An unmanned aerial vehicle inspection method, device, equipment and storage medium

Abstract

The present application relates to the technical field of unmanned aerial vehicle inspection, and particularly relates to an unmanned aerial vehicle inspection method, device, equipment and storage medium; an inspection area map is constructed according to inspection object information, inspection requirements and an inspection scene; the inspection area map is rasterized to obtain a rasterized map; a preset heuristic search algorithm is used to plan a path for the rasterized map to obtain an initial flight route; the initial flight route is optimized according to preset equipment performance to obtain an inspection flight route; the unmanned aerial vehicle is controlled to perform inspection work according to the inspection flight route; the scheme first collects multidimensional data to construct an integrated inspection area map, and the area is defined by rasterization; the heuristic search algorithm is used to plan an optimal initial flight route; the equipment performance is used to optimize the parameters, improve the data quality and the defect recognition rate, and adapt to multi-industry inspection, and the scheme has applicability and expansibility.

Description

Technical Field

[0001] This invention relates to the field of drone inspection technology, and in particular to a drone inspection method, apparatus, equipment and storage medium. Background Technology

[0002] In fields such as power line inspection and oil and gas pipeline monitoring, drone inspections are gradually replacing manual inspections due to their high efficiency and safety. However, existing drone inspection technologies have significant shortcomings: the path planning stage often only considers terrain obstacles, failing to fully utilize core requirements and complex scenario constraints, resulting in poor initial route adaptability; furthermore, route optimization is not precisely matched with the drone's endurance, flight speed, and other equipment performance characteristics, often leading to problems such as incomplete inspection coverage, insufficient endurance, or low efficiency, making it difficult to meet diverse and high-precision inspection needs. Therefore, there is an urgent need for an improved drone inspection solution that balances scenario adaptability and equipment compatibility. Summary of the Invention

[0003] In order to overcome the shortcomings of the prior art, the present invention aims to provide a method, apparatus, equipment and storage medium for unmanned aerial vehicle (UAV) inspection.

[0004] The first aspect of this invention provides a method for unmanned aerial vehicle (UAV) inspection, comprising: collecting inspection object information, inspection requirements, and inspection scenarios; constructing an inspection area map based on the inspection object information, inspection requirements, and inspection scenarios; performing rasterization processing on the inspection area map to obtain a rasterized map; performing path planning on the rasterized map according to a preset heuristic search algorithm to obtain an initial flight path; optimizing the initial flight path according to preset equipment performance to obtain an inspection flight path; and controlling the UAV to perform inspection work according to the inspection flight path.

[0005] Furthermore, the step of performing path planning on the raster map according to a preset heuristic search algorithm to obtain an initial route includes: extracting multiple boundary coordinates and inspection target coordinates from the raster map according to a preset geodetic coordinate system; generating a route type according to the inspection object information and the inspection scene; and performing path planning on the route type, multiple boundary coordinates, and inspection target coordinates according to the heuristic search algorithm to obtain the initial route.

[0006] Furthermore, the step of optimizing the initial route based on preset equipment performance to obtain the inspection route includes: acquiring environmental perception data and UAV status data; generating flight parameters based on inspection requirements and equipment performance; and optimizing the initial route based on the flight parameters, environmental perception data, and UAV status data to obtain the inspection route.

[0007] Furthermore, the step of controlling the UAV to perform inspection work according to the inspection route includes: dividing the inspection route according to preset division rules and multiple boundary coordinates to obtain multiple route segments; marking each route segment according to preset inspection priority rules to obtain multiple marked route segments; and controlling the UAV to perform inspection work according to the multiple marked route segments.

[0008] Furthermore, the step of generating an inspection report based on the inspection work status includes: generating positioning data based on the geodetic coordinate system and multiple marked flight line segments during the inspection work status; obtaining image data of the inspection area based on the positioning data; identifying the image data of the inspection area based on a preset defect detection model to obtain the defect location area and defect features; acquiring actual attitude data and UAV geodetic coordinate data, and analyzing the actual attitude data and defect location area to obtain defect location information; and generating an inspection report based on the defect location information, defect features, and defect location area.

[0009] Furthermore, the analysis of actual posture data and defect location area to obtain defect location information includes: analyzing the inspection area map based on the defect location area to obtain inspection terrain data; extracting coordinates of the defect location area based on a preset pixel coordinate system to obtain pixel coordinate information; analyzing the actual posture data based on a preset ideal vertical shooting posture and preset camera intrinsic parameters to obtain posture deviation; and generating defect location information based on the inspection terrain data, posture deviation, and pixel coordinate information.

[0010] Furthermore, the step of generating an inspection report based on defect location information, defect features, and defect location area includes: analyzing the inspection area map based on the defect location area to obtain LiDAR point cloud data; performing matching analysis on defect features based on a preset defect type feature library to obtain the defect type; determining multiple hovering waypoints on the inspection area map based on the defect type, inspection requirements, and LiDAR point cloud data; and generating an inspection report based on the defect location information and multiple hovering waypoints.

[0011] Furthermore, a drone inspection device includes: a data acquisition module for collecting inspection object information, inspection requirements, and inspection scenarios; a map construction module for constructing an inspection area map based on the inspection object information, inspection requirements, and inspection scenarios; a raster processing module for rasterizing the inspection area map to obtain a rasterized map; a path planning module for performing path planning on the rasterized map based on a preset heuristic search algorithm to obtain an initial flight path; a flight path optimization module for optimizing the initial flight path based on preset equipment performance to obtain an inspection flight path; and an inspection operation module for controlling the drone to perform inspection operations according to the inspection flight path.

[0012] Furthermore, a drone inspection device includes: a memory and at least one processor, wherein the memory stores instructions; at least one processor invokes the instructions in the memory to cause the drone inspection device to perform the various steps of the drone inspection method described above.

[0013] Furthermore, a computer-readable storage medium stores instructions that, when executed by a processor, implement the various steps of the unmanned aerial vehicle (UAV) inspection method described above.

[0014] In the technical solution of this invention, comprehensive information on inspection objects, inspection requirements, and scene data are collected to construct an integrated inspection area map containing spatial location, attribute constraints, and environmental features. Combined with rasterization processing, passable areas, obstacles, and inspection targets are accurately defined, laying a high-quality data foundation for planning. Heuristic search algorithms quickly plan initial routes that are fully covered, do not cross boundaries, and have the shortest mileage, reducing human error and planning costs. After equipment performance optimization, the routes are adapted to the extreme parameters of UAV speed and endurance, avoiding the risk of overload. Planning breakpoint resume nodes reduces operation interruptions, while ensuring data collection quality and improving the accuracy of defect identification. It is adaptable to multiple types of inspection objects and complex scenarios and can be widely used in routine inspections in multiple industries, combining applicability and scalability. Attached Figure Description

[0015] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the description of the embodiments taken in conjunction with the following drawings, in which: Figure 1 This is a first flowchart of a drone inspection method provided by an embodiment of the present invention; Figure 2 This is a second flowchart of a drone inspection method provided in an embodiment of the present invention; Figure 3 This is a third flowchart of a drone inspection method provided by an embodiment of the present invention; Figure 4 This is a fourth flowchart of a drone inspection method provided in an embodiment of the present invention; Figure 5 A fifth flowchart of a drone inspection method provided in an embodiment of the present invention; Figure 6 The sixth flowchart of a drone inspection method provided in this embodiment of the invention; Figure 7 A seventh flowchart of a drone inspection method provided in an embodiment of the present invention; Figure 8 This is a schematic diagram of the structure of a drone inspection device provided in an embodiment of the present invention; Figure 9 This is a schematic diagram of the structure of a drone inspection device provided in an embodiment of the present invention. Detailed Implementation

[0016] The terms "first," "second," "third," "fourth," etc. (if present) in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" or "having" and any variations thereof are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0017] For ease of understanding, the specific process of the embodiments of the present invention is described below. Please refer to [link / reference]. Figure 1 One embodiment of the unmanned aerial vehicle (UAV) inspection method of the present invention includes: 101. Collect information on inspection targets, inspection needs, and inspection scenarios; In this embodiment, the inspection object information includes object type (linear objects such as oil and gas pipelines and power transmission lines; three-dimensional objects such as power towers and bridges; planar objects such as photovoltaic module arrays), structural characteristics (such as tall, dense, and irregular), key inspection parts (such as top components and connection nodes); inspection scene analysis: identifying scene type (plains, mountains, urban building complexes, offshore platforms), and environmental constraints (such as obstacle distribution density and airspace restrictions). 102. Based on the inspection object information, inspection requirements, and inspection scenarios, an inspection area map is constructed. In this embodiment, a base geographic map (such as a GIS electronic map) that has been registered with a geodetic coordinate system (such as WGS-84) is selected as the base for the inspection area map. The collected inspection object information is labeled on the base geographic map according to type, clarifying the object location, structural outline, and coordinates of key parts. The key inspection area (such as the core equipment area) and the regular inspection area are divided according to the inspection requirements, and the accuracy requirements, collection parameters, and other constraints are marked. Scene environment information, including the three-dimensional coordinates and outlines of obstacles, terrain elevation data, airspace restrictions, meteorological sensitive areas, etc., are overlaid to verify the consistency (such as the matching of object location with terrain) and completeness (no omission of key areas or objects) of the information in each layer. Finally, an integrated inspection area map containing "spatial location, attribute constraints, and environmental characteristics" is formed. 103. Rasterize the inspection area map to obtain a rasterized map; In this embodiment, the inspection area map is divided into a regular square grid array according to a preset grid size (e.g., 10cm×10cm, which can be reduced to 5cm×5cm for key areas). Each grid is assigned a unique identifier (e.g., row and column number). Attribute information is labeled for each grid, including "accessible / inaccessible" (based on obstacles and no-fly zones), "inspection priority" (based on the key / regular area division of the demand layer), "scene adaptation parameters" (based on terrain and weather constraints of the scene layer), and "object association identifier" (whether it covers the inspection object or key parts). Invalid grids (e.g., grids that exceed the inspection area) are removed, and grids with consecutive identical attributes are merged (to improve computational efficiency). The final output is a rasterized map that the algorithm can directly recognize. 104. Perform path planning on the raster map according to the preset heuristic search algorithm to obtain the initial route; In this embodiment, the rasterized map clearly defines the passable area, obstacles, and inspection targets, providing accurate data support for planning; the heuristic search algorithm efficiently searches for the global optimal path, ensuring that the initial route fully covers the inspection targets, does not cross boundaries, and has the shortest mileage. The route balances completeness and efficiency, reduces manual planning errors and workload, lays a good foundation for subsequent equipment performance optimization, improves the efficiency and accuracy of inspection planning, adapts to inspection needs in multiple scenarios, and ensures the orderly progress of subsequent operations. 105. Optimize the initial route based on the preset equipment performance to obtain the inspection route; In this embodiment, by adapting to the performance limits of the UAV, such as maximum speed, endurance, and turning radius, the excess parts of the initial flight path are corrected to ensure the safety and feasibility of the flight path. The optimized flight path takes into account the endurance utilization rate, plans breakpoint resume flight nodes to reduce operation interruptions, and adapts to the effective working distance of the sensors to ensure the quality of data collection. This solution avoids the "disconnect between planning and execution", reduces the risk of overloaded flight, improves the landing efficiency of automated inspection, and provides highly adaptable, safe and efficient flight path support for the subsequent precise execution of UAVs. 106. Control the drone to carry out inspection work according to the inspection route; In this embodiment, comprehensive information on inspection objects, inspection requirements, and scenario data are collected to construct an integrated inspection area map containing spatial location, attribute constraints, and environmental features. Combined with rasterization processing, passable areas, obstacles, and inspection targets are accurately defined, laying a high-quality data foundation for planning. Heuristic search algorithms quickly plan initial routes that are fully covered, do not cross boundaries, and have the shortest mileage, reducing human error and planning costs. After equipment performance optimization, the routes are adapted to the extreme parameters of UAV speed and endurance, avoiding the risk of overload. Planned breakpoint resume nodes reduce operation interruptions while ensuring data collection quality and improving defect identification accuracy. It is adaptable to multiple types of inspection objects and complex scenarios and can be widely used in routine inspections in multiple industries, combining applicability and scalability.

[0018] Please see Figure 2 The second embodiment of the UAV inspection method in this invention specifically includes: 201. Extract multiple boundary coordinates and inspection target coordinates from the rasterized map according to the preset geodetic coordinate system; In this embodiment, a geodetic coordinate system (such as WGS-84) is used as the sole spatial reference, and a rasterized map (which has been georeferenced according to the geodetic coordinate system) is used as the data source for coordinate extraction. The advantage of the rasterized map is that it can quickly locate the boundary of the area and the target location, thus improving the extraction efficiency. The geodetic coordinates of two types of boundaries are identified and extracted from the rasterized map: the outer boundary of the inspection area (such as the four corner coordinates of a rectangular area and the contour vertex coordinates of an irregular area) and the inner boundary of the key inspection area (such as the range coordinates of high-voltage equipment area and bridge load-bearing structure area), forming an area range of "outer boundary constraint and inner boundary focus". The inspection objects marked in the rasterized map (i.e., inspection targets, such as power towers, bridge piers, and pipeline valves) are identified, and the core positioning coordinates of each object (such as the geodetic coordinates of the center point and the coordinates of key components) are extracted as the target points that the route must cover. The consistency (all conform to the geodetic coordinate system format) and completeness (no omission of key boundaries and no missing inspection targets) of all coordinates are verified. The coordinates are organized into a format that the algorithm can recognize (such as arrays and lists) to ensure the compatibility of subsequent path planning. 202. Generate route types based on inspection object information and inspection scenario; In this embodiment, the inspection object information is first decomposed to clarify the object type (linear / three-dimensional / area), structural characteristics, and key inspection parts; then the inspection scene is analyzed to identify the scene type (plain / mountain / city, etc.) and environmental constraints (obstacle density, airspace restrictions), and adaptation rules are established based on the two to map and generate targeted flight path types. For example, linear objects and open scenes correspond to straight-line round-trip flight paths, three-dimensional objects correspond to circular flight paths, and area objects correspond to grid scanning flight paths. Finally, basic parameters are configured for the flight path type to form a flight path type scheme that can be used for subsequent planning. 203. Based on the heuristic search algorithm, perform path planning for the route type, multiple boundary coordinates, and inspection target coordinates to obtain the initial route; In this embodiment, the route type, multiple boundary coordinates, and inspection target coordinates are used as constraint inputs to the algorithm. It is determined that the route must be within the boundary of the inspection area, cover all inspection targets, and conform to the specified route type (such as straight route / circular route / grid route). The heuristic search algorithm performs a global path search based on the constraints, plans the path with the optimal target coverage and the shortest mileage, and then smooths the turning trajectory through Bézier curves to check whether the path crosses the boundary, misses targets, and adapts to the performance of the UAV. Finally, the initial route is output. In this embodiment, using a unified geodetic coordinate system as the benchmark, multiple boundary coordinates and inspection target coordinates are efficiently extracted from the raster map to ensure that the flight path does not cross boundaries and has no missed inspections, laying a precise data foundation for planning. Based on the inspection object type (linear / three-dimensional / area) and scene constraints, a targeted flight path type is generated. With the help of a heuristic search algorithm, an initial path with optimal target coverage and shortest mileage is planned under multiple constraints. After trajectory smoothing and performance verification, it is adapted to the flight requirements of UAVs. This solution improves the efficiency and accuracy of flight path planning, reduces the cost of manual intervention, and adapts to the inspection needs of multiple industries and scenarios. It not only ensures the integrity of inspection and the optimality of the path, but also lays a solid foundation for subsequent flight path optimization and safe operation.

[0019] Please see Figure 3 The third embodiment of the UAV inspection method in this invention specifically includes: 301. Acquire environmental perception data and UAV status data; In this embodiment, environmental perception data is acquired synchronously through sensors (LiDAR, visual camera, and meteorological sensor) onboard the UAV and external systems (GIS terrain database and real-time meteorological platform). The core data includes: obstacle data (such as the 3D coordinates and outlines of power lines, buildings, and trees), terrain data (elevation, slope, and surface roughness), meteorological data (wind speed, visibility, and precipitation), and illumination data (light intensity, adapted for visual sensor operation). UAV status data is output in real-time by the flight control system, and the core data includes: remaining battery power, battery voltage, flight attitude (roll / pitch / yaw angles), sensor operating status (whether the camera and LiDAR are functioning correctly), and real-time flight speed and position. 302. Generate flight parameters based on inspection requirements and equipment performance; In this embodiment, inspection requirements include inspection accuracy, operational range (coverage area boundaries, distribution of key inspection targets), operational timeliness (inspection mileage to be completed in a single flight), and data acquisition requirements (e.g., simultaneous visible light / infrared acquisition, point cloud density ≥100 points / cm²). Equipment performance includes maximum flight speed (e.g., ≤8m / s), endurance (e.g., ≥30 minutes), sensor performance (camera focal length, lidar detection distance ≥100m), wind resistance level (e.g., ≤6), load limit, and safety threshold (minimum remaining battery power ≥30%). Based on the matching logic of "inspection requirements - equipment performance", a multi-dimensional... Flight parameters include basic parameters: flight speed (1m / s-2m / s in key areas, 3m / s-5m / s in general areas, not exceeding 80% of the equipment's maximum speed), flight altitude (3-5m from the inspection target to ensure effective sensor operating distance), shooting interval (0.1s / frame for high precision, 0.5s / frame-1s / frame for general requirements), sensor operating mode (multiple sensors operating simultaneously in key areas, single sensor operating in general areas), and safety control parameters: obstacle avoidance safety distance ≥1.5m, maximum permissible wind speed ≤5m / s, turning radius ≥5m (to avoid attitude instability). 303. Optimize the initial route based on flight parameters, environmental perception data, and UAV status data to obtain the inspection route; In this embodiment, the initial flight path is first checked to ensure it fully covers the inspection area and has no obvious logical errors (such as overlapping paths or omissions of key areas). Invalid path segments are eliminated to lay the foundation for optimization. Combined with environmental perception data (obstacle 3D coordinates, terrain elevation, and meteorological data), the flight path trajectory is adjusted using A or RRT obstacle avoidance algorithms to avoid obstacles such as buildings, trees, and high-voltage lines, while also avoiding areas with severe weather conditions such as strong winds and heavy rain, ensuring a safe distance of ≥1.5m between the flight path and obstacles. The generated flight parameters are embedded into segments according to the flight path, achieving "one parameter per segment" (e.g., low-speed, high-density shooting parameters are bound to key inspection target segments, while high-speed, regular shooting parameters are bound to regular area segments), ensuring accurate matching between parameters and the flight path scenario. This is combined with the use of drones... Status data (remaining power, battery voltage) is used to optimize the flight path to shorten the total distance (e.g., reducing turnarounds, optimizing turning paths). The flight path is also segmented based on endurance, and breakpoints for continued flight are planned (e.g., temporary return points when the battery drops to 30%) to avoid operational interruptions. Based on real-time UAV attitude data (roll / pitch / yaw angles), the slope and turning radius of the flight path are adjusted to ensure stable flight attitude (e.g., pitch angle ≤ 30°, turning radius ≥ 5m), avoiding sensor data distortion caused by rapid acceleration and sharp turns. Through simulation or short-distance test flights, the collision risk, parameter adaptability, and attitude stability of the flight path are verified. Unreasonable parts (e.g., insufficient obstacle avoidance, parameter compatibility conflicts) are corrected, and finally, a complete inspection flight path is output. In this embodiment, multi-dimensional data fusion and precise adaptation are used to optimize and upgrade inspection routes. Environmental perception data and UAV status data are collected comprehensively, and obstacle avoidance algorithms are used to avoid obstacles and severe weather, reducing safety risks such as collisions and attitude instability. Differentiated flight parameters are generated based on inspection needs and equipment performance, and segmented flight routes are embedded to ensure accuracy in key areas and efficiency in regular areas, improve data collection quality, optimize flight paths, plan breakpoint resume flights, shorten mileage and reduce operation interruptions, improve endurance utilization and overall efficiency. The solution is adaptable to complex scenarios, has a high degree of automation throughout the process, reduces human intervention and errors, is suitable for inspection in multiple industries, and combines safety, accuracy and applicability.

[0020] Please see Figure 4 The fourth embodiment of a drone inspection method according to the present invention specifically includes: 401. Divide the inspection route according to the preset division rules and multiple boundary coordinates to obtain multiple route segments; In this embodiment, the preset division rules include division by distance (e.g., every 500m is a segment, adapting to the drone's endurance and data transmission capabilities), division by inspection object (e.g., each independent inspection object (power tower, bridge pier) corresponds to a segment, adapting to object-oriented management), and division by area type (e.g., key inspection areas and regular inspection areas each constitute a segment, adapting to differentiated operational requirements). Based on the starting point of the complete inspection route, and combined with the division rules and boundary coordinates, the route is divided into multiple continuous and non-overlapping route segments. Each route segment must clearly define its core attributes: starting point coordinates, ending point coordinates, length, covered inspection objects / inspection areas, and included preset waypoints. Finally, route segment verification is performed to ensure that the coverage of all route segments completely covers the original inspection route, without omissions or overlaps. The length and complexity of a single route segment are adapted to the drone's operational capabilities (e.g., the length of a single segment does not exceed 80% of the drone's single-use endurance to avoid operational interruptions). 402. Mark each route segment according to the preset inspection priority rules to obtain multiple marked route segments; In this embodiment, the inspection priority rules include the importance of the inspection object (e.g., main facilities > branch facilities, core components > auxiliary components); historical defect risk (defects occurring ≥2 times in the past year are considered high risk, no defect records are considered low risk); and safety impact level (areas directly related to safe production > general areas). Priority levels are set based on dimension combinations (e.g., first-level priority / highest priority, second-level priority / regular priority, third-level priority / low priority). The coverage objects and regional characteristics of each route segment are compared one by one to assign a unique priority level to each route segment. Simultaneously, the priority is bound to the inspection object. Matched operational parameters (e.g., Level 1 priority: flight speed 1m / s-2m / s, shooting interval 0.1s / frame, multi-sensor synchronous acquisition; Level 2 priority: flight speed 3m / s-4m / s, shooting interval 0.5s / frame; Level 3 priority: flight speed 4m / s-5m / s, shooting interval 1s / frame) form "labeled flight line segments," which include structured data of basic attributes of the flight line segment (start / end coordinates, length), priority level, and operational parameters. Finally, a labeling consistency check is performed to ensure that the priority labeling standards are uniform for the same type of flight line segment (e.g., all trunk facility flight line segments). 403. Control the drone to carry out inspection work according to multiple marked flight path segments; In this embodiment, by dividing the flight path segments into multiple rules and prioritizing them, the system adapts to the operational capabilities of drones, enabling efficient and precise inspection. Based on the inspection priority rules, priority is marked and different operational parameters are bound to the system, allowing for refined inspection of key areas. Drones automatically execute the inspections according to the structured marked flight path segments, requiring minimal human intervention, thus improving the automation and orderliness of the operation. The defect detection rate in key areas is significantly increased, while resource allocation is optimized, reducing energy consumption and labor costs. This system is adaptable to inspection scenarios across multiple industries, enhancing the safety and adaptability of operations in complex environments.

[0021] Please see Figure 5 The fifth embodiment of the UAV inspection method in this invention further includes the following steps after the initial step: 501. During the inspection process, generate positioning data based on the geodetic coordinate system and multiple marked route segments; In this embodiment, a geodetic coordinate system (such as the WGS-84 coordinate system) is bound to multiple marked flight line segments, specifying the precise coordinates of the start, end, and intermediate waypoints of each marked flight line segment in the geodetic coordinate system (i.e., the coordinate attributes of the marked flight line segment). During the UAV inspection flight, two types of data are collected simultaneously: satellite positioning data (output from the GPS / BeiDou positioning module, providing the initial position in the geodetic coordinate system) and inertial measurement data (output from the IMU inertial measurement unit, providing the UAV's real-time attitude angles and acceleration). Based on the coordinate attributes of the marked flight line segments, the satellite positioning data and inertial measurement data are fused and calibrated: random noise from satellite positioning (such as position fluctuations caused by electromagnetic interference) is eliminated using a Kalman filter algorithm, and the cumulative error of inertial measurement is corrected using the preset path constraints of the marked flight line segments. Finally, positioning data containing "real-time geodetic coordinates of the UAV, flight speed, attitude angles (roll / pitch / yaw), the marked flight line segment number, and the preset waypoint coordinates of the marked flight line segment" is generated. The data update frequency is ≥20Hz to ensure real-time performance. 502. Obtain image data of the inspection area based on the positioning data; In this embodiment, the "real-time geodetic coordinates of the UAV" in the positioning data are compared with the preset waypoint coordinates of the marked flight line segment. When the UAV reaches the preset shooting waypoint (e.g., a shooting point every 50m) or enters the preset shooting area (e.g., the spatial range corresponding to the key inspection grid), image acquisition is automatically triggered. Two types of images are simultaneously acquired by the multi-sensor camera on the UAV: ​​① visible light images (used to identify surface defects, such as cracks and corrosion); ② infrared thermal imaging images (used to identify hidden defects, such as overheating of power equipment and pipeline leakage). The acquisition parameters (focal length, exposure time, shooting interval) are dynamically adjusted according to the inspection priority of the marked flight line segment (0.1s / image for first-priority flight line segments, 0.5s / image for second- and third-priority flight line segments). The current positioning data (UAV geodetic coordinates, attitude angle, and flight line segment number) are used as attribute tags and embedded into the acquired image data to form a one-to-one correspondence between "image content and location information". Finally, the inspection area image data is formed, which facilitates the rapid association of spatial coordinates when locating defects in the future. 503. Based on the preset defect detection model, identify the image data of the inspection area to obtain the defect location area and defect features; In this embodiment, the collected inspection area image data is input into a preset defect detection model (such as an improved YOLOv8 or Faster R-CNN model). This model needs to be pre-trained based on defect samples from specific inspection scenarios (such as power, bridges, and oil and gas). The defect detection model identifies the defective areas in the inspection area image data through feature extraction and target matching, and outputs the bounding box coordinates of the area in the pixel coordinate system (i.e., the defect location area, such as the upper left corner (u1, v1) and the lower right corner (u2, v2)). At the same time, the confidence level of the defect is marked (confidence threshold ≥ 0.8, areas below the threshold are considered as defects). Suspected defects need to be marked separately to form the identified defect location area. Finally, multi-dimensional defect features are extracted from the defect location area data. These features include visual features (texture in visible light images, such as mottled / linear, color, such as rust red / oil stain black; temperature anomalies and temperature distribution range in infrared images); morphological features (shape of the defect, such as linear / blocky / irregular, size, such as length / width / area, calculated based on pixel bounding boxes and camera parameters); and category features (defect type, such as crack, corrosion, loosening, leakage, category labels based on model training). This provides comprehensive support for the generation of subsequent inspection reports. 504. Obtain actual attitude data and UAV geodetic coordinate data, and analyze the actual attitude data and defect location area to obtain defect location information; In this embodiment, the "actual attitude data" corresponding to the image acquisition time of the defect location area is extracted from the UAV flight control system, including roll angle α, pitch angle β, and yaw angle γ (consistent with the attitude angles in the positioning data generated in step one to ensure time synchronization). Core coordinates (such as the center point pixel coordinates and boundary vertex pixel coordinates of the defect bounding box) are extracted from the defect location area as the basis for subsequent transformations. An attitude rotation matrix is ​​constructed using roll angle α, pitch angle β, yaw angle γ, and ideal vertical shooting attitude. This matrix is ​​used to describe the deviation relationship between the actual shooting attitude and the ideal vertical shooting attitude of the UAV. The core pixel coordinates (center point and boundary vertex) of the defect location area are extracted and combined with preset camera intrinsic parameters to construct a transformation link from the pixel coordinate system to the camera coordinate system. The attitude deviation is calculated using the attitude rotation matrix constructed from the attitude data. Finally, complete defect location information, including defect geodetic coordinate data, actual physical size, and the flight path segment to which it belongs, is obtained. 505. Generate an inspection report based on defect location information, defect characteristics, and defect location area; In this embodiment, high-precision positioning data is generated using the geodetic coordinate system as a unified benchmark to ensure accurate coverage of the inspection area image acquisition; relying on a scenario-based pre-trained defect detection model, defect identification and multi-dimensional feature extraction are achieved, improving identification efficiency and accuracy; by correcting shooting deviations through actual posture data and combining multi-coordinate system transformation, accurate spatial positioning of defects is achieved; finally, defect positioning information, defect features, and defect positioning area are integrated to generate a structured report, which combines visualization and maintenance guidance, supports rapid decision-making, ensures full-process data traceability, adapts to complex inspection scenarios, effectively improves the accuracy of inspection operations, and optimizes the value of the entire inspection process.

[0022] Please see Figure 6 The sixth embodiment of a drone inspection method according to the present invention specifically includes the following steps: 601. Analyze the inspection area map based on the defect location area to obtain inspection terrain data; In this embodiment, the inspection area map is pre-integrated with three-dimensional terrain information (such as GIS geographic data and LiDAR point cloud preprocessing results). Based on the defect location area (the defect range marked in the image), the corresponding inspection terrain data is accurately extracted from the inspection area map through spatial coordinate association. Specifically, this includes: the elevation of the defect location area, the surface slope (gentle / moderate / steep), the surface roughness (such as a smooth concrete surface / rugged mountain surface), the relative height of the inspection object to the ground, etc. These data directly reflect the terrain features around the defect, avoiding deviations between the actual location of the defect and the image imaging location due to terrain undulations. 602. Extract the coordinates of the defect location area according to the preset pixel coordinate system to obtain pixel coordinate information; In this embodiment, a preset pixel coordinate system (with the upper left corner of the inspection area map as the origin, the x-axis horizontal to the right, and the y-axis vertical downward) is used to extract core pixel coordinates from the defect location area (such as the defect bounding box output by the defect detection model). The pixel coordinates (u, v) of the center point of the defect bounding box are extracted first as the core positioning reference, and the pixel coordinates of the boundary vertices (such as the upper left corner (u1, v1) and the lower right corner (u2, v2)) are extracted to form complete pixel coordinate information. This pixel coordinate information directly reflects the position of the defect in the image and is the basis for subsequent calculation of the true spatial position of the defect. 603. Analyze the actual posture data based on the preset ideal vertical shooting posture and preset camera intrinsic parameters to obtain the posture deviation. In this embodiment, the actual attitude data (UAV roll angle α, pitch angle β, yaw angle γ) is synchronously output by the flight control system and is strictly time-synchronized with the image acquisition time corresponding to the defect location area. Secondly, an attitude rotation matrix is ​​constructed based on the roll angle α, pitch angle β, yaw angle γ, and the ideal vertical shooting attitude (the construction of the attitude rotation matrix must follow the right-hand coordinate system rule and be consistent with the definition of the UAV attitude parameters; the camera intrinsic parameters must be calibrated in advance using professional equipment to avoid distortion in the calculation of deviation due to focal length errors). This ideal vertical shooting attitude accurately describes the deviation relationship between the actual shooting attitude and the ideal vertical shooting attitude. Finally, combined with the preset camera intrinsic parameters (pre-calibrated fixed parameters, such as focal length and pixel size), the attitude deviation is calculated through the attitude rotation matrix, i.e., the angular deviation between the actual shooting optical axis and the ideal vertical optical axis (such as roll deviation Δα and pitch deviation Δβ), to clarify the quantified value of the imaging offset. 604. Generate defect location information based on the inspected terrain data, attitude deviation, and pixel coordinate information; In this embodiment, by accurately extracting the terrain data of the defect location area, covering key features such as altitude and slope, the positioning deviation caused by terrain undulation is effectively offset; the core coordinates of the defect are extracted using a pixel coordinate system, providing a high-precision image reference for positioning; an attitude rotation matrix is ​​constructed based on the ideal vertical shooting posture and camera intrinsic parameters to quantify the attitude deviation, correct the imaging offset caused by the tilt of the drone shooting, and improve the defect positioning accuracy. The entire process follows a standardized coordinate system and data processing rules to ensure the consistency and traceability of the positioning results, adapting to various inspection scenarios such as mountains and complex structures. The generated defect positioning information is complete and comprehensive, providing accurate spatial guidance for subsequent maintenance and reshooting, reducing on-site operation costs, and improving the efficiency and applicability of the entire inspection process.

[0023] Please see Figure 7 The seventh embodiment of a drone inspection method according to the present invention specifically includes the following steps: 701. Analyze the inspection area map based on the defect location area to obtain lidar point cloud data; In this embodiment, local point cloud data corresponding to the defect location area is accurately selected from the inspection area map to provide high-precision spatial data support for subsequent waypoint generation and 3D visualization of reports. The extracted local point cloud data is subjected to denoising processing (such as using statistical filtering algorithms to remove isolated noise points) and density enhancement (interpolating points to fill sparse areas) to ensure that the final lidar point cloud data can clearly and accurately reflect the 3D terrain of the defect location area and the structural features of the inspection object (such as the angle steel of power towers and the terrain around bridge cracks). 702. Perform matching analysis on defect features based on the preset defect type feature library to obtain the defect type; In this embodiment, the core component of the defect type feature library is the mapping relationship between "common industry defect types - multi-dimensional feature templates" (e.g., in the power industry: insulator corrosion is "mottled texture, surface depression point cloud, and slightly increased infrared temperature"; in the bridge industry: bridge deck cracks are "linear texture, depth gradient point cloud, and uniform infrared temperature"). The defect type feature library supports categorized access by industry (power, oil and gas, infrastructure). The extracted two-dimensional defect features and three-dimensional defect features are compared with the templates in the defect type feature library for similarity. A similarity threshold of ≥0.85 is set. If a match is successful, the accurate defect type is output (e.g., "crack, corrosion, loosening, leakage, deformation," etc.). 703. Based on the defect type, inspection requirements, and lidar point cloud data, determine multiple hovering waypoints on the inspection area map; In this embodiment, different defect types have different requirements for the location, height, and attitude of hovering waypoints (e.g., linear cracks require waypoints to be placed along the extension direction, while blocky corrosion requires waypoints to be placed around the cracks). The waypoint spacing and hovering height are determined according to inspection requirements (e.g., preset inspection accuracy (e.g., minimum defect identification size ≥ 0.1mm) and operation mode (only reshooting / close-range laser scanning). Based on LiDAR point cloud data, obstacles (e.g., power tower conductors, bridge crash barriers) and dangerous areas (e.g., high-voltage areas, flammable areas) around the defects are avoided to ensure the safety of hovering waypoints. Finally, multiple hovering waypoints are determined on the inspection area map based on the above three constraints (defect type, inspection requirements, and LiDAR point cloud data). 704. Generate an inspection report based on defect location information and multiple hovering waypoints; In this embodiment, LiDAR point cloud data of the defect location area is accurately extracted, providing high-quality 3D data support for subsequent analysis; relying on the defect type feature library classified by industry, accurate matching of defect types is achieved, improving the standardization and reliability of identification; differentiated hovering waypoints are planned by combining defect type, inspection requirements and LiDAR point cloud data, taking into account both coverage integrity and operational safety; finally, a report is generated by integrating defect location information and multiple hovering waypoints, which combines 3D visualization and practical guidance, improving inspection accuracy and operational efficiency, reducing labor costs, enhancing adaptability to complex scenarios, providing accurate support for subsequent reshooting and maintenance, and optimizing the value of the entire inspection process.

[0024] The above describes a drone inspection method according to an embodiment of the present invention. The following describes a drone inspection device according to an embodiment of the present invention. Please refer to [link / reference]. Figure 8 One embodiment of the unmanned aerial vehicle (UAV) inspection device of the present invention includes: Data acquisition module 1 is used to collect information on inspection objects, inspection requirements, and inspection scenarios; Map building module 2 is used to build an inspection area map based on inspection object information, inspection requirements and inspection scenarios; Raster processing module 3 is used to rasterize the inspection area map to obtain a rasterized map; Path planning module 4 is used to plan paths on the raster map according to a preset heuristic search algorithm to obtain an initial route. The route optimization module 5 is used to optimize the initial route based on the preset equipment performance to obtain the inspection route; Inspection module 6 is used to control the drone to carry out inspection work according to the inspection route. In this embodiment, comprehensive information on inspection objects, inspection requirements, and scenario data are collected to construct an integrated inspection area map containing spatial location, attribute constraints, and environmental features. Combined with rasterization processing, passable areas, obstacles, and inspection targets are accurately defined, laying a high-quality data foundation for planning. Heuristic search algorithms quickly plan initial routes that are fully covered, do not cross boundaries, and have the shortest mileage, reducing human error and planning costs. After equipment performance optimization, the routes are adapted to the extreme parameters of UAV speed and endurance, avoiding the risk of overload. Planned breakpoint resume nodes reduce operation interruptions while ensuring data collection quality and improving defect identification accuracy. It is adaptable to multiple types of inspection objects and complex scenarios and can be widely used in routine inspections in multiple industries, combining applicability and scalability.

[0025] Figure 9 This is a schematic diagram of the structure of a drone inspection device 900 provided in an embodiment of the present invention. The drone inspection device 900 can vary significantly due to different configurations or performance. It may include one or more central processing units (CPUs) 910 (e.g., one or more processors) and a memory 920, and one or more storage media 930 (e.g., one or more mass storage devices) for storing application programs 933 or data 932. The memory 920 and storage media 930 can be temporary or persistent storage. The program stored in the storage media 930 may include one or more modules (not shown in the diagram), each module may include a series of instruction operations on the drone inspection device 900. Furthermore, the processor 910 may be configured to communicate with the storage media 930 and execute a series of instruction operations in the storage media 930 on the drone inspection device 900 to implement the steps of the drone inspection method provided in the above-described method embodiments.

[0026] A drone inspection device 900 may further include one or more power supplies 940, one or more wired or wireless network interfaces 950, one or more input / output interfaces 960, and / or one or more operating systems 931, such as Windows Server, MacOSX, Unix, Linux, FreeBSD, etc. Those skilled in the art will understand that... Figure 9 The structure of the drone inspection equipment shown does not constitute a limitation on a drone inspection equipment. It may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0027] A computer-readable storage medium storing instructions that, when executed by a processor, implement the steps of the unmanned aerial vehicle (UAV) inspection method described above.

[0028] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the system, device, or unit described above can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0029] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0030] Finally, it should be noted that the above descriptions are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for unmanned aerial vehicle (UAV) inspection, characterized in that, include: Collect information on inspection targets, inspection needs, and inspection scenarios; An inspection area map is constructed based on the inspection object information, inspection requirements, and inspection scenarios. The inspection area map is rasterized to obtain a rasterized map; The raster map is used to plan routes based on a pre-defined heuristic search algorithm to obtain an initial flight path. The initial route is optimized based on the preset equipment performance to obtain the inspection route; The drone is controlled to carry out inspection work according to the inspection route, and an inspection report is generated based on the inspection work status.

2. The UAV inspection method as described in claim 1, characterized in that, The step of performing path planning on the raster map according to a preset heuristic search algorithm to obtain an initial flight path includes: Multiple boundary coordinates and inspection target coordinates are extracted from the rasterized map based on the preset geodetic coordinate system; Generate route types based on inspection object information and inspection scenario; The initial route is obtained by using a heuristic search algorithm to plan the route type, multiple boundary coordinates and inspection target coordinates.

3. The UAV inspection method as described in claim 1, characterized in that, The optimization of the initial route based on preset equipment performance to obtain the inspection route includes: Acquire environmental perception data and drone status data; Flight parameters are generated based on inspection requirements and equipment performance. The initial flight path is optimized based on flight parameters, environmental perception data, and UAV status data to obtain the inspection route.

4. The UAV inspection method as described in claim 2, characterized in that, The process of controlling the drone to perform inspection work according to the inspection route includes: The inspection route is divided according to the preset division rules and multiple boundary coordinates to obtain multiple route segments; Each route segment is marked according to the preset inspection priority rules to obtain multiple marked route segments; The drone is controlled to carry out inspection work based on multiple marked flight path segments.

5. The UAV inspection method as described in claim 4, characterized in that, The process of generating an inspection report based on the inspection work status includes: During the inspection process, positioning data is generated based on the geodetic coordinate system and multiple marked route segments; Image data of the inspection area is obtained based on the positioning data; The image data of the inspection area is identified according to the preset defect detection model in order to obtain the defect location area and defect features; Acquire actual attitude data and UAV geodetic coordinate data, and analyze the actual attitude data and defect location area to obtain defect location information; An inspection report is generated based on defect location information, defect characteristics, and defect location area.

6. The UAV inspection method as described in claim 5, characterized in that, The analysis of actual posture data and defect location areas to obtain defect location information includes: The inspection area map is analyzed based on the defect location area to obtain inspection terrain data; The coordinates of the defect location area are extracted according to the preset pixel coordinate system to obtain pixel coordinate information; The actual posture data is analyzed based on the preset ideal vertical shooting posture and preset camera intrinsic parameters to obtain the posture deviation. Defect location information is generated based on the inspection terrain data, attitude deviation, and pixel coordinate information.

7. The UAV inspection method as described in claim 5, characterized in that, The process of generating an inspection report based on defect location information, defect characteristics, and defect location area includes: The inspection area map is analyzed based on the defect location area to obtain lidar point cloud data; The defect features are matched and analyzed according to the preset defect type feature library to obtain the defect type; Based on the defect type, inspection requirements, and lidar point cloud data, determine multiple hovering waypoints on the inspection area map; An inspection report is generated based on defect location information and multiple hovering waypoints.

8. A drone inspection device, characterized in that, include: The data acquisition module is used to collect information on inspection objects, inspection requirements, and inspection scenarios. The map building module is used to build a map of the inspection area based on the inspection object information, inspection requirements and inspection scenarios. The raster processing module is used to rasterize the inspection area map to obtain a rasterized map. The path planning module is used to plan paths on the raster map according to a preset heuristic search algorithm to obtain an initial flight path. The route optimization module is used to optimize the initial route based on the preset equipment performance to obtain the inspection route; The inspection module is used to control the drone to carry out inspection work according to the inspection route.

9. A drone inspection device, characterized in that, include: A memory and at least one processor, wherein the memory stores instructions; At least one of the processors invokes the instructions in the memory to cause the unmanned aerial vehicle (UAV) inspection device to perform the steps of the UAV inspection method as described in any one of claims 1-7.

10. A computer-readable storage medium storing instructions thereon, characterized in that, When the instructions are executed by the processor, they implement the various steps of the UAV inspection method as described in any one of claims 1-7.

Images