Unmanned aerial vehicle out-of-communication return flight control method and system

Abstract

The invention provides an unmanned aerial vehicle lost flight return control method and system, and relates to the technical field of unmanned aerial vehicle flight control, and the method comprises the steps: starting the real-time sensing of the environment in front of the flight, obtaining the environment information of a dynamic obstacle, and generating a dynamic obstacle map after the unmanned aerial vehicle is lost; through the dynamic obstacle map, fusing pre-stored static topographic data and preset task route data to obtain a preliminary optimized return route; flying along the preliminarily optimized return path, selecting three specific high-voltage power transmission towers as fixed reference objects along the route, and establishing a spatial reference frame based on the spatial reference positions of the three high-voltage power transmission towers; and key feature point sets of the high-voltage power transmission tower body and the power transmission line in the spatial reference frame are extracted. According to the invention, safe obstacle avoidance, smooth path and accurate returning to the initial take-off and landing point are realized.

Description

Technical Field

[0001] This invention relates to the field of unmanned aerial vehicle (UAV) flight control technology, and in particular to a method and system for controlling the return of a lost UAV. Background Technology

[0002] As drones are increasingly used in complex scenarios such as power line inspection and mountain operations, the problem of losing contact occurs frequently. These problems are mostly caused by signal blockage, electromagnetic interference, or severe weather. Traditional return-to-home technologies rely on single inertial navigation, which is difficult to adapt to the needs of complex environments.

[0003] A power line inspection drone, while performing a mission in a mountainous area, experienced a signal interruption that triggered a return-to-home error. Traditional systems rely solely on preset paths and inertial navigation data, failing to perceive dynamic obstacles along the route in real time and lacking reliable ground references relevant to the operational scenario to correct positioning deviations. Ultimately, this resulted in the return path deviating from the predetermined route, failing to effectively avoid sudden obstacles, and struggling to accurately return to the initial take-off and landing points. This case highlights the core flaws of traditional methods: insufficient perception of the dynamic environment, easy accumulation of positioning errors, and a lack of sufficient integration with fixed infrastructure in the operational scenario to construct positioning references, leading to insufficient safety and accuracy in returning to home in complex environments. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide a method and system for controlling the return of a drone after losing contact, so as to achieve obstacle avoidance safety, smooth path and accurate return to the initial take-off and landing point.

[0005] To solve the above-mentioned technical problems, the technical solution of the present invention is as follows: Firstly, a method for controlling the return of a drone after losing contact, the method comprising: After the drone loses contact, it initiates real-time perception of the environment ahead of the flight, acquires environmental information of dynamic obstacles, and generates a dynamic obstacle map. By using a dynamic obstacle map and integrating pre-stored static terrain data with preset mission route data, a preliminary optimized return route is obtained. The aircraft flew along the initially optimized return route, selecting three specific high-voltage transmission towers as fixed reference points along the route. A spatial reference frame was established based on the spatial reference positions of the three high-voltage transmission towers. The key feature point set of the high-voltage transmission tower body and transmission line within the spatial reference frame was extracted. By performing coverage processing on the feature point set, a set of fixed-radius circles covering the key feature points is obtained; based on the coverage range, spatial location of the center of each circle and the distribution characteristics of the corresponding covered feature points in the fixed-radius circle set, the positioning correction coefficient of each region is calculated; simultaneously, the inherent characteristics of the high-voltage transmission lines along the route are identified through airborne sensors, and the identification results are obtained. The identification results are matched with a pre-stored feature geographic coordinate database, and the cumulative positioning error of inertial navigation is comprehensively corrected by the positioning correction coefficient to obtain the corrected accurate location information. Based on the corrected precise location information, the initially optimized return path is dynamically updated and the trajectory is replanned. The drone is controlled to follow the updated real-time return path until it lands precisely at the initial take-off and landing point.

[0006] Furthermore, after the drone loses contact, real-time perception of the environment ahead is initiated to acquire environmental information about dynamic obstacles and generate a dynamic obstacle map, including: By receiving the drone's disconnection status signal and verifying the duration of the disconnection status, the system confirms entry into the disconnection emergency mode when the duration exceeds a preset threshold. Based on the disconnection emergency mode, a real-time environmental perception activation command is sent to the drone, and the environmental data processing function is activated simultaneously to obtain the uploaded raw environmental perception data stream. Receive the uploaded raw environmental perception data stream, perform spatiotemporal alignment, noise filtering, and data integrity verification on the raw data to obtain a clean environmental perception dataset; Based on a clean environmental perception dataset, a dynamic obstacle feature vector is obtained by extracting the motion parameters, spatial distribution density, and threat level assessment values ​​of moving objects through a dynamic obstacle recognition algorithm. The feature vectors of dynamic obstacles are fused with a pre-stored geographic information database to construct a three-dimensional dynamic obstacle map that includes the real-time location of obstacles, predicted motion trajectories, and obstacle avoidance safety boundaries.

[0007] Furthermore, by using a dynamic obstacle map and integrating pre-stored static terrain data with preset mission route data, a preliminary optimized return route is obtained, including: By using a three-dimensional dynamic obstacle map and transforming the real-time location, predicted motion trajectory, and obstacle avoidance safety boundary of the obstacles into a spatial coordinate system, obstacle distribution data in a unified coordinate system is obtained. Based on obstacle distribution data in a unified coordinate system, the pre-stored static terrain database is retrieved to extract terrain elevation information, land cover type, and no-fly zone boundaries, thus obtaining static terrain constraints. Based on static terrain constraints, load the preset mission route data, extract the take-off and landing point coordinates, mission waypoint sequence and route altitude parameters to obtain the original mission route constraint information; The obstacle distribution data, static terrain constraints and original mission route constraints are fused from multiple sources. The obstacle avoidance safety coefficient, energy consumption optimization index and path smoothness parameter of each flight segment are calculated by the cost function to obtain the comprehensive optimization evaluation matrix. Based on the comprehensive optimization evaluation matrix, a preliminary optimized return path is obtained by planning the path node sequence in three-dimensional space.

[0008] Furthermore, the flight proceeds along the initially optimized return path, selecting three specific high-voltage transmission towers as fixed reference points along the route. A spatial reference frame is established based on the spatial reference positions of these three towers. Key feature points of the high-voltage transmission towers and transmission lines within the spatial reference frame are extracted, including: Receive preliminary optimized return route data, combine it with the pre-stored power facility geographic information database, perform spatial analysis on the distribution of high-voltage transmission towers along the route, and screen out all candidate high-voltage transmission towers located within the preset width range of the route; Based on the screening of candidate high-voltage transmission towers, three final high-voltage transmission towers are selected from the candidate set as fixed references according to tower height, structural stability, mutual spacing and visibility conditions, and reference selection decision data are obtained. Based on the reference selection decision data, the precise geographic coordinate information of the three selected high-voltage transmission towers is retrieved. The geographic coordinates are converted into a unified three-dimensional spatial coordinate system through a coordinate transformation algorithm, and a spatial reference frame based on the positions of the three towers is established. Based on the established spatial reference frame, geometric feature analysis is performed on the main structure of the high-voltage transmission tower and the connecting transmission lines within the frame. Tower corners, insulator suspension points, conductor connection points and line intersections are extracted as feature points to obtain a set of key feature points distributed in space.

[0009] Furthermore, by performing coverage processing on the feature point set, a set of fixed-radius circles covering the key feature points is obtained; based on the coverage area, spatial location of the center of each circle, and distribution characteristics of the corresponding covered feature points in the fixed-radius circle set, the positioning correction coefficient for each region is calculated; simultaneously, the inherent characteristics of the high-voltage transmission lines along the route are identified using airborne sensors, and the identification results are obtained, including: Based on the set of key feature points in spatial distribution, the spatial coordinates of the key feature points are normalized to obtain standardized feature point data. Based on standardized feature point data, a state compression dynamic programming algorithm with fixed radius circle coverage is used to iteratively calculate the coverage of feature points, thereby obtaining a set of fixed radius circles covering key feature points and the corresponding center coordinates of the circles. Based on the coverage area, spatial coordinates of the center of each circle in the set of fixed-radius circles, and density distribution characteristics of the corresponding coverage feature points, the geometric stability and feature saliency of the coverage area of ​​each circle are analyzed by weighted calculation method, and the positioning correction coefficient matrix of each area is obtained. While calculating the positioning correction coefficient, real-time high-voltage transmission line sensing data is received. Based on the established spatial reference frame, the sensing data is spatially positioned. The topology, insulator arrangement pattern, and conductor phase sequence characteristics of the transmission line are extracted through image processing algorithms to obtain the inherent feature recognition results of the high-voltage transmission line.

[0010] Furthermore, the identification results are matched with a pre-stored feature geographic coordinate database, and the cumulative positioning error of the inertial navigation is comprehensively corrected by combining the positioning correction coefficient to obtain the corrected accurate location information, including: Based on the comprehensive feature matching dataset, the inherent feature recognition results and the localization correction coefficient matrix are separated from it; Based on the identified inherent features, feature matching calculations are performed with the pre-stored feature geographic coordinate database. By comparing similarity, the current high-voltage transmission line section and corresponding reference geographic coordinates of the UAV are determined. Based on the reference geographic coordinates and combined with the positioning correction coefficient matrix, error compensation calculation is performed on the current position coordinates output by the UAV's inertial navigation to eliminate the cumulative positioning error caused by sensor drift and environmental interference, and obtain the coordinate data after error compensation. The error-compensated coordinate data is fused and weighted with the baseline geographic coordinates to obtain the corrected accurate location information.

[0011] Furthermore, based on the corrected precise location information, the initially optimized return path is dynamically updated and its trajectory is replanned. The drone is controlled to follow the updated real-time return path until it lands precisely at the initial take-off and landing point, including: Based on the corrected precise location information, the precise location information is compared with the preliminary optimized return path to calculate the spatial deviation between the current UAV position and the preset waypoint. Based on the spatial deviation, and combined with the latest position status of obstacles in the 3D dynamic obstacle map, the preliminary optimized return path is locally dynamically corrected to obtain the updated real-time return path node sequence. Based on the updated real-time return path node sequence, a smooth and continuous flight trajectory is obtained through spline curve interpolation algorithm, and the velocity constraints and attitude adjustment parameters on the trajectory are calculated to obtain complete trajectory replanning data. Based on trajectory replanning data, the flight status of the UAV along the updated return path is monitored in real time, and the flight trajectory is adjusted in a closed loop based on the received real-time position feedback data. As the drone approaches the initial take-off and landing point, it executes a high-precision landing control strategy based on precise position information and onboard visual sensor data. Through position fine-tuning and altitude descent control, the drone completes a precise automatic landing at the initial take-off and landing point.

[0012] Secondly, the drone loss return control system includes: The acquisition module is used to initiate real-time perception of the environment ahead of the drone after it loses contact, acquire environmental information of dynamic obstacles, and generate a dynamic obstacle map. The fusion module is used to merge pre-stored static terrain data with preset mission route data through a dynamic obstacle map to obtain a preliminary optimized return route; The extraction module is used to fly along the initially optimized return path, select three specific high-voltage transmission towers as fixed reference objects along the route, establish a spatial reference frame based on the spatial reference positions of the three high-voltage transmission towers, and extract the key feature point set of the high-voltage transmission tower body and transmission line within the spatial reference frame; The identification module is used to obtain a set of fixed-radius circles covering key feature points by performing coverage processing on the feature point set; based on the coverage range, spatial position of the center of each circle and the distribution characteristics of the corresponding covered feature points in the fixed-radius circle set, the positioning correction coefficient of each region is calculated; simultaneously, the inherent features of the high-voltage transmission lines along the route are identified by airborne sensors to obtain the identification results. The correction module is used to match the recognition results with the pre-stored feature geographic coordinate database, and combine the positioning correction coefficient to comprehensively correct the cumulative positioning error of inertial navigation, so as to obtain the corrected accurate location information. The processing module is used to dynamically update and replan the preliminary optimized return path based on the corrected precise location information, and control the UAV to follow the updated real-time return path until it lands precisely at the initial take-off and landing point.

[0013] Thirdly, a computing device, comprising: One or more processors; A storage device for storing one or more programs that, when executed by one or more processors, cause the one or more processors to implement the method.

[0014] Fourthly, a computer-readable storage medium storing a program that, when executed by a processor, implements the method.

[0015] The above-described solution of the present invention has at least the following beneficial effects: Because this invention employs dynamic environmental perception to construct a three-dimensional dynamic obstacle map, integrates dynamic obstacle data with static terrain and mission route data to plan a preliminary optimized return path, selects three high-voltage transmission towers as fixed reference objects to establish a spatial reference frame and extract key feature points, calculates positioning correction coefficients through a fixed-radius circle coverage algorithm and simultaneously identifies the inherent characteristics of transmission lines for matching dual positioning error correction, and uses precise location information combined with the latest obstacle status for dynamic path updates and flight closed-loop control, it effectively overcomes the technical problems of insufficient dynamic environmental perception, easy accumulation of positioning errors, and low return safety and accuracy due to reliance on satellite signals in traditional UAV disconnection return technology in complex operating scenarios. Thus, it achieves the technical effect of obstacle avoidance safety, energy consumption optimization, smooth path, and accurate return to the initial take-off and landing point after UAV disconnection in complex environments such as signal blockage and electromagnetic interference. Attached Figure Description

[0016] Figure 1 This is a flowchart illustrating the unmanned aerial vehicle (UAV) loss-of-connection return-to-home control method provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of a drone loss return control system provided in an embodiment of the present invention; Figure 3 This is a schematic diagram of a computing device. Detailed Implementation

[0017] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0018] like Figure 1 As shown, an embodiment of the present invention proposes a method for controlling the return of a drone after losing contact, the method comprising the following steps: Step 1: After the drone loses contact, initiate real-time perception of the environment ahead of the flight, obtain environmental information of dynamic obstacles, and generate a dynamic obstacle map; Step 2: By using a dynamic obstacle map, the pre-stored static terrain data and the preset mission route data are integrated to obtain a preliminary optimized return route; Step 3: Fly along the initially optimized return path, select three specific high-voltage transmission towers as fixed reference points along the route, and establish a spatial reference frame based on the spatial reference positions of the three high-voltage transmission towers; extract the key feature point set of the high-voltage transmission tower body and transmission line within the spatial reference frame; Step 4: By performing coverage processing on the feature point set, a set of fixed-radius circles covering the key feature points is obtained; based on the coverage range, spatial position of the center of each circle and the distribution characteristics of the corresponding covered feature points in the fixed-radius circle set, the positioning correction coefficient of each region is calculated; simultaneously, the inherent characteristics of the high-voltage transmission lines along the route are identified through airborne sensors to obtain the identification results. Step 5: Match the recognition results with the pre-stored feature geographic coordinate database, and combine the positioning correction coefficient to comprehensively correct the cumulative positioning error of inertial navigation to obtain the corrected accurate location information. Step 6: Based on the corrected precise location information, dynamically update and replan the preliminary optimized return path, and control the UAV to follow the updated real-time return path until it lands precisely at the initial take-off and landing point.

[0019] In this embodiment of the invention, because the invention employs the following technical means: real-time perception of the forward environment after the UAV loses contact to generate a dynamic obstacle map; integration of the dynamic obstacle map with pre-stored static terrain data and preset mission route data to plan a preliminary optimized return path; selection of three high-voltage transmission towers as fixed reference objects to establish a spatial reference frame and extract key feature point sets; calculation of positioning correction coefficients through fixed radius circle coverage processing and simultaneous identification and matching of inherent features of transmission lines; and dynamic updating of the return path based on the corrected accurate position information and closed-loop control of flight, the invention effectively overcomes the technical problems of poor dynamic environment adaptability, easy accumulation of positioning errors, and lack of scenario-based reliable positioning benchmarks in traditional UAV loss return technology, which lead to insufficient return safety and accuracy. This achieves the technical effect of safely avoiding dynamic obstacles, optimizing flight energy consumption, maintaining a smooth path, and accurately returning to the initial take-off and landing point without relying on satellite signals after the UAV loses contact in complex environments.

[0020] In a preferred embodiment of the present invention, step 1 above may include: Step 1.1 involves receiving the drone's disconnection status signal and verifying the duration of the disconnection. When the duration exceeds a preset threshold, the drone is confirmed to enter the emergency disconnection mode. Specifically, the drone receives flight status signals in real time through its onboard communication equipment, continuously monitoring parameters related to the communication link connection status and satellite positioning signal reception status. When the communication link connection status parameter fails to transmit for three consecutive frames, and the satellite positioning signal reception status parameter meets the criteria of fewer than four visible satellites and a positioning accuracy greater than 10 meters, both parameters are deemed to be in an abnormal state. This is recorded as the start time of the disconnection status, and the duration of disconnection is accumulated from this start time through the onboard timing module. Simultaneously, considering the common scenarios of instantaneous signal interruptions caused by mountainous terrain obstruction and electromagnetic interference from high-voltage lines during power line inspection operations, the criteria for determining the preset disconnection judgment threshold are clarified. One reference is... The time requirements for emergency response to communication interruptions are as follows: first, historical data on instantaneous signal interruptions in power inspection scenarios, typically lasting 1 to 2 seconds; second, the sensor warm-up and data initialization time after the drone's emergency mode is activated, approximately 1 second. Based on this, a reasonable threshold range for determining disconnection is preset to 3 to 5 seconds. Users can adjust the specific value according to the signal interference intensity of the specific inspection area. For example, 5 seconds can be set for mountainous areas with strong signal interference, while 3 seconds can be set for suburban plains inspection areas with relatively good signal conditions. The cumulative duration of disconnection is continuously compared with the preset threshold in real time. When the cumulative time exceeds the preset threshold and no normal transmission of the communication link is detected during the period, the uplink command transmission success rate is ≥95%, the downlink data feedback integrity is ≥98%, the satellite positioning signal meets the positioning accuracy requirements, the number of visible satellites is ≥6, and the calculation accuracy is ≤5 meters, the drone is officially confirmed to have entered the emergency mode for disconnection.

[0021] Step 1.2: Based on the out-of-connection emergency mode, send an environmental real-time perception activation command to the UAV and simultaneously activate the environmental data processing function to obtain the uploaded raw environmental perception data stream. Specifically, after confirming entry into the out-of-connection emergency mode, send an environmental real-time perception activation command to the UAV via a preset emergency communication link, using the 433MHz anti-interference frequency band with a communication distance ≥ 5 kilometers. This command includes specific instruction information such as sensor activation sequence and sampling parameter configuration, explicitly triggering the synchronous activation of the airborne lidar, visual sensor, millimeter-wave radar, and meteorological sensor. The preset sampling frequencies of each sensor are determined based on the needs of mountainous environmental detection: lidar sampling frequency is set to 100kHz, visual sensor sampling frequency is set to 30 frames per second, and millimeter-wave radar sampling frequency is set to... The sampling frequency of the meteorological sensor is set to 10Hz at 50Hz. At the same time, the built-in data processing unit of the UAV is activated to enter a high-speed operation state. Each sensor collects meteorological data such as terrain undulation data, various obstacle outline data, wind speed and direction, and visibility within a 100-meter range in front of the flight according to the preset sampling frequency. The collected raw data is packaged into a standardized data stream in the order of timestamps. The data frame format includes sensor identification, sampling time, data length, raw value, and check code. It is uploaded to the ground end in real time at a transmission rate of 1Mbps through the emergency communication link. The ground end simultaneously activates the data receiving and preprocessing functions, adopts a dual-buffer receiving mechanism to avoid data loss, and verifies the transmission integrity of each frame of data through the check code to ensure that the raw environmental perception data stream is received completely and without transmission loss.

[0022] Step 1.3: Receive the uploaded raw environmental sensing data stream and perform spatiotemporal alignment, noise filtering, and data integrity verification on the raw data to obtain a clean environmental sensing dataset. Specifically, after receiving the uploaded raw environmental sensing data stream, firstly, spatiotemporal alignment is performed on data from different sources based on the sampling timestamps of each sensor, so that data from multiple sensors at the same time and spatial location form a corresponding association. Then, an adaptive Kalman filter algorithm is used to filter the data for noise, removing invalid noise caused by mountain airflow disturbances, sensor errors, and electromagnetic interference. Next, the filtered data stream is subjected to frame-by-frame integrity verification, checking the field integrity, numerical rationality, and transmission check code consistency of each frame. For missing fields, interpolation between adjacent valid frames is used to complete the data. Invalid frames that cannot be repaired are marked and removed, ultimately forming a clean environmental sensing dataset that is temporally continuous, numerically reliable, and free of redundant noise.

[0023] Step 1.4: Based on a clean environmental perception dataset, extract the motion parameters, spatial distribution density, and threat level assessment values ​​of moving objects using a dynamic obstacle recognition algorithm to obtain dynamic obstacle feature vectors. Specifically, this includes: using a dynamic obstacle recognition algorithm that combines target detection and optical flow tracking based on a clean environmental perception dataset, accurately identifying and tracking moving objects in the data, extracting real-time motion parameters for each dynamic obstacle, including movement speed, acceleration, movement direction, and turning angular velocity; simultaneously, statistically analyzing the spatial distribution density of dynamic obstacles within a certain range ahead of the flight path according to a three-dimensional grid division method to clarify the obstacle density in different areas; and then, combining the obstacle's size, movement speed, and the probability of intersection with the UAV's return path, setting multi-dimensional threat assessment rules to obtain the threat level assessment value of each obstacle for the UAV's return; finally, integrating the motion parameters, spatial distribution location, and threat level assessment value of each dynamic obstacle into a standardized dynamic obstacle feature vector.

[0024] Step 1.5 involves fusing the dynamic obstacle feature vectors with a pre-stored geographic information database to construct a 3D dynamic obstacle map containing the real-time location of obstacles, predicted motion trajectories, and obstacle avoidance safety boundaries. Specifically, this includes: retrieving a pre-stored mountainous geographic information database, which contains basic geographic data such as terrain elevation, surface attachments, distribution of high-voltage transmission towers, and no-fly zones in the power inspection area; fusing the dynamic obstacle feature vectors with the geographic information database using a unified 3D spatial coordinate system; firstly, mapping the real-time location coordinates of the dynamic obstacles to the spatial framework of the geographic information database; then, calculating the predicted motion trajectory for the next 5 to 10 seconds based on the obstacle's motion parameters; and simultaneously setting a 3 to 5-meter obstacle avoidance safety boundary for each obstacle based on the UAV's fuselage size, flight speed, and obstacle avoidance response time. Finally, a 3D dynamic obstacle map is constructed that includes the real-time location of dynamic obstacles, predicted motion trajectories, and obstacle avoidance safety boundaries, and is accurately superimposed with the mountainous geographic environment and power facility distribution.

[0025] In this embodiment of the invention, because the invention employs the following technical means: first verifying the duration of the loss of contact to accurately trigger the emergency mode, then initiating real-time environmental perception and activating data processing functions, performing spatiotemporal alignment, noise filtering, and data integrity verification on the original environmental perception data stream, extracting multi-dimensional feature vectors of moving objects through a dynamic obstacle recognition algorithm, and fusing a pre-stored geographic information database to construct a three-dimensional dynamic obstacle map containing the real-time location of obstacles, predicted motion trajectories, and obstacle avoidance safety boundaries, the invention effectively overcomes the technical problems of traditional UAV loss of contact emergency triggering being prone to misjudgment, high noise and insufficient integrity of environmental perception data, incomplete extraction of dynamic obstacle information, and insufficient obstacle avoidance decision support due to the lack of three-dimensional spatial attributes and motion prediction capabilities in the obstacle map. This achieves the goals of accurately triggering the emergency response to loss of contact, obtaining high-quality environmental perception data, comprehensively understanding the status of dynamic obstacles, and constructing an accurate and reliable three-dimensional dynamic obstacle map.

[0026] In a preferred embodiment of the present invention, step 2 above may include: Step 2.1: Using a 3D dynamic obstacle map, transform the real-time position, predicted motion trajectory, and obstacle avoidance safety boundary of obstacles to obtain obstacle distribution data in a unified coordinate system. Specifically, this includes: extracting the real-time position coordinates and predicted motion trajectory data for the next 5 to 10 seconds from the constructed 3D dynamic obstacle map. The prediction time range is determined based on the UAV's return-to-home speed in mountainous areas (15 to 20 m / s), the maximum obstacle speed (30 m / s), and the obstacle avoidance response time (0.5 to 1 second), ensuring coverage of the obstacle avoidance decision lead and the preset obstacle avoidance safety boundary range. The original coordinate system for the real-time position coordinates is defined as 3D. The original coordinate system of the motion prediction trajectory data is the sensor local coordinate system, and the original coordinate system of the obstacle avoidance safety boundary is the obstacle relative coordinate system. A unified coordinate transformation rule is adopted, and the transformation parameters are obtained through pre-calibration with an error of ≤0.3 meters. All data are uniformly transformed to the three-dimensional spatial coordinate system. Through coordinate consistency verification, the coordinate deviation of the same obstacle after transformation in different coordinate systems is ≤0.5 meters. This ensures that the real-time position, motion trajectory, safety boundary and other information of the obstacle are consistent under the same spatial reference. Finally, complete obstacle distribution data under a unified coordinate system is formed, which includes obstacle ID, real-time coordinates, trajectory node coordinates and safety boundary size.

[0027] Step 2.2: Based on the obstacle distribution data under the unified coordinate system, retrieve the pre-stored static terrain database, extract terrain elevation information, land cover type, and no-fly zone boundaries to obtain static terrain constraints. Specifically, this includes: accurately retrieving the pre-stored static terrain database of the mountainous power inspection area based on the obstacle distribution data under the unified coordinate system; extracting terrain elevation information along the flight path, including the elevation data of valleys and ridges; land cover type, including the distribution of forest, rocky areas, and water bodies; and the coordinate range of no-fly zone boundaries, including the safety protection zone of transmission lines, residential areas, and geological disaster hazard areas. Organizing and integrating the extracted information, the static terrain constraints for the current return flight scenario are obtained.

[0028] Step 2.3: Based on the static terrain constraints, load the preset mission flight path data, extract the take-off and landing point coordinates, mission waypoint sequence, and flight path altitude parameters to obtain the original mission flight path constraint information. Specifically, this includes: combining the acquired static terrain constraints, loading the preset mission flight path data corresponding to this mountain power inspection, extracting the specific geographical coordinates of the initial take-off and landing point, i.e., the preset UAV take-off and landing position coordinates at the foot of the mountain. The location must meet the following conditions: flat terrain with a slope ≤ 5°, no tall obstructions, and the highest obstacle within 100 meters of the location must be more than 3 meters below the UAV's take-off and landing height; extracting the mission waypoint sequence, i.e., the coordinates of the UAV take-off and landing point along the route. The coordinates of the tower base center and the arrangement of waypoints for the high-voltage transmission towers to be inspected are 500 to 800 meters apart, determined based on the distribution density of transmission lines in mountainous areas. The flight path altitude parameters are extracted, and the safe flight altitude range is set at 5 to 10 meters above the highest point of the terrain and 3 to 5 meters above the transmission line conductors. The minimum flight altitude is not lower than 10 meters below the ground and the maximum is not higher than 120 meters, which complies with the regulations for the management of civil unmanned aerial vehicle flights. It is ensured that the extracted take-off and landing point coordinates, waypoint sequences, and altitude parameters are compatible with static terrain constraints, such as avoiding no-fly zones and adapting to terrain undulations, to form the original mission flight path constraint information.

[0029] Step 2.4 involves fusing obstacle distribution data, static terrain constraints, and original mission route constraints using multi-source data. The obstacle avoidance safety factor, energy consumption optimization index, and path smoothness parameter for each flight segment are calculated using a cost function to obtain a comprehensive optimization evaluation matrix. Specifically, this includes fusing obstacle distribution data, static terrain constraints, and original mission route constraints in a unified coordinate system according to spatial location association and logical correspondence. Reasonable evaluation criteria are then used to calculate the obstacle avoidance safety factor, energy consumption optimization index, and path smoothness parameter for each potential flight segment. The scores for all three indicators are uniformly standardized within a numerical range of 0 to 1. The weights are allocated as follows: obstacle avoidance safety factor accounts for 50%, energy consumption optimization index accounts for 30%, and path smoothness parameter accounts for 20%. First, the obstacle avoidance safety factor score of each flight segment is multiplied by 50% to obtain the weighted score of that index. Then, the energy consumption optimization index score of that flight segment is multiplied by 30% to obtain the weighted score of the corresponding index. Next, the path smoothness parameter score of that flight segment is multiplied by 20% to obtain the weighted score of that index. Finally, the weighted scores of the three indices are summed to obtain the overall optimization score of that flight segment. The overall optimization scores of all flight segments are arranged and integrated in order of flight route to form an overall optimization evaluation matrix.

[0030] Step 2.5: Based on the comprehensive optimization evaluation matrix, a preliminary optimized return path is obtained by planning a sequence of path nodes in three-dimensional space. Specifically, this includes: selecting the segment combination with the final comprehensive score based on the comprehensive optimization evaluation matrix; setting path nodes at reasonable intervals in three-dimensional space; ensuring that the node positions avoid obstacles and no-fly zones and conform to the terrain undulation trend; and connecting adjacent nodes in a smooth transition manner to ensure that the path formed by the node sequence meets the requirements of obstacle avoidance safety, energy consumption optimization, and path smoothness, while also taking into account the general direction of the original mission route, ultimately obtaining a preliminary optimized return path that is obstacle-avoiding safe and smooth and continuous.

[0031] In this embodiment of the invention, because the invention employs the following technical means: transforming the spatial coordinate system of obstacle-related information in a three-dimensional dynamic obstacle map to obtain obstacle distribution data in a unified coordinate system; retrieving a pre-stored static terrain database to extract static terrain constraints; loading preset task route data to extract original task route constraint information; fusing the three types of data from multiple sources and calculating a comprehensive optimization evaluation matrix for obstacle avoidance safety, energy consumption optimization, and path smoothness through a cost function; and planning the path node sequence in three-dimensional space based on the matrix, the invention effectively overcomes the technical problems of fusion deviation caused by inconsistent coordinate systems of multi-source data, lack of collaborative constraints between static terrain and task route, and poor path adaptability caused by failure to comprehensively consider multi-dimensional optimization indicators in traditional preliminary return path planning, thereby achieving the generation of a preliminary optimized return path.

[0032] In a preferred embodiment of the present invention, step 3 above may include: Step 3.1: Receive preliminary optimized return path data. Combined with a pre-stored power facility geographic information database, perform spatial analysis on the distribution of high-voltage transmission towers along the path, and filter out all candidate high-voltage transmission towers located within the preset width range of the path. Specifically, this includes: receiving preliminary optimized return path data planned for mountainous power inspection scenarios. This data includes the three-dimensional coordinate sequence of the path centerline, path curvature, and segment length information. Combined with a pre-stored power facility geographic information database, which contains detailed information such as the precise coordinates of the tower base center, tower model and specifications, height, and construction year of all high-voltage transmission towers in the inspection area, a detailed analysis of the distribution of high-voltage transmission towers along the path is performed using spatial distance calculation algorithms and range definition rules. The preset width is determined based on three aspects: firstly, it is compatible with the lateral detection range of UAV-borne lidar ≥ 40 meters and the lateral detection angle of visual sensors ≥ 60°. The lateral detection range is approximately 45 meters at a flight altitude of 50 meters. Secondly, the typical distribution spacing of high-voltage transmission towers in mountainous areas is 500 to 800 meters to ensure coverage of all effective towers along the route. Thirdly, a 10% redundancy is reserved to handle minor path deviations. Based on this, the preset width is set to 70 to 90 meters, extending 35 to 45 meters to each side from the path centerline. The upper limit of 45 meters is used in areas with strong signal interference and severe terrain obstruction, while the lower limit of 35 meters is used in areas with open terrain and dense tower distribution. The horizontal distance from the center of each high-voltage transmission tower's base to the path centerline is calculated. High-voltage transmission towers with a horizontal distance less than or equal to the preset width on one side and not severely obstructed by terrain, verified by a static terrain database, are selected. The terrain obstruction criteria are that the top elevation of the tower is more than 2 meters higher than the highest point of the surrounding terrain or there are no continuous obstructions within the sensor's detection angle range. Finally, a set of candidate high-voltage transmission towers is obtained.

[0033] Step 3.2: Based on the screening of candidate high-voltage transmission towers, three final high-voltage transmission towers are selected from the candidate set as fixed references according to tower height, structural stability, spacing, and visibility conditions. This yields reference selection decision data, specifically including: verifying the actual situation of each candidate tower based on the selected set of candidate high-voltage transmission towers; ensuring the tower height is higher than the surrounding terrain and vegetation to guarantee the observation field of view; confirming structural stability through database records that there are no damage, tilting, or loose foundations; controlling the spacing between towers within a reasonable range to construct a stable spatial reference frame, neither too dense nor too sparse; and ensuring visibility so that the tower images can be continuously captured by airborne sensors without continuous obstruction along the UAV's flight path. Finally, three qualified high-voltage transmission towers are selected from the candidate set as fixed references, and reference selection decision data containing the selected tower identifiers, location information, and selection criteria are compiled.

[0034] Step 3.3: Based on the reference selection decision data, retrieve the precise geographic coordinate information of the three selected high-voltage transmission towers. Use a coordinate transformation algorithm to convert the geographic coordinates into a unified three-dimensional spatial coordinate system, establishing a spatial reference frame based on the positions of the three towers. Specifically, this includes: retrieving the precise geographic coordinate information, including latitude, longitude, and elevation data, of the three selected high-voltage transmission towers from the power facility geographic information database, using a coordinate transformation algorithm to establish a unified three-dimensional spatial coordinate system. Figure 1 The coordinate transformation rule uses a coordinate transformation algorithm to uniformly transform geographic coordinates to a three-dimensional spatial coordinate system. A triangulation-based spatial reference frame is constructed using the transformed coordinates of three selected high-voltage transmission towers as reference points. The accuracy of the frame is calibrated by calculating the spatial distance and angular relationship between the three reference points, ensuring that the spatial reference frame has a stable spatial positioning reference capability.

[0035] Step 3.4: Based on the established spatial reference frame, perform geometric feature analysis on the main structure of the high-voltage transmission tower and the connected transmission lines within the frame. Extract tower corner points, insulator suspension points, conductor connection points, and line intersection points as feature points to obtain a set of key feature points distributed in space. Specifically, based on the established spatial reference frame, use UAV-borne visual sensors and LiDAR to perform comprehensive geometric feature analysis on the main structure of the three selected high-voltage transmission towers and the connected transmission lines within the frame. Extract the four corners of the tower base, the corner points at the connection of tower segments, and the key corner positions at the top of the tower for tower corner points. Extract the connection points between the insulator string and the tower crossarm, and the connection nodes of the insulator string itself for insulator suspension points. Extract the connection ends between the conductor and the insulator string, and the conductor segment connection joints for conductor connection points. Extract the specific spatial positions where different transmission lines intersect and overlap for line intersection points. Accurately record the three-dimensional coordinates of the feature points within the spatial reference frame, and finally form a set of key feature points with clear spatial distribution and well-defined geometric features.

[0036] In this embodiment of the invention, because the invention employs technical means such as combining a pre-stored power facility geographic information database to filter candidate high-voltage transmission towers within a preset path width range, selecting three high-voltage transmission towers as fixed reference objects based on multiple dimensions such as tower height and structural stability, converting the precise geographic coordinates of the selected towers into a unified three-dimensional spatial coordinate system to establish a spatial reference frame, and performing geometric feature analysis on the transmission tower bodies and lines within the frame to extract key feature point sets, it effectively overcomes the technical problems of traditional positioning references lacking scenario adaptability, arbitrary selection of reference objects, inconsistent spatial benchmarks, and inaccurate extraction of key features leading to unreliable basis for positioning correction. Thus, it achieves the construction of a stable spatial reference frame that is adapted to power inspection scenarios and obtains a precisely distributed set of key feature points.

[0037] In a preferred embodiment of the present invention, step 4 above may include: Step 4.1: Based on the set of key feature points in spatial distribution, normalize the spatial coordinates of the key feature points to obtain standardized feature point data. Specifically, this includes acquiring the set of key feature points for the high-voltage transmission tower body and transmission lines within the spatial reference frame. These feature points include the four corners of the tower bottom, corner points at tower segment connections, corner points where lightning rods are installed at the top of the tower, tower corner points, insulator suspension points such as the connection points between insulator strings and tower crossarms, and the connection nodes between the beginning and end of insulator strings; conductor connection points such as the crimp ends between conductors and insulator strings, and the ends of conductor segment splices; and line intersections such as spatial intersections where different circuit transmission lines intersect and overlap. Complete three-dimensional coordinate data is recorded for all feature points. The maximum and minimum x-axis values ​​and the maximum and minimum y-axis values ​​of all key feature points within the spatial reference frame are used as the basis for this process. The maximum and minimum values ​​of the z-axis are used as the normalization boundaries of each axis, respectively. The statistical range of the coordinate extrema is defined as the set of coordinates of all valid feature points without abnormal data. The preset unified numerical range is [0, 1]. The numerical range is adapted to the processing range of the subsequent fixed radius circle coverage algorithm to ensure data calculation efficiency. A linear scaling method is used to proportionally adjust the x-axis, y-axis and z-axis coordinates of each feature point. Specifically, the difference between each axis coordinate and the corresponding axis extremum is mapped to the range of axis extremum according to the proportion. During the scaling process, the spatial distance ratio and relative angle relationship between each feature point are strictly maintained to avoid distortion of the spatial topology due to scaling. Finally, standardized feature point data with a unified coordinate range in the range of [0, 1] and accurate relative positions of each feature point are formed.

[0038] Step 4.2: Based on standardized feature point data, a state compression dynamic programming algorithm with fixed radius circle coverage is used to iteratively calculate the coverage of feature points, obtaining a set of fixed radius circles covering key feature points and their corresponding center coordinates. Specifically, based on the standardized feature point data, the Euclidean distances between all adjacent feature points are first calculated, and the average of these distances is calculated. Considering the potential for localized density and overall dispersion of feature points in mountainous power inspection scenarios, a fixed circle coverage radius of 1.5 to 2 times the average distance between adjacent feature points is chosen. This radius is designed to cover a sufficient number of adjacent feature points to form effective clusters without excessive overlap or invalid space due to an excessively large radius. Iterative coverage calculations are then performed round by round using the state compression dynamic programming algorithm. The first round involves iteratively calculating the coverage of all standardized feature points. Spatial clustering analysis is performed on the feature points to identify the cluster region with the most feature points. Using the geometric center of the cluster region as the candidate center, circles of fixed radius are drawn, and the number of uncovered feature points covered is counted. If there are multiple candidate centers, the one with the most covered feature points is selected as the center of the first circle, and the standardized coordinates of the circle's center position and coverage area are recorded. In each subsequent round, the above clustering analysis, candidate center selection, and coverage count are repeated for the remaining uncovered feature points until all key feature points are included in the circle's coverage area. After drawing the circles in each round, it is verified whether the circle is a necessary cover, i.e., whether any feature points are exposed after removing the circle, to ensure that there are no redundant circles in the final set of fixed radius circles, and that each circle is indispensable. Finally, a set of fixed radius circles that covers all key feature points and has the fewest number of feature points is obtained.

[0039] Step 4.3: Based on the coverage area, spatial coordinates of the center, and density distribution characteristics of corresponding coverage feature points of each circle in the fixed radius circle set, the geometric stability and feature saliency of the coverage area of ​​each circle are analyzed using a weighted calculation method to obtain the positioning correction coefficient matrix for each area. Specifically, this includes: conducting multi-dimensional analysis for each circle in the fixed radius circle set; in terms of coverage area rationality analysis, determining whether the coverage area of ​​the circle is completely within the core positioning area of ​​the spatial reference frame, and whether it does not include invalid positioning areas such as steep mountainous terrain areas or areas with dense obstacles, ensuring that the coverage area is highly compatible with positioning requirements; in terms of center spatial stability analysis, by retrieving the pre-stored static terrain database, calculating the straight-line distance from the center to the surrounding terrain undulation change points, the larger the distance, the less the center is affected by terrain, and the higher the spatial stability; at the same time, checking whether there are electromagnetic interference source markers around the center, the absence of interference sources adds points to stability; in terms of coverage feature point density distribution analysis, calculating the number of feature points per unit area within the coverage area of ​​each circle, and summarizing... The uniformity of feature point distribution within the circle is calculated, i.e., the standard deviation of the distance from the feature point to the center of the circle. The smaller the standard deviation, the more uniform the distribution. Specific weights are assigned to three analysis dimensions: feature point density accounts for 40%, spatial stability of the center of the circle accounts for 40%, and coverage rationality accounts for 20%. The weight allocation is determined based on the degree of influence of each dimension on positioning accuracy. A comprehensive evaluation value for each circle's coverage area is calculated using a weighted summation formula. The evaluation value range is set from 0 to 100 points. The higher the score, the greater the positioning reference value of the area. The comprehensive evaluation value is converted into positioning correction coefficients in the three directions of x, y, and z axes according to a linear mapping relationship. The mapping range is from 0.8 to 1.2, where an evaluation value of 100 points corresponds to a correction coefficient of 1.2, an evaluation value of 60 points corresponds to a correction coefficient of 0.8, and intermediate scores correspond to corresponding correction coefficients proportionally. The larger the correction coefficient, the higher the reliability of the area's positioning data, and the higher the positioning weight should be assigned. Finally, a positioning correction coefficient matrix containing the three-dimensional correction coefficients of each area is formed according to the distribution order of the circle's coverage area within the spatial reference frame.

[0040] Step 4.4: While calculating the positioning correction coefficient, receive real-time collected high-voltage transmission line sensing data. Based on the established spatial reference frame, perform spatial positioning on the sensing data. Extract the topology, insulator arrangement pattern, and conductor phase sequence characteristics of the transmission line through image processing algorithms to obtain the inherent feature recognition results of the high-voltage transmission line. Specifically, this includes: while calculating the positioning correction coefficient, the UAV-mounted high-definition camera collects color image data of the high-voltage transmission line along the flight path at 30 frames per second, and the lidar collects three-dimensional point cloud data at a sampling frequency of 100kHz, covering the transmission line and surrounding environment 50 to 100 meters ahead of the flight path; based on the established spatial reference frame, first eliminate the installation deviation of the high-definition camera and lidar through equipment calibration parameters, and then use coordinate transformation formulas to map the collected sensing data from the equipment's own coordinate system to a unified three-dimensional spatial coordinate system, clarifying the actual geographic space corresponding to each frame of image pixels and each point cloud data. The process involves several steps: First, by extracting the contours of transmission line images and removing background noise, the line direction is fitted using Hough linear transform. Combined with the spatial connectivity of 3D point cloud data, branch nodes and connection paths of the transmission line are identified, forming complete topological information. Second, by separating insulator regions from the images, effective areas are selected based on the typical grayscale features of the insulators. The center distance and number of adjacent insulator regions are calculated. Combined with the size data of the 3D point cloud, the string arrangement and spacing parameters of the insulators are determined. Third, by using color component separation technology in the images, the phase sequence color marks on the conductor surface are identified. Simultaneously, by combining the z-axis coordinate values ​​of the 3D point cloud data, the spatial layering positions of conductors with different phase sequences are distinguished, clarifying the direction and relative positional relationship of each phase sequence conductor. Finally, the extracted transmission line topology, insulator arrangement patterns, and conductor phase sequence features are organized in a structured format, and ambiguous or abnormal feature data are removed, ultimately yielding comprehensive, accurate, and non-redundant identification results of the inherent features of high-voltage transmission lines.

[0041] In this embodiment of the invention, because the invention employs the following technical means: normalizing the spatial coordinates of key feature points to obtain standardized feature point data; using a state compression dynamic programming algorithm based on the standardized data to obtain a set of fixed-radius circles and their center coordinates; and using weighted calculation based on the circle's coverage area, center position, and feature point distribution characteristics to obtain the positioning correction coefficient matrix for each region; and simultaneously using a spatial reference frame to spatially locate the transmission line sensing data and extract inherent features such as the line topology, insulator arrangement pattern, and conductor phase sequence through image processing algorithms, this invention effectively overcomes the technical problems of traditional positioning correction, such as calculation deviations caused by the lack of standardized processing of feature point data, insufficient regional specificity of correction coefficients, asynchronous feature recognition and correction coefficient calculation, and incomplete extraction of inherent features of the line. This results in obtaining an accurate and regionally adaptable positioning correction coefficient matrix and simultaneously acquiring comprehensive and reliable identification results of the inherent features of the transmission line.

[0042] In a preferred embodiment of the present invention, step 5 above may include: Step 5.1: Based on the comprehensive feature matching dataset, separate the inherent feature recognition results and the positioning correction coefficient matrix. Specifically, the comprehensive feature matching dataset contains two core data types: the inherent feature recognition results of high-voltage transmission lines and the positioning correction coefficient matrix of each region. The inherent feature recognition results are associated with image analysis information such as the transmission line topology, insulator arrangement pattern, and conductor phase sequence. The positioning correction coefficient matrix corresponds to the three-dimensional correction values ​​of the x-axis, y-axis, and z-axis of each circle-covered area. The dataset is classified and split according to the differences in data type identification and storage format. The structured information describing the characteristics of the transmission line and the numerical set containing the positioning correction parameters are extracted separately to ensure that the two types of data are not confused and the original details are completely preserved.

[0043] Step 5.2: Based on the separated inherent feature recognition results, feature matching calculations are performed with the pre-stored feature geographic coordinate database. The high-voltage transmission line segment where the UAV is currently located and its corresponding reference geographic coordinates are determined through similarity comparison. Specifically, the pre-stored feature geographic coordinate database contains feature templates and precise geographic coordinates of all high-voltage transmission line segments within the mountainous power inspection area. The feature templates cover the standard topology, insulator arrangement patterns, and conductor phase sequence configurations of each transmission line segment. Based on the separated inherent feature recognition results, the consistency of the connection relationship between the currently identified topology and the templates of each segment in the database, the consistency of the spacing of the insulator arrangement patterns, and the consistency of the spatial distribution of conductor phase sequence features are compared one by one. The similarity value is obtained by accumulating the multi-dimensional feature overlap. The preset similarity standard value is 85 points. When the similarity value exceeds 85 points, the match is considered successful, the high-voltage transmission line segment where the UAV is currently located is locked, and the precise geographic coordinates of the preset key high-voltage transmission towers within the high-voltage transmission line segment are retrieved as the reference geographic coordinates.

[0044] Step 5.3: Based on the reference geographic coordinates and combined with the positioning correction coefficient matrix, perform error compensation calculation on the current position coordinates output by the UAV's inertial navigation system to eliminate the cumulative positioning error caused by sensor drift and environmental interference, and obtain the error-compensated coordinate data. Specifically, this includes: first, obtaining the current position coordinates output by the UAV's inertial navigation system in real time. The current position coordinates have accumulated errors due to sensor drift and airflow electromagnetic interference in the complex mountainous environment. Determine the current spatial area of ​​the UAV based on the reference geographic coordinates, retrieve the x-axis, y-axis, and z-axis positioning correction coefficients corresponding to the area from the positioning correction coefficient matrix, calculate the difference between the coordinates output by the inertial navigation system and the reference geographic coordinates to obtain the original error data, and perform proportional adjustment and orientation calibration on the original error data in combination with the correction coefficients, focusing on eliminating drift errors caused by long-term operation of the inertial navigation sensor and random errors caused by mountainous environmental interference. Obtain the error-compensated coordinate data after eliminating the cumulative deviation through axis-by-axis correction.

[0045] Step 5.4 involves fusing and weighting the error-compensated coordinate data with the reference geographic coordinates to obtain the corrected precise location information. Specifically, this includes: the error-compensated coordinate data reflecting the dynamic changes in the UAV's real-time position; the reference geographic coordinates providing a stable spatial reference; and, based on the dynamic response characteristics of inertial navigation data in mountainous environments and the static stability of the reference coordinates, assigning 30% to 40% weight to the error-compensated coordinate data to meet real-time requirements, and assigning 60% to 70% weight to the reference geographic coordinates to meet stability requirements. Weighted calculations are then performed separately for the x, y, and z axes. The x-axis coordinate value after error compensation is multiplied by the corresponding weight to obtain the dynamic weighted value of the x-axis. Then, the x-axis coordinate value of the reference geographic coordinate is multiplied by the corresponding weight to obtain the static weighted value of the x-axis. The two weighted values ​​of the x-axis are added together to obtain the fused accurate x-axis coordinate value. The fused accurate y-axis coordinate value and accurate z-axis coordinate value are calculated in the same way. After the three-dimensional coordinate fusion is completed, the rationality is checked in combination with the spatial distribution range of the high-voltage transmission line and the terrain undulation characteristics. Abnormal values ​​that exceed the normal geographic range or are seriously inconsistent with the terrain characteristics are removed during the fusion process. Finally, the corrected accurate location information with both real-time performance and accuracy is obtained.

[0046] In this embodiment of the invention, because the invention employs the following technical means: separating the inherent feature recognition result and the positioning correction coefficient matrix from the comprehensive feature matching dataset; matching the inherent feature recognition result with the pre-stored feature geographic coordinate database to determine the reference geographic coordinates; combining the reference geographic coordinates and the positioning correction coefficient matrix to perform error compensation calculation on the position coordinates output by inertial navigation; and fusing and weighting the error-compensated coordinate data with the reference geographic coordinates, the invention effectively overcomes the technical problems of traditional inertial navigation, such as the difficulty in accurately eliminating accumulated errors, the lack of coordination between feature matching and error compensation, and the significant impact of sensor drift and environmental interference on position correction. Thus, the invention achieves the goal of reducing the accumulated positioning error of inertial navigation and outputting high-precision and stable and reliable position information.

[0047] In a preferred embodiment of the present invention, step 6 above may include: Step 6.1: Based on the corrected precise location information, compare the precise location information with the preliminary optimized return path to calculate the spatial deviation between the current UAV position and the preset waypoints. Specifically, this includes: obtaining the precise UAV position information after positioning correction, which includes x-axis, y-axis, and z-axis coordinates in three-dimensional space and has eliminated accumulated positioning errors; comparing these coordinates with the coordinates of each preset waypoint in the preliminary optimized return path; the preset waypoints are set according to the following values ​​and are evenly distributed along the preliminary optimized return path with a spacing of 2... The coordinate accuracy of the nodes associated with the high-voltage transmission tower is ±0.3 meters horizontally and ±0.5 meters vertically, while the coordinate accuracy of the terrain adaptation nodes is ±0.5 meters horizontally and ±0.8 meters vertically. The total number of waypoints is determined based on the total length of the return path to ensure that each path segment has a corresponding reference node. The deviation values ​​between the current UAV position and the corresponding waypoint in the horizontal, vertical, and spatial straight lines are calculated respectively. At the same time, the elevation deviation in the z-axis direction is checked in particular, taking into account the undulating characteristics of the mountainous terrain. Finally, the spatial deviation that fully reflects the position deviation is obtained.

[0048] Step 6.2: Based on the spatial deviation and combined with the latest position status of obstacles in the 3D dynamic obstacle map, perform local dynamic correction on the initially optimized return path to obtain the updated real-time return path node sequence. Specifically, this includes: comparing the calculated spatial deviation with the preset deviation threshold. The preset deviation threshold values ​​are set as follows: the horizontal deviation threshold is 5 to 8 meters, the vertical deviation threshold is 8 to 12 meters, the spatial straight line deviation threshold is 10 to 15 meters, the upper limit of each threshold is taken in mountainous areas with complex terrain and dense obstacles, and the lower limit of each threshold is taken in areas with open terrain and no obvious obstruction. When the deviation exceeds the corresponding threshold or new dynamic obstacles appear in the 3D dynamic obstacle map, or the positions of existing obstacles change significantly, a preliminary optimization of the return path is initiated for local dynamic correction. The correction focuses on the segment composed of three waypoints before and after the current flight phase. The spatial coordinates of the waypoints are adjusted with reference to the latest real-time position, trajectory, and obstacle avoidance safety boundary of the obstacle. This ensures that the new nodes avoid the obstacle area and conform to the contours of the mountainous terrain, while maintaining a safe distance from high-voltage power transmission lines. This ensures that the corrected node sequence meets the obstacle avoidance requirements while being as close as possible to the original return path, ultimately resulting in the updated real-time return path node sequence.

[0049] Step 6.3: Based on the updated real-time return path node sequence, a smooth and continuous flight trajectory is obtained through spline curve interpolation algorithm. Speed ​​constraints and attitude adjustment parameters on the trajectory are calculated to obtain complete trajectory replanning data. Specifically, this includes: for the updated real-time return path node sequence, selecting a spline curve interpolation algorithm adapted to the undulating characteristics of mountainous terrain, performing smooth fitting processing on the paths between adjacent nodes, eliminating sharp corners at node connections to ensure a continuous and smooth flight trajectory; setting corresponding speed constraints based on the terrain slope, obstacle distance, and remaining UAV battery power for each segment; maintaining an economical speed on flat roads; appropriately reducing speed when approaching obstacles on uphill sections; and decelerating when approaching obstacles. Simultaneously, calculating the pitch, roll, and yaw angle adjustment parameters for each trajectory segment clarifies the attitude control range of the UAV in scenarios such as turning, climbing, and obstacle crossing, ensuring smooth attitude adjustment without exceeding safety limits, ultimately forming trajectory replanning data containing complete flight trajectory speed and attitude parameters.

[0050] Step 6.4: Based on the trajectory replanning data, monitor the flight status of the UAV along the updated return path in real time, and adjust the flight trajectory in a closed loop based on the received real-time position feedback data. Specifically, this includes: sending instructions to the UAV flight control unit based on the trajectory replanning data; monitoring the UAV's flight status in real time, including actual flight position, speed, attitude, and remaining energy consumption; collecting real-time position feedback data at a fixed frequency through onboard sensors and transmitting it to the processing end; continuously comparing the feedback data with the replanned trajectory; and when a deviation between the actual position and the planned trajectory is found to be beyond the allowable range, quickly adjusting the control signal in conjunction with possible interference from changes in mountainous airflow to correct the UAV's flight speed and attitude parameters so that the UAV re-fits the planned trajectory. This forms a closed-loop control logic of instruction sending, status monitoring, and deviation correction, ensuring stable flight along the updated return path throughout the entire process.

[0051] Step 6.5: As the UAV approaches the initial take-off and landing point, a high-precision landing control strategy is executed based on precise position information and onboard visual sensor data. Through position fine-tuning and gradual altitude descent control, the UAV completes a precise automatic landing at the initial take-off and landing point. Specifically, when the UAV flies to a preset distance range of 100 to 150 meters from the initial take-off and landing point, the real-time relative position of the take-off and landing point is determined by combining the corrected precise position information. At the same time, the onboard high-definition visual sensor is activated to capture images of the take-off and landing point area, identify preset take-off and landing markers on the ground and the flatness of the surrounding terrain, and execute the high-precision landing control strategy. In the horizontal direction, the remaining position deviation is corrected by fine-tuning the flight attitude to align the UAV with the take-off and landing center point. In the vertical direction, the altitude descent control is implemented in a gradient manner. When moving away from the ground, a moderate descent speed is maintained, and the descent rate is gradually reduced when approaching the ground. At the same time, the visual sensor detects the altitude above the ground in real time to ensure a smooth descent. Finally, the UAV completes a precise automatic landing in the flat area of ​​the preset initial take-off and landing point in the mountainous area.

[0052] In this embodiment of the invention, because the invention employs the following technical means: spatial comparison based on the corrected precise location information and the preliminary optimized return path to calculate the deviation; local dynamic correction of the path based on the latest state of obstacles in the three-dimensional dynamic obstacle map; generation of a smooth and continuous flight trajectory through spline curve interpolation algorithm and calculation of speed constraints and attitude adjustment parameters; real-time monitoring of flight status and closed-loop control adjustment based on trajectory replanning data; and execution of a high-precision landing control strategy based on precise location information and airborne visual sensor data when approaching the initial take-off and landing point, the invention effectively overcomes the technical problems of poor safety and reliability of return in complex environments caused by traditional return path updates being lagging, not adapting to dynamic changes in obstacles, unsmooth trajectories, lack of closed-loop adjustment in flight control, and insufficient landing accuracy. This achieves the technical effect of a return path that can adapt to environmental changes in real time, a smooth and stable flight trajectory, and precise and controllable flight control, ultimately enabling the UAV to land automatically with high precision at the initial take-off and landing point.

[0053] like Figure 2 As shown, embodiments of the present invention also provide a drone loss return-to-home control system, including: The acquisition module is used to initiate real-time perception of the environment ahead of the drone after it loses contact, acquire environmental information of dynamic obstacles, and generate a dynamic obstacle map. The fusion module is used to merge pre-stored static terrain data with preset mission route data through a dynamic obstacle map to obtain a preliminary optimized return route; The extraction module is used to fly along the initially optimized return path, select three specific high-voltage transmission towers as fixed reference objects along the route, establish a spatial reference frame based on the spatial reference positions of the three high-voltage transmission towers, and extract the key feature point set of the high-voltage transmission tower body and transmission line within the spatial reference frame; The identification module is used to obtain a set of fixed-radius circles covering key feature points by performing coverage processing on the feature point set; based on the coverage range, spatial position of the center of each circle and the distribution characteristics of the corresponding covered feature points in the fixed-radius circle set, the positioning correction coefficient of each region is calculated; simultaneously, the inherent features of the high-voltage transmission lines along the route are identified by airborne sensors to obtain the identification results. The correction module is used to match the recognition results with the pre-stored feature geographic coordinate database, and combine the positioning correction coefficient to comprehensively correct the cumulative positioning error of inertial navigation, so as to obtain the corrected accurate location information. The processing module is used to dynamically update and replan the preliminary optimized return path based on the corrected precise location information, and control the UAV to follow the updated real-time return path until it lands precisely at the initial take-off and landing point.

[0054] The control system according to embodiments of the present invention can correspond to performing the methods described in the embodiments of the present invention, and the above and other operations and / or functions of each module of the control system are respectively for implementing Figure 1 The corresponding process of the method in the illustrated embodiment will not be described in detail here for the sake of brevity.

[0055] This application also provides a computing device. This computing device can utilize a server.

[0056] like Figure 3 As shown in the figure, this is a schematic diagram of a computing device provided in an embodiment of this application. The computing device 700 includes a bus 701, a processor 702, a communication interface 703, and a memory 704. The processor 702, the memory 704, and the communication interface 703 communicate with each other via the bus 701.

[0057] The 701 bus can be a peripheral component interconnect standard, such as the PCI bus, or an extended industry standard architecture, such as the EISA bus. Buses can be categorized as address buses, data buses, and control buses. For ease of representation, Figure 3 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0058] The processor 702 can be any one or more of the following: a central processing unit (CPU), a graphics processing unit (GPU), a microprocessor (MP), a digital signal processor (DSP), etc.

[0059] Communication interface 703 is used for external communication. Memory 704 may include volatile memory, such as random access memory (RAM). Memory 704 may also include non-volatile memory, such as read-only memory (ROM), flash memory, hard disk drive, HDD, or solid state drive (SSD). Memory 704 stores executable code, which processor 702 executes to perform the aforementioned UAV loss-of-connection return control method.

[0060] Specifically, in implementing the UAV loss-of-connection and return-to-home control system described in the above embodiments, and where each module or unit of the UAV loss-of-connection and return-to-home control system described in the above embodiments is implemented by software, the software or program code required to execute the functions of each module / unit in the UAV loss-of-connection and return-to-home control system described in the above embodiments can be partially or stored in the memory 704. The processor 702 executes the program code corresponding to each unit stored in the memory 704 to execute the aforementioned UAV loss-of-connection and return-to-home control method.

[0061] This application also provides a computer-readable storage medium. The computer-readable storage medium can be any available medium that a computing device can store, or a data storage device such as a data center containing one or more available media. The available medium can be a magnetic medium such as a floppy disk, hard disk, magnetic tape, optical medium such as a DVD, or a semiconductor medium such as a solid-state drive. The computer-readable storage medium includes instructions that instruct the computing device to execute the aforementioned UAV loss-of-connection return control method.

[0062] This application also provides a computer program product comprising one or more computer instructions. When these computer instructions are loaded and executed on a computing device, they may partially generate the processes or functions described in this application.

[0063] The computer instructions may be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions may be transmitted from one website, computer, or data center to another website, computer, or data center via wired means such as coaxial cable, optical fiber, digital subscriber line, or wireless means such as infrared, wireless, microwave, etc.

[0064] When the computer program product is executed by a computer, the computer executes any of the aforementioned methods of the UAV loss-of-connection and return-to-home control method. The computer program product can be a software installation package; when any of the aforementioned methods of the UAV loss-of-connection and return-to-home control method is required, the computer program product can be downloaded and executed on the computer.

[0065] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for controlling the return of a lost drone, characterized in that, The method includes: After the drone loses contact, it initiates real-time perception of the environment ahead of the flight, acquires environmental information of dynamic obstacles, and generates a dynamic obstacle map. By using a dynamic obstacle map and integrating pre-stored static terrain data with preset mission route data, a preliminary optimized return route is obtained. The aircraft flew along the initially optimized return route, selecting three specific high-voltage transmission towers as fixed reference points along the route. A spatial reference frame was established based on the spatial reference positions of the three high-voltage transmission towers. The key feature point set of the high-voltage transmission tower body and transmission line within the spatial reference frame was extracted. By performing coverage processing on the feature point set, a set of fixed-radius circles covering the key feature points is obtained; based on the coverage range, spatial location of the center of each circle and the distribution characteristics of the corresponding covered feature points in the fixed-radius circle set, the positioning correction coefficient of each region is calculated; simultaneously, the inherent characteristics of the high-voltage transmission lines along the route are identified through airborne sensors, and the identification results are obtained. The identification results are matched with a pre-stored feature geographic coordinate database, and the cumulative positioning error of inertial navigation is comprehensively corrected by the positioning correction coefficient to obtain the corrected accurate location information. Based on the corrected precise location information, the initially optimized return path is dynamically updated and the trajectory is replanned. The drone is controlled to follow the updated real-time return path until it lands precisely at the initial take-off and landing point.

2. The method for controlling the return of a lost UAV according to claim 1, characterized in that, After the drone loses contact, real-time perception of the environment ahead is initiated to acquire environmental information about dynamic obstacles and generate a dynamic obstacle map, including: By receiving the drone's disconnection status signal and verifying the duration of the disconnection status, the system confirms entry into the disconnection emergency mode when the duration exceeds a preset threshold. Based on the disconnection emergency mode, a real-time environmental perception activation command is sent to the drone, and the environmental data processing function is activated simultaneously to obtain the uploaded raw environmental perception data stream. Receive the uploaded raw environmental perception data stream, perform spatiotemporal alignment, noise filtering, and data integrity verification on the raw data to obtain a clean environmental perception dataset; Based on a clean environmental perception dataset, a dynamic obstacle feature vector is obtained by extracting the motion parameters, spatial distribution density, and threat level assessment values ​​of moving objects through a dynamic obstacle recognition algorithm. The feature vectors of dynamic obstacles are fused with a pre-stored geographic information database to construct a three-dimensional dynamic obstacle map that includes the real-time location of obstacles, predicted motion trajectories, and obstacle avoidance safety boundaries.

3. The UAV loss-of-connection return control method according to claim 2, characterized in that, By using a dynamic obstacle map and integrating pre-stored static terrain data with preset mission route data, a preliminary optimized return route is obtained, including: By using a three-dimensional dynamic obstacle map and transforming the real-time location, predicted motion trajectory, and obstacle avoidance safety boundary of the obstacles into a spatial coordinate system, obstacle distribution data in a unified coordinate system is obtained. Based on obstacle distribution data in a unified coordinate system, the pre-stored static terrain database is retrieved to extract terrain elevation information, land cover type, and no-fly zone boundaries, thus obtaining static terrain constraints. Based on static terrain constraints, load the preset mission route data, extract the take-off and landing point coordinates, mission waypoint sequence and route altitude parameters to obtain the original mission route constraint information; The obstacle distribution data, static terrain constraints and original mission route constraints are fused from multiple sources. The obstacle avoidance safety coefficient, energy consumption optimization index and path smoothness parameter of each flight segment are calculated by the cost function to obtain the comprehensive optimization evaluation matrix. Based on the comprehensive optimization evaluation matrix, a preliminary optimized return path is obtained by planning the path node sequence in three-dimensional space.

4. The UAV loss-of-connection return control method according to claim 3, characterized in that, Fly along the initially optimized return route, select three specific high-voltage transmission towers along the route as fixed reference points, and establish a spatial reference frame based on the spatial reference positions of the three high-voltage transmission towers; Extract the key feature point set of the high-voltage transmission tower body and transmission line within the spatial reference frame, including: Receive preliminary optimized return route data, combine it with the pre-stored power facility geographic information database, perform spatial analysis on the distribution of high-voltage transmission towers along the route, and screen out all candidate high-voltage transmission towers located within the preset width range of the route; Based on the screening of candidate high-voltage transmission towers, three final high-voltage transmission towers are selected from the candidate set as fixed references according to tower height, structural stability, mutual spacing and visibility conditions, and reference selection decision data are obtained. Based on the reference selection decision data, the precise geographic coordinate information of the three selected high-voltage transmission towers is retrieved. The geographic coordinates are converted into a unified three-dimensional spatial coordinate system through a coordinate transformation algorithm, and a spatial reference frame based on the positions of the three towers is established. Based on the established spatial reference frame, geometric feature analysis is performed on the main structure of the high-voltage transmission tower and the connecting transmission lines within the frame. Tower corners, insulator suspension points, conductor connection points and line intersections are extracted as feature points to obtain a set of key feature points distributed in space.

5. The UAV loss-of-connection return control method according to claim 4, characterized in that, By performing coverage processing on the feature point set, a set of fixed-radius circles covering the key feature points is obtained; based on the coverage range, spatial location of the center of each circle and the distribution characteristics of the corresponding covered feature points in the fixed-radius circle set, the positioning correction coefficient of each region is calculated. Simultaneously, the inherent characteristics of high-voltage transmission lines along the flight path are identified using airborne sensors, yielding identification results, including: Based on the set of key feature points in spatial distribution, the spatial coordinates of the key feature points are normalized to obtain standardized feature point data. Based on standardized feature point data, a state compression dynamic programming algorithm with fixed radius circle coverage is used to iteratively calculate the coverage of feature points, thereby obtaining a set of fixed radius circles covering key feature points and the corresponding center coordinates of the circles. Based on the coverage area, spatial coordinates of the center of each circle in the set of fixed-radius circles, and density distribution characteristics of the corresponding coverage feature points, the geometric stability and feature saliency of the coverage area of ​​each circle are analyzed by weighted calculation method, and the positioning correction coefficient matrix of each area is obtained. While calculating the positioning correction coefficient, real-time high-voltage transmission line sensing data is received. Based on the established spatial reference frame, the sensing data is spatially positioned. The topology, insulator arrangement pattern, and conductor phase sequence characteristics of the transmission line are extracted through image processing algorithms to obtain the inherent feature recognition results of the high-voltage transmission line.

6. The UAV loss-of-connection return control method according to claim 5, characterized in that, The identification results are matched with a pre-stored feature geographic coordinate database, and the cumulative positioning error of the inertial navigation is comprehensively corrected by combining positioning correction coefficients to obtain the corrected accurate location information, including: Based on the comprehensive feature matching dataset, the inherent feature recognition results and the localization correction coefficient matrix are separated from it; Based on the identified inherent features, feature matching calculations are performed with the pre-stored feature geographic coordinate database. By comparing similarity, the current high-voltage transmission line section and corresponding reference geographic coordinates of the UAV are determined. Based on the reference geographic coordinates and combined with the positioning correction coefficient matrix, error compensation calculation is performed on the current position coordinates output by the UAV's inertial navigation to eliminate the cumulative positioning error caused by sensor drift and environmental interference, and obtain the coordinate data after error compensation. The error-compensated coordinate data is fused and weighted with the baseline geographic coordinates to obtain the corrected accurate location information.

7. The method for controlling the return of a lost UAV according to claim 6, characterized in that, Based on the corrected precise location information, the initially optimized return path is dynamically updated and replanned. The drone is controlled to follow the updated real-time return path until it lands precisely at the initial take-off and landing point, including: Based on the corrected precise location information, the precise location information is compared with the preliminary optimized return path to calculate the spatial deviation between the current UAV position and the preset waypoint. Based on the spatial deviation, and combined with the latest position status of obstacles in the 3D dynamic obstacle map, the preliminary optimized return path is locally dynamically corrected to obtain the updated real-time return path node sequence. Based on the updated real-time return path node sequence, a smooth and continuous flight trajectory is obtained through spline curve interpolation algorithm, and the velocity constraints and attitude adjustment parameters on the trajectory are calculated to obtain complete trajectory replanning data. Based on trajectory replanning data, the flight status of the UAV along the updated return path is monitored in real time, and the flight trajectory is adjusted in a closed loop based on the received real-time position feedback data. As the drone approaches the initial take-off and landing point, it executes a high-precision landing control strategy based on precise position information and onboard visual sensor data. Through position fine-tuning and altitude descent control, the drone completes a precise automatic landing at the initial take-off and landing point.

8. A drone loss-of-connection return control system, wherein the system implements the method as described in any one of claims 1 to 7, characterized in that, include: The acquisition module is used to initiate real-time perception of the environment ahead of the drone after it loses contact, acquire environmental information of dynamic obstacles, and generate a dynamic obstacle map. The fusion module is used to merge pre-stored static terrain data with preset mission route data through a dynamic obstacle map to obtain a preliminary optimized return route; The extraction module is used to fly along the initially optimized return path, select three specific high-voltage transmission towers as fixed reference objects along the route, establish a spatial reference frame based on the spatial reference positions of the three high-voltage transmission towers, and extract the key feature point set of the high-voltage transmission tower body and transmission line within the spatial reference frame; The identification module is used to obtain a set of fixed-radius circles covering key feature points by performing coverage processing on the feature point set; based on the coverage range, spatial position of the center of each circle and the distribution characteristics of the corresponding covered feature points in the fixed-radius circle set, the positioning correction coefficient of each region is calculated; simultaneously, the inherent features of the high-voltage transmission lines along the route are identified by airborne sensors to obtain the identification results. The correction module is used to match the recognition results with the pre-stored feature geographic coordinate database, and combine the positioning correction coefficient to comprehensively correct the cumulative positioning error of inertial navigation, so as to obtain the corrected accurate location information. The processing module is used to dynamically update and replan the preliminary optimized return path based on the corrected precise location information, and control the UAV to follow the updated real-time return path until it lands precisely at the initial take-off and landing point.

9. A computing device, characterized in that, include: One or more processors; A storage device for storing one or more programs, which, when executed by one or more processors, cause the one or more processors to implement the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a program that, when executed by a processor, implements the method as described in any one of claims 1 to 7.

Images