initial collision detection 3D mesh to obtain the collision detection 3D mesh.

[0022] In one embodiment, the step of selecting multiple grid cells not identified as obstacle grids as candidate waypoints from the collision detection 3D grid based on the location information of the inspection target point and the preset camera parameters of the UAV includes:

[0023] The optimal inspection distance is determined based on the preset camera parameters of the drone;

[0024] Based on the location information of the inspection target point and the optimal inspection distance, rays are emitted in a hemispherical shape into the free-flying grid adjacent to the inspection target point;

[0025] All free-flying grids that intersect with the free-flying grid but do not intersect with the obstacle grid are selected as candidate waypoints.

[0026] In one embodiment, the step of selecting the optimal observation waypoint for the inspection target point from all the candidate waypoints based on the spatial geometric relationship between each candidate waypoint and the inspection target point, thereby obtaining the optimal observation waypoint set, includes:

[0027] All candidate waypoints are mapped to a three-dimensional polar coordinate system with the inspection target point as the origin, and the three-dimensional polar coordinate system is divided into several view sectors according to a preset observation view.

[0028] Within each view sector, based on the spatial geometric relationship between each candidate waypoint and the inspection target point within the view sector, the distance deviation and angle deviation of each candidate waypoint are determined, and based on the weighted calculation result of the distance deviation and the angle deviation, the comprehensive inspectability of each candidate waypoint is determined;

[0029] Based on the overall inspectability of each candidate waypoint, the candidate waypoint with the highest overall inspectability is selected in each view sector as the optimal observation waypoint in the view sector.

[0030] The optimal observation waypoint set is constructed based on the optimal observation waypoints within all the aforementioned view sectors.

[0031] In one embodiment, the heuristic cost algorithm, combined with the collision detection 3D mesh, plans a globally optimal inspection path connecting each of the optimal observation waypoints in the optimal observation waypoint set for each of the inspection target points, including:

[0032] Under the constraints of the collision detection 3D grid, the path nodes of the preset path start point, preset path end point, and the optimal observation waypoint set of each inspection target point are sorted to obtain the inspection waypoint path node sequence.

[0033] Based on a preset heuristic cost function, the cost of each node in the inspection waypoint path node sequence is evaluated, and the path of the inspection waypoint path node sequence is optimized based on the cost evaluation results to generate the globally optimal inspection path that minimizes the cost evaluation results; wherein, the heuristic cost function is constructed by weighted fusion based on the flight distance between each waypoint, the turning angle, and the spatial distance penalty between the waypoint and the threat device.

[0034] Secondly, the present invention also provides a substation unmanned aerial vehicle (UAV) inspection path planning system, comprising:

[0035] The patrol point acquisition module is used to acquire a three-dimensional vector model of the target substation, extract the location information of each inspection target point based on the three-dimensional vector model, and construct an inspection target point set.

[0036] The octree mesh generation module is used to perform multi-scale three-dimensional mesh generation on the global space of the target substation based on the octree structure, and to mark the mesh cells as obstacles to obtain a collision detection three-dimensional mesh.

[0037] The candidate waypoint determination module is used to select multiple grid cells that are not identified as obstacle grids as candidate waypoints for each of the inspection target points in the inspection target point set, based on the location information of the inspection target point and the preset camera parameters of the UAV;

[0038] The optimal waypoint filtering module is used to select the optimal observation waypoint of the inspection target point from all the candidate waypoints from each of the inspection target points based on the spatial geometric relationship between each candidate waypoint and the inspection target point, so as to obtain the optimal observation waypoint set.

[0039] The inspection path planning module is used to plan a globally optimal inspection path connecting each of the optimal observation waypoints in the set of optimal observation waypoints for each of the inspection target points, based on a heuristic cost algorithm and in combination with the collision detection 3D mesh.

[0040] Thirdly, the present invention also provides an electronic device, which includes a memory and a processor. The memory stores a computer program, and when the computer program is executed by the processor, the processor performs the steps of the substation UAV inspection path planning method as described in the first aspect.

[0041] Fourthly, the present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed, implements the steps of the substation unmanned aerial vehicle inspection path planning method as described in the first aspect.

[0042] Fifthly, the present invention also provides a computer program product, the computer program product comprising a computer program stored on a non-transitory computer-readable storage medium, the computer program comprising program instructions, wherein, when the program instructions are executed by a computer, the computer performs the steps of the substation UAV inspection path planning method as described in the first aspect.

[0043] As can be seen from the above technical solutions, this invention acquires a three-dimensional vector model of the target substation and extracts the position information of each inspection target point to construct an inspection target point set, thereby accurately expressing the positional relationship of the inspection objects in three-dimensional space. By dividing the global space of the target substation into a multi-scale three-dimensional mesh based on an octree structure and marking the mesh units as obstacle meshes, a collision detection three-dimensional mesh is obtained. The octree structure can adaptively represent meshes of different coarseness according to the spatial complexity, improving the recognition accuracy of complex structures and small obstacles such as busbar bridges, gantry frames, insulators, surge arresters, and conductor connection areas while ensuring computational efficiency. This effectively improves the accuracy of collision detection. Furthermore, it filters out multiple meshes not marked as obstacles from the collision detection three-dimensional mesh. Using grid cells as candidate waypoints ensures that these waypoints are generated only within safe and accessible spaces. Furthermore, based on the location information of the inspection target point, the UAV's preset camera parameters, and the spatial geometric relationship between the candidate waypoints and the inspection target point, the optimal observation waypoint for the inspection target point under each preset observation viewpoint is selected from the candidate waypoints. This reduces duplicate shooting or missed inspections caused by unreasonable viewpoints and avoids the large number of invalid and redundant waypoints generated by fixed-distance grid sampling in existing technologies. This reduces the scale of subsequent path search and improves waypoint planning efficiency. Furthermore, based on a heuristic cost algorithm combined with a collision detection 3D grid, connecting paths are sequentially planned from the optimal observation waypoint set for each inspection target point. This enables global inspection path optimization under obstacle avoidance constraints, improving the efficiency of UAV inspection operations. Attached Figure Description

[0044] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0045] Figure 1 This is an application environment diagram of a substation unmanned aerial vehicle (UAV) inspection path planning method provided in an embodiment of the present invention.

[0046] Figure 2A flowchart of a substation unmanned aerial vehicle (UAV) inspection path planning method provided in an embodiment of the present invention;

[0047] Figure 3 This is a schematic diagram illustrating the principle of reverse visibility analysis.

[0048] Figure 4 This is a schematic diagram of a substation unmanned aerial vehicle (UAV) inspection path planning system provided in an embodiment of the present invention;

[0049] Figure 5 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0050] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0051] Currently, there is no mature solution for 3D spatial path planning in power scenarios. Drone path planning is fundamental to the flight safety and mission completion of drones, requiring extremely high reliability and practicality. Existing automatic route planning methods typically use simple methods such as "normal vector offset based on equipment surface" or "grid sampling at fixed distances" to generate waypoints. Faced with the intricate electrical connections, densely packed insulators, and surge arresters within substations, this traditional forward planning method generates a massive number of redundant waypoints and is highly prone to collisions at complex topologies, and is particularly susceptible to safety accidents when traversing complex structures (such as busbars and gantry cranes).

[0052] Furthermore, existing methods that rely on extracting inspection points from raw point clouds are limited by the sparsity and noise of the point clouds, making it impossible to automatically identify equipment types. This necessitates manual point marking, resulting in low cost-effectiveness and insufficient robustness. Existing algorithms often employ forward ray-casting from the UAV's viewpoint to the target surface for visibility analysis. In the vast candidate waypoint space of a substation, this calculation explodes exponentially, forcing the planner to sacrifice waypoint flexibility and generate a large number of redundant "surface-hugging" waypoints, significantly increasing the UAV's flight energy consumption and inspection time. Substation equipment components are numerous and heavily interdependent. Traditional planning often only considers the local normal vectors of individual devices for waypoint inverse calculation, ignoring the global topological relationships between devices and lacking efficient three-dimensional spatial mesh management. This makes it highly susceptible to safety accidents when traversing complex structures (such as busbars and gantry cranes).

[0053] To address the aforementioned issues, this application proposes a method for planning unmanned aerial vehicle (UAV) inspection paths for substations. By acquiring a 3D vector model of the target substation and extracting the location information of each inspection target point, a set of inspection target points is constructed, accurately representing the positional relationships of the inspection objects in 3D space. A multi-scale 3D mesh is created based on an octree structure to divide the global space of the target substation into a grid, and obstacle meshes are labeled on the mesh units to obtain a collision detection 3D mesh. The octree structure can adaptively represent meshes of varying coarseness according to spatial complexity, improving the accuracy of identifying complex structures and small obstacles such as busbars, gantry frames, insulators, surge arresters, and conductor connection areas while maintaining computational efficiency. This effectively enhances the accuracy of collision detection. Furthermore, the method filters out unlabeled obstacles from the collision detection 3D mesh. Multiple grid cells of the obstacle grid serve as candidate waypoints, ensuring that these waypoints are generated only within a safe and passable space. Furthermore, based on the location information of the inspection target point, the UAV's preset camera parameters, and the spatial geometric relationship between the candidate waypoints and the inspection target point, the optimal observation waypoint for the inspection target point under each preset observation viewpoint is selected from the candidate waypoints. This reduces duplicate shooting or missed inspections caused by unreasonable viewpoints and avoids the large number of invalid and redundant waypoints generated by fixed-distance grid sampling in existing technologies. This reduces the scale of subsequent path search and improves waypoint planning efficiency. Furthermore, based on a heuristic cost algorithm combined with a collision detection 3D grid, connecting paths are sequentially planned from the optimal observation waypoint set for each inspection target point. This enables global inspection path optimization under obstacle avoidance constraints, improving the efficiency of UAV inspection operations.

[0054] Therefore, this invention effectively solves the technical problems in existing automatic route planning methods, such as the large number of redundant waypoints, the susceptibility to collisions in complex topologies, and insufficient safety when traversing complex structures, caused by the use of simple normal vector offset or fixed-distance grid sampling. It realizes safe, efficient, and intelligent planning of UAV inspection routes in complex substation scenarios and has high engineering application value.

[0055] In a specific example, the substation UAV inspection path planning method provided in this application embodiment can be applied to, for example, Figure 1The application environment shown is illustrated. Terminal 101 communicates with server 102 via a network. A data storage system can store the data that server 102 needs to process. The data storage system can be integrated onto server 102, or it can be located in the cloud or on another network server. Terminal 101 or server 102 executes a substation UAV inspection path planning method, which includes: acquiring a three-dimensional vector model of the target substation; extracting the position information of each inspection target point based on the three-dimensional vector model to construct an inspection target point set; dividing the global space of the target substation into a multi-scale three-dimensional mesh based on an octree structure, and marking the divided mesh units as obstacle meshes to obtain a collision detection three-dimensional mesh; for each inspection target point in the inspection target point set, selecting multiple mesh units not marked as obstacle meshes as candidate waypoints in the collision detection three-dimensional mesh based on the position information of the inspection target point and the preset camera parameters of the UAV; for each inspection target point, selecting the optimal observation waypoint of the inspection target point under each preset observation view from all candidate waypoints based on the spatial geometric relationship between each candidate waypoint and the inspection target point to obtain an optimal observation waypoint set; and planning a globally optimal inspection path connecting each optimal observation waypoint in the optimal observation waypoint set of each inspection target point based on a heuristic cost algorithm and the collision detection three-dimensional mesh.

[0056] Terminal 101 can be, but is not limited to, various personal computers, laptops, smartphones, and tablets.

[0057] Server 102 can be a standalone physical server, a server cluster or distributed system consisting of multiple physical servers, or a cloud server that provides cloud computing services.

[0058] like Figure 2 As shown in the figure, this application provides a method for planning the path of a substation unmanned aerial vehicle (UAV) inspection, which is applied to... Figure 1 Taking terminal 101 or server 102 as an example, the explanation includes the following steps S1 to S5. Wherein:

[0059] Step S1: Obtain the three-dimensional vector model of the target substation, extract the location information of each inspection target point based on the three-dimensional vector model, and construct the inspection target point set.

[0060] Among them, the three-dimensional vector model is obtained by laser point cloud scanning and three-dimensional modeling of all equipment in the target substation. The attributes of the three-dimensional vector model include semantic information, geometric information, topological relationship information and surface attributes. Semantic information is the explicit labeling of equipment semantic tags, such as transformers, circuit breakers, insulators, surge arresters, bushings, busbars, etc. Geometric information is the precise shape, size and spatial coordinates of the equipment. Topological relationship information includes the connection relationship and spatial relative position between equipment. Surface attributes include the normal vector, curvature and boundary of each equipment surface.

[0061] Inspection target points are determined based on inspection requirements and tasks, and the location information of each inspection target point, such as spatial coordinates and normal vector direction, is extracted from the three-dimensional vector model. This ensures that the three-dimensional spatial coordinates and geometric topological attributes of all inspection target points are obtained, forming a high-precision set of inspection target points with semantic interpretability.

[0062] Step S2: Based on the octree structure, perform multi-scale three-dimensional meshing of the global space of the target substation, and mark the obstacle meshes of the divided mesh cells to obtain the collision detection three-dimensional mesh.

[0063] This application employs a digital grid method to efficiently compress the global space of the substation. The octree structure transforms the continuous three-dimensional space of the substation into a discrete 3D grid managed by an octree, accurately marking no-fly (collision) zones and safe flight zones to obtain a collision detection three-dimensional grid. This grid precisely maps equipment entities and safety margin expansion areas, ensuring the safe passage of UAVs through gaps in complex electrical equipment. Simultaneously, unoccupied free space is marked as a safe flight grid, providing a high-fidelity spatial constraint basis for path planning.

[0064] Step S3: For each inspection target point in the inspection target point set, based on the location information of the inspection target point and the preset camera parameters of the UAV, select multiple grid cells that are not marked as obstacle grids as candidate waypoints in the collision detection 3D grid.

[0065] The preset camera parameters for the drone include focal length, sensor size, pixel resolution, and ground sample distance (GSD).

[0066] Understandably, in order to clearly distinguish minor defects in equipment by capturing images from drones, this application performs reverse visibility analysis by using the location information of the inspection target point and the preset camera parameters of the drone. The inspection point is used as the emission source to project rays in the reverse direction. Combined with GSD constraints, the optimal observation distance is dynamically calculated in reverse. Free-flying grid cells that are not obscured by obstacles and meet the imaging resolution requirements are accurately selected as candidate waypoints.

[0067] Step S4: For each inspection target point, based on the spatial geometric relationship between each candidate waypoint and the inspection target point, select the optimal observation waypoint for the inspection target point from all candidate waypoints under each preset observation perspective, and obtain the optimal observation waypoint set.

[0068] Due to the massive number of candidate waypoints and their extensive overlap, this application, based on better imaging quality and diverse perspectives, assesses the inspectability of each candidate waypoint under each preset observation perspective by considering the spatial geometric relationship (including distance and angle deviation) between each candidate waypoint and the target point under each preset observation perspective. Redundant waypoints with single perspectives and non-optimal distances are eliminated first, greatly reducing the scale of the planning problem. Based on this mechanism, hundreds or thousands of dense and redundant waypoints can be sparsed (pruned) into a set of single-digit high-quality, multi-view key waypoints.

[0069] Step S5: Based on the heuristic cost algorithm and combined with the collision detection 3D mesh, plan the globally optimal inspection path connecting each optimal observation waypoint in the optimal observation waypoint set of each inspection target point.

[0070] Among them, the heuristic cost algorithm is an intelligent search strategy that integrates physical risk perception and multi-objective trade-offs, such as the improved A* or Theta* algorithm. In the embodiments of this application, a heuristic cost function that considers the shortest flight distance, the smallest turning angle, and avoidance of strong electromagnetic interference areas will be considered. Under the premise of ensuring safety constraints that it does not intersect with all obstacles in the collision detection three-dimensional grid, the connection order between waypoints will be dynamically optimized, and the flight trajectory of all optimal observation waypoints will be connected in sequence to realize the dynamic global optimization of the inspection path. Finally, a smooth, safe, and efficient automatic inspection route within the substation will be output.

[0071] This application embodiment acquires a three-dimensional vector model of the target substation and extracts the location information of each inspection target point to construct an inspection target point set, enabling accurate representation of the positional relationships of the inspection objects in three-dimensional space. It then uses an octree structure to perform multi-scale three-dimensional mesh division of the global space of the target substation and identifies obstacle meshes in the mesh units, resulting in a collision detection three-dimensional mesh. The octree structure can adaptively represent meshes of varying coarseness according to spatial complexity, improving the accuracy of identifying complex structures and small obstacles such as busbars, gantry frames, insulators, surge arresters, and conductor connection areas while maintaining computational efficiency. This effectively enhances the accuracy of collision detection. Furthermore, it filters out multiple mesh units not identified as obstacle meshes from the collision detection three-dimensional mesh. Candidate waypoints are generated only within safe and accessible spaces. Further, based on the location information of the inspection target point, the UAV's preset camera parameters, and the spatial geometric relationship between the candidate waypoints and the inspection target point, the optimal observation waypoint for the inspection target point under each preset observation viewpoint is selected from the candidate waypoints. This reduces duplicate shooting or missed inspections caused by unreasonable viewpoints and avoids the large number of invalid and redundant waypoints generated by fixed-distance grid sampling in existing technologies. This reduces the scale of subsequent path search and improves waypoint planning efficiency. Furthermore, based on a heuristic cost algorithm combined with a collision detection 3D grid, connecting paths are sequentially planned from the optimal observation waypoint set for each inspection target point. This enables global inspection path optimization under obstacle avoidance constraints, improving the efficiency of UAV inspection operations.

[0072] In some embodiments, the location information of each inspection target point is extracted based on the three-dimensional vector model to construct an inspection target point set, including:

[0073] Step S101: Obtain the geometric features and semantic labels of each device object in the 3D vector model.

[0074] Among them, equipment objects refer to power equipment with independent structure and function within the substation, such as transformers, circuit breakers, disconnect switches, surge arresters, current transformers, etc.; semantic tags represent the equipment category of each equipment object in the substation, and geometric features are the spatial coordinates, surface planes and curvature, normal vectors and other geometric attributes of the equipment objects.

[0075] Step S102: Filter multiple equipment objects under the preset inspection equipment type according to semantic tags, and identify the inspection target surface of the equipment object according to the geometric features of the filtered equipment objects.

[0076] Among them, multiple equipment objects under the preset inspection equipment type are filtered by semantic tags. The inspection equipment type is determined according to the inspection requirements, such as transformer bushings, circuit breaker contacts, insulator skirts, bolts / fasteners, terminals, disconnector contacts, etc.

[0077] Then, due to the power inspection standards, the equipment surface to be photographed is located on each component to be inspected. Therefore, the inspection target surface of the equipment object is identified by geometric features. The inspection target surface is a photographable area with typical defect features on the equipment surface defined in the preset substation inspection rule library.

[0078] Step S103: Determine the location information of each inspection target point based on the geometric center and normal vector direction of the inspection target surface.

[0079] The geometric center and normal vector direction of the target surface can be determined by using geometric information and surface properties. The normal vector direction is used to characterize the orientation of the target surface, ensuring that the camera's principal optical axis is perpendicular to the surface to obtain the best imaging quality.

[0080] Step S104: Construct a set of inspection target points based on the location information of each inspection target point.

[0081] The set of inspection target points is denoted as: Each inspection target point Includes its three-dimensional spatial coordinates and the unit normal vector of the surface. .

[0082] In some embodiments, the global space of the target substation is divided into a multi-scale three-dimensional mesh based on an octree structure, and obstacle meshes are marked on the divided mesh cells to obtain a collision detection three-dimensional mesh, including:

[0083] Step S201: Determine the global three-dimensional spatial bounding box of the target substation based on the three-dimensional vector model of the target substation.

[0084] Specifically, by using the three-dimensional vector model of the target substation, the maximum / minimum coordinates of the substation's X / Y / Z axes are extracted, and the global bounding box of the entire inspection scene is defined, which is the global three-dimensional spatial bounding box of the target substation.

[0085] Step S202: Use the global 3D spatial bounding box as the root node of the octree, and recursively divide the root node of the octree into octree child nodes to obtain a multi-scale 3D mesh.

[0086] By defining multi-scale voxels, a finer voxel resolution is used for densely populated equipment areas (busbars, insulators, circuit breakers), while a coarse-grained voxel is used for open areas. The global 3D spatial bounding box is used as the root node of an octree, and the root node is evenly divided into 8 child nodes according to the XYZ axes. In areas with dense equipment and complex obstacles, further subdivision is carried out, while subdivision is stopped in open and unobstructed areas. Subdivision is stopped until the grid size is less than or equal to the minimum safe distance of the UAV, or there are no obstacles in the grid, forming a non-uniform multi-scale grid. This significantly reduces the total number of grids and the computational cost of path planning while ensuring the accuracy of inspection coverage.

[0087] Step S203: Based on the three-dimensional vector model of the target substation, extract the geometric envelope of all obstacles, and perform safe distance expansion processing on the geometric envelope of the obstacles to generate the obstacle expansion body.

[0088] The process involves extracting the geometric envelope of all obstacles from the three-dimensional vector model of the target substation, determining the safe expansion distance based on the sum of the maximum contour radius of the UAV and the legally permissible safe flight distance, and using this safe expansion distance to perform safe distance expansion processing on the geometric envelope of the obstacles to generate an expanded obstacle body, thereby preventing the UAV from colliding with the equipment.

[0089] Step S204: Map the obstacle expansion body to a multi-scale three-dimensional mesh, identify all mesh cells that intersect with the obstacle expansion body as obstacle meshes, and identify the remaining mesh cells in the multi-scale three-dimensional mesh other than all obstacle meshes as free-flight meshes to obtain the initial collision detection three-dimensional mesh.

[0090] Specifically, obstacle expansions are mapped to a multi-scale 3D mesh using voxel rasterization. If a mesh cell intersects with or is contained within an obstacle expansion, it is identified as an obstacle mesh. If a mesh cell is completely outside the expansion body and has no overlap, it is identified as a free-flying mesh. This refers to areas where drones can safely traverse.

[0091] Step S205: Assign a unique octree code to all grid cells in the initial collision detection 3D mesh to obtain the collision detection 3D mesh.

[0092] To achieve rapid localization and retrieval, a unique octree code is assigned to all grid cells in the initial collision detection 3D grid. This code is generated based on the octree hierarchical structure, with the high-order bits representing the root node and higher-level parent nodes, and the low-order bits corresponding to the leaf node positions. This supports grid access and collision query. Finally, the collision detection 3D grid after octree coding is obtained, which significantly improves the real-time performance and accuracy of obstacle identification in path planning.

[0093] In some embodiments, based on the location information of the inspection target point and the preset camera parameters of the UAV, multiple grid cells not identified as obstacle grids are selected as candidate waypoints in the collision detection 3D grid, including:

[0094] Step S301: Determine the optimal inspection distance based on the preset camera parameters of the drone.

[0095] In order to ensure that the images captured by the drone can clearly distinguish the details of the substation equipment (such as insulator cracks and loose screws), the optimal inspection distance is calculated based on the camera performance parameters of the drone. :

[0096]

[0097] In the formula, Set the required ground target resolution (e.g., 0.5 cm / pixel to meet power defect identification requirements). The physical focal length for a camera mounted on a drone. This refers to the horizontal pixel resolution of the camera sensor (e.g., 5472 pixels). This refers to the physical width of the camera sensor (e.g., 13.2 mm).

[0098] Step S302: Based on the location information of the inspection target point and the optimal inspection distance, emit rays in a hemispherical shape into the free-flying grid adjacent to the inspection target point.

[0099] Step S303: Select all free-flying grids that intersect with the free-flying grid but do not intersect with the obstacle grid as candidate waypoints.

[0100] In this application, after obtaining the optimal inspection distance, a reverse visibility analysis strategy from the inspection target point to the viewpoint is introduced: using the extracted equipment inspection point as the emission source, detection rays are emitted outward in a fan shape. The area where the ray intersects with the collision detection grid in the multi-scale grid and is not occluded and falls within the safe flight zone grid is defined as the observability zone of the inspection point.

[0101] In the specific implementation process, the inspection target point is... For the center of the ball, A free-flying grid with radius [missing information] The interior emits rays in a hemispherical shape. If the rays do not intersect the obstacle grid... If they intersect, then the free-flying grid is... That is, marked as a candidate waypoint ,like Figure 3As shown, the camera light is projected back from the candidate observation point (the overlapping position of the pentagram mark) based on the voxel on the equipment surface (the location of the inspection target point). Combined with the upsampled field of view, it is determined whether the light is blocked by the voxel already occupied (the equipment entity or the obstacle). In this way, candidate waypoints that can cover the target surface voxel without obstruction and are located in the safe area are selected, providing effective observation positions that are collision-free and can be clearly imaged for subsequent path planning.

[0102] In some embodiments, based on the spatial geometric relationship between each candidate waypoint and the inspection target point, the optimal observation waypoint for the inspection target point under each preset observation viewpoint is selected from all candidate waypoints to obtain the optimal observation waypoint set, including:

[0103] Step S401: Map all candidate waypoints to a three-dimensional polar coordinate system with the inspection target point as the origin, and divide the three-dimensional polar coordinate system into several view sectors according to the preset observation view.

[0104] In this process, all candidate waypoints are mapped to a three-dimensional polar coordinate system according to their azimuth and pitch angles relative to the inspection target point, and the space is divided into several view sectors according to the preset view resolution, with each view sector corresponding to a local view coverage area.

[0105] Step S402: Within each view sector, based on the spatial geometric relationship between each candidate waypoint and the inspection target point within the view sector, determine the distance deviation and angle deviation of each candidate waypoint, and determine the comprehensive inspectability of each candidate waypoint based on the weighted calculation results of the distance deviation and angle deviation.

[0106] Among them, for any candidate waypoint For inspection points The image quality is mainly affected by the deviation of the actual distance and the deviation of the incident angle. Therefore, a comprehensive inspectability is constructed by weighted fusion of the distance deviation and the angle deviation, which ensures that the imaging resolution meets the constraints and also ensures the diversity of viewpoints and geometric robustness.

[0107] Among them, the distance deviation is evaluated through the distance evaluation function. To determine the distance evaluation function Using a Gaussian attenuation model, the actual distance The greater the deviation from the optimal distance (Right now The lower the score:

[0108]

[0109] In the formula, This represents the standard deviation of the distance tolerance.

[0110] Angular deviation is evaluated through an angular evaluation function. To determine the optimal effect (0° angle), the drone's lens optical axis is completely opposite to the normal vector of the inspection point (perpendicular illumination). This leads to the introduction of the incident angle. Gaussian penalty:

[0111] )

[0112] In the formula, Ideally, the angle should be 0 degrees, meaning shooting directly at the camera. The standard deviation is the tolerance for angle.

[0113] Overall inspectability Then it is:

[0114]

[0115] Step S403: Based on the overall inspectability of each candidate waypoint, select the candidate waypoint with the highest overall inspectability in each view sector as the optimal observation waypoint in the view sector.

[0116] Step S404: Construct the optimal observation waypoint set based on the optimal observation waypoints in all view sectors.

[0117] Within each view sector, only the overall score is retained. The highest and only waypoint is used as the optimal observation waypoint for that viewpoint. In this way, the original hundreds or thousands of dense and redundant waypoints are sparsified (pruned) into a set of single-digit high-quality, multi-view key waypoints. .

[0118] Among them, the analysis based on reverse visibility can solve the computational bottleneck problem of traditional ray forward detection, greatly improving the efficiency of visibility solution in complex substations. Furthermore, by evaluating the rationality of the spatial distribution of waypoints in the polar coordinate system and eliminating redundant waypoints, it can also improve the geometric robustness and view coverage completeness of the inspection path.

[0119] In some embodiments, based on a heuristic cost algorithm and combined with a collision detection 3D mesh, a globally optimal inspection path connecting each optimal observation waypoint is planned from the optimal observation waypoint set for each inspection target point, including:

[0120] Step S501: Under the constraints of the collision detection 3D mesh, sort the path nodes of the preset path start point, preset path end point and the optimal observation waypoint set of each inspection target point to obtain the inspection waypoint path node sequence.

[0121] The inspection path must be absolutely safe, meaning that all waypoints are located in the free space of the octree collision detection grid. Therefore, under the constraints of the collision detection 3D grid, it is necessary to strictly avoid obstacle grids. The preset path start point, preset path end point, and the optimal observation waypoint set of each inspection target point are pre-sorted to determine the order of waypoint visits and obtain the inspection waypoint path node sequence.

[0122] Step S502: Based on the preset heuristic cost function, perform cost evaluation on each node of the inspection waypoint path node sequence, and optimize the path of the inspection waypoint path node sequence based on the cost evaluation results to generate a globally optimal inspection path that satisfies the requirement of minimizing the cost evaluation results; wherein, the heuristic cost function is constructed by weighted fusion based on the flight distance between each waypoint, the turning angle, and the spatial distance penalty between the waypoint and the threat device.

[0123] In substation scenarios, drones not only need the shortest path but also must stay away from areas with strong electromagnetic interference and high-voltage cables. The definition starts from the current waypoint node. Move to the next node Heuristic cost function for:

[0124]

[0125] In the formula, This represents the actual physical distance already traveled. The estimated straight-line distance (Euclidean distance) to the next target waypoint. This is a penalty for space physics risks. Based on an octree mesh model, the penalty value increases exponentially the closer the location is to the boundary of high-voltage equipment or when passing under a complex busbar bridge. These are the corresponding weighting coefficients.

[0126] Preferably, in the process of finding the next node n′ from the current node n, there are many choices for the next node. The node with the minimum cost evaluation is always selected as the position for the next flight, and finally a minimum local optimum path is found. In addition, the weights α, β, γ are often not fixed, but dynamically adjusted according to the current task and the local environment.

[0127] This application's embodiments achieve fully automated, semantic-level waypoint extraction based on a precise 3D vector model. By leveraging inverse visibility analysis and octree multi-scale grid technology, 3D spatial visibility and obstacle avoidance calculations are significantly simplified, enabling the planning algorithm to achieve second-level planning for large substations on ordinary microcomputers. By introducing a perspective diversity-driven sparse optimization mechanism, the number of UAV waypoints is reduced to less than 30% of traditional grid-based planning methods, significantly extending the single inspection range of the UAV and saving manpower and time costs.

[0128] Based on the same inventive concept, this application also provides a substation drone inspection path planning system for implementing the above-mentioned substation drone inspection path planning method.

[0129] The solution provided by this system is similar to the solution described in the above method. Therefore, the specific limitations of one or more embodiments of the substation UAV inspection path planning system provided below can be found in the limitations of the substation UAV inspection path planning method described above, and will not be repeated here.

[0130] like Figure 4 As shown in the figure, this application provides a substation unmanned aerial vehicle (UAV) inspection path planning system, including:

[0131] The patrol point acquisition module 100 is used to acquire the three-dimensional vector model of the target substation, extract the location information of each inspection target point based on the three-dimensional vector model, and construct a set of inspection target points.

[0132] Octree mesh generation module 200 is used to perform multi-scale three-dimensional mesh generation of the global space of the target substation based on octree structure, and to mark the obstacle meshes of the generated mesh cells to obtain a collision detection three-dimensional mesh.

[0133] The candidate waypoint determination module 300 is used to select multiple grid cells that are not identified as obstacle grids as candidate waypoints for each inspection target point in the inspection target point set, based on the location information of the inspection target point and the preset camera parameters of the UAV;

[0134] The optimal waypoint selection module 400 is used to select the optimal observation waypoint of the inspection target point from all candidate waypoints based on the spatial geometric relationship between each candidate waypoint and the inspection target point, and obtain the optimal observation waypoint set.

[0135] The inspection path planning module 500 is used to plan the globally optimal inspection path connecting each optimal observation waypoint from the set of optimal observation waypoints for each inspection target point, based on a heuristic cost algorithm and combined with a collision detection 3D mesh.

[0136] In some embodiments, the cruise point acquisition module 100 is used for:

[0137] Obtain the geometric features and semantic labels of each device object in the 3D vector model;

[0138] Based on semantic tags, multiple equipment objects under the preset inspection equipment type are filtered, and the inspection target surface of the equipment objects is identified based on the geometric features of the filtered equipment objects.

[0139] The location information of each inspection target point is determined based on the geometric center and normal vector direction of the inspection target surface;

[0140] Based on the location information of each inspection target point, a set of inspection target points is constructed.

[0141] In some embodiments, the octree mesh partitioning module 200 is used for:

[0142] Based on the three-dimensional vector model of the target substation, determine the global three-dimensional spatial bounding box of the target substation;

[0143] The global 3D spatial bounding box is used as the root node of an octree, and the root node of the octree is recursively divided into octree child nodes to obtain a multi-scale 3D mesh.

[0144] Based on the three-dimensional vector model of the target substation, the geometric envelopes of all obstacles are extracted, and the geometric envelopes of the obstacles are subjected to safe distance expansion processing to generate obstacle expansion bodies;

[0145] The obstacle expansion is mapped onto a multi-scale 3D mesh, and all mesh cells that intersect with the obstacle expansion are identified as obstacle meshes. The remaining mesh cells in the multi-scale 3D mesh, excluding all obstacle meshes, are identified as free-flight meshes, thus obtaining the initial collision detection 3D mesh.

[0146] A unique octree code is assigned to all grid cells in the initial collision detection 3D mesh to obtain the collision detection 3D mesh.

[0147] In some embodiments, the candidate waypoint determination module 300 is configured to:

[0148] Determine the optimal inspection distance based on the drone's preset camera parameters;

[0149] Based on the location information of the inspection target point and the optimal inspection distance, rays are emitted in a hemispherical shape into the free-flying grid adjacent to the inspection target point;

[0150] All free-flying grids that intersect with the free-flying grid but do not intersect with the obstacle grid are selected as candidate waypoints.

[0151] In some embodiments, the optimal waypoint selection module 400 is used for:

[0152] All candidate waypoints are mapped to a three-dimensional polar coordinate system with the inspection target point as the origin, and the three-dimensional polar coordinate system is divided into several view sectors according to the preset observation view.

[0153] Within each view sector, based on the spatial geometric relationship between each candidate waypoint and the inspection target point, the distance deviation and angle deviation of each candidate waypoint are determined, and the comprehensive inspectability of each candidate waypoint is determined based on the weighted calculation results of the distance deviation and angle deviation.

[0154] Based on the overall inspectability of each candidate waypoint, the candidate waypoint with the highest overall inspectability is selected in each view sector as the optimal observation waypoint in the view sector.

[0155] Construct an optimal set of observation waypoints based on the optimal observation waypoints within all view sectors.

[0156] In some embodiments, the inspection path planning module 500 is used for:

[0157] Under the constraints of the collision detection 3D mesh, the path nodes are sorted according to the preset path start point, preset path end point, and the optimal observation waypoint set of each inspection target point to obtain the inspection waypoint path node sequence.

[0158] Based on a preset heuristic cost function, the cost of each node in the inspection waypoint path node sequence is evaluated, and the path of the inspection waypoint path node sequence is optimized based on the cost evaluation results to generate a globally optimal inspection path that minimizes the cost evaluation results. The heuristic cost function is constructed by weighted fusion based on the flight distance between each waypoint, the turning angle, and the spatial distance penalty between the waypoint and the threat device.

[0159] like Figure 5 As shown, this application embodiment provides an electronic device. The electronic device 10 includes a memory 20 and a processor 30. The memory 20 stores a computer program. When the computer program is executed by the processor 30, the processor 30 performs the following steps:

[0160] Obtain a 3D vector model of the target substation, extract the location information of each inspection target point based on the 3D vector model, and construct a set of inspection target points;

[0161] The global space of the target substation is divided into a multi-scale three-dimensional mesh based on an octree structure, and obstacle meshes are marked on the divided mesh cells to obtain a collision detection three-dimensional mesh.

[0162] For each inspection target point in the inspection target point set, based on the location information of the inspection target point and the preset camera parameters of the UAV, multiple grid cells that are not marked as obstacle grids are selected as candidate waypoints in the collision detection 3D grid;

[0163] For each inspection target point, based on the spatial geometric relationship between each candidate waypoint and the inspection target point, the optimal observation waypoint for the inspection target point under each preset observation view is selected from all candidate waypoints, and the optimal observation waypoint set is obtained.

[0164] Based on the heuristic cost algorithm and combined with the collision detection 3D mesh, a globally optimal inspection path connecting each optimal observation waypoint is planned from the optimal observation waypoint set of each inspection target point.

[0165] In some embodiments, when the computer program is executed by the processor 30, the processor 30 further performs the following steps:

[0166] Obtain the geometric features and semantic labels of each device object in the 3D vector model;

[0167] Based on semantic tags, multiple equipment objects under the preset inspection equipment type are filtered, and the inspection target surface of the equipment objects is identified based on the geometric features of the filtered equipment objects.

[0168] The location information of each inspection target point is determined based on the geometric center and normal vector direction of the inspection target surface;

[0169] Based on the location information of each inspection target point, a set of inspection target points is constructed.

[0170] In some embodiments, when the computer program is executed by the processor 30, the processor 30 further performs the following steps:

[0171] Based on the three-dimensional vector model of the target substation, determine the global three-dimensional spatial bounding box of the target substation;

[0172] The global 3D spatial bounding box is used as the root node of an octree, and the root node of the octree is recursively divided into octree child nodes to obtain a multi-scale 3D mesh.

[0173] Based on the three-dimensional vector model of the target substation, the geometric envelopes of all obstacles are extracted, and the geometric envelopes of the obstacles are subjected to safe distance expansion processing to generate obstacle expansion bodies;

[0174] The obstacle expansion is mapped onto a multi-scale 3D mesh, and all mesh cells that intersect with the obstacle expansion are identified as obstacle meshes. The remaining mesh cells in the multi-scale 3D mesh, excluding all obstacle meshes, are identified as free-flight meshes, thus obtaining the initial collision detection 3D mesh.

[0175] A unique octree code is assigned to all grid cells in the initial collision detection 3D mesh to obtain the collision detection 3D mesh.

[0176] In some embodiments, when the computer program is executed by the processor 30, the processor 30 further performs the following steps:

[0177] Determine the optimal inspection distance based on the drone's preset camera parameters;

[0178] Based on the location information of the inspection target point and the optimal inspection distance, rays are emitted in a hemispherical shape into the free-flying grid adjacent to the inspection target point;

[0179] All free-flying grids that intersect with the free-flying grid but do not intersect with the obstacle grid are selected as candidate waypoints.

[0180] In some embodiments, when the computer program is executed by the processor 30, the processor 30 further performs the following steps:

[0181] All candidate waypoints are mapped to a three-dimensional polar coordinate system with the inspection target point as the origin, and the three-dimensional polar coordinate system is divided into several view sectors according to the preset observation view.

[0182] Within each view sector, based on the spatial geometric relationship between each candidate waypoint and the inspection target point, the distance deviation and angle deviation of each candidate waypoint are determined, and the comprehensive inspectability of each candidate waypoint is determined based on the weighted calculation results of the distance deviation and angle deviation.

[0183] Based on the overall inspectability of each candidate waypoint, the candidate waypoint with the highest overall inspectability is selected in each view sector as the optimal observation waypoint in the view sector.

[0184] Construct an optimal set of observation waypoints based on the optimal observation waypoints within all view sectors.

[0185] In some embodiments, when the computer program is executed by the processor 30, the processor 30 further performs the following steps:

[0186] Under the constraints of the collision detection 3D mesh, the path nodes are sorted according to the preset path start point, preset path end point, and the optimal observation waypoint set of each inspection target point to obtain the inspection waypoint path node sequence.

[0187] Based on a preset heuristic cost function, the cost of each node in the inspection waypoint path node sequence is evaluated, and the path of the inspection waypoint path node sequence is optimized based on the cost evaluation results to generate a globally optimal inspection path that minimizes the cost evaluation results. The heuristic cost function is constructed by weighted fusion based on the flight distance between each waypoint, the turning angle, and the spatial distance penalty between the waypoint and the threat device.

[0188] This application provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed, it performs the following steps:

[0189] Obtain a 3D vector model of the target substation, extract the location information of each inspection target point based on the 3D vector model, and construct a set of inspection target points;

[0190] The global space of the target substation is divided into a multi-scale three-dimensional mesh based on an octree structure, and obstacle meshes are marked on the divided mesh cells to obtain a collision detection three-dimensional mesh.

[0191] For each inspection target point in the inspection target point set, based on the location information of the inspection target point and the preset camera parameters of the UAV, multiple grid cells that are not marked as obstacle grids are selected as candidate waypoints in the collision detection 3D grid;

[0192] For each inspection target point, based on the spatial geometric relationship between each candidate waypoint and the inspection target point, the optimal observation waypoint for the inspection target point under each preset observation view is selected from all candidate waypoints, and the optimal observation waypoint set is obtained.

[0193] Based on the heuristic cost algorithm and combined with the collision detection 3D mesh, a globally optimal inspection path connecting each optimal observation waypoint is planned from the optimal observation waypoint set of each inspection target point.

[0194] In some embodiments, when a computer program is executed, it also performs the following steps:

[0195] Obtain the geometric features and semantic labels of each device object in the 3D vector model;

[0196] Based on semantic tags, multiple equipment objects under the preset inspection equipment type are filtered, and the inspection target surface of the equipment objects is identified based on the geometric features of the filtered equipment objects.

[0197] The location information of each inspection target point is determined based on the geometric center and normal vector direction of the inspection target surface;

[0198] Based on the location information of each inspection target point, a set of inspection target points is constructed.

[0199] In some embodiments, when a computer program is executed, it also performs the following steps:

[0200] Based on the three-dimensional vector model of the target substation, determine the global three-dimensional spatial bounding box of the target substation;

[0201] The global 3D spatial bounding box is used as the root node of an octree, and the root node of the octree is recursively divided into octree child nodes to obtain a multi-scale 3D mesh.

[0202] Based on the three-dimensional vector model of the target substation, the geometric envelopes of all obstacles are extracted, and the geometric envelopes of the obstacles are subjected to safe distance expansion processing to generate obstacle expansion bodies;

[0203] The obstacle expansion is mapped onto a multi-scale 3D mesh, and all mesh cells that intersect with the obstacle expansion are identified as obstacle meshes. The remaining mesh cells in the multi-scale 3D mesh, excluding all obstacle meshes, are identified as free-flight meshes, thus obtaining the initial collision detection 3D mesh.

[0204] A unique octree code is assigned to all grid cells in the initial collision detection 3D mesh to obtain the collision detection 3D mesh.

[0205] In some embodiments, when a computer program is executed, it also performs the following steps:

[0206] Determine the optimal inspection distance based on the drone's preset camera parameters;

[0207] Based on the location information of the inspection target point and the optimal inspection distance, rays are emitted in a hemispherical shape into the free-flying grid adjacent to the inspection target point;

[0208] All free-flying grids that intersect with the free-flying grid but do not intersect with the obstacle grid are selected as candidate waypoints.

[0209] In some embodiments, when a computer program is executed, it also performs the following steps:

[0210] All candidate waypoints are mapped to a three-dimensional polar coordinate system with the inspection target point as the origin, and the three-dimensional polar coordinate system is divided into several view sectors according to the preset observation view.

[0211] Within each view sector, based on the spatial geometric relationship between each candidate waypoint and the inspection target point, the distance deviation and angle deviation of each candidate waypoint are determined, and the comprehensive inspectability of each candidate waypoint is determined based on the weighted calculation results of the distance deviation and angle deviation.

[0212] Based on the overall inspectability of each candidate waypoint, the candidate waypoint with the highest overall inspectability is selected in each view sector as the optimal observation waypoint in the view sector.

[0213] Construct an optimal set of observation waypoints based on the optimal observation waypoints within all view sectors.

[0214] In some embodiments, when a computer program is executed, it also performs the following steps:

[0215] Under the constraints of the collision detection 3D mesh, the path nodes are sorted according to the preset path start point, preset path end point, and the optimal observation waypoint set of each inspection target point to obtain the inspection waypoint path node sequence.

[0216] Based on a preset heuristic cost function, the cost of each node in the inspection waypoint path node sequence is evaluated, and the path of the inspection waypoint path node sequence is optimized based on the cost evaluation results to generate a globally optimal inspection path that minimizes the cost evaluation results. The heuristic cost function is constructed by weighted fusion based on the flight distance between each waypoint, the turning angle, and the spatial distance penalty between the waypoint and the threat device.

[0217] This application provides a computer program product, which includes a computer program stored on a non-transitory computer-readable storage medium. The computer program includes program instructions, wherein when the program instructions are executed by a computer, the computer performs the following steps:

[0218] Obtain a 3D vector model of the target substation, extract the location information of each inspection target point based on the 3D vector model, and construct a set of inspection target points;

[0219] The global space of the target substation is divided into a multi-scale three-dimensional mesh based on an octree structure, and obstacle meshes are marked on the divided mesh cells to obtain a collision detection three-dimensional mesh.

[0220] For each inspection target point in the inspection target point set, based on the location information of the inspection target point and the preset camera parameters of the UAV, multiple grid cells that are not marked as obstacle grids are selected as candidate waypoints in the collision detection 3D grid;

[0221] For each inspection target point, based on the spatial geometric relationship between each candidate waypoint and the inspection target point, the optimal observation waypoint for the inspection target point under each preset observation view is selected from all candidate waypoints, and the optimal observation waypoint set is obtained.

[0222] Based on the heuristic cost algorithm and combined with the collision detection 3D mesh, a globally optimal inspection path connecting each optimal observation waypoint is planned from the optimal observation waypoint set of each inspection target point.

[0223] In some embodiments, when program instructions are executed by a computer, the computer also performs the following steps:

[0224] Obtain the geometric features and semantic labels of each device object in the 3D vector model;

[0225] Based on semantic tags, multiple equipment objects under the preset inspection equipment type are filtered, and the inspection target surface of the equipment objects is identified based on the geometric features of the filtered equipment objects.

[0226] The location information of each inspection target point is determined based on the geometric center and normal vector direction of the inspection target surface;

[0227] Based on the location information of each inspection target point, a set of inspection target points is constructed.

[0228] In some embodiments, when program instructions are executed by a computer, the computer also performs the following steps:

[0229] Based on the three-dimensional vector model of the target substation, determine the global three-dimensional spatial bounding box of the target substation;

[0230] The global 3D spatial bounding box is used as the root node of an octree, and the root node of the octree is recursively divided into octree child nodes to obtain a multi-scale 3D mesh.

[0231] Based on the three-dimensional vector model of the target substation, the geometric envelopes of all obstacles are extracted, and the geometric envelopes of the obstacles are subjected to safe distance expansion processing to generate obstacle expansion bodies;

[0232] The obstacle expansion is mapped onto a multi-scale 3D mesh, and all mesh cells that intersect with the obstacle expansion are identified as obstacle meshes. The remaining mesh cells in the multi-scale 3D mesh, excluding all obstacle meshes, are identified as free-flight meshes, thus obtaining the initial collision detection 3D mesh.

[0233] A unique octree code is assigned to all grid cells in the initial collision detection 3D mesh to obtain the collision detection 3D mesh.

[0234] In some embodiments, when program instructions are executed by a computer, the computer also performs the following steps:

[0235] Determine the optimal inspection distance based on the drone's preset camera parameters;

[0236] Based on the location information of the inspection target point and the optimal inspection distance, rays are emitted in a hemispherical shape into the free-flying grid adjacent to the inspection target point;

[0237] All free-flying grids that intersect with the free-flying grid but do not intersect with the obstacle grid are selected as candidate waypoints.

[0238] In some embodiments, when program instructions are executed by a computer, the computer also performs the following steps:

[0239] All candidate waypoints are mapped to a three-dimensional polar coordinate system with the inspection target point as the origin, and the three-dimensional polar coordinate system is divided into several view sectors according to the preset observation view.

[0240] Within each view sector, based on the spatial geometric relationship between each candidate waypoint and the inspection target point, the distance deviation and angle deviation of each candidate waypoint are determined, and the comprehensive inspectability of each candidate waypoint is determined based on the weighted calculation results of the distance deviation and angle deviation.

[0241] Based on the overall inspectability of each candidate waypoint, the candidate waypoint with the highest overall inspectability is selected in each view sector as the optimal observation waypoint in the view sector.

[0242] Construct an optimal set of observation waypoints based on the optimal observation waypoints within all view sectors.

[0243] In some embodiments, when program instructions are executed by a computer, the computer also performs the following steps:

[0244] Under the constraints of the collision detection 3D mesh, the path nodes are sorted according to the preset path start point, preset path end point, and the optimal observation waypoint set of each inspection target point to obtain the inspection waypoint path node sequence.

[0245] Based on a preset heuristic cost function, the cost of each node in the inspection waypoint path node sequence is evaluated, and the path of the inspection waypoint path node sequence is optimized based on the cost evaluation results to generate a globally optimal inspection path that minimizes the cost evaluation results. The heuristic cost function is constructed by weighted fusion based on the flight distance between each waypoint, the turning angle, and the spatial distance penalty between the waypoint and the threat device.

[0246] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, electronic devices, computer storage media, and computer program products described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0247] In the embodiments of this application, the terms "exemplary" or "for example" are used to indicate that something is an example, illustration, or description. Any embodiment or design that is described as "exemplary" or "for example" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or design. Specifically, the use of the terms "exemplary" or "for example" is intended to present the relevant concepts in a specific manner.

[0248] It should be noted that the terms "comprising" and "having" and any variations thereof in the specification, claims and accompanying drawings of this invention are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are explicitly listed, but may include other steps or units that are not explicitly listed or that are inherent to such processes, methods, products or devices.

[0249] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0250] In the several embodiments provided by this invention, it should be understood that the disclosed systems, electronic devices, computer storage media, computer program products, and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces, indirect coupling or communication connection between devices or units, and may be electrical, mechanical, or other forms.

[0251] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0252] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0253] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for executing all or part of the steps of the methods described in the various embodiments of the present invention through a computer device (which may be a personal computer, a server, or a network device, etc.). The aforementioned storage medium includes: USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, optical disks, and other media capable of storing program code.

[0254] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for planning unmanned aerial vehicle (UAV) inspection paths for substations, characterized in that, include: Obtain a three-dimensional vector model of the target substation, extract the location information of each inspection target point based on the three-dimensional vector model, and construct an inspection target point set; The global space of the target substation is divided into a multi-scale three-dimensional mesh based on an octree structure, and obstacle meshes are marked on the divided mesh cells to obtain a collision detection three-dimensional mesh. For each of the inspection target points in the set of inspection target points, based on the location information of the inspection target point and the preset camera parameters of the UAV, multiple grid cells that are not identified as obstacle grids are selected as candidate waypoints in the collision detection three-dimensional grid. For each inspection target point, based on the spatial geometric relationship between each candidate waypoint and the inspection target point, the optimal observation waypoint for the inspection target point under each preset observation view is selected from all the candidate waypoints to obtain the optimal observation waypoint set; Based on the heuristic cost algorithm and combined with the collision detection 3D mesh, a globally optimal inspection path connecting each of the optimal observation waypoints is planned from the optimal observation waypoint set of each of the inspection target points.

2. The substation UAV inspection path planning method according to claim 1, characterized in that, The step of extracting the location information of each inspection target point based on the three-dimensional vector model and constructing an inspection target point set includes: Obtain the geometric features and semantic tags of each device object in the three-dimensional vector model; Based on the semantic tags, multiple equipment objects under the preset inspection equipment type are filtered, and based on the geometric features of the filtered equipment objects, the inspection target surface of the equipment objects is identified. The position information of each inspection target point is determined based on the geometric center and normal vector direction of the inspection target surface; Based on the location information of each inspection target point, the set of inspection target points is constructed.

3. The substation UAV inspection path planning method according to claim 1, characterized in that, The process involves dividing the global space of the target substation into a multi-scale three-dimensional mesh based on an octree structure, and marking the divided mesh cells with obstacle meshes to obtain a collision detection three-dimensional mesh, including: Based on the three-dimensional vector model of the target substation, determine the global three-dimensional spatial bounding box of the target substation; The global 3D spatial bounding box is used as the root node of an octree, and the root node of the octree is recursively divided into octree child nodes to obtain a multi-scale 3D mesh. Based on the three-dimensional vector model of the target substation, the geometric envelopes of all obstacles are extracted, and the geometric envelopes of the obstacles are subjected to safety distance expansion processing to generate obstacle expansion bodies; The obstacle expansion is mapped onto the multi-scale three-dimensional mesh, and all mesh cells that intersect with the obstacle expansion are identified as obstacle meshes. The remaining mesh cells in the multi-scale three-dimensional mesh, excluding all obstacle meshes, are identified as free-flight meshes, thus obtaining the initial collision detection three-dimensional mesh. A unique octree code is assigned to all grid cells in the initial collision detection 3D mesh to obtain the collision detection 3D mesh.

4. The substation UAV inspection path planning method according to claim 3, characterized in that, The step of selecting multiple grid cells not identified as obstacle grids as candidate waypoints from the collision detection 3D grid based on the location information of the inspection target point and the preset camera parameters of the UAV includes: The optimal inspection distance is determined based on the preset camera parameters of the drone; Based on the location information of the inspection target point and the optimal inspection distance, rays are emitted in a hemispherical shape into the free-flying grid adjacent to the inspection target point; All free-flying grids that intersect with the free-flying grid but do not intersect with the obstacle grid are selected as candidate waypoints.

5. The substation UAV inspection path planning method according to claim 1, characterized in that, The optimal observation waypoint set is obtained by selecting the optimal observation waypoint from all the candidate waypoints based on the spatial geometric relationship between each candidate waypoint and the inspection target point, under each preset observation viewpoint, including: All candidate waypoints are mapped to a three-dimensional polar coordinate system with the inspection target point as the origin, and the three-dimensional polar coordinate system is divided into several view sectors according to a preset observation view. Within each view sector, based on the spatial geometric relationship between each candidate waypoint and the inspection target point within the view sector, the distance deviation and angle deviation of each candidate waypoint are determined, and based on the weighted calculation result of the distance deviation and the angle deviation, the comprehensive inspectability of each candidate waypoint is determined; Based on the overall inspectability of each candidate waypoint, the candidate waypoint with the highest overall inspectability is selected in each view sector as the optimal observation waypoint in the view sector. The optimal observation waypoint set is constructed based on the optimal observation waypoints within all the aforementioned view sectors.

6. The substation UAV inspection path planning method according to claim 1, characterized in that, The heuristic cost algorithm, combined with the collision detection 3D mesh, plans a globally optimal inspection path connecting each of the optimal observation waypoints in the optimal observation waypoint set for each of the inspection target points, including: Under the constraints of the collision detection 3D grid, the path nodes of the preset path start point, preset path end point, and the optimal observation waypoint set of each inspection target point are sorted to obtain the inspection waypoint path node sequence. Based on a preset heuristic cost function, the cost of each node in the inspection waypoint path node sequence is evaluated, and the path of the inspection waypoint path node sequence is optimized based on the cost evaluation results to generate the globally optimal inspection path that minimizes the cost evaluation results; wherein, the heuristic cost function is constructed by weighted fusion based on the flight distance between each waypoint, the turning angle, and the spatial distance penalty between the waypoint and the threat device.

7. A substation unmanned aerial vehicle (UAV) inspection path planning system, characterized in that, include: The patrol point acquisition module is used to acquire a three-dimensional vector model of the target substation, extract the location information of each inspection target point based on the three-dimensional vector model, and construct an inspection target point set. The octree mesh generation module is used to perform multi-scale three-dimensional mesh generation on the global space of the target substation based on the octree structure, and to mark the mesh cells as obstacles to obtain a collision detection three-dimensional mesh. The candidate waypoint determination module is used to select multiple grid cells that are not identified as obstacle grids as candidate waypoints for each of the inspection target points in the inspection target point set, based on the location information of the inspection target point and the preset camera parameters of the UAV; The optimal waypoint filtering module is used to select the optimal observation waypoint of the inspection target point from all the candidate waypoints from each of the inspection target points based on the spatial geometric relationship between each candidate waypoint and the inspection target point, so as to obtain the optimal observation waypoint set. The inspection path planning module is used to plan a globally optimal inspection path connecting each of the optimal observation waypoints in the set of optimal observation waypoints for each of the inspection target points, based on a heuristic cost algorithm and in combination with the collision detection 3D mesh.

8. An electronic device, characterized in that, The electronic device includes a memory and a processor. The memory stores a computer program. When the computer program is executed by the processor, the processor performs the steps of the substation UAV inspection path planning method as described in any one of claims 1-6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed, it implements the steps of the substation UAV inspection path planning method as described in any one of claims 1-6.

10. A computer program product, characterized in that, The computer program product includes a computer program stored on a non-transitory computer-readable storage medium, the computer program including program instructions, wherein when the program instructions are executed by a computer, the computer performs the steps of the substation UAV inspection path planning method as described in any one of claims 1-6.

Images