Unmanned aerial vehicle supervision method, device and equipment

Abstract

The invention provides an unmanned aerial vehicle supervision method, device and equipment, and relates to the technical field of unmanned aerial vehicle control. The method comprises the following steps: acquiring associated data of a to-be-supervised operation site; for each obstacle, based on the associated data, determining a plurality of bypassing exits meeting safe flight constraints through multi-exit guiding obstacle bypassing analysis; based on each detour exit, constructing a candidate detour route for avoiding the obstacle and a candidate flight route which is corresponding to the candidate detour route and flies from the detour exit to the target position of the unmanned aerial vehicle; for the candidate detour route and the candidate flight route, taking the shortest path length, the shortest flight time and the minimum obstacle approaching risk coefficient as targets, performing joint cost optimization to obtain an optimal route; and controlling the unmanned aerial vehicle to fly to the unmanned aerial vehicle target position along the optimal route, and supervising the to-be-supervised operation site. According to the invention, the efficiency of supervising the operation site by the unmanned aerial vehicle can be improved.

Description

Technical Field

[0001] This invention relates to the field of unmanned aerial vehicle (UAV) control technology, and in particular to a UAV inspection method, apparatus, and equipment. Background Technology

[0002] With the large-scale expansion and remote distribution of work sites in fields such as power and infrastructure, drones, with their advantages of flexibility and mobility, have been gradually applied to work site safety supervision scenarios, becoming an important means to replace manual labor in completing high-risk, wide-area supervision tasks.

[0003] Current drone inspection methods mainly rely on manual judgment of the location of the work site based on experience, manual setting of flight routes, or the system generating fixed routes based on simple path planning algorithms to guide the drone to the target inspection area.

[0004] However, work sites are often densely covered with various obstacles, such as power transmission lines, towers, construction machinery, and mountains. Existing methods can only adopt a uniform, conservative detour strategy for multiple obstacles to ensure that the drone will not collide with any of them. This results in redundancy and detours in the detour paths, significantly increasing the ineffective flight distance and ultimately causing low efficiency in the drone's flight to the target location, which affects the overall efficiency of drone supervision. Summary of the Invention

[0005] This invention provides a method, apparatus, and equipment for drone supervision to solve the problem of low efficiency when drones are used to supervise work sites in traditional methods.

[0006] In a first aspect, embodiments of the present invention provide a method for drone supervision, comprising: acquiring associated data of the work site to be supervised, the associated data including the drone target position, the drone initial position, and the positional and geometric parameters of obstacles; for each obstacle, based on the associated data, determining multiple detour exits that meet safe flight constraints through multi-exit guided obstacle detour analysis; constructing candidate detour routes to avoid obstacles based on each detour exit, and candidate flight routes corresponding to the candidate detour routes from the detour exits to the drone target position; for the candidate detour routes and candidate flight routes, performing joint cost optimization with the objectives of minimizing path length, flight time, and the risk coefficient of proximity to obstacles, to obtain the optimal route; controlling the drone to fly along the optimal route to the drone target position for supervision of the work site to be supervised.

[0007] In one possible implementation, for each obstacle, based on associated data, multiple detour exits satisfying safe flight constraints are determined through multi-exit guided obstacle detour analysis. This includes: for each obstacle, projecting a polygonal cross-section of the obstacle onto the plane of the UAV's flight based on the obstacle's position and geometric parameters; determining the UAV's left and right initial tangent directions based on the UAV's initial position and the two vertices closest to the UAV's initial position in the polygonal cross-section; controlling the UAV to virtually attach to the boundary of the obstacle and fly along the left and right initial tangent directions; determining potential detour exits based on whether the ray formed by the current boundary point and the UAV's target position intersects with the obstacle after each preset distance traveled; repeating the boundary flight and potential detour exit determination steps to obtain multiple detour exits satisfying safe flight constraints.

[0008] In one possible implementation, after advancing a preset distance, a potential detour exit is determined based on whether the ray formed by the current boundary point and the UAV target position intersects with an obstacle. This includes: pausing virtual flight after advancing a preset distance and recording the coordinates of the current boundary point; constructing a ray with the current boundary point coordinates as the starting point and the UAV target position as the ending point; if the ray intersects with any obstacle, continuing virtual flight along the obstacle boundary and repeating the steps of advancing the preset distance and constructing the ray; if the ray does not intersect with any obstacle, the current boundary point is determined as a potential detour exit.

[0009] In one possible implementation, for candidate detour routes and candidate flight routes, the optimal route is obtained by optimizing the joint cost with the objectives of minimizing path length, flight time, and obstacle proximity risk coefficient. This includes: calculating the detour cost for each candidate detour route based on path length, estimated flight time, and obstacle proximity risk coefficient; calculating the flight cost for each candidate flight route based on path length, estimated flight time, and obstacle proximity risk coefficient; calculating the joint cost for each candidate detour route and its corresponding candidate flight route based on the detour cost and flight cost; and ranking the joint costs of all combinations of candidate detour routes and candidate flight routes, with the combination having the smallest joint cost being the optimal route.

[0010] In one possible implementation, before sorting the joint costs corresponding to all candidate detour routes and candidate flight routes and selecting the combination with the smallest joint cost as the optimal route, the following steps are included: for the combination with the smallest joint cost, determining a splicing node based on the end point of the candidate detour route and the start point of the candidate flight route; using a Bézier curve interpolation algorithm, performing a smooth transition processing on the trajectory segment at the end of the candidate detour route and the trajectory segment at the start of the candidate flight route at the splicing node to obtain a continuous transition trajectory segment; fusing the transition trajectory segment with the candidate detour route and the candidate flight route to obtain a fused trajectory; and obtaining the optimal route based on the fused trajectory.

[0011] In one possible implementation, before acquiring the associated data of the work site to be inspected, the following steps are also included: calculating the geographic coordinates of the core work point corresponding to the central area of ​​the control sphere's image based on the location information and basic parameters reported in real time by the control sphere; calculating the observation radius of the control sphere using the geographic coordinates of the core work point as the spatial reference origin; calculating the observation radius of the UAV based on the observation radius of the control sphere and a preset safety offset; constructing a collaborative observation ring with the geographic coordinates of the core work point as the center and the UAV's observation radius as the radius, and determining the point on the ring corresponding to the shortest path from the UAV's initial position to this collaborative observation ring as the UAV's target position.

[0012] In one possible implementation, controlling the drone to fly along the optimal flight path to the drone's target location for on-site inspection includes: controlling the drone to start from its initial position and fly along the optimal flight path; if an unknown obstacle is detected intruding into the preset safe airspace, replanning the path to the drone's target location based on the current position and controlling the drone to continue flying; after the drone arrives at the target location, hovering and locking the position to conduct on-site inspection.

[0013] In one possible implementation, after the drone arrives at the target location, it hovers and locks its position. Before conducting the inspection of the work site, the process includes: calculating the pitch angle and yaw angle of the core work point falling into the center of the lens's field of view based on the drone's target location and the geographic coordinates of the core work point to be inspected; adjusting the drone's lens attitude based on the pitch angle and yaw angle, while matching the lens focal length, and then conducting the inspection of the work site.

[0014] Secondly, embodiments of the present invention provide a drone inspection device, comprising: a communication module for acquiring associated data of the work site to be inspected, the associated data including the drone target position, the drone initial position, and the positional and geometric parameters of obstacles; a processing module for determining multiple detour exits that meet safe flight constraints based on the associated data and through multi-exit guided obstacle detour analysis; constructing candidate detour routes to avoid obstacles based on each detour exit, and candidate flight routes corresponding to the candidate detour routes from the detour exits to the drone target position; performing joint cost optimization on the candidate detour routes and candidate flight routes with the objectives of minimizing path length, flight time, and the risk coefficient of approaching obstacles, to obtain the optimal route; and controlling the drone to fly along the optimal route to the drone target position for inspection of the work site to be inspected.

[0015] Thirdly, embodiments of the present invention provide an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the method described in the first aspect or any possible implementation thereof.

[0016] This invention first generates multiple safe detour exits based on obstacle locations and geometric parameters, ensuring the detour path closely follows the obstacle, resulting in a compact detour and reducing ineffective detour distances. Then, it analyzes the candidate detour routes constructed based on these exits, as well as candidate flight routes from these exits to the UAV's target location. Finally, for each candidate detour route and candidate flight route, it performs joint cost optimization based on path length, flight time, and obstacle proximity risk coefficient, selecting the optimal route with the minimum joint cost. This shortens the total distance the UAV travels to the target location, reduces travel time, and ensures controllable flight safety, thereby improving the efficiency of UAVs reaching the target location and enhancing the overall efficiency of UAV supervision. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating the implementation of the drone inspection method provided in this embodiment of the invention. Figure 2 This is a schematic diagram of the structure of the drone inspection device provided in an embodiment of the present invention; Figure 3 This is a schematic diagram of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0018] The embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0019] See Figure 1 The document illustrates a flowchart of the drone inspection method provided in an embodiment of the present invention, which is described in detail below: Step 101: Obtain the associated data of the work site to be inspected. The associated data includes the target position of the UAV, the initial position of the UAV, and the position and geometric parameters of the obstacles.

[0020] In some embodiments, the drone target location is the specific spatial coordinate point (including latitude, longitude, and altitude information) that the drone ultimately needs to reach and hover at. It is the core observation location for the drone to perform its inspection mission. The drone target location serves as the endpoint benchmark for the drone's flight, clearly defining the drone's final deployment location and ensuring that the drone can accurately target the core area of ​​the work site.

[0021] In some embodiments, the initial position of the UAV is the starting spatial coordinate point (including latitude, longitude, and altitude information) when the UAV initiates its flight mission, serving as the starting reference for flight path planning. The initial position of the UAV, as the initial reference point for flight path planning, is used to calculate the flight path from the starting point to the target location, and is the basis for evaluating costs such as path length and flight time. The initial position of the UAV is the real-time coordinates obtained by the UAV through its own GPS / BeiDou positioning module at takeoff.

[0022] In some embodiments, obstacle location parameters are key data describing the spatial location of obstacles. These parameters include the obstacle's geographic coordinates (latitude and longitude, altitude range), boundary vertex coordinates, etc., used to define the spatial distribution range of the obstacle. Obstacle location parameters clearly define the obstacle's specific location in geographic space, providing a basis for determining whether a flight path conflicts with an obstacle and generating obstacle avoidance paths, thus avoiding collision risks during drone flight.

[0023] In some embodiments, the geometric parameters of an obstacle are key data describing the shape and size of the obstacle.

[0024] For example, the geometric parameters of an obstacle include its three-dimensional contour (such as a polygonal cross-section or a solid envelope), side length, height, and width, which are used to accurately model the physical form of the obstacle. By reconstructing the actual size and shape of the obstacle through geometric modeling, obstacle avoidance algorithms (such as the improved Bug algorithm with multi-exit guidance) can calculate the detour direction, detour path length, and safe distance from the obstacle, ensuring the feasibility and safety of the obstacle avoidance path.

[0025] Step 102: For each obstacle, based on the associated data, multiple detour exits that meet the safe flight constraints are determined through multi-exit guided obstacle detour analysis.

[0026] In some embodiments, multi-exit guided obstacle detour analysis is a method for obstacle detour path planning and analysis. Its core principle is not limited to a single detour direction (left / right detour), but actively explores multiple potential exit points along the obstacle boundary. By determining whether the path from the exit point to the target location is unobstructed, multiple detour exits that meet safety constraints are selected, providing multiple candidate solutions for subsequent global optimal route selection. Multi-exit guided obstacle detour analysis breaks through the traditional binary obstacle detour logic, generating multiple sets of feasible detour exits, avoiding problems such as lengthy paths and high risks caused by a single detour direction.

[0027] In some embodiments, safe flight constraints are the safety rules and physical limitations that must be followed during drone flight. Safe flight constraints include collision-free constraints, flight performance constraints, and operational specification constraints. Flight safety constraints serve as screening criteria for detour exits and detour routes, ensuring that all candidate solutions meet the safety baseline, eliminating paths that may lead to collisions, violations, or operational failures, and mitigating flight risks and the risk of inspection mission failure from the source.

[0028] For example, collision-free constraints include not conflicting with obstacles, no-fly zones, and height-restricted zones.

[0029] For example, flight performance constraints include meeting dynamic parameters such as the minimum turning radius and maximum flight speed of the UAV.

[0030] For example, operational constraints include maintaining a safe working distance from the surveillance sphere and meeting the airspace requirements for power operation supervision.

[0031] In some embodiments, a detour exit is a specific spatial coordinate point where, during the UAV's detour along the obstacle boundary, the path from that point to the target location is free of obstacles. It is the end point of the detour segment and the starting point of the subsequent flight path to the target location. As the connecting node between the obstacle detour path and the subsequent main flight path, each detour exit corresponds to a complete detour segment and a combination of remaining segments. It is the core unit for joint cost calculation and determines the feasibility of the obstacle detour path and the efficiency of the overall flight path.

[0032] As one possible implementation, step 102 can be specifically implemented as steps A11-A15.

[0033] A11: For each obstacle, on the plane of the UAV's flight, the polygonal cross-section of each obstacle is obtained by projecting based on the obstacle's position and geometric parameters.

[0034] A12: Based on the initial position of the UAV and the two vertices in the polygonal cross section that are closest to the initial position of the UAV, determine the left and right directions of the UAV around the initial tangent.

[0035] A13: Control the drone to virtually attach to the boundary of the obstacle, and fly along the left and right directions around the initial tangent.

[0036] A14: After each preset distance forward, determine potential detour exits based on whether the ray formed by the current boundary point and the drone target position intersects with obstacles.

[0037] In some embodiments, the plane in which the UAV flies is a two-dimensional horizontal plane (ignoring the vertical height dimension and retaining only the latitude and longitude coordinates) corresponding to the specific altitude layer in which the UAV is currently in the flight mission. This plane serves as the reference plane for obstacle projection modeling and detour direction calculation. By simplifying obstacles in three-dimensional space into polygonal cross-sections on a two-dimensional plane, the computational complexity of obstacle modeling and path planning is reduced, making calculations such as detour direction (left / right detour) and boundary following path easier to perform, thus meeting the requirements for stable flight of UAVs at specific altitude layers.

[0038] In some embodiments, the polygonal cross-section of an obstacle is a two-dimensional polygonal profile obtained by vertically projecting the obstacle (such as a tower, mountain, or temporary crane) in three-dimensional space onto the plane of the UAV's flight. This profile is used to characterize the horizontal occupancy of the obstacle at that altitude level, precisely defining the boundary range of the obstacle within the flight plane.

[0039] In some embodiments, the two vertices closest to the UAV's initial position in the polygonal cross-section are the two vertices on the two-dimensional polygonal cross-sectional profile of the obstacle that are closest to the UAV's initial position, selected through spatial distance calculation. These two vertices are the key geometric feature points defining the initial direction of the detour. Serving as the pointing targets for the initial tangential directions of the UAV's left and right detours, by drawing rays from the initial position to these two vertices, the two basic directions for detours around the obstacle are clearly defined, avoiding blind exploration of detour paths and improving the efficiency of detour analysis.

[0040] In some embodiments, the left-side initial tangent direction is the ray direction from the UAV's initial position to the nearest left vertex of the obstacle's polygonal cross-section, and is the initial path direction for the UAV to circumvent the obstacle's left boundary. The left-side initial tangent direction provides one of the basic circumvention direction options, which, together with the right-side initial tangent direction, constitutes the starting point for bidirectional exploration, ensuring that the circumvention path covers the main feasible directions and avoiding missing the optimal circumvention path.

[0041] In some embodiments, the initial right-side tangent direction is the ray direction from the UAV's initial position to the nearest right vertex of the obstacle's polygonal cross-section, representing the initial path direction of the UAV around the right boundary of the obstacle. Complementing the initial left-side tangent direction, it provides another basic bypass direction option, ensuring that bypass analysis covers potential feasible paths on both sides of the obstacle, laying the foundation for subsequent multi-exit detection.

[0042] In some embodiments, virtually attaching to the boundary of an obstacle and flying along the initial tangent direction to the left or right is a virtual flight process in an algorithm simulation environment. This involves keeping the drone close to the boundary of the polygonal cross-section of the obstacle (maintaining a safe distance) and continuously moving along the initial tangent direction to the left or right. This is the core action for detecting potential detour exits. Simulating the actual obstacle-avoiding trajectory of the drone, by gradually exploring along the boundary, provides continuous boundary points for detecting exits at each preset distance of advancement, ensuring comprehensive detection of all potential feasible exits on the obstacle boundary.

[0043] As one possible implementation, step A14 can be specifically implemented as steps B11-B14.

[0044] B11: Pause virtual flight after advancing a preset distance and record the coordinates of the current boundary point.

[0045] B12: Construct a ray starting from the current boundary point coordinates and ending at the UAV target position.

[0046] B13: If the ray intersects with any obstacle, it continues to fly forward virtually along the obstacle boundary, repeating the steps of advancing the preset distance and constructing the ray.

[0047] B14: If the ray does not intersect with any obstacles, the current boundary point is identified as a potential detour exit.

[0048] In some embodiments, the preset distance is a step size (e.g., 5 meters) pre-set in the algorithm when the UAV virtually attaches to the obstacle boundary during flight, and is the interval distance for triggering potential exit detection. The preset distance controls the detection frequency of potential exits, avoiding both excessively large step sizes that may cause the optimal exit to be missed, and excessively small step sizes that may increase computational load and reduce real-time performance, thus balancing the comprehensiveness of detection with algorithm efficiency.

[0049] In some embodiments, the current boundary point is the position (including planar coordinates) where the UAV pauses after each preset distance of flight while virtually attaching to the obstacle boundary. It serves as a reference point for detecting potential exits. The current boundary point, as the starting point of a ray, is used to determine whether the target location can be reached without obstacles from that point, and is a core reference point for determining potential detour exits.

[0050] In some embodiments, a ray is a straight line constructed from the current boundary point to the target location of the UAV, used to determine whether the boundary point has unobstructed access to the target location. As a screening tool for potential exits, the ray quickly determines whether the current boundary point is a feasible exit by detecting whether the ray intersects with obstacles, serving as a key criterion for multi-exit detection.

[0051] In some embodiments, a potential detour exit is a current boundary point that satisfies the condition that the ray does not intersect with any obstacle, and is a candidate exit point with the ability to fly unimpeded to the target location from that point. As candidate endpoints of the detour segment, multiple potential detour exits form the basis for subsequent candidate routes, providing sufficient candidate solutions for joint cost optimization and ensuring the optimality of the final route.

[0052] As one possible implementation, embodiments of the present invention can accurately identify potential detour exits that meet the requirements through a closed-loop logic of interval pausing, ray detection, and conditional screening, ensuring the feasibility and safety of the exits and providing reliable candidate samples for subsequent candidate route construction and joint cost optimization.

[0053] A15: Repeat the boundary flight and determine potential detour exit steps to obtain multiple detour exits that meet the safe flight constraints.

[0054] As one possible implementation, this invention explicitly projects 3D obstacles onto the UAV's flight plane as polygonal cross-sections, transforming the complex 3D obstacle avoidance problem into a 2D path planning problem. This simplification preserves the core area occupied by the obstacle at the flight altitude (horizontal boundary) while avoiding the significant computational overhead of 3D modeling. This makes subsequent operations such as nearest vertex calculation and tangent direction determination easier to execute, adapting to the real-time requirements of UAV edge computing devices. By finding the two vertices in the polygonal cross-section closest to the UAV's initial position, the initial tangent direction for left / right bypass is determined, providing a clear starting direction for the obstacle avoidance path. Compared to traditional unguided obstacle avoidance algorithms, this design can quickly lock the core bypass paths on both sides of the obstacle, preventing the UAV from blindly exploring around the obstacle, significantly improving the detection efficiency of bypass exits, and ensuring the rapid finding of feasible bypass directions in complex power operation environments (such as multi-tower, corridor-type obstacles). The standardized multi-exit detection process ensures that the UAV always explores along the safe edge of the obstacle, avoiding collisions, through virtual boundary attachment flight. On the other hand, the preset distance interval detection balances the comprehensiveness of detection with algorithm efficiency (neither missing the optimal exit nor increasing the computational load due to excessive detection). Finally, through the ray collision-free screening condition, it ensures that all potential detour exits are feasible to fly from that point to the target location without obstacles, providing sufficient and safe candidate samples for subsequent joint cost optimization.

[0055] Step 103: Based on each detour exit, construct candidate detour routes to avoid obstacles, and candidate flight routes corresponding to the candidate detour routes to fly from the detour exit to the target position of the UAV.

[0056] In some embodiments, the candidate detour route is a path scheme specifically designed to safely bypass the current obstacle, starting from the UAV's initial position and ending at a detour exit. It is a feasible obstacle bypass path generated through multi-exit guided obstacle bypass analysis and must meet safe flight constraints (no collision, conforming to UAV dynamic parameters). The candidate detour route provides multiple obstacle bypass path options to avoid the problems of excessively long distances and high risks that may exist with a single obstacle bypass path. It provides sufficient obstacle bypass path samples for joint cost optimization, ensuring the safety and flexibility of the obstacle bypass process.

[0057] In some embodiments, a candidate flight path is a collision-free flight path scheme that starts at a detour exit and ends at the UAV target location, meeting safety constraints. It is a connecting path that bypasses obstacles and directly reaches the target location. The candidate flight path connects the detour exit and the target location, completing a complete closed loop of path bypassing obstacles and directly reaching the target. Each candidate flight path, combined with its corresponding candidate detour path, constitutes a complete set of feasible route schemes from the starting point to the target, providing a complete sample for joint cost optimization.

[0058] Step 104: For candidate detour routes and candidate flight routes, with the objectives of minimizing path length, flight time, and risk coefficient of approaching obstacles, perform joint cost optimization to obtain the optimal route.

[0059] As one possible implementation, step 104 can be specifically implemented as steps A21-A24.

[0060] A21: For each candidate detour route, the detour cost of each candidate detour route is calculated based on the path length, estimated flight time, and obstacle proximity risk coefficient.

[0061] A22: For each candidate flight route, the flight cost is calculated based on the path length, estimated flight time, and obstacle proximity risk coefficient.

[0062] A23: For each candidate detour route and the corresponding candidate flight route, the joint cost is calculated based on the detour cost and the flight cost.

[0063] A24: Sort the combined costs of all candidate detour routes and candidate flight routes, and select the combination with the lowest combined cost as the optimal route.

[0064] In some embodiments, path length is the total geometric trajectory length of a flight path (candidate detour path or candidate flight path), and is a core indicator for measuring the distance cost of a flight path. As a fundamental parameter for cost calculation, it directly affects the energy consumption and flight efficiency of a flight path. Shorter paths generally have lower energy consumption and faster arrival speeds, making it a key consideration in joint cost optimization.

[0065] In some embodiments, the estimated flight time is the total time the UAV takes to fly along the flight path, estimated based on the UAV's dynamics model (such as maximum flight speed and acceleration constraints) and the path length and curvature. It is a core indicator for measuring the time cost of the flight path. The shorter the estimated flight time, the faster the UAV can reach the target location to perform the inspection task, making it a key parameter for balancing efficiency in joint cost optimization.

[0066] In some embodiments, the obstacle proximity risk coefficient is a numerical indicator that quantifies the risk of a flight path being too close to an obstacle (the closer to the obstacle, the higher the risk coefficient; if there is no close contact, the risk coefficient approaches 0), and it is a core indicator for measuring the safety cost of a flight path. Ensuring flight safety and avoiding collision risks caused by flight paths being too close to obstacles is a key parameter in joint cost optimization that balances safety, ensuring that the ultimately selected flight path meets safety requirements while being highly efficient.

[0067] In some embodiments, the detour cost is a comprehensive evaluation value obtained by weighting the distance cost, time cost, and safety cost of the candidate detour route. Transforming the multi-dimensional performance (distance, time, and safety) of the candidate detour route into a single quantitative indicator facilitates a direct comparison of the advantages and disadvantages of different obstacle avoidance paths, providing a unified evaluation standard for obstacle avoidance segments in the joint cost calculation.

[0068] In some embodiments, the flight cost is a comprehensive evaluation value obtained by weighting the distance cost, time cost, and safety cost of the candidate flight route. Quantifying the overall performance of the candidate flight route and adding it to the detour cost forms the joint cost of the complete route, ensuring a comprehensive evaluation of the entire process of obstacle avoidance and direct access.

[0069] In some embodiments, the joint cost is a comprehensive evaluation value of a combination of candidate detour routes and candidate flight routes, equal to the sum of the detour cost and flight cost in that combination. It is a core indicator for measuring the overall merits of a complete route. Transforming the performance of multiple complete route options into a single, directly comparable value provides a clear basis for global optimization, ensuring that the final selected route balances obstacle avoidance safety, flight efficiency, and overall feasibility. This allows for direct comparison of performance across different dimensions, avoiding confusion in the optimization logic caused by differing indicator dimensions.

[0070] In some embodiments, the optimal route is the complete path solution with the minimum joint cost among all candidate detour routes and combinations of candidate flight routes. It is the globally optimal solution that balances shortest path, least time, and lowest risk. As the basis for the UAV's final flight mission, it directly serves the core objectives of this invention: improving inspection efficiency and ensuring flight safety, ensuring that the UAV reaches the target location in the optimal way to perform inspections. Through joint cost calculation, the obstacle avoidance process and subsequent flight process are evaluated as a whole, ensuring that the finally selected route achieves a comprehensive balance of distance, time, and risk throughout the entire process.

[0071] As one possible implementation, the embodiments of the present invention can solve the key problem of how to find a balance between safety, efficiency and distance by integrating multi-dimensional quantitative evaluation and full-process cost, ensuring that the drone route not only meets the safety flight constraints, but also completes the supervision task with optimal efficiency, while realizing the automation and standardization of route selection.

[0072] Step 105: Control the drone to fly along the optimal flight path to the drone target location and conduct on-site inspection of the operation to be inspected.

[0073] As one possible implementation, step 105 can be specifically implemented as steps A31-A33.

[0074] A31: Control the drone to start from its initial position and fly along the optimal route.

[0075] A32: If an unknown obstacle is detected encroaching on the preset safe airspace, the path to the drone's target location will be replanned based on the current location, and the drone will be controlled to continue flying.

[0076] A33: After the drone arrives at the target location, it hovers and locks onto the position to conduct inspections of the work site to be inspected.

[0077] In some embodiments, unknown obstacles are obstacles that are not pre-stored in the background airspace map (an enhanced version of a single geographic information map for power grids) and suddenly appear during the drone's flight (such as temporary construction cranes, birds, suddenly moving equipment, etc.). These are dynamic or temporary obstacles that exceed the initial flight path planning expectations. Unknown obstacles are the key triggering condition for the dynamic obstacle avoidance and path replanning mechanism, demonstrating the invention's adaptability to complex and changing operating environments and preventing flight accidents or mission interruptions caused by unforeseen obstacles.

[0078] In some embodiments, a preset safe airspace is a specific spatial range (including horizontal safe distance and vertical safe altitude) defined around the optimal flight path. It serves as a safe buffer zone for UAV flight and must meet the requirements of not conflicting with known obstacles and complying with power operation safety regulations. The preset safe airspace acts as a detection benchmark for unknown obstacles, clearly defining the safe flight boundary of the UAV. When an unknown obstacle intrudes into this airspace, obstacle avoidance actions are immediately triggered to ensure that the UAV always flies within a safe range, avoiding the risk of collision.

[0079] In some embodiments, the current location is the real-time spatial coordinates (including latitude, longitude, and altitude) of the UAV when it detects an unknown obstacle intruding into a preset safe airspace, completes emergency avoidance, and hovers. It serves as the new starting point for resuming flight after a break. The current location acts as the starting reference for replanning the flight path, replacing the original initial location. This ensures that path replanning can be based on the UAV's current actual location, enabling resuming flight after a break rather than returning to the original flight path, thus improving mission execution efficiency and continuity.

[0080] In some embodiments, hovering and locking the position refers to the state in which the UAV, after arriving at the target location, maintains a stable spatial position (latitude, longitude, and altitude remain unchanged) and locks its current attitude (gimbal angle and flight direction are fixed) through its own flight control system. This provides a stable observation platform for subsequent automatic gimbal alignment with the center of the work area and dynamic target tracking, avoiding blurred images and target loss due to UAV position drift, and ensuring the stability and standardization of supervision and evidence collection.

[0081] In some embodiments, the work site to be inspected is a power operation area (such as a tower construction area or a line maintenance site) that includes inspected personnel, construction machinery, and power transmission line facilities, and is the core object of the drone inspection task. The work site to be inspected clearly defines the inspection scope and target of the drone, and all flight, aiming, and tracking actions are carried out around this site to ensure that the drone can accurately cover the inspected object after reaching the target location and complete the safety supervision and evidence collection tasks.

[0082] As one possible implementation, embodiments of the present invention can ensure that drones can safely, efficiently, and continuously complete inspection tasks in complex power operation environments through standardized processes such as optimal route execution, handling of unknown obstacles, resuming flight after interruption, and arrival for inspection. This directly supports the core objectives of the present invention: full-process automation and strong system robustness.

[0083] This invention first generates multiple safe detour exits based on obstacle locations and geometric parameters, ensuring the detour path closely follows the obstacle, resulting in a compact detour and reducing ineffective detour distances. Then, it analyzes the candidate detour routes constructed based on these exits, as well as candidate flight routes from these exits to the UAV's target location. Finally, for each candidate detour route and candidate flight route, it performs joint cost optimization based on path length, flight time, and obstacle proximity risk coefficient, selecting the optimal route with the minimum joint cost. This shortens the total distance the UAV travels to the target location, reduces travel time, and ensures controllable flight safety, thereby improving the efficiency of UAVs reaching the target location and enhancing the overall efficiency of UAV supervision.

[0084] As one possible implementation, steps A41-A44 can be performed before step A24.

[0085] A41: For the combination with the minimum joint cost, determine the splicing node based on the endpoint of the candidate detour route and the starting point of the candidate flight route.

[0086] A42: By using the Bézier curve interpolation algorithm, the end trajectory segment of the candidate detour route and the beginning trajectory segment of the candidate flight route at the splicing node are smoothly transitioned to obtain a continuous transition trajectory segment.

[0087] A43: The transition trajectory segment is merged with the candidate detour route and the candidate flight route to obtain the merged trajectory.

[0088] A44: Based on the fused trajectory, the optimal route is obtained.

[0089] In some embodiments, the combination with the minimum joint cost is the combination of routes with the smallest comprehensive evaluation value obtained through joint cost calculation among all combinations of candidate detour routes and candidate flight routes. It is the route scheme with the best overall distance, time, and risk in the initial screening. The combination with the minimum joint cost serves as the core object of subsequent trajectory smoothing processing, clarifying the target route combination that needs to be optimized, and ensuring that the final output optimal route has both a global cost advantage and solves the problem of smooth trajectory connection.

[0090] In some embodiments, the endpoint of a candidate detour route is the spatial coordinate point at the end of a candidate detour route, i.e., the selected detour exit, which is the natural connection point between the obstacle bypass segment and the subsequent direct target segment. The endpoint of the candidate detour route serves as the preceding connection point for trajectory splicing, clearly defining the end position of the obstacle bypass segment and providing reference coordinates for subsequent calculation of splicing nodes and achieving a seamless transition between the two route segments.

[0091] In some embodiments, the starting point of a candidate flight path is the initial spatial coordinate point of a candidate flight path, which shares the same coordinate point as the ending point of the corresponding candidate detour path, representing the starting position of the direct route to the target. The starting point of the candidate flight path serves as the final connection point for trajectory splicing, jointly defining the splicing node with the ending point of the candidate detour path. This ensures precise connection between the two flight paths, providing a clear start and end range for a smooth transition.

[0092] In some embodiments, the splicing node is the spatial coordinate point where the end point of the candidate detour route coincides with the starting point of the candidate flight route. It is the core node connecting the two route segments and a key area for trajectory smoothing. The splicing node clarifies the specific location for trajectory optimization, taking the connection points that might otherwise have bends as the optimization object, and using a smoothing algorithm to eliminate abrupt changes in motion between the two route segments, ensuring the continuity of the UAV's flight attitude.

[0093] In some embodiments, the Bézier curve interpolation algorithm is a mathematical algorithm for generating smooth curves. By controlling the shape of the curve defined by the vertices, it can generate continuous, smooth transition curves that conform to motion constraints between two known trajectory segments. It is commonly used for trajectory smoothing optimization in path planning. The Bézier curve interpolation algorithm is a core trajectory optimization tool used to eliminate hard angles at the junction of candidate detour routes and candidate flight routes, generating smooth transition trajectories that conform to the dynamic characteristics of UAVs (such as minimum turning radius), and avoiding attitude instability or flight risks caused by sudden trajectory changes in the UAV.

[0094] In some embodiments, the final trajectory segment of the candidate detour route is the last trajectory segment (usually a short-distance trajectory near the endpoint) in the candidate detour route, and is a key part that connects with the subsequent flight route.

[0095] In some embodiments, the initial trajectory segment of the candidate flight path is the first segment of the trajectory (usually a short-distance trajectory near the starting point) in the candidate flight path, which is the key part that connects to the obstacle bypass segment.

[0096] In some embodiments, the fused trajectory is a complete trajectory that integrates continuous transitional trajectory segments with the original candidate detour routes and candidate flight routes. It is the final trajectory scheme after cost optimization and smoothing optimization. It has the dual advantages of optimal global cost and controllable trajectory smoothness. It retains the core advantage of minimum joint cost and solves the technical defects of trajectory connection. It is the direct source of the final optimal route.

[0097] As one possible implementation, the embodiments of the present invention can optimize the trajectory of the route combination with the minimum joint cost, solve the problem of hard connection between two route segments, and achieve continuous control of the route through smooth transition processing, ensuring the safety, stability and standardization of UAV flight and supervision and evidence collection. It is a key technology supplement to the globally optimal route.

[0098] As one possible implementation, steps A51-A54 can be performed before step 101.

[0099] A51: Based on the location information and basic parameters reported in real time by the surveillance camera, the geographical coordinates of the core operation point corresponding to the central area of ​​the surveillance camera screen are calculated.

[0100] A52: Calculate the observation radius of the control sphere using the geographical coordinates of the core operation point as the spatial reference origin.

[0101] A53: The observation radius of the UAV is calculated based on the observation radius of the control sphere and the preset safety offset.

[0102] A54: Construct a collaborative observation ring with the geographic coordinates of the core operation point as the center and the observation radius of the UAV as the radius. The point on the ring corresponding to the shortest path from the initial position of the UAV to this collaborative observation ring is determined as the target position of the UAV.

[0103] In some embodiments, the PTZ (Planet Sphere) is a ground-based intelligent monitoring device deployed at the power operation site. It possesses location positioning, image acquisition, and parameter reporting functions, and serves as the core carrier for the dynamic spatial anchor point and task trigger in this invention. As the core data source for UAV collaborative supervision, the PTZ provides real-time location information and relevant parameters from the operation site, providing a basis for UAV target position calculation and flight path planning. It is a key device for realizing intelligent collaboration between UAVs and the PTZ.

[0104] In some embodiments, the real-time location information reported by the drone is spatial coordinate data (including latitude, longitude, and installation height) collected and uploaded to the remote control system in real time by the drone's own positioning module (such as GPS / BeiDou). This real-time location information serves as core data for dynamic spatial anchor points, providing a positional reference for calculating the geographic coordinates of key operational points and deriving the drone's observation radius. This ensures that the drone's flight path planning always revolves around the operational site, avoiding deviation from the inspection target.

[0105] In some embodiments, the basic parameters are the hardware performance and operational status parameters of the control sphere, including sensor size, lens focal length, lens height, gimbal pitch angle, and horizontal orientation angle. These basic parameters support the inversion calculation of the geographic coordinates of the core operational points. By combining these parameters with the pixel coordinates of the real-time images from the control sphere, the true geographic coordinates of the core operational area can be accurately derived, providing data support for the planning of UAV observation points.

[0106] In some embodiments, the central region of the PTZ camera image is the image range corresponding to the core area of ​​the work site (such as personnel gathering points or construction machinery locations) in the real-time images acquired by the PTZ camera. It serves as an intermediate carrier for the transformation from image pixel coordinates to real geographic coordinates. The central region of the PTZ camera image, as the image reference range for the core work point, is mapped to real-world geographic coordinates through perspective geometric backprojection calculations. This achieves the correlation between image information and spatial location, providing target anchor points for precise UAV observation.

[0107] In some embodiments, the geographic coordinates of the core operation point are real-world latitude and longitude coordinates calculated by perspective geometric back-projection using the pixel coordinates of the central area of ​​the PTZ camera image, combined with the PTZ camera's internal and external parameters (focal length, altitude, and pitch angle). This serves as the common origin for all subsequent spatial calculations. The geographic coordinates of the core operation point act as the spatial reference benchmark for the entire collaborative observation system, unifying the observation coordinate systems of the PTZ camera and the UAV, ensuring complementary perspectives and blind-spot-free monitoring, and are the core coordinate points for achieving collaborative intelligence.

[0108] In some embodiments, the spatial reference origin is a benchmark point used for unified spatial calculations. In this invention, it is the geographic coordinate of the core operation point, and all calculations related to the observation range and UAV position are carried out around this point. The spatial reference origin eliminates the positional deviation between the control sphere and the UAV, providing a unified benchmark for calculating the observation radius of the control sphere and the UAV, ensuring accurate connection between their observation ranges, and avoiding omissions or overlaps in supervision.

[0109] In some embodiments, the observation radius of the control sphere is the planar projection distance from the control sphere's installation point to the core operation point, centered on the core operation point. The observation radius of the control sphere defines its effective observation range, provides a basis for setting the observation radius of the UAV, and ensures that the UAV's observation range complements that of the control sphere, covering areas that the control sphere cannot reach.

[0110] In some embodiments, the preset safety offset is a fixed distance set to ensure the flight safety of the UAV and avoid conflict with the observation range of the surveillance sphere. It is the incremental value of the UAV's observation radius relative to the observation radius of the surveillance sphere. The preset safety offset ensures that the UAV and the surveillance sphere maintain a safe cooperative distance, avoiding flight conflicts while allowing the UAV to obtain a wider observation angle, achieving automatic perspective complementarity, and supporting the invention goal of eliminating blind spots in supervision.

[0111] In some embodiments, the UAV observation radius is the radius of the circle containing the UAV target hovering point, centered on the core operation point. The UAV observation radius constrains the UAV's observation range, ensuring that the UAV's hovering position covers the core operation area while maintaining a safe distance from the control sphere; it is a key parameter for UAV target location selection.

[0112] In some embodiments, the collaborative observation ring is a circular trajectory centered on the core operation point and with the UAV's observation radius as its radius. It represents the selectable range of UAV target hovering points. The collaborative observation ring provides candidate areas for UAV hovering points, ensuring that all potential UAV observation positions revolve around the core operation point and form a collaborative observation relationship with the control sphere, thus providing a target range for subsequent shortest path selection.

[0113] As one possible implementation, embodiments of the present invention can construct a spatial coordinate reference system for collaborative observation of the control ball and UAVs, automatically determine the optimal target position of the UAVs, and provide accurate destination basis for subsequent route planning and flight execution. This is a core prerequisite for achieving collaborative intelligence and full-process automation.

[0114] As one possible implementation, steps A61-A62 can be performed before step A33.

[0115] A61: Based on the target location of the UAV and the geographical coordinates of the core operation point to be inspected, the pitch angle and yaw angle of the core operation point falling into the center of the camera's field of view are calculated.

[0116] A62: Based on pitch and yaw angles, adjust the drone's camera attitude and match the lens focal length to conduct on-site inspections.

[0117] In some embodiments, the pitch angle is the vertical rotation angle of the UAV gimbal, calculated using the arctangent function from the height difference and horizontal distance between the UAV target position and the core operation point. The pitch angle controls the vertical tilt attitude of the UAV lens, ensuring that the lens optical axis is precisely pointed vertically to the core operation point, solving the problem of the core area being out of the field of view due to lens elevation deviation, and providing an angular basis for clear observation.

[0118] In some embodiments, the yaw angle is the rotation angle of the UAV gimbal in the horizontal direction, determined by the azimuth angle calculated from the projected coordinates of the UAV target position and the core operation point on the horizontal plane. The yaw angle controls the left and right turning attitude of the UAV lens, ensuring that the lens is accurately aligned with the core operation point in the horizontal direction, avoiding lens directional deviation that causes the core area to deviate from the frame, and achieving accurate positioning of the horizontal viewpoint.

[0119] In some embodiments, lens attitude refers to the spatial attitude state of the UAV camera lens, which is determined by both the pitch angle (vertical direction) and the yaw angle (horizontal direction), directly affecting the lens's observation direction and field of view coverage. By adjusting the lens attitude, the core point of the operation is accurately placed in the center of the lens's field of view, providing an initial alignment basis for subsequent dynamic target tracking and standardized evidence collection, and avoiding target offset or loss in the image due to attitude deviation.

[0120] In some embodiments, the lens focal length is an optical parameter of the drone camera lens, determining the lens's magnification and field of view (the longer the focal length, the greater the magnification and the narrower the field of view; conversely, the shorter the focal length, the wider the field of view). By matching the lens focal length, the imaging size of the core operational area in the image is adjusted to ensure that details in the core area are clearly discernible (such as personnel operating actions and mechanical operating status), avoiding situations where the image is too far away (blurred details) or too close (insufficient field of view) due to improper focal length, thus supporting the standardization of evidence collection.

[0121] In some embodiments, the work site to be inspected is a power operation area (such as a tower maintenance site or a line construction area) that includes inspected personnel, construction machinery, and power transmission line facilities, and is the core object of the UAV inspection mission. Clearly defining the final service goal of adjusting the lens attitude and focal length at the work site to be inspected ensures that all operations revolve around a clear and comprehensive inspection of the safety situation at the work site, providing a clear scene scope for subsequent target tracking and evidence collection.

[0122] As one possible implementation, the embodiments of the present invention can achieve precise alignment of the observation perspective and optimization of image quality after the UAV arrives at the target location. By quantitatively calculating the lens attitude parameters and matching the focal length, it lays the foundation for subsequent dynamic target tracking and standardized evidence collection, directly supporting the core objectives of the present invention of standardized evidence collection and full-process automation.

[0123] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0124] The above embodiments are in Figure 1 Based on the method shown, each step will be discussed in detail. To facilitate understanding of the complete execution process, the overall method flow will be discussed below with reference to an embodiment.

[0125] In this invention, the PTZ (Planet Sphere) uploads its real-time location information to the remote control system. Based on the drone's and PTZ's positions, the system creates the optimal flight path for the drone to reach the work site using "automatic spatial positioning guidance technology based on the PTZ's real-time coordinates," and then sends the flight path to the drone control tool. During the drone's journey to the work site, real-time images captured by the drone's camera and sensor data are synchronized to the remote control tool. Relying on the obstacle avoidance algorithm within the remote control tool, the drone is automatically guided to fly around obstacles. Upon arrival at the work site, the drone automatically aligns itself with the work area using "automatic adjustment of the drone's gimbal observation angle based on spatial coordinates." On-site supervisors select the target to be tracked via the remote control screen, activate the target tracking model within the remote control, and automatically track the target at the work site.

[0126] 1. Spatial Position Automatic Guidance Technology Based on Real-Time Coordinate Position of Controlled Ball The coordinates reported in real time by the operational surveillance sphere are used as a "dynamic spatial anchor point." The system can automatically identify the activation status of this anchor point and use it as a mission trigger condition. Without manual input of coordinates, the system can automatically use this anchor point as the center, combined with safe flight rules (such as no-fly zones and altitude-restricted zones) and the on-site environment (such as preset safe approach angles and distances), to intelligently calculate and generate an optimal temporary guidance route, including one or more approach waypoints.

[0127] 1.1 Static safety planning route generation The core of the system's solution is to find the path with the shortest flight time while ensuring absolute safety. The specific solution process is as follows: 1.1.1 Establish a safe flight strategy Instead of flying directly towards the control sphere itself (which may be at a very low altitude), the system calculates an "aerial approach point" located at a safe distance (20m away) from the control sphere and at a sufficiently safe altitude (e.g., 100m) above the ground as the flight destination, thus avoiding long-distance flight of the drone in complex low-altitude environments.

[0128] 1.1.2 Method for setting the flight destination The system can acquire basic parameters of the surveillance sphere, either built-in or remotely, including: sensor size, lens focal length (f), lens height, gimbal pitch angle (θ_ball), and horizontal orientation angle (α_ball). Based on these parameters and real-time image analysis, the system first calculates the central region of the current image from the surveillance sphere.

[0129] 1.1.3 Calculation of UAV observation points (1) The system determines the core operation area (such as the gathering point of construction machinery or the work point of personnel) in the real-time image of the control ball, and calculates the geographic coordinates of the latitude and longitude coordinates (P_focus) of the core operation point in the real world by combining the pixel coordinates of this area in the image with the internal and external parameters (focal length, height, pitch angle) of the control ball through perspective geometric back projection calculation. This point is the spatial origin for subsequent calculations. The inverted core operation point (P_focus) is used as the common spatial reference origin. Control ball observation ring: Connect P_focus with the latitude and longitude coordinates (P_ball) of the control ball installation point, and calculate the distance R_ball=distance(P_focus, P_ball) (the distance between the two points projected on the plane, not the three-dimensional spatial distance). This distance is the observation radius of the control ball for the core operation point. The effective field of view of the control ball can be modeled as a fan-shaped ring with P_focus as the center and R_ball as the radius.

[0130] (2) The latitude and longitude (P_uav) of the UAV's target hovering point is constrained to another concentric circle centered on P_focus, but its observation radius R_uav must be greater than R_ball, and a fixed safety offset ΔR must be maintained (e.g., ΔR = 20 meters), i.e., R_uav = R_ball + ΔR. Calculate the three-dimensional spatial distance from the UAV's current takeoff point or real-time position (P_start) to each candidate point on this spatial ring. The system will automatically select the candidate point with the shortest distance to P_start as the final target hovering point (P_uav). This optimization ensures that the UAV can reach the cooperative observation position with the shortest flight path, the least time, and the least energy consumption.

[0131] 1.1.4 Static Safety Route Planning The background operating system includes a dynamically updated "airspace map," which is an enhanced version of a dedicated geographic information map for the power grid. It not only includes general no-fly zones and height-restricted zones, but also deeply integrates power grid asset data, such as: precise 3D models of transmission line corridors, tower coordinates, conductor sag envelopes, and substation restricted areas, forming a high-precision 3D geofencing for power facility obstacles.

[0132] After the drone observation point calculates and generates the initial flight path, it will compare it with the "airspace map" in the background. If there are no known no-fly zones (such as airports, military zones), height-restricted zones (such as city centers), or fixed obstacles (such as high-voltage power line corridors, mountains), the background will automatically send the flight path to the drone remote controller.

[0133] If the system traverses these areas in a straight line, it will automatically invoke a fast local path planner to generate N feasible obstacle avoidance detours. For each candidate detour and its exit, the algorithm performs a global task cost evaluation. For each exit point, the algorithm quickly simulates and calculates the remaining optimal path from that point to the collaborative observation ring. The final selected exit point minimizes the sum of the detour cost from the starting point to the exit point plus the remaining flight cost from the exit point to the observation ring, and the corresponding path is generated. The specific implementation is as follows: 1.1.4.1 Local Path Planner The local path planner in this invention aims to quickly generate multiple safe, smooth, and friable alternative flight segments that bypass sudden or known obstacles. It employs a "multi-exit guided improved Bug algorithm," which is lightweight, real-time, and particularly suitable for running on edge computing devices.

[0134] ①Detailed step-by-step explanation of the algorithm Step 1: Obstacle Modeling and Tangent Direction Calculation When the planner detects an intersection between the flight path and an obstacle (represented as a polygon or a 3D envelope in the "Power Grid Map"), it projects the obstacle in two dimensions at the drone's current flight altitude.

[0135] The algorithm calculates the two vertices V_left and V_right on the obstacle projection polygon that are closest to the current drone position P_start.

[0136] Draw rays from P_start to V_left and V_right respectively. These two rays define two initial tangential directions for circumventing the obstacle (circumventing to the left and to the right).

[0137] Step 2: Generate multiple candidate detour routes The planner doesn't offer two options; instead, it actively explores multiple possible exits along the boundary of obstacles.

[0138] Main directional detour: Along the aforementioned left and right tangent directions, the drone virtually "attaches" to the obstacle boundary (following the boundary following mode of the Bug algorithm). Every time it advances a certain distance (e.g., 5 meters), the algorithm attempts to calculate an escape ray that points directly from the current boundary point B_i to the original target direction (i.e., towards the P_focus direction).

[0139] Detecting exits: If the escape ray no longer intersects any obstacle, the current boundary point B_i is considered a potential exit point E_candidate. The algorithm records the path Path_boundary from P_start along the boundary to E_candidate.

[0140] Repeated detection: During the boundary following process, the algorithm will detect multiple such E_candidates, thus forming a set of candidate exits and corresponding boundary detour paths.

[0141] Step 3: Path Smoothing and Preliminary Cost Assessment For each original polyline path Path_boundary obtained by following the boundary, smoothing is performed using Bézier curves or spline interpolation to generate a continuous curve Path_smooth_i that meets the minimum turning radius constraint of the UAV.

[0142] Simultaneously, a preliminary cost estimate C_rough(Path_smooth_i) for each smooth path is quickly calculated, primarily considering path length and the degree of curvature in turns (curvature integral). This value can serve as a reference for higher-level optimization or as a pre-filtering condition.

[0143] Step 4: Output the set of alternative flight segments Finally, the local planner outputs a set containing N alternative detour routes: text { Path_detour_1: {Sequence of trajectory points, exit point E_1, initial cost C_rough_1}, Path_detour_2: {Sequence of trajectory points, exit point E_2, initial cost C_rough_2}, ... Path_detour_N: {Sequence of trajectory points, exit point E_N, initial cost C_rough_N} } This set is provided to the upper-level global optimization algorithm to compute the complete joint cost J_i.

[0144] 1.1.4.2 Detour Path Generation Method Step 1: Generate a set of candidate detour routes After detecting an obstacle, the system invokes a fast local path planner to generate N feasible obstacle avoidance detour segments. Each segment, Path_detour_i (i=1,2,...,N), satisfies the following conditions: it starts at the UAV's current planned starting point P_start; it safely bypasses the currently conflicting obstacle; and it terminates at an obstacle-free segment exit position E_i. These exit positions {E_i} constitute the candidate point set for subsequent decisions.

[0145] Step 2: Define and calculate the joint cost function For each candidate detour segment Path_detour_i and its exit E_i, the algorithm performs a global task cost evaluation and calculates its joint cost J_i: J_i=C_detour(Path_detour_i)+C_remain(E_i,C_ring) In the formula: C_detour(Path_detour_i) is the cost of executing the i-th detour segment itself. Its calculation comprehensively considers: C_detour = ω_L*L_i + ω_T*T_i + ω_R*R_i, where L_i is the geometric length of the detour segment, T_i is the flight time estimated based on the UAV dynamics model, and R_i is the risk coefficient of the segment approaching obstacles (the closer the distance, the higher the risk value). ω_L, ω_T, and ω_R are normalized weighting coefficients.

[0146] C_remain(E_i,C_ring) is the remaining optimal path cost from exit E_i to the collaborative observation ring C_ring.

[0147] Remaining path search: Taking E_i as the new starting point and the entire circular ring C_ring as the target area, the system replans a collision-free remaining optimal path Path_remain_i under the constraint of "one power grid map". The endpoint of this path is an optimal access point P_ring_i∈C_ring on the circular ring.

[0148] Cost calculation: C_remain is the estimated cost of path Path_remain_i (the calculation method is the same as C_detour, but can be simplified to mainly consider the length L_remain_i).

[0149] Step 3: Global Optimization and Path Combination The system compares the joint costs of all candidate solutions: The optimal solution index k = argmin{J_i|i=1 to N} That is, select the scheme k that minimizes the joint cost J_i.

[0150] The final generated complete safe flight path Path_final is composed of two seamlessly joined parts: Path_final=Path_detour_k⊕Path_remain_k Meanwhile, the final precise hovering observation point of the drone was determined to be: P_uav_actual=P_ring_k 1.1.5 Output Route After the above safety filtering and adjustment, the system finally outputs a collision-free route that complies with regulations and has the shortest possible total flight time. This is the "optimal temporary guidance route" defined in this invention, and it is then sent to the UAV.

[0151] 1.2 Dynamic Safe Obstacle Avoidance Path Planning Building upon initial guidance based on the coordinates of the control sphere, a safety mechanism is added during flight. When the drone flies along the planned route, and its built-in obstacle avoidance module suddenly detects an unforeseen obstacle ahead (such as a temporary crane or birds), it automatically performs obstacle avoidance. 1.2.1 Automatic Emergency Avoidance The drone flight control immediately took over, controlling the drone to apply emergency braking, and then automatically ascending to avoid obstacles, prioritizing safety.

[0152] 1.2.2 Interruption Reporting and New Starting Point Confirmation After completing the obstacle avoidance maneuver, the drone hovers at the new location and immediately reports "flight interruption" and its new precise location coordinates to the ground control system.

[0153] 1.2.3 Resume Flight After Disconnection After receiving the report, the ground control system will not order the drone to return to the original flight path. Instead, it will take the drone's current hovering position as the new starting point and the original "aerial approach point" as the end point, and re-execute the above "intelligent calculation" process in real time to generate a brand new "continued flight path" from the current point to the target point.

[0154] After receiving a new "flight route," the drone automatically continues flying along the new route. This process can occur multiple times until the drone safely reaches the target location. This ensures high robustness and a high completion rate of the flight mission when facing unknown obstacles.

[0155] 2. Automatic adjustment of the observation angle of the UAV gimbal based on spatial location coordinates Automatic gimbal alignment and interactive target tracking: ① Automatic alignment: After the drone reaches the optimal observation point, the system automatically calculates the gimbal yaw and pitch angles based on the spatial relationship between the drone, the control sphere, and the center of the work area, ensuring the lens optical axis is precisely aligned with the center of the work area. ② Interactive tracking: An AI recognition frame is overlaid on the remote control screen while the gimbal remains stationary. When the user selects the target frame, the system activates tracking mode, dynamically controlling the gimbal rotation and camera zoom to keep the target always centered in the image and maintaining its image size at a preset ratio (e.g., ≥1 / 3 of the screen height), achieving standardized evidence collection. The specific execution process is as follows: 2.1 Automatic alignment with the center of the work surface The gimbal yaw and pitch angles are calculated based on the geometric relationship between the control sphere, the operation position center, and the UAV space. The system first performs initial alignment. Pitch angle (θ_gimbal) calculation: Based on the altitude difference (ΔH) and horizontal distance (D_horizontal) between P_uav and P_focus, the required pitch angle of the gimbal is calculated in real time using the arctangent function: θ_gimbal = arctan(ΔH / D_horizontal). This ensures that the camera optical axis accurately points in the vertical direction of P_focus. Yaw angle (ψ_gimbal) calculation: Based on the projected coordinates of P_uav and P_focus on the horizontal plane, the azimuth angle is calculated as the basis for the gimbal yaw angle, ensuring that the lens is horizontally aligned with the target.

[0156] 2.2 Dynamic Target Tracking After initial alignment, the system enters dynamic tracking mode. Dynamic tracking is achieved through the combined use of target detection and target tracking models in the UAV remote controller.

[0157] 2.2.1. Dual-objective closed-loop control strategy The electronic forensics standard is quantified into a control loop setpoint S_set, which coordinates with the target location control loop, fundamentally ensuring the standardized and high-quality output of forensic footage. After locking onto the target, the system simultaneously activates two parallel closed-loop control loops, both with setpoints derived from the forensics standard. Loop 1: Position control loop (PTZ control) Control target: Make the pixel offset (Δx,Δy) between the geometric center (Cx_t,Cy_t) of the target recognition box and the center (Cx_img,Cy_img) of the camera image approach zero.

[0158] Control law: A proportional-integral (PI) controller is used. The control commands are the gimbal's yaw rate ω_yaw and pitch rate ω_pitch.

[0159] ω_yaw=Kp_x*Δx+Ki_x*∫Δxdt ω_pitch = Kp_y * Δy + Ki_y * ∫Δy dt Smoothing process: Apply acceleration limit and low-pass filtering to the output angular velocity command to ensure the "gentle and smooth" movement of the pan-tilt head, and avoid sudden changes in the image.

[0160] Loop 2: Scale control loop (zoom control) Control objective: Make the ratio S = H_box / H_img of the height H_box of the target recognition frame to the total height H_img of the image stably maintain at the preset optimal evidence-taking ratio S_set (for example, S_set = 1 / 3).

[0161] Control law: Also use a PI controller. The control command is the zoom speed v_zoom of the camera (or directly the zoom magnification increment).

[0162] v_zoom = Kp_s * (S_set - S) + Ki_s * ∫(S_set - S) dt Direction logic: When S < S_set (the target is too small), the command is to zoom in; when S > S_set (the target is too large), the command is to zoom out.

[0163] 2.2.2. Cooperative coupling and anti-saturation strategy The cooperative control strategy for the two strongly coupled actuators of the pan-tilt head and zoom is clearly designed, and an intelligent mode switching strategy in extreme cases such as zoom saturation is proposed, ensuring the robustness of the system.

[0164] ① Compensation for the influence of zoom on the position loop: The zoom operation changes the camera's field of view angle, thus affecting the pixel displacement speed of the target in the image. When the system calculates the error (Δx, Δy) of the position loop, it will estimate and compensate for the natural drift of the target pixel position caused by the current zoom speed v_zoom to prevent interference and oscillation between the two loops.

[0165] ② Control command priority and anti-saturation: When the target approaches or moves away from the UAV rapidly, it may cause the zoom to reach the optical limit (such as the maximum magnification) and unable to maintain the scale S_set. At this time, the output of the scale loop saturates. The system will trigger a mode switch, temporarily switching the control objective of the position loop from "aligning the center of the frame" to "ensuring that the target does not leave the image" to prioritize ensuring tracking without loss. After the target movement becomes stable, the dual-loop cooperative control is restored.

[0166] 2.2.3 Feedforward control based on target motion prediction Based on the traditional feedback control, feedforward control based on target motion estimation is introduced, significantly improving the tracking smoothness and stability for fast-moving targets (such as walking personnel, rotating machinery) at the operation site.

[0167] ① Motion state estimation: Using a Kalman filter, the image plane motion velocity (Vx_t, Vy_t) and acceleration of the target are estimated in real time based on the historical position sequence of the target recognition box.

[0168] ② Feedforward command generation: The estimated motion velocity is fed forward to the gimbal control loop, providing additional angular velocity commands (ω_yaw_ff, ω_pitch_ff), enabling the gimbal to respond to target motion in advance and reduce tracking lag. ω_yaw += Kff * Vx_t, ω_pitch += Kff * Vy_t.

[0169] This invention links discrete "inspection points" with specific "inspection targets," forming a complete autonomous task chain. The system pre-sets not only spatial coordinates (points) but also an intelligent task sequence that includes "what identification and tracking tasks are expected to be performed at each point." When the inspection task at one point is completed (e.g., operator confirmation of completion or reaching the preset tracking duration), the drone can automatically fly to the next preset point without human intervention, activating the corresponding lightweight AI recognition model and preparing to execute a new round of the "identification-confirmation-lock-tracking" process.

[0170] This invention features full-process automation, from task triggering, route planning, flight alignment to target tracking, greatly reducing human intervention and reliance on pilot experience. It boasts intelligent collaboration, pioneering a geometric model-based UAV-control sphere collaborative observation method that achieves automatic perspective complementarity and eliminates blind spots in supervision. Its innovative planning mechanism proposes a "task-fidelity-preserving intelligent detour" algorithm, strictly ensuring the achievement of the ultimate mission objective while avoiding obstacles, achieving globally optimal paths. Standardized evidence collection ensures that the collected video evidence meets the requirements of safety supervision through interactive locking and automatic zoom tracking. The system is robust, combining static airspace planning, dynamic real-time obstacle avoidance, and breakpoint resuming flight mechanisms to adapt to complex field power grid operation environments.

[0171] The following are device embodiments of the present invention. For details not described in detail, please refer to the corresponding method embodiments described above.

[0172] Figure 2 A schematic diagram of the structure of the drone inspection device provided in an embodiment of the present invention is shown. For ease of explanation, only the parts related to the embodiment of the present invention are shown, and are described in detail below: like Figure 2 As shown, the drone inspection device 2 includes: Communication module 21 is used to acquire associated data of the work site to be inspected. The associated data includes the target position of the UAV, the initial position of the UAV, and the position and geometric parameters of the obstacles. The processing module 22 is used to determine multiple detour exits that meet safe flight constraints based on the correlation data and through multi-exit guided obstacle detour analysis for each obstacle; based on each detour exit, it constructs candidate detour routes to avoid obstacles, as well as candidate flight routes from the detour exits to the target position of the UAV; for the candidate detour routes and candidate flight routes, it performs joint cost optimization with the objectives of minimizing path length, flight time, and risk coefficient of approaching obstacles, to obtain the optimal route; and controls the UAV to fly along the optimal route to the target position of the UAV for on-site supervision of the operation to be supervised.

[0173] Figure 3 This is a schematic diagram of an electronic device provided in an embodiment of the present invention. For example... Figure 3 As shown, the electronic device 3 of this embodiment includes a processor 30 and a memory 31. The memory 31 stores a computer program 32. When the processor 30 executes the computer program 32, it implements the steps in the various method embodiments described above. Alternatively, when the processor 30 executes the computer program 32, it implements the functions of each module / unit in the various device embodiments described above.

[0174] For example, computer program 32 may be divided into one or more modules / units, which are stored in memory 31 and executed by processor 30 to complete the present invention. The one or more modules / units may be a series of computer program instruction segments capable of performing a specific function, which describe the execution process of computer program 32 in electronic device 3.

[0175] Electronic device 3 may include, but is not limited to, processor 30 and memory 31. Those skilled in the art will understand that... Figure 3 This is merely an example of electronic device 3 and does not constitute a limitation on electronic device 3. It may include more or fewer components than shown, or combine certain components, or different components. For example, electronic device 3 may also include input / output devices, network access devices, buses, etc.

[0176] For the sake of simplicity and clarity, only the above-described functional modules / units are used as examples. In practical applications, the functions described above can be assigned to different functional modules / units as needed. These modules / units can be implemented in hardware, software, or a combination of both.

[0177] In the above embodiments, the descriptions of each embodiment have their own emphasis. Parts not detailed or described in a particular embodiment can be referred to in the relevant descriptions of other embodiments. Unless otherwise specified or in conflict with logic, the terminology and / or descriptions between different embodiments are consistent and can be referenced interchangeably. Technical features in different embodiments can be combined to form new embodiments based on their inherent logical relationships.

[0178] The above-described 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, and should all be included within the protection scope of the present invention.

Claims

1. A method for drone supervision, characterized in that, include: Acquire relevant data from the site to be inspected, including the target position of the UAV, the initial position of the UAV, and the position and geometric parameters of obstacles; For each obstacle, based on the aforementioned correlation data, multiple detour exits that meet safe flight constraints are determined through multi-exit guided obstacle detour analysis; Based on each detour exit, candidate detour routes to avoid obstacles are constructed, as well as candidate flight routes from the detour exits to the target location of the UAV. For the candidate detour route and the candidate flight route, the optimal route is obtained by joint cost optimization with the objectives of minimizing the path length, flight time and the risk coefficient of approaching obstacles. Control the drone to fly along the optimal flight path to the drone's target location for on-site inspection of the operation to be supervised.

2. The drone inspection method according to claim 1, characterized in that, For each obstacle, based on the associated data, multiple detour exits that meet safe flight constraints are determined through multi-exit guidance obstacle detour analysis, including: For each obstacle, a polygonal cross-section of the obstacle is obtained by projecting it onto the plane of the UAV's flight based on the obstacle's position and geometric parameters; Based on the initial position of the UAV and the two vertices in the polygonal cross section that are closest to the initial position of the UAV, the left and right directions of the UAV around the initial tangent are determined. The drone is controlled to virtually attach to the boundary of the obstacle and fly along the left and right directions around the initial tangent. For each preset distance advanced, potential detour exits are determined based on whether the ray formed by the current boundary point and the drone target position intersects with obstacles; Repeat the boundary flight and potential detour exit steps to obtain multiple detour exits that meet the safe flight constraints.

3. The drone inspection method according to claim 2, characterized in that, For each preset distance traveled, potential detour exits are determined based on whether the ray formed by the current boundary point and the UAV target position intersects with an obstacle, including: The virtual flight pauses after advancing a preset distance to record the coordinates of the current boundary point. Construct a ray with the current boundary point coordinates as the starting point and the UAV target position as the ending point; If the ray intersects any obstacle, it continues to virtually fly forward along the obstacle boundary, repeating the steps of advancing the preset distance and constructing the ray. If the ray does not intersect with any obstacles, the current boundary point is identified as the potential detour exit.

4. The drone inspection method according to claim 1, characterized in that, The optimal route is obtained by jointly optimizing the candidate detour route and the candidate flight route, with the objectives of minimizing path length, flight time, and risk coefficient of approaching obstacles. This includes: For each of the candidate detour routes, the detour cost of each candidate detour route is calculated based on the path length, estimated flight time and obstacle proximity risk coefficient. For each of the candidate flight routes, the flight cost is calculated based on the path length, estimated flight time, and obstacle proximity risk coefficient. For each candidate detour route and the corresponding candidate flight route, the joint cost is calculated based on the detour cost and the flight cost. The combined costs of all candidate detour routes and candidate flight route combinations are sorted, and the combination with the smallest combined cost is taken as the optimal route.

5. The drone inspection method according to claim 4, characterized in that, Before sorting the joint costs corresponding to all candidate detour routes and candidate flight route combinations, and selecting the combination with the minimum joint cost as the optimal route, the process also includes: For the combination with the minimum joint cost, the splicing node is determined based on the endpoint of the candidate detour route and the starting point of the candidate flight route; By using the Bézier curve interpolation algorithm, the end trajectory segment of the candidate detour route and the beginning trajectory segment of the candidate flight route at the splicing node are smoothly transitioned to obtain a continuous transition trajectory segment. The transition trajectory segment is fused with the candidate detour route and the candidate flight route to obtain the fused trajectory; The optimal route is obtained based on the fused trajectory.

6. The drone inspection method according to claim 1, characterized in that, Before obtaining the relevant data from the work site to be inspected, the following steps are also included: Based on the location information and basic parameters reported in real time by the surveillance sphere, the geographical coordinates of the core operation point corresponding to the central area of ​​the surveillance sphere screen are calculated. Using the geographic coordinates of the core operation point as the spatial reference origin, calculate the observation radius of the control sphere; The observation radius of the UAV is calculated based on the observation radius of the control sphere and the preset safety offset. A collaborative observation ring is constructed with the geographic coordinates of the core operation point as the center and the observation radius of the UAV as the radius. The point on the ring corresponding to the shortest path from the initial position of the UAV to the collaborative observation ring is determined as the target position of the UAV.

7. The drone inspection method according to claim 1, characterized in that, The control of the drone to fly along the optimal route to the target location, and the on-site supervision of the operation to be supervised, includes: Control the drone to start from its initial position and fly along the optimal flight path; If an unknown obstacle is detected encroaching on the preset safe airspace, the path to the drone's target location is replanned based on the current location, and the drone is controlled to continue flying. After the drone arrives at the target location, it hovers and locks onto the position to conduct an inspection of the site to be inspected.

8. The drone inspection method according to claim 7, characterized in that, After the drone arrives at the target location, it hovers and locks onto the position. Before conducting on-site inspections, the process also includes: Based on the target location of the drone and the geographical coordinates of the core operation point to be inspected, the pitch angle and yaw angle of the core operation point falling into the center of the camera's field of view are calculated. Based on the pitch angle and yaw angle, the camera attitude of the UAV is adjusted, and the lens focal length is matched to conduct inspections of the work site to be inspected.

9. A drone inspection device, characterized in that, include: The communication module is used to acquire relevant data of the work site to be inspected. The relevant data includes the target position of the UAV, the initial position of the UAV, and the position and geometric parameters of the obstacles. The processing module is used to determine multiple detour exits that meet safe flight constraints based on the associated data for each obstacle through multi-exit guide obstacle detour analysis; Based on each detour exit, candidate detour routes to avoid obstacles are constructed, as well as candidate flight routes from the detour exits to the target position of the UAV. For the candidate detour routes and the candidate flight routes, joint cost optimization is performed with the objectives of minimizing path length, flight time, and risk coefficient of approaching obstacles to obtain the optimal route. The UAV is controlled to fly along the optimal route to the target position of the UAV for on-site supervision of the operation to be supervised.

10. An electronic device, characterized in that, It includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the method as described in any one of claims 1 to 8.

Images