Automatic route planning method of unmanned aerial vehicle for electric power inspection

Abstract

The invention discloses an unmanned aerial vehicle route automatic planning method for electric power inspection, and relates to the technical field of unmanned aerial vehicle inspection. Comprising the following steps: starting an unmanned aerial vehicle, carrying out environment perception initialization, calculating the total cruise mileage, evaluating the interference risk, carrying out real-time obstacle avoidance, dynamically optimizing an inspection route, carrying out energy monitoring management, generating a return flight strategy, recording an inspection task and carrying out adaptive learning. Through dynamic electromagnetic interference modeling, multi-modal fusion perception, self-adaptive risk decision and cloud collaborative learning, the problem of insufficient adaptability of a traditional electric power inspection unmanned aerial vehicle in a complex electromagnetic environment and a dynamic obstacle scene is solved, the safety and the inspection efficiency are improved, the robustness is enhanced, and the method is suitable for popularization and application. Intelligent upgrading is carried out through continuous learning and multi-machine cooperation, the overall operation and maintenance cost of the system is reduced, a high-reliability and full-automatic inspection solution is provided for intelligent power grid construction, and the industrial application value is remarkable.

Description

Technical Field

[0001] The present invention relates to the technical field of drone inspection, and particularly to an automatic route planning method for drones used in power inspection. Background Art

[0002] The drone inspection of power system poles and towers is in a stage of rapid development and wide application, and the overall situation is good. With the improvement of drone performance and the development of artificial intelligence technology, multi-sensor fusion (such as equipped with high-definition cameras, thermal imagers, lidar, etc.) realizes multi-dimensional detection of pole and tower equipment. AI intelligent analysis can automatically identify defects such as insulator breakage and conductor strand breakage, greatly improving the detection accuracy and efficiency. The autonomous inspection of drones is widely popularized in the power industry, achieving large-area coverage of line poles and towers in airworthiness areas. In the future, it will develop towards higher intelligence and autonomy, deepen the integration with emerging technologies, and continuously improve the quality and efficiency of power pole and tower inspection.

[0003] However, the existing power inspection drones have problems of insufficient adaptability to electromagnetic interference, as well as resulting limitations in the ability to avoid dynamic obstacles and low efficiency in real-time path optimization. Since traditional drone inspection systems usually adopt a fixed safety distance threshold without considering the dynamic changes of electromagnetic interference, and the electromagnetic field intensity around high-voltage transmission lines changes in real time with voltage fluctuations and equipment load changes, the existing methods cannot dynamically adjust the safety distance, resulting in a risk of crashing due to insufficient safety redundancy in strong interference areas, or reducing the inspection efficiency due to excessive conservatism in low interference areas. And most of the existing obstacle avoidance algorithms are based on the assumption of static obstacles and cannot effectively deal with dynamic obstacles such as birds and temporary construction equipment. The single-modal perception of vision or lidar is easily interfered by the environment, resulting in missed detection or misdetection of obstacles, poor robustness of path planning, and lack of real-time response ability to sudden environmental changes such as sudden increase in electromagnetic interference and sudden appearance of obstacles.

[0004] In view of the above technical defects, a solution is proposed now. Summary of the Invention

[0005] The purpose of the present invention is to solve the problems of insufficient adaptability of existing power inspection drones to electromagnetic interference, as well as resulting limitations in the ability to avoid dynamic obstacles and low efficiency in real-time path optimization.

[0006] To achieve the above purpose, the present invention adopts the following technical solutions: An automatic route planning method for drones used in power inspection, comprising the following steps: S1. Start the drone and initialize environmental perception: Start the drone, execute the on-board system self-check, collect environmental data in real time, construct a three-dimensional dynamic electromagnetic interference model, and generate a safety distance correction coefficient in real time; S2. Evaluate interference risks and calculate the total cruising mileage: Dynamically correct the preset reference safety distance by collecting the coordinates and voltage levels of the poles and towers to be inspected and combining with the safety distance correction coefficient; Obtain the environmental risk coefficient by combining electromagnetic interference with meteorological parameters, and combine it with the electromagnetic interference intensity to dynamically correct the preset redundancy coefficient; Then obtain the initial path from the cruising starting coordinates and the pole and tower coordinates, calculate the detour path increment through the safety distance and obstacle coordinates, and then combine the initial path, detour path increment and redundancy coefficient to calculate the total cruising mileage; S3. Avoid obstacles in real time and dynamically optimize the inspection route: Detect dynamic obstacles in real time through lidar ranging and visual sensors, construct a prediction model of the obstacle movement trajectory, then generate a local obstacle avoidance path through the improved A* algorithm, and reassign the priority of the inspection tasks; S4. Energy monitoring and management and generate a return strategy: Set the safety mileage threshold Ls, obtain the real-time risk coefficient through weighted fusion of the electromagnetic interference level and the obstacle density, and dynamically refresh the safety mileage threshold, so as to control the return of the drone; S5. Record inspection tasks and perform adaptive learning.

[0007] Furthermore, the specific process of step one is as follows: Start the drone, execute the on-board system self-check, and confirm that the positioning device, electromagnetic interference sensor, lidar, image acquisition device and communication module are operating normally; If the self-check fails, trigger a fault alarm and terminate the task; Collect environmental data in real time, including the electromagnetic interference intensity distribution map, three-dimensional coordinates of obstacles and meteorological parameters, and the meteorological parameters include wind speed and temperature; Divide the electromagnetic interference level area through the electromagnetic data monitored by the electromagnetic interference sensor, and dynamically generate the correction coefficient of the electromagnetic safety distance.

[0008] Furthermore, the generation process of the electromagnetic interference intensity distribution map is as follows: Arrange a distributed electromagnetic sensor network within the flight range of the drone to transmit the electric field intensity of each area in real time; Use the Kriging interpolation algorithm to construct a three-dimensional electromagnetic field intensity model, and then divide the three-dimensional electromagnetic field intensity model into n0 flight areas; Set the electric field intensity threshold of the flight area, and mark the flight areas exceeding the threshold as high-risk avoidance areas.

[0009] Furthermore, the specific process of step two is as follows: S2-1, collect the coordinates and voltage levels of the poles and towers to be inspected; Set the reference value D0 of the safety distance, obtain the electromagnetic interference intensity E through the normalization of the division results of the electromagnetic interference level regions, and dynamically correct the reference safety distance D0 by combining the correction coefficient α of the electromagnetic safety distance, and mark the corrected safety distance as D1; S2-2, calculate the initial total cruising mileage: Generate a shortest path topology map through the cruising starting coordinates and the coordinates of the poles and towers, and obtain the initial path L0; Then calculate the detour path increment L1 through the safety distance D1 corrected by electromagnetic interference and the obstacle coordinate distribution; Obtain the environmental risk coefficient k through the weighted fusion of the obstacle density and meteorological parameters of the environment; Set the initial value β0 of the redundancy coefficient, and correct and refresh the initial value of the redundancy coefficient through the environmental risk coefficient k and the electromagnetic interference intensity E.

[0010] Furthermore, the specific process of step three is as follows: Real-time detect dynamic obstacles through the fusion positioning technology of lidar ranging and visual SLAM, and construct a prediction model for the movement trajectory of obstacles; Generate a local obstacle avoidance path through the improved A* algorithm, combining the electromagnetic interference intensity, the predicted trajectory of obstacles and the remaining cruising mileage; Through the prioritization of inspection tasks, select an alternative pole and tower sequence when re-planning the global path, and update the total cruising mileage of the UAV; Obtain through the weighted fusion of the voltage level and the historical failure probability, and re-allocate the priorities of the inspection tasks.

[0011] Furthermore, the specific steps for constructing a prediction model for the movement trajectory of obstacles are as follows: Synchronously collect multi-sensor data through lidar and visual sensors: Scan the environment at a fixed frequency through the lidar to obtain high-precision three-dimensional point cloud data, and mark the distance, azimuth and contour information of the obstacles; Synchronously capture environmental images through the visual sensor, obtain visual data and extract feature points and depth information to construct a dense point cloud; Strictly align the timestamps of the data of the lidar and the visual sensor; Data preprocessing and feature extraction: The lidar point cloud processing is to remove noise points through the voxel filtering and statistical outlier removal algorithms, then segment the point cloud through the Euclidean clustering algorithm, extract potential obstacle clusters, and calculate the centroid coordinates, bounding boxes and movement speeds of the obstacle clusters through the adjacent frame point cloud matching; Visual data processing detects dynamic targets in images in real time through the YOLOv7 algorithm, outputs the target category, bounding box, and confidence level, and generates a three-dimensional position estimate of the target; Model and predict the movement trajectory of dynamic obstacles: Estimate the motion state through Kalman filtering, obtain the position, velocity, and acceleration of the target obstacle, and output the predicted position and confidence interval of the obstacle within the next m seconds.

[0012] Furthermore, the specific content of improving the A* algorithm includes: Mark the actual cost of the UAV from the starting point to node i as G(i); mark the heuristic estimated cost as H(i); mark the electromagnetic interference intensity penalty term of node i as E(i); Thus, obtain the cost function F(i), and generate a local detour path in real time by minimizing the cost function F(i).

[0013] Furthermore, the specific process of step four is: Calculate the remaining endurance mileage Ly and the return flight requirement mileage Lx in real time; Set a safety mileage threshold Ls. When the difference between the remaining endurance mileage Ly and the return flight requirement mileage Lx is lower than the safety mileage threshold Ls, immediately interrupt the inspection task and return along the optimal path; Obtain the real-time risk coefficient G through weighted fusion of the electromagnetic interference level and the obstacle density; Then, dynamically refresh the initial preset value L0 of the safety mileage threshold through the real-time risk coefficient G to obtain the refreshed safety mileage threshold Ls.

[0014] Furthermore, the specific process of step five is: After the task is completed, upload the cruise flight data to the cloud. The cruise flight data includes the obstacle avoidance path, electromagnetic interference distribution, and algorithm decision records. By comparing the lengths and durations of the obstacle avoidance paths of different tasks, identify efficient obstacle avoidance patterns and achieve the adaptive learning of the UAV system.

[0015] In summary, due to the adoption of the above technical solutions, the beneficial effects of the present invention are: The present invention improves the inspection efficiency through dynamic electromagnetic interference modeling and safety distance optimization, generates differentiated safety strategies according to the tower voltage level, and adapts to complex power grid environments; through multi-modal fusion perception and dynamic obstacle avoidance, fuses lidar and visual SLAM for positioning, conducts obstacle target detection and trajectory prediction, realizes high-precision identification and motion modeling of dynamic obstacles, and improves the detection accuracy and obstacle avoidance success rate; The present invention dynamically optimizes the return path through adaptive environmental risk assessment and UAV energy management, and integrates and shares the obstacle avoidance experience of UAVs through cloud collaboration and adaptive learning, thereby identifying efficient obstacle avoidance modes, realizing the adaptive learning of the UAV system, and improving the obstacle avoidance efficiency and electromagnetic adaptability of subsequent tasks. In summary, the present invention solves the problem of insufficient adaptability of traditional power inspection UAVs in complex electromagnetic environments and dynamic obstacle scenarios through dynamic electromagnetic interference modeling, multi-modal fusion perception, adaptive risk decision-making, and cloud collaborative learning, improves safety and inspection efficiency, enhances robustness, and through continuous learning and multi-aircraft collaboration for intelligent upgrading, reduces the overall operation and maintenance cost of the system, provides a highly reliable and fully automatic inspection solution for the construction of smart grids, and has significant industrial application value. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 The schematic diagram of the steps showing the overall process of the present invention; Figure 2 The schematic diagram of the process of step two of the present invention is shown. DETAILED DESCRIPTION OF THE INVENTION

[0017] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0018] Embodiment 1: As Figure 1-2 shown, a method for automatically planning the flight path of a UAV for power inspection includes the following steps: S1. Start the UAV and initialize environmental perception: By starting the UAV and performing self-check of the on-board system, then collecting environmental data in real time, and constructing a three-dimensional model of dynamic electromagnetic interference, and generating a safety distance correction coefficient in real time. The specific process is as follows: S1-1. Start the UAV and perform self-check of the on-board system to confirm that the positioning device, electromagnetic interference sensor, lidar, image acquisition device, and communication module are operating normally; if the self-check fails, trigger a fault alarm and terminate the task. S1-2. Collect environmental data in real time, including the electromagnetic interference intensity distribution map, three-dimensional coordinates of obstacles, and meteorological parameters. Among them, the meteorological parameters include wind speed and temperature. The generation process of the electromagnetic interference intensity distribution map is as follows: Arrange a distributed electromagnetic sensor network within the flight range of the unmanned aerial vehicle (UAV) to transmit the electric field intensity of each area in real time; use the Kriging interpolation algorithm to construct a three-dimensional electromagnetic field intensity model, and then divide the three-dimensional electromagnetic field intensity model into n0 flight areas; set the electric field intensity threshold for the flight area, and mark the flight area exceeding the threshold as a high-risk avoidance area; S1-3, divide the electromagnetic interference level area through the electromagnetic data monitored by the electromagnetic interference sensor, and dynamically generate the correction coefficient of the electromagnetic safety distance; Among them, a dynamic three-dimensional electromagnetic interference model is constructed by the Kriging interpolation algorithm, and the safety distance correction coefficient α is generated in real time.

[0019] S2, calculate the total cruise mileage and evaluate the interference risk: dynamically correct the preset reference safety distance by collecting the coordinates and voltage levels of the towers to be inspected and combining with the safety distance correction coefficient; obtain the environmental risk coefficient by combining electromagnetic interference with meteorological parameters, and combine it with the electromagnetic interference intensity to dynamically correct the preset redundancy coefficient; then obtain the initial path through the cruise starting coordinate and the tower coordinate, calculate the detour path increment through the safety distance and the obstacle coordinate, and then combine the initial path, the detour path increment and the redundancy coefficient to calculate the total cruise mileage. The specific process is as follows: S2-1, collect the coordinates, voltage levels of the towers to be inspected and the preset reference safety distance; Dynamically correct the reference safety distance D0 through the division result of the electromagnetic interference level area, and mark the corrected safety distance as D1: , where E is the electromagnetic interference intensity, which is a normalized value of the electromagnetic interference intensity in the current area; D0 is the reference safety distance, which is preset in advance through past inspection experience; S2-2, calculate the initial total cruise mileage: Generate the shortest path topology diagram through the cruise starting coordinate and the tower coordinate, and obtain the initial path L0; Then calculate the detour path increment L1 through the safety distance D1 corrected by electromagnetic interference and the obstacle coordinate distribution; Calculate the total cruise mileage Lz: ; Among them, β is the redundancy coefficient, and the initial preset value β0 of the redundancy coefficient is corrected and refreshed through the environmental risk coefficient k and the electromagnetic interference intensity E to obtain the refreshed redundancy coefficient β: ; The environmental risk coefficient k is obtained through weighted assessment of the obstacle density, wind speed, and temperature in the cruise environment. The calculation formula needs to satisfy the logical relationship between the parameters, that is, the higher the amplitude of the obstacle density, wind speed, and temperature exceeding the preset parameter standard threshold, the higher the environmental risk coefficient k. And the preset weight factor for weighting here needs to make the calculated result k fall within .

[0020] Through dynamic electromagnetic interference modeling and safety distance optimization, the present invention normalizes the electromagnetic interference intensity and dynamically corrects the safety distance, thereby automatically increasing the safety distance in strong interference areas, reducing the out-of-control risk caused by electromagnetic interference, reducing redundancy in weak interference areas, improving the inspection efficiency, and generating differentiated safety strategies according to the tower voltage level to adapt to complex power grid environments.

[0021] S3. Real-time obstacle avoidance and dynamic optimization of the inspection route: Dynamically detect obstacles in real time through lidar ranging and visual sensors, construct a prediction model of the obstacle movement trajectory, and then generate a local obstacle avoidance path through an improved A* algorithm and reassign the priority of the inspection tasks. The specific process is as follows: S3-1. Through the fusion positioning technology of lidar ranging and visual SLAM, dynamically detect obstacles in real time, such as birds, so as to construct a prediction model of the obstacle movement trajectory. The specific steps are as follows: S3-101. Synchronously collect multi-sensor data through lidar and visual sensors: Scan the environment at a fixed frequency through lidar to obtain high-precision three-dimensional point cloud data, and mark the distance, azimuth, and contour information of the obstacles. For example, use solid-state or mechanical lidar; Synchronously capture environmental images through visual sensors to obtain visual data and extract feature points and depth information to construct a dense point cloud. Among them, the visual sensor is, for example, a binocular camera or an RGB-D camera; Strictly align the timestamps of the data of the lidar and the visual sensor, and control the error within the millisecond level; Through multi-modal fusion perception and dynamic obstacle avoidance, the present invention fuses and locates the lidar and visual SLAM, performs obstacle target detection and trajectory prediction, realizes high-precision recognition and motion modeling of dynamic obstacles, and comprehensively weighs distance, energy consumption, and electromagnetic interference when generating a bypass path, improving the detection accuracy and obstacle avoidance success rate.

[0022] S3-102. Data preprocessing and feature extraction: Lidar point cloud processing is to remove noise points through voxel filtering and statistical outlier removal algorithms, then segment the point cloud through the Euclidean clustering algorithm, extract potential obstacle clusters, and calculate the centroid coordinates, bounding boxes, and motion speeds of the obstacle clusters through adjacent frame point cloud matching; Visual data processing is to detect dynamic targets in images in real time through the YOLOv7 algorithm, output the target category, bounding box and confidence level, and generate a three-dimensional position estimate of the target.

[0023] S3-103, model and predict the motion trajectory of dynamic obstacles: Estimate the motion state through Kalman filtering to obtain the target state estimate, including the position, speed, and acceleration of the target obstacle, and output the predicted position and confidence interval of the obstacle within the next m seconds.

[0024] S3-2, by improving the A* algorithm, combining the electromagnetic interference intensity, the predicted trajectory of the obstacle, and the remaining battery life, generate a local obstacle avoidance path; when the detected electromagnetic interference intensity exceeds the threshold or when it is detected that the obstacle cannot be avoided, trigger a global path replanning; The specific content of improving the A* algorithm includes: Mark the actual cost of the UAV from the starting point to node i as G(i); Among them, the actual cost G(i) may be related to the flight distance, energy consumption, and time. When the UAV is flying, the cost of each path can be accumulated. For example, from the starting point to node i, it passes through several paths, and the cost of each path is the distance multiplied by the cost per unit distance, such as the energy consumption coefficient. By recording the flown path and calculating the accumulated actual consumption, G(i) can be obtained; Mark the heuristic estimated cost as H(i); Among them, the heuristic function H(i) is an estimate of the current node to the target node, calculated by the Euclidean distance, and the straight-line distance from node i on the map to the end point is multiplied by an estimated coefficient to obtain the shortest path estimate value from node i to the end point; Mark the electromagnetic interference intensity penalty term of node i as E(i); Among them, the electromagnetic interference intensity of each node is extracted through the electromagnetic interference intensity distribution map to obtain the electromagnetic interference intensity penalty term of node i; Thus, the cost function F(i) is obtained: ; Among them, γ is the weight coefficient of electromagnetic interference and γ is dynamically adjusted through the remaining battery life; Generate a local detour path in real time by minimizing the cost function F(i); it shows that during the path search process, nodes with low electromagnetic interference and sparse obstacles are preferentially selected for expansion.

[0025] S3-3, through the inspection task priority sorting, select an alternative tower sequence during the global path replanning to avoid high-risk areas, and then update the total cruise mileage of the UAV, thereby reallocating the inspection task priority; The priority of the inspection task is obtained by weighted fusion of the voltage level and the historical failure probability. The higher the voltage level and the historical failure probability of the tower, the higher the priority of the inspection task for that tower.

[0026] S4. Energy monitoring and management and generating a return strategy: Set a safety mileage threshold Ls. Through weighted fusion of the electromagnetic interference level and the obstacle density, obtain the real-time risk coefficient and dynamically refresh the safety mileage threshold, so as to control the UAV to return. The specific process is as follows: By calculating the remaining endurance mileage Ly and the return demand mileage Lx in real time; Set the safety mileage threshold Ls. When the difference between the remaining endurance mileage Ly and the return demand mileage Lx is lower than the safety mileage threshold Ls, immediately interrupt the inspection task and return along the optimal path; Through weighted fusion of the electromagnetic interference level and the obstacle density, obtain the real-time risk coefficient G; Preset the weight factors for weighting the electromagnetic interference level and the obstacle density, and it is required to satisfy that the calculated result G takes values within ; Then, dynamically refresh the initial preset value L0 of the safety mileage threshold through the real-time risk coefficient G to obtain the refreshed safety mileage threshold Ls: , where ε is an adjustment coefficient and is preset through historical cruise experience. When the environmental danger level is higher, the preset value of the adjustment coefficient increases, so as to automatically increase the redundancy coefficient and extend the total cruise mileage reserve in a high-risk environment; The present invention avoids the delay of global replanning through local path real-time optimization, and through adaptive environmental risk assessment and energy management, thereby dynamically refreshing the redundancy coefficient and the safety mileage threshold, adjusting the return strategy in real time, dynamically optimizing the return path, and avoiding the risk of staying in the air caused by misjudgment of the battery power; Continuously monitor electromagnetic interference and obstacles during the return process, and dynamically adjust the UAV landing coordinates.

[0027] S5. Record the inspection task and perform adaptive learning: When the task is over, upload the cruise flight data to the cloud. The cruise flight data includes the obstacle avoidance path, electromagnetic interference distribution, and algorithm decision records. By comparing the lengths and durations of the obstacle avoidance paths of different tasks, identify efficient obstacle avoidance modes, realize the adaptive learning of the UAV system, improve the obstacle avoidance efficiency and electromagnetic adaptability of subsequent tasks, and provide a highly reliable and fully automatic inspection solution for the construction of smart grids, with significant industrial application value; Through cloud collaboration and adaptive learning, the present invention integrates and shares the obstacle avoidance experience of drones, automatically identifies fault modes, generates preventive maintenance strategies, and through the collaborative design of anti-interference hardware and algorithms, ensures stable operation in a strong interference environment, guarantees the integrity rate of data transmission, and reduces the hardware failure rate.

[0028] In summary, the beneficial effects of the present invention are as follows: Through dynamic electromagnetic interference modeling and safety distance optimization, the present invention improves the inspection efficiency, generates differentiated safety strategies according to the tower voltage level, and adapts to complex power grid environments; through multi-modal fusion perception and dynamic obstacle avoidance, it fuses lidar and visual SLAM for positioning, conducts obstacle target detection and trajectory prediction, realizes high-precision recognition and motion modeling of dynamic obstacles, and improves the detection accuracy and obstacle avoidance success rate. Through adaptive environmental risk assessment and drone energy management, the present invention dynamically optimizes the return path. Through cloud collaboration and adaptive learning, it integrates and shares the obstacle avoidance experience of drones, thereby identifying efficient obstacle avoidance modes, realizing the adaptive learning of the drone system, and improving the obstacle avoidance efficiency and electromagnetic adaptability of subsequent tasks. Through dynamic electromagnetic interference modeling, multi-modal fusion perception, adaptive risk decision-making, and cloud collaborative learning, the present invention solves the problem of insufficient adaptability of traditional power inspection drones in complex electromagnetic environments and dynamic obstacle scenarios, improves safety and inspection efficiency, enhances robustness, and through continuous learning and multi-aircraft collaboration for intelligent upgrading, reduces the overall system operation and maintenance cost, providing a highly reliable and fully automatic inspection solution for the construction of smart grids, and having significant industrial application value.

[0029] The above embodiments can be implemented in whole or in part by software, hardware, firmware, or any other combination. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. Those of ordinary skill in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution.

[0030] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place, or may be distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0031] The above data processing is to remove the dimension and take its numerical calculation. The settings of the interval and the threshold size are for the convenience of comparison. Regarding the threshold size, it depends on the amount of sample data and the number of base quantities set by those skilled in the art for each group of sample data, as long as it does not affect the proportional relationship between the parameters and the quantified values. The preset parameters are set by those skilled in the art according to the actual situation.

[0032] The above is only a preferred specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention, according to the technical solution and inventive concept of the present invention, makes equivalent substitutions or changes, and should be covered by the protection scope of the present invention.

Claims

1. An automatic route planning method for drones used in power inspection, characterized in that: It includes the following steps: S1. Start the drone and initialize environmental perception: Start the drone and perform self-check of the on-board system, then collect environmental data in real time, construct a dynamic three-dimensional model of electromagnetic interference, and generate a safety distance correction coefficient in real time; S2. Evaluate interference risks and calculate the total cruise mileage: Dynamically correct the preset reference safety distance by collecting the coordinates and voltage levels of the towers to be inspected and combining with the safety distance correction coefficient; Obtain the environmental risk coefficient by combining electromagnetic interference with meteorological parameters, and combine it with the electromagnetic interference intensity to dynamically correct the preset redundancy coefficient; Then obtain the initial path from the cruise starting coordinates and the tower coordinates, calculate the detour path increment through the safety distance and obstacle coordinates, and then combine the initial path, detour path increment and redundancy coefficient to calculate the total cruise mileage; S3. Avoid obstacles in real time and dynamically optimize the inspection route: Detect dynamic obstacles in real time through lidar ranging and vision sensors, construct a prediction model of the obstacle movement trajectory, then generate a local obstacle avoidance path through the improved A* algorithm, and reassign the priorities of inspection tasks; S4. Energy monitoring and management and generate a return strategy: Set the safety mileage threshold Ls, obtain the real-time risk coefficient through weighted fusion of the electromagnetic interference level and obstacle density, and dynamically refresh the safety mileage threshold, so as to control the return of the drone; S5. Record inspection tasks and perform adaptive learning.

2. The automatic flight path planning method for an unmanned aerial vehicle used in power inspection according to claim 1, wherein: The specific process of step one is as follows: Start the drone, perform self-check of the on-board system, and confirm that the positioning device, electromagnetic interference sensor, lidar, image acquisition device and communication module are operating normally; If the self-check fails, trigger a fault alarm and terminate the task; Collect environmental data in real time, including the electromagnetic interference intensity distribution map, three-dimensional coordinates of obstacles and meteorological parameters, and the meteorological parameters include wind speed and temperature; Divide the electromagnetic interference level area through the electromagnetic data monitored by the electromagnetic interference sensor, and dynamically generate the correction coefficient of the electromagnetic safety distance.

3. The automatic flight path planning method for an unmanned aerial vehicle used in power inspection according to claim 2, wherein: The generation process of the electromagnetic interference intensity distribution map is as follows: Arrange a distributed electromagnetic sensor network within the flight range of the drone to transmit the electric field strength of each area in real time; Use the Kriging interpolation algorithm to construct a three-dimensional electromagnetic field strength model, and then divide the three-dimensional electromagnetic field strength model into n0 flight areas; Set the electric field strength threshold of the flight area, and mark the flight area exceeding the threshold as a high-risk avoidance area.

4. The automatic route planning method for an unmanned aerial vehicle used in power inspection according to claim 3, wherein: The specific process of step two is as follows: S2-1. Collect the coordinates and voltage levels of the towers to be inspected; Set the reference value D0 of the safety distance, obtain the electromagnetic interference intensity E through normalization of the division results of the electromagnetic interference level area, and combine it with the correction coefficient α of the electromagnetic safety distance to dynamically correct the reference safety distance D0, and mark the corrected safety distance as D1; S2-2. Calculate the initial total cruise mileage: Generate a shortest path topology map from the cruise starting coordinates and the tower coordinates to obtain the initial path L0; Then calculate the detour path increment L1 through the safety distance D1 corrected by electromagnetic interference and the obstacle coordinate distribution; Obtain the environmental risk coefficient k through weighted fusion of the obstacle density and meteorological parameters of the environment; Set the initial value of the redundancy coefficient β0, and correct and refresh the initial value of the redundancy coefficient through the environmental risk coefficient k and the electromagnetic interference intensity E.

5. The automatic route planning method for an unmanned aerial vehicle used for power inspection according to claim 4, characterized in that: The specific process of step three is as follows: Through the fusion technology of lidar ranging and visual positioning SLAM, dynamically detect obstacles in real time and build a prediction model for the movement trajectory of obstacles; Through the improved A* algorithm, combined with the electromagnetic interference intensity, the predicted trajectory of obstacles, and the remaining battery life, generate a local obstacle avoidance path; Through the prioritization of inspection tasks, select an alternative tower sequence during the global path replanning to update the total cruising mileage of the UAV; Obtained through the weighted fusion of voltage levels and historical failure probabilities, and reallocate the priorities of inspection tasks.

6. The automatic flight path planning method for a drone used in power inspection according to claim 5, characterized in that: The specific steps for building a prediction model for the movement trajectory of obstacles are as follows: Synchronously collect multi-sensor data through lidar and visual sensors: Scan the environment at a fixed frequency through lidar to obtain high-precision three-dimensional point cloud data, and mark the distance, azimuth, and contour information of obstacles; Synchronously capture environmental images through visual sensors, obtain visual data, extract feature points and depth information, and build a dense point cloud; Strictly align the timestamps of the data from the lidar and the visual sensor; Data preprocessing and feature extraction: Lidar point cloud processing is to remove noise points through voxel filtering and statistical outlier removal algorithms, then segment the point cloud through the Euclidean clustering algorithm, extract potential obstacle clusters, and calculate the centroid coordinates, bounding boxes, and movement speeds of obstacle clusters through adjacent frame point cloud matching; Visual data processing is to use the YOLOv7 algorithm to detect dynamic targets in images in real time, output the target category, bounding box, and confidence level, and generate a three-dimensional position estimate of the target; Model and predict the movement trajectory of dynamic obstacles: Through Kalman filtering for motion state estimation, obtain the position, speed, and acceleration of the target obstacle, and output the predicted position and confidence interval of the obstacle within the next m seconds.

7. A method for automatically planning the flight path of an unmanned aerial vehicle for power inspection according to claim 6, characterized in that: The specific content of the improved A* algorithm includes: Mark the actual cost of the UAV from the starting point to node i as G(i); mark the heuristic estimated cost as H(i); mark the electromagnetic interference intensity penalty term of node i as E(i); Thus obtain the cost function F(i), and generate a local detour path in real time by minimizing the cost function F(i).

8. A method for automatically planning the flight path of an unmanned aerial vehicle for power inspection according to claim 7, characterized in that: The specific process of step four is as follows: By calculating the remaining battery life Ly and the return demand mileage Lx in real time; Set a safety mileage threshold Ls. When the difference between the remaining battery life Ly and the return demand mileage Lx is lower than the safety mileage threshold Ls, immediately interrupt the inspection task and return along the optimal path; Through the weighted fusion of the electromagnetic interference level and the obstacle density, obtain the real-time risk coefficient G; Then dynamically refresh the initial preset value L0 of the safety mileage threshold through the real-time risk coefficient G to obtain the refreshed safety mileage threshold Ls.

9. The automatic route planning method for an unmanned aerial vehicle used in power inspection according to claim 8, wherein: The specific process of step five is as follows: After the task is completed, upload the cruise flight data to the cloud. The cruise flight data includes the obstacle avoidance path, electromagnetic interference distribution, and algorithm decision records. By comparing the lengths and time-consuming of the obstacle avoidance paths of different tasks, identify efficient obstacle avoidance modes and achieve the adaptive learning of the UAV system.

Images