Unmanned aerial vehicle high-voltage line automatic patrol method and system

Abstract

The invention relates to an unmanned aerial vehicle high-voltage line automatic patrol method, device and equipment and a medium. The method comprises the following steps: acquiring multi-source data; the multi-source data comprises a tower position, point cloud data, visual data, I MU inertial navigation data and positioning data; fusing the multi-source data to obtain an environment map corresponding to the location of the unmanned aerial vehicle; based on the environment map, utilizing an improved RRT * algorithm to generate a safe waypoint sequence of the location of the unmanned aerial vehicle; based on the multi-source data, identifying objects in the location of the unmanned aerial vehicle, and determining an obstacle list; performing dynamic obstacle avoidance based on the safe waypoint sequence and the obstacle list, and generating a real-time flight path; the real-time flight path is used for indicating the unmanned aerial vehicle to execute an automatic line patrol task. The method can reduce the consumption of human resources, and guarantees the line patrol safety of the high-voltage line.

Description

Technical Field

[0001] The present invention belongs to the field of waterway control, and in particular relates to a method and system for automatically inspecting high-voltage lines using an unmanned aerial vehicle (UAV). Background Art

[0002] With the continuous advancement of high-voltage transmission technology, high-voltage line inspections have traditionally been conducted manually. Traditional manual inspections typically involve teams of two to three patrollers, armed with binoculars, infrared thermometers, and notebooks, patrolling along transmission corridors on foot or by car. During these inspections, personnel must visually and telescopically observe defects such as broken conductor strands and damaged insulators from the ground. They must also climb to heights of 30-50 meters to manually inspect key towers for corrosion, loose bolts, and other hardware. This process relies heavily on personnel experience, covering only 3-5 kilometers of line per day, and poses safety risks such as falls and electric shock. Summary of the Invention

[0003] Based on this, it is necessary to provide a method and system for automatically inspecting high-voltage lines using drones, which can reduce human resource consumption and ensure the safety of high-voltage line inspections, in response to the above technical problems.

[0004] In a first aspect, the present application provides a method for automatically inspecting high-voltage lines using a drone, comprising:

[0005] Acquire multi-source data; multi-source data includes tower position, point cloud data, visual data, IMU inertial navigation data and positioning data;

[0006] Fuse multi-source data to obtain an environmental map corresponding to the location of the drone;

[0007] Based on the environmental map, the improved RRT* algorithm is used to generate a safe waypoint sequence for the UAV's location;

[0008] Based on multi-source data, identify objects in the drone's location and determine the obstacle list;

[0009] Dynamic obstacle avoidance is performed based on the safe waypoint sequence and obstacle list to generate a real-time flight path; the real-time flight path is used to instruct the drone to perform automatic line patrol tasks.

[0010] Furthermore, based on the environmental map, the improved RRT* algorithm is used to generate a safe waypoint sequence for the UAV’s location, including:

[0011] Based on the environment map, the wire parameterization model is calculated using the following formula:

[0012]

[0013] P i =(x i ,yi ,z i +Δh i )

[0014]

[0015] Among them, P(t) is the conductor parameterized model, which dynamically describes the conductor shape through the cubic B-spline basis function and the sag correction formula, N i,3 (t) is the ith cubic B-spline basis function, P i is the i-th control point, t is the curve parameter, i is the control point index, Δh i is the height adjustment of the i-th control point due to the sag of the conductor, calculated by the sag formula, where w is the weight per unit length of the conductor, L is the span length, and T is the horizontal tension of the conductor;

[0016] Based on the conductor parameterized model and the tower coordinate set, a discrete waypoint sequence is obtained by improving the RRT* model. The tower coordinate set represents the location set of high-voltage line towers and is obtained from the environmental map.

[0017] Based on the constraints, the discrete waypoint sequence is adjusted to obtain the global waypoint sequence of the mission area;

[0018] Based on the global waypoint sequence and terrain elevation data, four-dimensional space-time collision detection is performed to generate safety detection results;

[0019] Based on the safety detection results, unsafe waypoints are eliminated to obtain a safe waypoint sequence at the location of the drone.

[0020] Furthermore, based on the parameterized model of the conductor and the tower coordinate set, the discrete waypoint sequence is obtained by improving the RRT* model, including:

[0021] Based on the tower coordinate set, auxiliary nodes are generated around the tower using the following formula:

[0022]

[0023] Among them, t j is the coordinate of the j-th tower, R is the radius of the auxiliary nodes around the tower, θ is the horizontal angle, and Δh is the vertical height offset;

[0024] Based on the parametric model of the conductor, random sampling is performed in the area around the conductor to obtain sampling points;

[0025] Calculate the distance from the sampling point to the wire, determine whether the distance exceeds the threshold, eliminate points whose distance exceeds the threshold, and obtain safe sampling points;

[0026] Search for safe sampling points and select the sampling point that minimizes the total value of the path as the target node;

[0027] The total cost is calculated using the following formula:

[0028]

[0029] Among them, α, β, γ, λ are weight coefficients, q j ,q i is the path node coordinate, θ j is the sensor coverage angle of the jth node, Δh i is the absolute value of the height change between adjacent nodes, k i is the curvature of the path segment;

[0030] When the safe sampling point is extended to the end point where the drone is located, the original path node sequence is obtained by tracing back to the starting point through the target node;

[0031] Determine whether the number of auxiliary nodes contained in the original path node sequence is greater than a threshold. If it is insufficient, insert auxiliary nodes based on the curvature between adjacent sampling points and the climb rate of the UAV.

[0032] The original path node sequence is sampled dynamically at intervals according to the path curvature to generate a discrete waypoint sequence.

[0033] Furthermore, based on the constraints, the discrete waypoint sequence is adjusted to obtain the global waypoint sequence, including:

[0034] The seventh-order polynomial path is obtained by fitting the global waypoint sequence with a seventh-order polynomial using the following formula:

[0035] q(s)=a0+a1s+a2s 2 +…+a7s 7 (s∈[0,1])

[0036] Where q(s) is the seventh-degree polynomial path and s is the normalized path length;

[0037] Based on the constraints, the septad polynomial path is subjected to sequential quadratic programming to generate an optimized continuous path;

[0038] Among them, after sequential quadratic programming, the comprehensive cost of the path is calculated using the following formula:

[0039]

[0040] Among them, k1, k2 are weight coefficients, v i is the linear velocity, w i is the angular velocity; the path with the minimum comprehensive cost is determined as the optimized continuous path;

[0041] Conduct feasibility verification on the optimized continuous path to obtain path verification data;

[0042] Based on the path verification data, the optimized continuous path is adjusted to obtain the global waypoint sequence.

[0043] Furthermore, the constraints are adjusted for the seventh-order polynomial path using the following formula:

[0044] Safety corridor constraint formula:

[0045]

[0046] Where q(s) is the coordinate of the optimized path at parameter s, and P(t(s)) is the point on the wire closest to q(s);

[0047] Heading angle alignment constraint formula:

[0048]

[0049] Where v(s) is the first-order derivative of the path, i.e., the velocity direction, t(s) is the unit tangent vector of the wire, and v(s)·t(s) is the cosine of the angle between the velocity direction of the path and the wire direction.

[0050] Sensor coverage constraint formula:

[0051]

[0052] Among them, δ j is a binary variable indicating whether the jth sensor is enabled, θ j (s) represents the coverage angle of the jth sensor at parameter s.

[0053] Furthermore, dynamic obstacle avoidance is performed based on the safe waypoint sequence and obstacle list to generate a real-time flight path, including:

[0054] With the global path as the center, a safety area with variable width is expanded to both sides to obtain a dynamic safety corridor;

[0055] Based on the dynamic safety corridor and obstacle list, a virtual repulsive force is created for each obstacle to obtain a resultant force field map;

[0056] Based on the resultant force field map, search for alternative paths along the direction of the resultant force field to obtain candidate paths;

[0057] Mark each candidate path with the estimated energy consumption to obtain the energy-optimized candidate path;

[0058] Generate a safe path list based on energy-optimized candidate paths and obstacle collision time;

[0059] Based on the aircraft performance, the safe path list is smoothed to obtain the real-time flight path.

[0060] Furthermore, after generating a safe path list based on energy-optimized candidate paths and obstacle collision times, the following steps are also included:

[0061] When the safe path list is empty, a recalculation signal is generated; the recalculation signal is used to instruct to recalculate the global path;

[0062] The global route is recalculated based on the recalculated signals and obstacle list to generate a new global reference path.

[0063] Furthermore, based on multi-source data, objects in the drone's location are identified and an obstacle list is determined, including:

[0064] Align the spatiotemporal information of multi-source data to obtain spatiotemporally aligned multi-source data;

[0065] Based on the multi-source data aligned in time and space, the dynamic objects at the location of the drone are identified through the YOLO model to obtain 2D dynamic objects;

[0066] Calculate the three-dimensional coordinates of the 2D dynamic object based on the binocular depth map to obtain the 3D dynamic object;

[0067] Based on the morphological features of 3D dynamic objects, 3D dynamic objects are classified to obtain a candidate list of dynamic obstacles;

[0068] Based on the LiDAR point cloud in the spatiotemporally aligned multi-source data, the LiDAR point cloud is Euclidean clustered to obtain independent object clusters;

[0069] Calculate the volume, height, and aspect ratio of independent object clusters, identify and classify static objects based on the rule base, and obtain a list of static obstacles;

[0070] Predict the motion trajectory of the dynamic obstacle candidate list and the static obstacle list to obtain the obstacle motion trajectory;

[0071] Calculate the collision time based on the obstacle trajectory and the drone control status;

[0072] Generate an obstacle list based on the dynamic obstacle candidate list, static obstacle list and collision time.

[0073] Furthermore, after performing dynamic obstacle avoidance based on the safe waypoint sequence and obstacle list and generating a real-time flight path, the following steps are also included:

[0074] Based on multi-source data, extract positioning data and IMU inertial navigation data to obtain positioning coordinates;

[0075] Generate flight control instructions based on positioning coordinates and real-time flight path;

[0076] Control the motor drive system and attitude adjustment mechanism based on flight control instructions to perform automatic line patrol tasks;

[0077] Among them, the motor drive system is used to control the flight speed of the UAV, and the attitude adjustment mechanism is used to control the flight attitude of the UAV.

[0078] In a second aspect, the present application also provides a drone high-voltage line automatic patrol system, comprising:

[0079] Source module, used to obtain multi-source data; multi-source data includes tower position, point cloud data, visual data, IMU inertial navigation data and positioning data;

[0080] The map module is used to fuse multi-source data to obtain the environmental map corresponding to the location of the drone;

[0081] The global waypoint module is used to generate a safe waypoint sequence for the UAV's location based on the environmental map using the improved RRT* algorithm;

[0082] The obstacle avoidance module is used to identify objects in the drone's location and determine the obstacle list based on multi-source data;

[0083] The flight module is used to perform dynamic obstacle avoidance based on a safe waypoint sequence and obstacle list, and generate a real-time flight path; the real-time flight path is used to instruct the drone to perform automatic line patrol missions.

[0084] The above-mentioned method and system for automatically patrolling high-voltage lines using a drone acquires multi-source data, including tower locations, point cloud data, visual data, IMU inertial navigation data, and positioning data. This multi-source data is then integrated to obtain an environmental map corresponding to the drone's location. Based on this environmental map, an improved RRT* algorithm is used to generate a safe waypoint sequence for the drone's location. Based on the multi-source data, objects in the drone's location are identified and an obstacle list is determined. Dynamic obstacle avoidance is performed based on the safe waypoint sequence and obstacle list, generating a real-time flight path. This real-time flight path is used to instruct the drone to perform automatic patrol tasks. Through these technical means, a method is provided that reduces human resource consumption and allows for the rapid and safe execution of patrol tasks. BRIEF DESCRIPTION OF THE DRAWINGS

[0085] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the following briefly introduces the drawings required for use in the embodiments or related technical descriptions. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0086] Figure 1This is a flow chart of a method for automatically inspecting high-voltage lines using a drone according to the present invention;

[0087] Figure 2 This is a diagram of an automatic high-voltage line inspection system for a drone of the present invention. DETAILED DESCRIPTION

[0088] In order to make the purpose, technical solutions and advantages of this application more clearly understood, the present application is further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0089] The embodiment of the present application provides a method for automatically inspecting high-voltage lines using a drone, which can be used in an application environment where high-voltage lines are automatically inspected.

[0090] In one embodiment, Figure 1 As shown, a method for automatically inspecting high-voltage lines using a drone is provided. This embodiment uses the method applied to a controller as an example. It is understood that the method can also be applied to a server, or to a system including a controller and a server, and implemented through interaction between the controller and the server. In this embodiment, the method includes the following steps:

[0091] Step 101, obtaining multi-source data; the multi-source data includes tower position, point cloud data, visual data, IMU inertial navigation data and positioning data.

[0092] Specifically, tower locations are the geographic coordinates of the steel towers or concrete structures supporting power transmission lines. These coordinates are manually entered in advance and serve as core reference points for line inspection missions, ensuring the drone's flight along the power lines. Point cloud data is a collection of three-dimensional spatial points captured by lidar. Visual data is collected by cameras. IMU inertial navigation data is acceleration and angular velocity data provided by the inertial measurement unit. Positioning data is the drone's geographic coordinates obtained via satellite systems such as GPS or Beidou.

[0093] Step 102: fuse the multi-source data to obtain an environmental map corresponding to the location of the UAV.

[0094] Specifically, multi-source data fusion involves integrating data from different sensors into unified information using algorithms (such as Kalman filtering and SLAM). The environmental map is a model of the drone's surroundings generated by fusing multi-source data. It can include a 3D point cloud map or a 2D grid map.

[0095] Step 103: Based on the environmental map, a safe waypoint sequence of the UAV's location is generated using the improved RRT* algorithm.

[0096] Specifically, the improved RRT* algorithm is an optimized path planning algorithm based on the rapidly expanding random tree (RRT*), which improves efficiency through heuristic search or dynamic weight adjustment. A safe waypoint sequence is a series of collision-free intermediate target points in the drone's flight path, forming a global flight path in the form of a series of coordinate sequences.

[0097] Step 104: Based on the multi-source data, identify objects in the location of the drone and determine an obstacle list.

[0098] Specifically, the obstacle list is a collection of objects that pose a threat to flight safety, identified through sensor data. Objects in the environment are classified using computer vision (such as the YOLO model) or point cloud segmentation technology.

[0099] Step 105 , dynamically avoid obstacles based on the safe waypoint sequence and the obstacle list, and generate a real-time flight path; the real-time flight path is used to instruct the UAV to perform the automatic line patrol mission.

[0100] Specifically, dynamic obstacle avoidance adjusts the flight path based on real-time sensor data to avoid unexpected obstacles, ensuring flight safety and responding to dynamic environmental changes. The real-time flight path is a continuous trajectory generated by combining global safe waypoints and local obstacle avoidance, typically implemented using PID control or model predictive control.

[0101] This embodiment provides a method for automatically patrolling high-voltage lines using a drone. The method obtains multi-source data, including tower locations, point cloud data, visual data, IMU inertial navigation data, and positioning data. This multi-source data is then integrated to create an environmental map corresponding to the drone's location. Based on the environmental map, an improved RRT* algorithm is used to generate a safe waypoint sequence for the drone's location. Based on the multi-source data, objects in the drone's location are identified and an obstacle list is determined. Dynamic obstacle avoidance is performed based on the safe waypoint sequence and obstacle list, generating a real-time flight path. This real-time flight path is used to instruct the drone to perform automatic patrol tasks. This method reduces human resource consumption and allows for the rapid and safe execution of patrol tasks.

[0102] In one embodiment, based on the environmental map, an improved RRT* algorithm is used to generate a safe waypoint sequence for the location of the drone, including:

[0103] Step 201: Based on the environment map, the parameterized model of the conductor is calculated using the following formula:

[0104]

[0105] P i =(x i ,y i ,z i +Δh i )

[0106]

[0107] Among them, P(t) is the conductor parameterized model, which dynamically describes the conductor shape through the cubic B-spline basis function and the sag correction formula, N i,3 (t) is the ith cubic B-spline basis function, P i is the i-th control point, t is the curve parameter, i is the control point index, Δh i is the height adjustment of the i-th control point due to the sag of the conductor, calculated by the sag formula, w is the weight of the conductor per unit length, L is the span length, and T is the horizontal tension of the conductor.

[0108] Specifically, the conductor parameterized model dynamically describes the three-dimensional spatial shape of power conductors through mathematical formulas, combining B-spline curves with conductor sag correction. A cubic (third-order) B-spline basis function is used to smoothly interpolate control points, ensuring the curve is continuous and differentiable. The control points are the coordinates of the control points in the conductor model and include the sag-corrected height values. The sag formula is the sag height Δh caused by the conductor's deadweight and tension.

[0109] Step 202: Based on the conductor parameterized model and the tower coordinate set, a discrete waypoint sequence is obtained by improving the RRT* model; wherein the tower coordinate set represents the location set of high-voltage line towers and is obtained from the environmental map.

[0110] Specifically, the tower coordinate set is the geographic coordinates of all towers in the power line, extracted from the environmental map. The improved RRT* model builds on the classic rapidly expanding random tree algorithm by adding heuristic search and dynamic weighting. This model, combined with a parameterized conductor model, guides the search direction, reduces randomness, and balances path length, obstacle avoidance distance, and computational efficiency. The discrete waypoint sequence is the set of waypoints output by the improved RRT* algorithm, representing the intermediate locations that the drone must pass through.

[0111] Step 203: Based on the constraint conditions, the discrete waypoint sequence is adjusted to obtain a global waypoint sequence of the mission area.

[0112] Specifically, the constraints are the physical and mission restrictions that the UAV must meet during flight. The global waypoint sequence is a set of adjusted continuous waypoints that covers the entire patrol mission area.

[0113] For example, the constraints may include: safety distance, maximum climb rate, mission priority, etc.

[0114] Step 204 : Based on the global waypoint sequence and terrain elevation data, four-dimensional space-time collision detection is performed to generate a safety detection result.

[0115] Specifically, four-dimensional space-time collision detection superimposes the time dimension (t) on the three-dimensional space (x, y, z) to detect whether the drone's future trajectory will collide with the ground. Terrain elevation data is surface elevation information obtained through a digital elevation model. The safety check result is a Boolean value indicating whether the waypoint is safe.

[0116] Step 205: Based on the safety detection results, unsafe waypoints are eliminated to obtain a safe waypoint sequence for the location of the UAV.

[0117] Specifically, unsafe waypoints are waypoints marked as dangerous and may cause collisions or violations. The safe waypoint sequence is the set of waypoints remaining after removing unsafe waypoints, forming the global flight path.

[0118] This embodiment uses a parameterized conductor model to accurately model the power line geometry and combines it with an improved RRT* algorithm to generate a preliminary path. Waypoints are adjusted based on physical constraints, and then dangerous nodes are eliminated through four-dimensional space-time collision detection. Ultimately, a safe waypoint sequence is output. This improves flight path accuracy, balances energy consumption with mission costs, and enhances mission safety.

[0119] In one embodiment, based on the parameterized model of the conductor and the tower coordinate set, a discrete waypoint sequence is obtained by improving the RRT* model, including:

[0120] Step 301, based on the tower coordinate set, generate auxiliary nodes around the tower using the following formula:

[0121]

[0122] Among them, t j is the coordinate of the j-th tower, R is the radius of the auxiliary nodes around the tower, θ is the horizontal angle, and Δh is the vertical height offset.

[0123] Specifically, auxiliary nodes are additional path nodes generated around the tower coordinates to enhance path planning's obstacle avoidance and coverage capabilities within the tower area. Three horizontal angles are used to form a triangular distribution around the tower to prevent side collisions between drones and the tower. Vertical nodes are used to adapt to different flight altitudes.

[0124] Step 302: Based on the parameterized model of the conductor, random sampling is performed in the area around the conductor to obtain sampling points.

[0125] Specifically, random sampling points are generated in an area around the wire, such as a cylindrical space with a radius of 5 meters.

[0126] Step 303: Calculate the distance from the sampling point to the conductor, determine whether the distance exceeds a threshold, eliminate points whose distance exceeds the threshold, and obtain safe sampling points.

[0127] Specifically, the threshold is a dual threshold mechanism that introduces a minimum safety distance and a maximum safety distance, eliminating sampling points whose distance to the conductor is less than the minimum safety distance or greater than the maximum safety distance, and only retaining points whose distance is between the thresholds.

[0128] Step 304: Search for safe sampling points, and select the sampling point that minimizes the total value of the path as the target node.

[0129] The total cost is calculated using the following formula:

[0130]

[0131] Among them, α, β, γ, λ are weight coefficients, q j ,q i is the path node coordinate, θ j is the sensor coverage angle of the jth node, Δh i is the absolute value of the height change between adjacent nodes, k i is the curvature of the path segment.

[0132] Specifically, the total cost formula is used to evaluate a function of the comprehensive cost of a path, balancing path length, sensor coverage, altitude change, and curvature.

[0133] Step 305: When the safety sampling point is extended to the end point where the UAV is located, the original path node sequence is obtained by tracing back to the starting point through the target node.

[0134] Specifically, the path nodes are traversed in reverse from the end point to the starting point to form a preliminary path sequence.

[0135] Step 306 : Determine whether the number of auxiliary nodes included in the original path node sequence is greater than a threshold. If not, insert auxiliary nodes based on the curvature between adjacent sampling points and the climb rate of the UAV.

[0136] Specifically, if the original path contains insufficient auxiliary nodes (e.g., fewer than three), new nodes are inserted. Curvature Limit: If the curvature between adjacent nodes exceeds the drone's maximum turning capability (e.g., 0.5 rad / m), nodes are inserted to smooth the path. Climb Rate Limit: If the altitude change rate exceeds the drone's climb rate (e.g., 5 m / s), nodes are inserted to reduce the ascent and descent speed.

[0137] Step 307: dynamically sample the original path node sequence according to the path curvature to generate a discrete waypoint sequence.

[0138] Specifically, the waypoint density is adjusted according to the path curvature, with dense sampling in areas with large curvature and sparse sampling in straight areas.

[0139] This embodiment generates candidate paths by sampling around auxiliary nodes of towers and conductors, screens for safe nodes, optimizes path costs, and ultimately generates a discrete waypoint sequence tailored to the drone's capabilities. This improves the drone's flight stability and route safety.

[0140] In one embodiment, based on the constraint conditions, the discrete waypoint sequence is adjusted to obtain the global waypoint sequence, including:

[0141] Step 401: Fit the global waypoint sequence with a seventh-order polynomial using the following formula to obtain a seventh-order polynomial path:

[0142] q(s)=a0+a1s+a2s 2 +…+a7s 7 (s∈[0,1])

[0143] Where q(s) is the seventh-degree polynomial path and s is the normalized path length.

[0144] Specifically, a seventh-order polynomial path is a seventh-order polynomial function constructed using eight coefficients, describing the drone's path's continuous position, velocity, acceleration, and other high-order motion parameters. This ensures the continuity of the path's third-order derivatives, preventing the drone from vibrating or losing control due to sudden changes in motion.

[0145] Step 402 : Based on the constraints, the septad polynomial path is subjected to sequential quadratic programming to generate an optimized continuous path.

[0146] Among them, after sequential quadratic programming, the comprehensive cost of the path is calculated using the following formula:

[0147]

[0148] Among them, k1, k2 are weight coefficients, v i is the linear velocity, w i is the angular velocity; the path with the minimum comprehensive cost is determined as the optimized continuous path.

[0149] Specifically, sequential quadratic programming is an iterative optimization algorithm that solves a nonlinear constraint problem by breaking it down into multiple quadratic programming subproblems. Constraints can include maximum velocity, acceleration, angular velocity, and minimum distance to a wire or obstacle.

[0150] Step 403: Perform feasibility verification on the optimized continuous path to obtain path verification data.

[0151] Specifically, feasibility verification checks whether the optimized path meets the drone's actual flight capabilities and environmental safety requirements. This includes: dynamic feasibility (whether the speed and acceleration exceed the drone's performance limits); obstacle avoidance feasibility (whether the path will collide with obstacles or wires); and energy feasibility (whether the estimated energy consumption is within the battery capacity). This ensures that the mathematical optimality of the path is consistent with practical feasibility to prevent theoretical optimization results from being unimplementable.

[0152] Step 404 : Based on the path verification data, the optimized continuous path is adjusted to obtain a global waypoint sequence.

[0153] Specifically, we re-optimize the path segments that failed validation using SQP. We relax some constraints, such as allowing slightly higher speeds, in exchange for feasibility. We also insert auxiliary waypoints in areas with excessive curvature to reduce motion complexity.

[0154] In this embodiment, through seventh-order polynomial fitting and sequential quadratic programming optimization, the system converts a discrete waypoint sequence into a dynamically feasible, safe and smooth continuous path, significantly improving the autonomy and reliability of the drone in complex line patrol tasks.

[0155] In one embodiment, the constraint condition is adjusted by the following formula for the seventh-degree polynomial path:

[0156] Safety corridor constraint formula:

[0157]

[0158] Where q(s) is the coordinate of the optimized path at parameter s, and P(t(s)) is the point on the wire closest to q(s).

[0159] Specifically, this prevents the drone from approaching conductors, potentially causing collisions or high-voltage arcing. It also ensures that the drone is within the effective operating range of the sensors. This formula must be satisfied at all locations along the path to avoid local violations caused by undulations or bends in the conductors. This hard constraint is embedded in the optimization algorithm to enforce global compliance along the path.

[0160] Heading angle alignment constraint formula:

[0161]

[0162] Where v(s) is the first-order derivative of the path, i.e., the velocity direction, t(s) is the unit tangent vector of the wire, and v(s)·t(s) is the cosine of the angle between the velocity direction of the path and the wire direction.

[0163] Specifically, the deviation between the drone's flight direction and the direction of the wire is limited to no more than 15° to ensure that the sensor continues to aim at the target.

[0164] Sensor coverage constraint formula:

[0165]

[0166] Among them, δ j is a binary variable indicating whether the jth sensor is enabled, θ j (s) represents the coverage angle of the jth sensor at parameter s.

[0167] Specifically, the drone ensures continuous monitoring of the wires and surrounding environment from both sides. This allows for wide-angle overlay of multiple sensors. Sensors are selectively enabled through binary variables, reducing energy consumption.

[0168] Through the combined effect of three types of constraints, this embodiment achieves a balance between the drone's patrol path in terms of safety distance, heading stability, and perception coverage, ensuring efficient and reliable execution of tasks in complex environments.

[0169] In one embodiment, dynamic obstacle avoidance is performed based on a safe waypoint sequence and an obstacle list to generate a real-time flight path, including:

[0170] Step 501 : Taking the global path as the center, expand safety areas of variable width to both sides to obtain a dynamic safety corridor.

[0171] Specifically, the dynamic safety corridor is a variable-width safety area centered on the global path and expanding to both sides based on real-time environmental data.

[0172] Step 502: Based on the dynamic safety corridor and the obstacle list, a virtual repulsive force is created for each obstacle to obtain a resultant force field map.

[0173] Specifically, the virtual repulsive force is the simulated repulsive force generated by obstacles on the drone path, and its magnitude is inversely proportional to the distance. The resultant force field map is the vector sum of the repulsive forces of all obstacles.

[0174] Step 503: Based on the resultant force field map, search for alternative paths along the resultant force field direction to obtain candidate paths.

[0175] Specifically, a gradient descent method or an improved A* algorithm is used to search for alternative paths along the direction of the resultant force field. Paths are preferentially expanded along the direction of the negative gradient of the resultant force field, i.e., the direction in which the repulsive force decreases.

[0176] Step 504 : Mark the estimated energy consumption of each candidate path to obtain an energy-optimized candidate path.

[0177] Specifically, the energy consumption of each candidate path is annotated to select low-power routes and extend battery life. Energy consumption is related to physical quantities such as speed, acceleration, and altitude change.

[0178] Step 505 : Generate a safe path list based on the energy-optimized candidate paths and obstacle collision times.

[0179] Specifically, the obstacle collision time is the time when the closest distance between the predicted path and the obstacle reaches a threshold. Paths with obstacle collision times less than the safe time are eliminated. The safe path list is a collection of compliant paths sorted in ascending order of energy consumption.

[0180] Step 506 : Based on the aircraft performance, the safe path list is smoothed to obtain a real-time flight path.

[0181] Specifically, cubic spline interpolation or B-spline optimization is applied to the path nodes, limiting the maximum curvature and climb rate to ensure curvature continuity. This generates a smooth, continuous flight path that adapts to the drone's maneuverability and avoids sharp turns or sudden stops.

[0182] This embodiment responds to environmental changes in real time through safety corridors and resultant force fields, reduces energy consumption of inspection tasks, improves endurance, and ensures flight safety through collision time assessment and smoothing.

[0183] In one embodiment, after generating a safe path list based on energy-optimized candidate paths and obstacle collision times, the method further includes:

[0184] Step 601: When the safe path list is empty, a recalculation signal is generated; the recalculation signal is used to instruct to recalculate the global path.

[0185] Specifically, the recalculation signal is a system instruction triggered when the safe path list is empty (i.e., all candidate paths fail to meet safety or feasibility constraints), indicating that the global path needs to be replanned. This can occur when all candidate paths are eliminated due to collision risk, excessive energy consumption, or insufficient sensor coverage. It can also occur when a new obstacle renders the original global path invalid.

[0186] Step 602 : recalculate the global route based on the recalculated signals and the obstacle list to generate a new global reference path.

[0187] Specifically, based on the latest obstacle list and environment map, the process of regenerating the global reference path can include expanding the search space and relaxing constraints. The new global reference path is a new path framework that avoids the current infeasible area.

[0188] This embodiment avoids the drone from hovering or losing control due to path planning failure by recalculating the path, thereby ensuring mission continuity.

[0189] In one embodiment, identifying objects in a location of a drone based on multi-source data and determining an obstacle list includes:

[0190] Step 701: aligning the spatiotemporal information of multi-source data to obtain spatiotemporally aligned multi-source data.

[0191] Specifically, spatiotemporal alignment is to unify the data from different sensors in terms of timestamp and spatial coordinate system to ensure data consistency.

[0192] Step 702: Based on the spatiotemporally aligned multi-source data, the YOLO model is used to identify dynamic objects at the location of the drone to obtain 2D dynamic objects.

[0193] Specifically, the YOLO model is a real-time object detection model based on convolutional neural networks. It inputs an image and outputs the object category and 2D bounding box. 2D dynamic objects are objects whose position or shape changes in consecutive image frames.

[0194] Step 703 : Calculate the three-dimensional coordinates of the 2D dynamic object based on the binocular depth map to obtain the 3D dynamic object.

[0195] Specifically, the scene depth information is calculated through dual-camera parallax, and 2D dynamic objects are converted to 3D spatial coordinates to support motion prediction.

[0196] Step 704 : Classify the 3D dynamic objects based on their morphological features to obtain a list of dynamic obstacle candidates.

[0197] Specifically, morphological features are the geometric and motion properties of an object.

[0198] Step 705 : Based on the LiDAR point cloud in the spatiotemporally aligned multi-source data, the LiDAR point cloud is Euclidean clustered to obtain independent object clusters.

[0199] Specifically, Euclidean clustering is based on the Euclidean distance of points in a point cloud, classifying points with a distance less than a threshold as the same object. An independent object cluster is a collection of point clouds that represent independent entities in the environment.

[0200] Step 706 , calculate the volume, height, and aspect ratio of the independent object clusters, identify and classify static objects based on the rule base, and obtain a static obstacle list.

[0201] Specifically, the rule base is a collection of classification rules based on expert experience.

[0202] Step 707: Predict motion trajectories for the dynamic obstacle candidate list and the static obstacle list to obtain obstacle motion trajectories.

[0203] Specifically, Kalman filtering or LSTM is used to predict future trajectories.

[0204] Step 708: Calculate the collision time based on the obstacle's trajectory and the drone's control state.

[0205] Step 709: Generate an obstacle list based on the dynamic obstacle candidate list, the static obstacle list, and the collision time.

[0206] Specifically, the obstacle list contains information such as the type, location, speed, collision time, and threat level of each obstacle.

[0207] This embodiment generates an obstacle list through multi-source data alignment, dynamic and static obstacle detection, and trajectory prediction, providing core input for autonomous obstacle avoidance and path planning of drones.

[0208] In one embodiment, after performing dynamic obstacle avoidance based on the safe waypoint sequence and obstacle list and generating a real-time flight path, the process further includes:

[0209] Step 801: Extract positioning data and IMU inertial navigation data based on multi-source data to obtain positioning coordinates.

[0210] Specifically, the positioning coordinates are the precise three-dimensional coordinates of the drone after integrating satellite positioning, IMU and vision / LiDAR data.

[0211] Step 802: Generate flight control instructions based on the positioning coordinates and the real-time flight path.

[0212] Specifically, flight control instructions convert the target path into low-level control signals that can be executed by the drone. These may include target forward speed, climb rate, target pitch angle, roll angle, and yaw angle.

[0213] Step 803: Control the motor drive system and the attitude adjustment mechanism based on the flight control command to perform the automatic line patrol task.

[0214] Among them, the motor drive system is used to control the flight speed of the UAV, and the attitude adjustment mechanism is used to control the flight attitude of the UAV.

[0215] Specifically, the motor drive system is a power unit consisting of an electronic speed controller (ESC) and a brushless motor. It controls the thrust of the drone by adjusting the motor speed. The attitude adjustment mechanism adjusts the drone's attitude through motor differential speed (for multi-rotor aircraft) or servos (for fixed-wing aircraft).

[0216] In this embodiment, the UAV realizes the fully automatic line patrol mission through positioning coordinate solution, control instruction generation and execution mechanism coordination, and has the ability of high-precision path tracking, anti-interference and adaptability to complex environments.

[0217] It should be understood that, although the various steps in the flowcharts involved in the various embodiments described above are displayed in sequence according to the instructions of the arrows, these steps are not necessarily executed in sequence in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and these steps can be executed in other orders. Moreover, at least a portion of the steps in the flowcharts involved in the various embodiments described above can include multiple steps or multiple stages, and these steps or stages are not necessarily executed and completed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a portion of steps or stages in other steps.

[0218] Based on the same inventive concept, the present application also provides an unmanned aerial vehicle (UAV) high-voltage line automatic patrol system for implementing the aforementioned unmanned aerial vehicle (UAV) high-voltage line automatic patrol method. The solution provided by this system is similar to the solution described in the aforementioned method. Therefore, the specific limitations of one or more of the following unmanned aerial vehicle (UAV) high-voltage line automatic patrol system embodiments can be found in the aforementioned limitations of the unmanned aerial vehicle (UAV) high-voltage line automatic patrol method, and will not be further elaborated here.

[0219] In an exemplary embodiment, Figure 2 As shown, a UAV high-voltage line automatic inspection system 900 is provided, comprising:

[0220] Source module 901 is used to obtain multi-source data; the multi-source data includes tower position, point cloud data, visual data, IMU inertial navigation data and positioning data;

[0221] Map module 902, used to fuse multi-source data to obtain an environmental map corresponding to the location of the UAV;

[0222] The global waypoint module 903 is used to generate a safe waypoint sequence of the UAV's location based on the environment map using the improved RRT* algorithm;

[0223] The obstacle avoidance module 904 is used to identify objects in the drone's location based on multi-source data and determine an obstacle list;

[0224] The flight module 905 is used to perform dynamic obstacle avoidance based on the safe waypoint sequence and obstacle list and generate a real-time flight path; the real-time flight path is used to instruct the UAV to perform the automatic line patrol mission.

[0225] Furthermore, the global waypoint module 903 is further configured to:

[0226] Based on the environment map, the wire parameterization model is calculated using the following formula:

[0227]

[0228] P i =(x i ,y i ,z i +Δh i )

[0229]

[0230] Among them, P(t) is the conductor parameterized model, which dynamically describes the conductor shape through the cubic B-spline basis function and the sag correction formula, N i,3 (t) is the ith cubic B-spline basis function, P i is the i-th control point, t is the curve parameter, i is the control point index, Δh i is the height adjustment of the i-th control point due to the sag of the conductor, calculated by the sag formula, where w is the weight per unit length of the conductor, L is the span length, and T is the horizontal tension of the conductor;

[0231] Based on the conductor parameterized model and the tower coordinate set, a discrete waypoint sequence is obtained by improving the RRT* model. The tower coordinate set represents the location set of high-voltage line towers and is obtained from the environmental map.

[0232] Based on the constraints, the discrete waypoint sequence is adjusted to obtain the global waypoint sequence of the mission area;

[0233] Based on the global waypoint sequence and terrain elevation data, four-dimensional space-time collision detection is performed to generate safety detection results;

[0234] Based on the safety detection results, unsafe waypoints are eliminated to obtain a safe waypoint sequence at the location of the drone.

[0235] Furthermore, based on the parameterized model of the conductor and the tower coordinate set, the discrete waypoint sequence is obtained by improving the RRT* model, including:

[0236] Based on the tower coordinate set, auxiliary nodes are generated around the tower using the following formula:

[0237]

[0238] Among them, t j is the coordinate of the j-th tower, R is the radius of the auxiliary nodes around the tower, θ is the horizontal angle, and Δh is the vertical height offset;

[0239] Based on the parametric model of the conductor, random sampling is performed in the area around the conductor to obtain sampling points;

[0240] Calculate the distance from the sampling point to the wire, determine whether the distance exceeds the threshold, eliminate points whose distance exceeds the threshold, and obtain safe sampling points;

[0241] Search for safe sampling points and select the sampling point that minimizes the total value of the path as the target node;

[0242] The total cost is calculated using the following formula:

[0243]

[0244] Among them, α, β, γ, λ are weight coefficients, q j ,q i is the path node coordinate, θ j is the sensor coverage angle of the jth node, Δh i is the absolute value of the height change between adjacent nodes, k i is the curvature of the path segment;

[0245] When the safe sampling point is extended to the end point where the drone is located, the original path node sequence is obtained by tracing back to the starting point through the target node;

[0246] Determine whether the number of auxiliary nodes contained in the original path node sequence is greater than a threshold. If it is insufficient, insert auxiliary nodes based on the curvature between adjacent sampling points and the climb rate of the UAV.

[0247] The original path node sequence is sampled dynamically at intervals according to the path curvature to generate a discrete waypoint sequence.

[0248] Furthermore, based on the constraints, the discrete waypoint sequence is adjusted to obtain the global waypoint sequence, including:

[0249] The seventh-order polynomial path is obtained by fitting the global waypoint sequence with a seventh-order polynomial using the following formula:

[0250] q(s)=a0+a1s+a2s 2 +…+a7s 7 (s∈[0,1])

[0251] Where q(s) is the seventh-degree polynomial path and s is the normalized path length;

[0252] Based on the constraints, the septad polynomial path is subjected to sequential quadratic programming to generate an optimized continuous path;

[0253] Among them, after sequential quadratic programming, the comprehensive cost of the path is calculated using the following formula:

[0254]

[0255] Among them, k1, k2 are weight coefficients, v iis the linear velocity, w i is the angular velocity; the path with the minimum comprehensive cost is determined as the optimized continuous path;

[0256] Conduct feasibility verification on the optimized continuous path to obtain path verification data;

[0257] Based on the path verification data, the optimized continuous path is adjusted to obtain the global waypoint sequence.

[0258] Furthermore, the constraints are adjusted for the seventh-order polynomial path using the following formula:

[0259] Safety corridor constraint formula:

[0260]

[0261] Where q(s) is the coordinate of the optimized path at parameter s, and P(t(s)) is the point on the wire closest to q(s);

[0262] Heading angle alignment constraint formula:

[0263]

[0264] Where v(s) is the first-order derivative of the path, i.e., the velocity direction, t(s) is the unit tangent vector of the wire, and v(s)·t(s) is the cosine of the angle between the velocity direction of the path and the wire direction.

[0265] Sensor coverage constraint formula:

[0266]

[0267] Among them, δ j is a binary variable indicating whether the jth sensor is enabled, θ j (s) represents the coverage angle of the jth sensor at parameter s.

[0268] Furthermore, the flight module 905 is also used to:

[0269] With the global path as the center, a safety area with variable width is expanded to both sides to obtain a dynamic safety corridor;

[0270] Based on the dynamic safety corridor and obstacle list, a virtual repulsive force is created for each obstacle to obtain a resultant force field map;

[0271] Based on the resultant force field map, search for alternative paths along the direction of the resultant force field to obtain candidate paths;

[0272] Mark each candidate path with the estimated energy consumption to obtain the energy-optimized candidate path;

[0273] Generate a safe path list based on energy-optimized candidate paths and obstacle collision time;

[0274] Based on the aircraft performance, the safe path list is smoothed to obtain the real-time flight path.

[0275] Furthermore, after generating a safe path list based on energy-optimized candidate paths and obstacle collision times, the following steps are also included:

[0276] When the safe path list is empty, a recalculation signal is generated; the recalculation signal is used to instruct to recalculate the global path;

[0277] The global route is recalculated based on the recalculated signals and obstacle list to generate a new global reference path.

[0278] Furthermore, the obstacle avoidance module 904 is further configured to:

[0279] Align the spatiotemporal information of multi-source data to obtain spatiotemporally aligned multi-source data;

[0280] Based on the multi-source data aligned in time and space, the dynamic objects at the location of the drone are identified through the YOLO model to obtain 2D dynamic objects;

[0281] Calculate the three-dimensional coordinates of the 2D dynamic object based on the binocular depth map to obtain the 3D dynamic object;

[0282] Based on the morphological features of 3D dynamic objects, 3D dynamic objects are classified to obtain a candidate list of dynamic obstacles;

[0283] Based on the LiDAR point cloud in the spatiotemporally aligned multi-source data, the LiDAR point cloud is Euclidean clustered to obtain independent object clusters;

[0284] Calculate the volume, height, and aspect ratio of independent object clusters, identify and classify static objects based on the rule base, and obtain a list of static obstacles;

[0285] Predict the motion trajectory of the dynamic obstacle candidate list and the static obstacle list to obtain the obstacle motion trajectory;

[0286] Calculate the collision time based on the obstacle trajectory and the drone control status;

[0287] Generate an obstacle list based on the dynamic obstacle candidate list, static obstacle list and collision time.

[0288] Furthermore, the system further includes an execution module for:

[0289] Based on multi-source data, extract positioning data and IMU inertial navigation data to obtain positioning coordinates;

[0290] Generate flight control instructions based on positioning coordinates and real-time flight path;

[0291] Control the motor drive system and attitude adjustment mechanism based on flight control instructions to perform automatic line patrol tasks;

[0292] Among them, the motor drive system is used to control the flight speed of the UAV, and the attitude adjustment mechanism is used to control the flight attitude of the UAV.

[0293] In one embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the steps of the above-mentioned method for automatic high-voltage line inspection using a drone are implemented.

[0294] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0295] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to the partial description of the method embodiments. The device embodiments described above are merely illustrative, wherein the components described as separate parts may or may not be physically separated, and the parts displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the disclosed solution. A person of ordinary skill in the art can understand and implement it without expending creative work.

[0296] The above-described embodiments merely represent several implementation methods of the embodiments of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the patent application. It should be noted that a person skilled in the art may make various modifications and improvements without departing from the concept of the embodiments of the present application, and these modifications and improvements fall within the scope of protection of the embodiments of the present application.

Claims

1. A method for automatically inspecting high-voltage lines using a drone, characterized in that: The method comprises: Acquire multi-source data; the multi-source data includes tower position, point cloud data, visual data, IMU inertial navigation data and positioning data; Fusing the multi-source data to obtain an environmental map corresponding to the location of the UAV; Based on the environmental map, a safe waypoint sequence of the location of the UAV is generated using an improved RRT* algorithm; Based on the multi-source data, identify objects in the location of the drone and determine an obstacle list; Dynamic obstacle avoidance is performed based on the safe waypoint sequence and the obstacle list to generate a real-time flight path; the real-time flight path is used to instruct the UAV to perform an automatic line patrol mission.

2. The method for automatically inspecting high-voltage lines using a drone according to claim 1, characterized in that: The method of generating a safe waypoint sequence for the location of the UAV based on the environmental map using an improved RRT* algorithm includes: Based on the environment map, the wire parameterization model is calculated using the following formula: P i =(x i ,y i ,z i +Δh i ) Among them, P(t) is the conductor parameterized model, which dynamically describes the conductor shape through the cubic B-spline basis function and the sag correction formula, N i,3 (t) is the ith cubic B-spline basis function, P i is the i-th control point, t is the curve parameter, i is the control point index, Δh i is the height adjustment of the i-th control point due to the sag of the conductor, calculated by the sag formula, where w is the weight per unit length of the conductor, L is the span length, and T is the horizontal tension of the conductor; Based on the conductor parameterized model and the tower coordinate set, a discrete waypoint sequence is obtained by improving the RRT* model; wherein the tower coordinate set represents the location set of high-voltage line towers and is obtained from an environmental map; Based on the constraint conditions, the discrete waypoint sequence is adjusted to obtain a global waypoint sequence of the mission area; Based on the global waypoint sequence and terrain elevation data, four-dimensional spatiotemporal collision detection is performed to generate a safety detection result; Based on the safety detection results, unsafe waypoints are eliminated to obtain a safe waypoint sequence at the location of the drone.

3. The method for automatically inspecting high-voltage lines using an unmanned aerial vehicle according to claim 2, characterized in that: The discrete waypoint sequence is obtained by improving the RRT* model based on the conductor parameterized model and the tower coordinate set, including: Based on the tower coordinate set, auxiliary nodes are generated around the tower using the following formula: Among them, t j is the coordinate of the j-th tower, R is the radius of the auxiliary nodes around the tower, θ is the horizontal angle, and Δh is the vertical height offset; Based on the conductor parameterized model, randomly sampling the area around the conductor to obtain sampling points; Calculating the distance from the sampling point to the wire, determining whether the distance exceeds a threshold, eliminating points whose distance exceeds the threshold, and obtaining safe sampling points; Searching for the safe sampling point, and selecting the sampling point that minimizes the total value of the path as the target node; The total cost is calculated using the following formula: Among them, α, β, γ, λ are weight coefficients, q j ,q i is the path node coordinate, θ j is the sensor coverage angle of the jth node, Δh i is the absolute value of the height change between adjacent nodes, k i is the curvature of the path segment; When the safety sampling point is extended to the end point where the drone is located, the original path node sequence is obtained by tracing back to the starting point through the target node; Determine whether the number of auxiliary nodes included in the original path node sequence is greater than a threshold; if not, insert auxiliary nodes based on the curvature between adjacent sampling points and the climb rate of the UAV; The original path node sequence is sampled at dynamic intervals according to the path curvature to generate the discrete waypoint sequence.

4. The method for automatically inspecting high-voltage lines using an unmanned aerial vehicle according to claim 2, characterized in that: The step of adjusting the discrete waypoint sequence based on the constraint condition to obtain a global waypoint sequence includes: The seventh-order polynomial path is obtained by fitting the global waypoint sequence to a seventh-order polynomial using the following formula: q(s)=a0+a1s+a2s 2 +…+a7s 7 (s∈[0,1]) Where q(s) is the seventh-degree polynomial path and s is the normalized path length; Based on the constraint conditions, the septad polynomial path is subjected to sequential quadratic programming to generate an optimized continuous path; Among them, after sequential quadratic programming, the comprehensive cost of the path is calculated using the following formula: Among them, k1, k2 are weight coefficients, v i is the linear velocity, w i is the angular velocity; the path with the minimum comprehensive cost is determined as the optimized continuous path; Performing feasibility verification on the optimized continuous path to obtain path verification data; Based on the path verification data, the optimized continuous path is adjusted to obtain the global waypoint sequence.

5. The method for automatically inspecting high-voltage lines using an unmanned aerial vehicle according to claim 4, characterized in that: The constraint condition is to adjust the seventh-order polynomial path by the following formula: Safety corridor constraint formula: Where q(s) is the coordinate of the optimized path at parameter s, and P(t(s)) is the point on the wire closest to q(s); Heading angle alignment constraint formula: Where v(s) is the first-order derivative of the path, i.e., the velocity direction, t(s) is the unit tangent vector of the wire, and v(s)·t(s) is the cosine of the angle between the velocity direction of the path and the wire direction. Sensor coverage constraint formula: Among them, δ j is a binary variable indicating whether the jth sensor is enabled, θ j (s) represents the coverage angle of the jth sensor at parameter s.

6. The method for automatically inspecting high-voltage lines using an unmanned aerial vehicle according to claim 1, characterized in that: The performing dynamic obstacle avoidance based on the safe waypoint sequence and the obstacle list to generate a real-time flight path includes: With the global path as the center, a safety area with variable width is expanded to both sides to obtain a dynamic safety corridor; Based on the dynamic safety corridor and the obstacle list, creating a virtual repulsive force for each obstacle to obtain a resultant force field map; Based on the resultant force field map, searching for alternative paths along the resultant force field direction to obtain candidate paths; Marking the estimated energy consumption of each candidate path to obtain an energy-optimized candidate path; generating a safe path list based on the energy-optimized candidate paths and obstacle collision times; Based on the aircraft performance, the safe path list is smoothed to obtain the real-time flight path.

7. The method for automatically inspecting high-voltage lines using an unmanned aerial vehicle according to claim 6, characterized in that: After generating a safe path list based on the energy-optimized candidate paths and obstacle collision times, the method further includes: When the safe path list is empty, a recalculation signal is generated; the recalculation signal is used to instruct to recalculate the global path; The global route is recalculated based on the recalculation signal and the obstacle list to generate a new global reference path.

8. The method for automatically inspecting high-voltage lines using an unmanned aerial vehicle according to claim 1, characterized in that: The step of identifying objects in the location of the drone based on the multi-source data and determining an obstacle list includes: Aligning the spatiotemporal information of the multi-source data to obtain spatiotemporally aligned multi-source data; Based on the spatiotemporal aligned multi-source data, the dynamic objects at the location of the drone are identified using the YOLO model to obtain 2D dynamic objects; Calculate the three-dimensional coordinates of the 2D dynamic object based on the binocular depth map to obtain a 3D dynamic object; classifying the 3D dynamic objects based on their morphological features to obtain a candidate list of dynamic obstacles; Based on the LiDAR point cloud in the spatiotemporally aligned multi-source data, Euclidean clustering of the LiDAR point cloud is performed to obtain independent object clusters; Calculating the volume, height, and aspect ratio of the independent object cluster, identifying and classifying static objects based on a rule base, and obtaining a static obstacle list; Predicting motion trajectories of the dynamic obstacle candidate list and the static obstacle list to obtain obstacle motion trajectories; Calculating the collision time based on the obstacle's motion trajectory and the drone's control state; The obstacle list is generated based on the dynamic obstacle candidate list, the static obstacle list, and the collision time.

9. The method for automatically inspecting high-voltage lines using an unmanned aerial vehicle according to claim 1, characterized in that: After performing dynamic obstacle avoidance based on the safe waypoint sequence and the obstacle list to generate a real-time flight path, the method further includes: Extracting the positioning data and the IMU inertial navigation data based on the multi-source data to obtain positioning coordinates; generating flight control instructions based on the positioning coordinates and the real-time flight path; Controlling the motor drive system and the attitude adjustment mechanism based on the flight control instructions to perform the automatic line patrol task; The motor drive system is used to control the flight speed of the UAV, and the attitude adjustment mechanism is used to control the flight attitude of the UAV.

10. An automatic high-voltage line inspection system using a drone, characterized in that: The system comprises: A source module is used to obtain multi-source data; the multi-source data includes tower position, point cloud data, visual data, IMU inertial navigation data and positioning data; A map module is used to fuse the multi-source data to obtain an environmental map corresponding to the location of the UAV; A global waypoint module, configured to generate a safe waypoint sequence for the location of the UAV based on the environmental map using an improved RRT* algorithm; an obstacle avoidance module, configured to identify objects in the location of the drone based on the multi-source data and determine an obstacle list; The flight module is used to perform dynamic obstacle avoidance based on the safe waypoint sequence and the obstacle list, and generate a real-time flight path; the real-time flight path is used to instruct the UAV to perform an automatic line patrol mission.

Images