An unmanned aerial vehicle inspection path planning method, system, device and storage medium

Abstract

The present application relates to the technical field of unmanned aerial vehicle navigation, and relates to an unmanned aerial vehicle inspection path planning method, system, device and storage medium; wherein the unmanned aerial vehicle inspection path planning method comprises: obtaining a digital elevation model, a digital surface model of a plateau forest area and laser point cloud data of the plateau forest area; determining a three-dimensional voxel space of the plateau forest area based on the digital elevation model, the digital surface model and the laser point cloud data, and determining an unmanned aerial vehicle accessible region by performing accessibility determination on the three-dimensional voxel space based on a vertical safety margin; performing three-dimensional path search from a starting point to an ending point in the unmanned aerial vehicle accessible region by using an improved algorithm; performing smoothing processing on the searched three-dimensional path, and imposing a maximum turning radius constraint and a maximum climb rate constraint to generate a final three-dimensional track. The present application can improve the safety and executability of the planned route of the unmanned aerial vehicle.

Description

Technical Field

[0001] This invention relates to the field of unmanned aerial vehicle (UAV) navigation technology, and in particular to a method, system, device, and storage medium for UAV inspection path planning. Background Technology

[0002] High-altitude forest regions are characterized by high altitude, significant topographic relief, closed canopies, and dense understory obstacles, posing significant challenges to UAVs flying at low altitudes near and under the forest canopy. Complex mountainous terrain fragments navigable areas, and elevation differences limit the safe flight altitude of UAVs. Dense canopies cause GNSS (Global Navigation Satellite System) signal attenuation and multipath errors, leading to unstable positioning. The thin air at high altitudes reduces propeller lift, making it difficult for aircraft to perform sharp turns and rapid climbs. Simultaneously, frequent gusts and turbulence in high-altitude regions dynamically change the boundaries of navigable understory paths, further increasing flight uncertainty. In this environment, UAV path planning not only needs to perform basic obstacle avoidance but also must strictly control climb rate and turning radius while allowing sufficient safety margins, placing extremely high demands on the spatial understanding, efficiency, and robustness of the planning algorithm.

[0003] Existing 3D path planning algorithms, such as traditional Algorithms, Dijkstra's algorithm, RRT, etc., have significant shortcomings when applied to such environments. Traditional Algorithms typically rely solely on Euclidean distance to design heuristic functions, lacking consideration for vertical factors such as terrain undulation and canopy thickness. This often results in planned paths that "skim over obstacles," involve numerous sharp turns, and have low search efficiency, failing to meet the stringent requirements of high-altitude environments for flight safety margins, turning radii, and climb rates. Dijkstra's algorithm, with its low search efficiency, is unsuitable for online applications. Random sampling algorithms like RRT often generate circuitous, coarse paths with poor safety. Therefore, a three-dimensional path planning scheme that balances search efficiency, path smoothness, safety margins, and task adaptability is urgently needed to support safe and efficient ecological monitoring tasks performed by UAVs in high-altitude forest areas. Summary of the Invention

[0004] This application aims to at least solve the technical problems existing in the prior art and provide a method, system, device and storage medium for unmanned aerial vehicle (UAV) inspection path planning.

[0005] In a first aspect, the present invention provides a method for planning unmanned aerial vehicle (UAV) inspection paths, the method comprising: Obtain digital elevation models, digital surface models, and data of the plateau forest region; The three-dimensional voxel space of the plateau forest area is determined based on digital elevation model, digital surface model and laser point cloud data, and the accessibility of the three-dimensional voxel space is determined based on vertical safety margin to determine the area accessible by UAV. Within the drone's reach area, utilizing based on The heuristic search algorithm performs a three-dimensional path search from the starting point to the destination; the algorithm based on The heuristic search algorithm uses a terrain-aware heuristic function to generate target flight nodes when performing 3D path search, and uses a multi-factor cost function that includes corner penalty and obstacle proximity penalty to calculate the inter-node cost between the current position node and the target flight node. The 3D path of the UAV is obtained based on the target flight node and the inter-node cost. The obtained 3D path is smoothed and constraints on the maximum turning radius and maximum climb rate are applied to generate the final 3D track.

[0006] Optionally, the vertical safety margin is a parameter representing the flight safety status of the UAV, and the vertical safety margin is determined based on the UAV's basic safety altitude, wind-induced sway amplitude, and UAV positioning error.

[0007] Optionally, the step of determining the three-dimensional voxel space of the plateau forest area based on digital elevation models, digital surface models, and laser point cloud data, and determining the reachability of the three-dimensional voxel space based on vertical safety margin to determine the reachable area of ​​the UAV includes: The canopy thickness of the plateau forest area was calculated based on the digital elevation model and digital surface model, and the canopy thickness of the plateau forest area was locally corrected using laser point cloud data to obtain the corrected canopy thickness information. The three-dimensional voxel space of the plateau forest area was determined based on laser point cloud data and corrected canopy thickness information. Accessibility is determined in three-dimensional voxel space based on vertical safety margin to identify the reachable area of ​​the UAV.

[0008] Optionally, the expression for the terrain-aware heuristic function is: ; in, This represents the terrain-aware heuristic function. This represents the three-dimensional Euclidean distance from the current position to the target point. Indicates a high penalty weight. This indicates taking the maximum value of the two parameters within the parentheses. This represents the height value of the current position. This represents the height value of the target point.

[0009] Optionally, the expression for the multi-factor cost function is: , in, The adjacency distance, The weighting coefficient for the corner penalty term. Penalty for turning corners; The weighting coefficient for the obstacle proximity penalty. Penalty for proximity to obstacles; This represents the index of a 3D mesh node.

[0010] Optionally, cubic B-splines can be used to smooth the obtained 3D path.

[0011] Optionally, the three-dimensional path search employs jump point search and / or bidirectional expansion strategy and / or adaptive step size expansion strategy.

[0012] Secondly, the present invention provides a drone inspection path planning system, the system comprising: The acquisition module is used to acquire digital elevation models, digital surface models, and laser point cloud data of the plateau forest area. The processing module is used to determine the three-dimensional voxel space of the plateau forest area based on the digital elevation model, digital surface model and laser point cloud data, and to determine the reachability of the three-dimensional voxel space based on the vertical safety margin to determine the reachable area of ​​the UAV. The path search module is used to search within the drone's reach area using a path-based approach. The heuristic search algorithm performs a three-dimensional path search from the starting point to the destination; the algorithm based on The heuristic search algorithm uses a terrain-aware heuristic function to generate target flight nodes when performing 3D path search, and uses a multi-factor cost function that includes corner penalty and obstacle proximity penalty to calculate the inter-node cost between the current position node and the target flight node. The 3D path of the UAV is obtained based on the target flight node and the inter-node cost. The trajectory generation module is used to smooth the searched 3D path and apply maximum turning radius constraints and maximum climb rate constraints to generate the final 3D trajectory.

[0013] Thirdly, the present invention provides an electronic device, the electronic device comprising: At least one processor; and, A memory communicatively connected to the at least one processor; wherein, The memory stores a computer program that can be executed by the at least one processor, which enables the at least one processor to perform the UAV inspection path planning method described above.

[0014] Fourthly, the present invention also provides a computer-readable storage medium storing at least one computer program, which is executed by a processor in an electronic device to implement the above-described UAV inspection path planning method.

[0015] In summary, this application includes the following beneficial technical effects: Based on the vertical safety margin, the reachability of the UAV is determined by the three-dimensional voxel space. The obstacle proximity penalty term of the multi-factor cost function enables the UAV to actively stay away from obstacles when planning its path, which improves the overall safety of the UAV flight path and reduces the possibility of the UAV flying close to obstacles. The corner penalty term of the multi-factor cost function and the operation of smoothing the searched three-dimensional path make the generated UAV path have gentle corners, which are consistent with the dynamic characteristics of UAV in the thin air of the plateau and improve the feasibility of the flight path. By using terrain-aware heuristics to guide the search direction, the algorithm significantly reduces the number of node visits and planning time while ensuring path quality, thus improving algorithm efficiency. Attached Figure Description

[0016] Figure 1 This is a flowchart illustrating a method for planning unmanned aerial vehicle (UAV) inspection paths according to an embodiment of the present invention. Figure 2 Simulation rendering of forest land on steep slopes in a plateau; Figure 3 For tradition Algorithms and based on The heuristic search algorithm generates a 3D path projected onto the XY plane, and the traditional Algorithms and based on The heuristic search algorithm generates a 3D path that is projected onto the XZ plane. Figure 4 A comparative diagram of 3D paths generated by multiple algorithms in a high-altitude steep slope forest scene; Figure 5 Based on A comparative diagram of 3D paths generated by different combinations of heuristic search algorithms; Figure 6 For discrete stand density and canopy thickness combinations based on Heuristic search algorithms and traditional A diagram showing the performance changes of the algorithms; Figure 7 Based on the baseline canopy Heuristic search algorithms and traditional A schematic diagram illustrating the stand density sensitivity of the algorithm; Figure 8A schematic diagram of the results of the plateau topography-forest stand coupling adaptation experiment; Figure 9 This is a schematic diagram of the structure of an electronic device for implementing the UAV inspection path planning method according to an embodiment of the present invention.

[0017] Reference numerals: 10, processor; 11, memory; 12, communication bus; 13, communication interface.

[0018] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0019] Embodiments of the present invention are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0020] In the description of this invention, it should be understood that the terms "longitudinal", "lateral", "up", "down", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0021] In the description of this invention, unless otherwise specified and limited, it should be noted that the terms "installation", "connection" and "linking" should be interpreted broadly. For example, they can refer to mechanical or electrical connections, or internal connections between two components. They can be direct connections or indirect connections through an intermediate medium. Those skilled in the art can understand the specific meaning of the above terms according to the specific circumstances.

[0022] Reference Figure 1 The diagram shown is a flowchart illustrating a drone inspection path planning method according to an embodiment of the present invention. In this embodiment, the drone inspection path planning method includes: S1. Obtain digital elevation model, digital surface model and laser point cloud data of the plateau forest area.

[0023] Digital Elevation Model (DEM) and Digital Surface Model (DSM) are two core data models used in Geographic Information Systems (GIS) to represent surface elevation. Their main difference lies in whether or not they include height information of surface objects.

[0024] The full name of a Digital Elevation Model is Digital Elevation Model, abbreviated as DEM. A DEM is a dataset consisting of the planar coordinates (X, Y) and elevation (Z) of regular grid points within a certain area. It only describes the exposed topographic relief of the ground and does not include the height information of any surface objects such as buildings, vegetation, and bridges. A DEM is generated by filtering out features from point cloud data or by interpolating from topographic maps.

[0025] The full name of a digital surface model is Digital Surface Model, abbreviated as DSM. It refers to a three-dimensional model that contains the elevation information of the highest point of the top of all objects on the earth's surface (such as buildings, trees, power lines, bridges, etc.), and reflects the complete and realistic earth's surface morphology from the ground to the top of the earth's surface cover.

[0026] S2. Based on digital elevation model, digital surface model and laser point cloud data, determine the three-dimensional voxel space of the plateau forest area, and determine the reachability of the three-dimensional voxel space based on vertical safety margin to determine the reachable area of ​​the UAV.

[0027] Vertical safety margin is a parameter representing the flight safety status of a drone. It is determined based on the drone's basic safe altitude, wind-induced sway amplitude, and drone positioning error.

[0028] Specifically, a three-dimensional voxel space for the plateau forest area is determined based on digital elevation models, digital surface models, and laser point cloud data. Accessibility of this three-dimensional voxel space is then assessed based on vertical safety margins to determine the reachable area for the UAV, including: S21. Calculate the canopy thickness of the plateau forest area based on the digital elevation model and digital surface model, and use laser point cloud data to locally correct the canopy thickness of the plateau forest area to obtain the corrected canopy thickness information. S22. Determine the three-dimensional voxel space of the plateau forest area based on the laser point cloud data and the corrected canopy thickness information of the plateau forest area; S23. Based on the vertical safety margin, determine the accessibility of the three-dimensional voxel space to identify the reachable area of ​​the UAV.

[0029] For plateau forest areas characterized by large topographic relief, thick canopy and dense obstacles, the traditional A* algorithm is systematically improved from five aspects: accessibility modeling, heuristic design, cost function, parameter adaptation and path smoothing, so that the generated three-dimensional track has both safety, smoothness and monitoring adaptability.

[0030] Canopy thickness is constructed based on DEM / DSM difference model. The formula for calculating the canopy thickness of nodes in three-dimensional voxel space is as follows: By combining point cloud data with local canopy thickness correction, the vertical structure of the forest canopy can be obtained.

[0031] Further define the vertical safety margin for drone flight: ; That Based on the safety height, Indicates the amplitude of wind-induced swaying. This represents the amplification term of the positioning error in the Global Navigation Satellite System. The node reachability criterion is defined as: A passageway is considered passable only if the voxels of a node in the three-dimensional voxel space satisfy this condition. In this way, the model can simultaneously reflect the impact of canopy thickness, surface undulation, and uncertain disturbances in a plateau environment on the safe passage.

[0032] S3. Within the drone's reach area, utilize based on The heuristic search algorithm performs a three-dimensional path search from the starting point to the end point.

[0033] First, regarding the traditional The algorithm is introduced here. In UAV 3D path planning, the workspace is usually discretized into a 3D voxel grid structure, denoted as an undirected graph. Where V is the set of nodes and E is the set of edges. In a voxel grid, the center point of each voxel is considered a node. If two adjacent voxels are both free space, then a walkable edge is considered to connect the two nodes. In a 3D mesh, to ensure path connectivity, a 26-neighborhood expansion method is typically used to achieve spatial search in all directions. The cost between adjacent nodes is generally taken as the Euclidean distance: ; in and These are the coordinates of the two nodes. The cumulative cost of the path can be expressed as: ; The algorithm selects the current expansion node using the following formula: ; in From the starting point to the current node The actual cost, For spatial heuristics, used to estimate from The minimum possible cost to reach the target node. When When sufficient permissibility (not exceeding the actual cost) and consistency conditions are met, The algorithm guarantees that it finds the optimal path relative to the defined cost. A commonly used baseline heuristic in 3D path planning is the 3D Euclidean distance or its decomposition: ; In this embodiment, the improved When performing 3D path search, the algorithm uses a terrain-aware heuristic function to generate target flight nodes and a multi-factor cost function that includes corner penalty and obstacle proximity penalty to calculate the inter-node cost between the current position node and the target flight node. The 3D path of the UAV is obtained based on the target flight node and the inter-node cost.

[0034] The expression for the terrain-aware heuristic function is: ; in, This represents the terrain-aware heuristic function. This represents the three-dimensional Euclidean distance from the current position to the target point. Indicates a high penalty weight. This indicates taking the maximum value of the two parameters within the parentheses. This represents the height value of the current position. This represents the height value of the target point.

[0035] To balance path smoothness and safety, this application is based on The heuristic search algorithm also improves the cost function; the expression for the multi-factor cost function is as follows: , in, The adjacency distance, The weighting coefficient for the corner penalty term. This is a corner penalty used to suppress sharp turns. ; The weighting coefficient for the obstacle proximity penalty. The penalty is based on proximity to the obstacle; the closer to the obstacle, the greater the penalty. This represents the index of the 3D mesh node; through this cost function, the algorithm can avoid obstacle-hitting flight and generate a smooth trajectory.

[0036] In addition, to adapt the algorithm to different forest stand densities and topographical conditions, improvements were made. The algorithm also introduces adaptive weight mapping, which adaptively adjusts the weight coefficients of the corner penalty term and the obstacle proximity penalty term: , ; in, This refers to the local canopy thickness (the thicker the canopy, the further away it needs to be from obstacles). The width of the local passageway is considered; the narrower the local passageway, the more penalties are required for obstacles. and This is a constant coefficient. The more complex the environment (i.e., thick canopy, narrow channel), the more conservative the algorithm tends to be, increasing the safety margin.

[0037] The 3D path search employs jump point search (JPS) and / or bidirectional expansion strategies and / or adaptive step size expansion strategies to reduce the number of nodes to be expanded and improve search efficiency.

[0038] S4. Smooth the obtained 3D path and apply maximum turning radius constraints and maximum climb rate constraints to generate the final 3D track.

[0039] Tradition The output discrete path is a piecewise linear sequence and cannot be directly used for flight control. In the preferred embodiment of this example, cubic B-splines are used to smooth the searched three-dimensional path.

[0040] The formula for cubic B-spline processing is: ; The smoothed 3D path, Describes the B-spline basis functions. These are base function indices, used to indicate different corresponding base functions; The parameter used to determine the position on the curve takes values ​​within a certain range; These are control points, which form a set of points. The positions of these control points determine the approximate shape of the cubic B-spline curve. These represent different control points.

[0041] By applying the maximum curvature constraint to the maximum turning radius constraint, and then applying both the maximum turning radius constraint and the maximum climb rate constraint, the optimized path satisfies aerodynamic characteristics while ensuring smoothness and safety, resulting in better performance.

[0042] In practical implementation, reachable voxels can first be labeled using DEM / DSM and point cloud data, and then improved operations can be performed on this reachable space. The search yields discrete path points, which are then smoothed and constrained.

[0043] To evaluate the proposal based on this application The applicability of the heuristic search 3D path planning algorithm to UAV flight missions in plateau forest areas, referencing Figure 2 A three-dimensional simulation environment of plateau forest land combining typical slope topography features and understory vegetation structure was constructed. This environment comprehensively reflects common elements in plateau mountain forest areas at altitudes of 1500-2000m, such as slope aspect, topographic relief, tree density, and canopy shading, and can simulate the spatial constraints of drone flight under the forest in real-world scenarios.

[0044] The simulation area was set at 500m × 500m, corresponding to the length of a typical multi-rotor UAV's inspection path (300-600m) completed in a high-altitude region at a speed of 3-5 m / s and a flight time of 10-15 minutes. This scale ensures that the path length in the experiment is representative while covering a sufficiently rich variety of terrain and vegetation types within an acceptable computational cost. To avoid oversimplifying the path planning problem by using only a small-scale environment, this study uses a medium-sized local slope segment, allowing the UAV's path to undergo multiple terrain changes, obstacle avoidance decisions, and altitude adjustments, thereby more comprehensively verifying the algorithm's performance.

[0045] Plateau and mountainous regions typically exhibit distinct slope aspects and significant elevation differences. To reflect this characteristic, this study employs a linear slope along the main direction and multi-scale spatially correlated noise to generate a digital elevation model (DEM). A linear slope increasing along the X-direction is established, resulting in an overall elevation difference of approximately 120m within a 500m range, corresponding to an average slope of 13°–25°, consistent with common slope shapes in plateau and mountainous forests. Superimposed multi-scale noise is used to simulate natural textures such as surface roughness, terraces, gullies, and small-scale subsidence, ensuring the slope does not appear as a simplified plane but rather reflects realistic surface variations. Gaussian filtering is also considered for smoothing to avoid abrupt topographical changes and ensure continuous surface elevation distribution and natural slope transitions. The generated DEM includes both large-scale slope and surface texture variations, with the main slope consistent with typical plateau and mountainous landforms.

[0046] To construct a typical narrow forest passage environment, vegetation elements at different levels, including trees, shrubs, and ground cover, were deployed on the DEM. At least 450 trees were randomly generated using a Poisson distribution, with the following parameters: tree height 12-30m, crown radius 3-7m, trunk position completely random and non-overlapping, vertical tree trunk height 25%-35% of tree height, and crown height 60%-70%. This structure simulates the characteristics of uneven tree height and canopy distribution in mixed coniferous and broadleaf forests in plateau regions. 2000 shrub clumps, 1-3m in height, were randomly generated with a significantly higher density than trees, forming areas that act as forest obstacles, passage boundaries, and localized shading, interfering with the flight altitude of drones. Grass texture rendering was used to simulate ground cover on the slope, providing a natural background.

[0047] In summary, the simulation environment constructed in this study has typical terrain features, vegetation layers, and spatial complexity of a plateau forest area, which can support the comprehensive verification of UAV three-dimensional path planning algorithms.

[0048] To verify the basis proposed in this application To explore the application value of heuristic search algorithms in complex slope environments of plateau forest areas, this application sets up a baseline comparison experiment, comparing it with traditional... A comparison was made in the same scenario. This application is based on... The core idea of ​​the heuristic search algorithm is: in classical The framework introduces constraints and preferences more suited to the realities of high-altitude forest environments, including a terrain-aware heuristic function oriented towards uphill cost, a multi-factor cost function that balances trajectory smoothness and obstacle avoidance safety, and a search acceleration strategy suitable for large-scale grid environments. After the search is complete, cubic B-spline smoothing and maximum climb rate constraints are used to improve the executability of the trajectory at the flight control level. This contrasts with traditional methods that rely solely on Euclidean distance as a heuristic, use path length as the single cost, and employ unidirectional expansion. In comparison, this application is based on The heuristic search algorithm aims to simultaneously obtain smoother path morphology, more sufficient obstacle safety margin, and higher search efficiency in plateau slope forest land.

[0049] Baseline contrast experiment in Figure 2 The simulation was conducted in a high-altitude steep slope forest environment, and both shared completely identical external conditions: using the same three-dimensional occupancy grid (determined by DEM, vegetation point cloud and safety margin), the same spatial resolution and neighborhood structure, and the same start and end point settings, so that the UAV had to fly from the middle and lower part of the slope to the middle and upper part, traversing multiple understory passages and slope change ranges.

[0050] Table 1. Quantitative results of baseline comparison experiment Table 1 shows the quantitative results of the baseline comparison experiment. As can be seen from the table, based on... The heuristic search algorithm has a path length that is only slightly shorter than the traditional one. With a slight increase in algorithm performance, the average turning angle was significantly reduced while the minimum safety margin was significantly improved, indicating that the improved strategy effectively improved the smoothness of the flight path and the safety of flight in the forest without significantly sacrificing path cost. Figure 3 The shown 3D path and XY and XZ plane projections can be observed intuitively, unlike traditional methods. The algorithm tends to follow the "geometric shortest path" through the forest stand, with local paths close to the tree canopy and shrub edges, exhibiting distinct zigzag inflection points; while based on The heuristic search algorithm, under the combined effect of corner penalty and obstacle penalty, actively chooses to bypass the dense area of ​​local obstacles, resulting in a smoother and more continuous overall trajectory, and maintaining a more sufficient spatial distance from obstacles.

[0051] Table 1 also shows that, based on The heuristic search algorithm has significantly higher node visit count and planning time than the traditional one. The algorithm demonstrates that, after incorporating terrain-aware heuristics and multi-factor costs, the search process becomes more conservative and the state space exploration is more thorough. This result indicates that in mission scenarios with complex environments and high flight risks, such as high-altitude forest areas, smoother and obstacle-free paths have higher engineering value. However, in applications with relatively simple environments or stronger real-time requirements, the traditional A* algorithm can still serve as a low-computational-cost alternative.

[0052] To further evaluate based on Compared to the overall performance of different types of path planning methods, Dijkstra's heuristic search algorithm and traditional methods were selected. algorithm, With based Four heuristic search algorithms were compared under the same high-altitude steep-slope forest simulation environment and start-end point settings as the baseline experiment. The four algorithms shared a unified grid map, occupancy determination, and neighborhood structure. Dijkstra's algorithm served as the non-heuristic global search baseline. As a classic example of heuristic search, RRT* represents a global planning method based on random sampling. Heuristic search algorithms in traditional The framework incorporates mechanisms such as terrain-aware heuristics, multi-factor costs, and obstacle-related penalties.

[0053] The path distribution of the multi-algorithm comparison experiment is as follows: Figure 4 As shown, the quantitative indicators are shown in Table 2.

[0054] It can be seen that, under the same high-altitude steep slope forest environment and start-end point settings, Dijkstra's method and the traditional method... The paths obtained by the algorithms largely overlap in the XY plane, indicating that both algorithms can converge to the same geometric shortest path when their cost functions are completely identical. Table 2 shows that Dijkstra's node visit count and planning time are significantly higher than those of the traditional algorithm. The algorithm verifies that heuristic search significantly improves search efficiency in large-scale grid environments. The path of the algorithm exhibits a clear "wandering" characteristic in the figure. Compared with the other three algorithms, it detours further, has greater local sway, a significantly longer path length, and the largest average turning angle. Although it has the shortest planning time, the geometry of the flight path is more tortuous, which is not conducive to the attitude stability control and mission repeatability of the UAV.

[0055] based on The path of the heuristic search algorithm in the graph is generally between the shortest line and... Between lengthy paths, relative to Dijkstra and traditional The path deviates significantly from the canopy boundary when traversing areas densely populated with trees and shrubs, tending to bypass areas with relatively sparse obstacles and gentler slopes. The path exhibits fewer broken segments, gentler turning angles, and maintains a more reasonable buffer distance from most of the canopy. The results in Table 2 further demonstrate that, based on… The heuristic search algorithm effectively reduces the average turning angle and path geometric complexity with only a slight increase in path length, making the trajectory superior to traditional methods in terms of smoothness and executability. Meanwhile, the planning time and node access count are significantly lower than Dijkstra's. Overall, in high-altitude forest environments characterized by steep slopes and high-density vegetation barriers, based on... Compared to Dijkstra's and traditional heuristic search algorithms, and It achieves a better balance between path quality, flight safety and computational efficiency, and is more suitable as the main algorithm for path planning of UAVs (Unmanned Aerial Vehicles) in plateau forest areas.

[0056] For analysis based on To assess the contribution of each submodule of the heuristic search algorithm to the overall performance, this application designs four strategy combinations for ablation comparison under the same plateau steep slope forest scenario and start-end point settings as the aforementioned experiments. Improved It includes three key improvements: terrain awareness heuristics (H_{terrain}); corner penalty (C_{turn}); and obstacle clear penalty (C_{clear}). Based on this, it will be based on... The heuristic search algorithm is configured with four modes: Table 3. Illustration of Improved A* Configuration Mode Note: Mean(M) is the average distance from all sampling points on the path to the nearest obstacle; Q10(M) is the 10th percentile of this distance distribution.

[0057] The quantitative results for each configuration are shown in Table 3, and the XY plane projections of the corresponding paths in typical scenarios are shown in Table 3. Figure 5 As shown in the table, configuration C, which removes the corner penalty, significantly increases the average corner angle, reaching several times that of the full configuration A, even with little change in path length. Figure 5 The corresponding flight path exhibits a distinct "zigzag" pattern, indicating that C_{turn} is a key module for ensuring geometrical smoothness of the path and suppressing sharp turns. In contrast, the complete configuration A, as well as B and D with C_{turn} retained, significantly outperform configuration C in terms of average turning angle, and their flight paths are more continuous and smoother on the plane, better meeting the requirements of UAV attitude control.

[0058] From the perspective of safety margin-related indicators, the four configurations show significant differences in the average safety margin (Mean(M)) and the 10th percentile safety margin (Q10(M)). Configuration D is significantly lower than the complete configuration A in both of these indicators, while its path length is significantly shorter. This indicates that C_{clear} prompts the algorithm to maintain a larger spatial distance from trees and shrubs for most flight segments, reducing the risk of prolonged obstacle-hugging flight, rather than simply improving the distance to a very few of the most dangerous points; Figure 4 As can be seen, the path in configuration D is closer to the treetop boundary, while the path in full configuration A tends to be closer to areas with sparser obstacles. The terrain-aware concept H_{terrain} has a relatively limited impact on planar geometric parameters such as path length and turning angle in the current scenario. The full configuration A and configuration B (without H_{terrain}) show little difference in L, θ, and average safety margin, but there are some differences in planning time and search behavior. H_{terrain} helps suppress ineffective expansion into areas with significant steep slopes, and its advantages are mainly reflected in the optimization of the slope ascent process and height profile, which will be further analyzed in subsequent plateau sensitivity experiments.

[0059] In summary, C_{turn} primarily improves path smoothness, C_{clear} primarily enhances the overall safety margin, and H_{terrain} improves the algorithm's adaptability to steep terrain. These three strategies are functionally complementary. Under the combined effect of these three strategies, the complete configuration A achieves a relatively balanced trade-off between smoothness, safety margin, and computational cost with only a slight increase in path length, validating the improved algorithm proposed in this application. The rationality and necessity of the design of each sub-module.

[0060] Sensitivity analysis of stand density and canopy thickness: The geometric feasibility and safety margin of UAV understory passages in high-altitude forest areas are largely controlled by stand density and canopy thickness. To assess improvements... To assess the adaptability and robustness of the algorithm under different forest stand structures, based on a constructed 500 m × 500 m plateau steep slope DEM, the algorithm performance was analyzed from two levels: discrete combination scanning and fine-grained density scanning, to examine the changes in forest stand parameters. Forest stand density was represented by the proportionality coefficient kρ relative to the baseline stand number, and canopy thickness was represented by the canopy radius scaling factor kc: when kρ = 1.0 and kc = 1.0, it corresponds to the baseline stand; kρ < 1 indicates a sparse stand, and kρ > 1 indicates a dense stand; kc < 1 indicates a sparse canopy, and kc > 1 indicates a dense canopy.

[0061] According to forestry requirements, stand density and canopy thickness were discretized into three levels: kρ = {0.7, 1.0, 1.3} (low, medium, high density) and kc = {0.8, 1.0, 1.2} (sparse, baseline, relatively thick), forming 9 combinations. For each group (kρ, kc), while keeping the DEM unchanged, the number of trees was randomly selected or expanded to a multiple of kρ based on the tree position and canopy radius distribution of the baseline stand, and the canopy radius was scaled proportionally to kc, generating 20 forest land samples with the same statistical characteristics but different tree positions; the shrub layer remained consistent with the baseline scenario. The start and end points in each sample were set the same as in the previous experiment, using a uniform occupied grid structure and safe radius constraints, and traditional methods were run respectively. With improvement Two algorithms were used to record indicators such as path length L, average turning angle θ, average safety margin Mean(M), 10th percentile safety margin Q10(M), and planning time Tplan, and the mean and standard deviation were calculated over 20 samples. The main results of the discrete scan are shown in Table 4, and the corresponding heatmaps for path length increment, average safety margin improvement, and planning time multiplier are shown in [Table 4]. Figure 6 .

[0062] Table 4 Comparison of path planning performance of two A* algorithms under discrete combinations of stand density and canopy thickness. Discrete combination results show that, within the currently set parameter range, both algorithms can successfully plan feasible paths with 100% success rate in low-, medium-, and sparse canopy scenarios; in high-density combinations, the traditional... The success rate is 65%-90%, while improved A* (i.e.) can maintain 80%-95%. Traditional The path length remained relatively constant within a narrow range of 556-564 m across all combinations, showing little sensitivity to changes in forest stand structure and exhibiting a "geometric shortest path priority" tendency. In contrast, the improved... The path length is generally longer than traditional The path length is slightly longer, with an increase of approximately 3%-6%, and it rises slowly with increasing stand density and canopy thickness. This indicates that the algorithm actively increases the detour distance when the stand is denser or the canopy is more developed, in exchange for a greater safety margin and a smoother path. Safety indicators show that, under all (kρ, kc) combinations, the improved path... Both Mean(M) and Q10(M) are significantly better than traditional methods. Among them, in the combination of medium and high density and canopy baseline or thicker, the improvement The average safety margin is approximately 1.9-2.2 times that of conventional A*, and the 10th percentile safety margin is approximately 1.5-2.0 times; under high-density + baseline or thicker canopy conditions, conventional... The success rate dropped to about 65%, while improvements The success rate can still reach 80%-95%, indicating that the improved algorithm has significant advantages in both safety margin and success rate when approaching the critical point of forest passage blockage. Regarding planning time, the improved algorithm... Overall, compared to traditional The time factor is 1.5-5.5 times slower. In low-density, sparse canopy combinations with relatively wide passageways, the time factor is mostly around 2 times. Improvements are only seen in some high-density combinations with relatively thick canopies. To maintain safety and success rates, a more thorough search was conducted, resulting in a slightly longer planning time, but it remains in the tens of seconds range, which is acceptable for route planning scenarios.

[0063] Based on discrete combinatorial analysis, to more precisely characterize the sensitivity of path characteristics to stand density, the stand density coefficient kρ was further refined from 0.7 to 1.6 in increments of 0.1, with a total of 10 density levels, at a baseline canopy thickness kc=1.0. For each kρ level, approximately kρ times the baseline number of trees were uniformly and randomly generated within the entire 500 m × 500 m area. The canopy radii were sampled with replacement from the statistical distribution of the baseline stand to maintain consistency in canopy morphology and size; the shrub layer remained unchanged. Several forest plots were randomly generated for each density level, and traditional methods were used to analyze the results. and improvements Route planning was performed, and the above indicators were calculated to obtain curves showing the continuous variation of route length, average safety margin, and planning time with stand density (see...). Figure 7 ).

[0064] Fine-grained scanning results show that as kρ increases from 0.7 to approximately 1.2, the path length of both algorithms exhibits a slow upward trend, while the traditional... The length was increased from approximately 556m to around 564m, indicating an improvement. The path length increased from approximately 572m to around 590m, with the latter consistently being 15-25m longer than the former, corresponding to approximately 3%-6% redundant path length. When kρ ≥ 1.3, some forest walkways tended to become congested, and the average path length and planning time of both algorithms significantly increased, accompanied by a certain proportion of planning failures; in these high-density areas, the traditional... The success rate dropped to a minimum of 62.5%, an improvement. It can be maintained at 75%-87.5%. The average safety margin Mean(M) and Q10(M) generally decrease with increasing stand density, but improvements... It is consistently significantly higher than traditional density across the entire density range. When kρ=0.7, the traditional The Mean (M) is approximately 3.5m, while the improved It can reach about 5m; in the extremely high density section, kρ=1.5-1.6, traditional The Mean(M) was reduced to around 2.5m, an improvement. It can still maintain a speed above 4m, showing a smoother performance degradation curve. Regarding planning time, before medium density, the time of both algorithms changes relatively smoothly with density, showing improvement. The planning time is approximately the traditional Twice as high; in extremely high-density areas, traditional The increase over time was more pronounced, causing the time multiplier curve to drop slightly at the high-density end.

[0065] Combining the discrete combination and fine-grained density scanning experiments, it can be seen that within a wide range of stand density and canopy thickness variations, the improved... It consistently outperforms traditional methods in terms of path smoothness and overall safety margin. Furthermore, the additional cost of path length is controlled within 3%-6%, and the increase in computational overhead is mostly about 2 times, with improvements only seen in extremely high-density, thick canopy combinations. As the forest stand gradually approaches the traversable-impassable critical state of understory pathways, the performance of both algorithms inevitably degrades, but improvements... The degradation in safety margin, success rate and planning time is more stable, demonstrating strong structural adaptability, making it more suitable as a basic algorithm for UAV three-dimensional path planning under complex forest stand conditions in plateau forest areas.

[0066] To further evaluate improvements The applicability of the algorithm under conditions of multiple elevations and forest types in plateau forest areas was investigated by constructing a sensitivity experiment on the coupling of topographic elevation, slope, and forest stand structure based on a slope DEM. First, the baseline DEM was zero-mean-normalized to obtain the undulating field DEM0, and then... Five baseline elevations (Hbase) of 2800, 3000, 3200, 3400, and 3600 m were set, along with three slope factors (ks) of 0.5, 1.0, and 1.5, corresponding to three levels of undulation: gentle slope, baseline slope, and steep slope, respectively. The starting and ending points remained unchanged, with only the overall elevation and undulation of the terrain altered.

[0067] Considering the distinct altitudinal differentiation of vegetation in plateau regions, this application simultaneously adjusts stand density and individual tree crown width across different elevation zones to approximate the overall trend of slightly sparser stands and larger crown widths at low altitudes, and denser stands and smaller individual trees at high altitudes. Specifically, stand density factor kρ(H) and crown scaling factor kc(H) are introduced, as shown in Table 5: Table 5. Stand density and canopy parameters under elevation zones Each (Hbase, ks) combination regenerates the tree canopy and shrub occupancy grid according to the table, thus synchronously reflecting the joint changes in topography and forest stand in the simulation.

[0068] This application uses a scenario at an altitude of 3200m and ks=1.0 as the baseline and evaluates it in the analysis using relative change. Figure 8 Table 5 shows the trends of path length and average safety margin with slope factor at different elevation zones. It can be seen that within the medium elevation range (2800-3400 m), regardless of the slope factor, the improved... All algorithms consistently found feasible paths, with path lengths (L) primarily ranging from 568 to 590 meters, and variations relative to the baseline generally less than 3%. Mean (M) remained between 4.5 and 5.3 meters, and Q10 (M) was mostly 2.83 meters, decreasing to 2.24 meters only in a few high-density forest stands, still exceeding the minimum safety threshold of 2.0 meters set in this study. Overall, Tplan increased slightly with increasing slope, while path length and safety margin changed little, indicating that the algorithm primarily absorbed terrain disturbances by locally fine-tuning flight altitude and obstacle avoidance strategies when undulations increased, without significantly lengthening the total flight distance.

[0069] Table 6 Representative working conditions for the plateau topography-forest stand coupling adaptation experiment The comparison between different elevation zones reveals the comprehensive impact of altitude-stand coupling on planning results: In the low-altitude (2800 m) scenario, due to the lower stand density and larger canopy width, the improvement... The algorithm tends to utilize relatively open forest corridors, achieving shorter paths and a higher average safety margin (Mean(M)≈5.26 m) under gentle slope conditions. As the altitude increases to 3200-3400m, the forest stands gradually become denser and the canopy width shrinks, resulting in a slight increase in path length. However, Mean(M) remains around 4.5-5.1m, indicating that the algorithm can maintain a similar safety margin under more complex canopy distributions. When the altitude further increases to 3600m, the forest stands corresponding to density factor kρ=1.2 and canopy factor ks=0.80 are already quite dense. Under gentle slope and baseline slope conditions (ks=0.5, 1.0), the algorithm can still find feasible pathways, with path lengths increasing to approximately 582-597m, and Mean(M) remaining basically around 5m. However, in the most extreme combination of high altitude and steep slope, no feasible path appeared in the simulation. This result is not due to numerical divergence in the algorithm, but rather because after the terrain undulations are magnified and superimposed with high-density forest stands, the available safe passages are completely blocked in the occupied grid, and the three-dimensional safe distance constraint cannot be met. Therefore, the planner correctly returns "no solution".

[0070] Based on the above results, it can be concluded that, within the main plateau forest belt at an altitude of 2800-3400m and under certain topographical variations, the improvement... The algorithm exhibits good robustness to joint changes in altitude and forest stand structure: path length and planning time fluctuate only slightly, while the average and quantile safety margins remain stable or even slightly improve. When altitude and slope approach extreme conditions, the algorithm can mark areas without safe navigation conditions through "no path" feedback, providing safety envelope information for subsequent task planning in a plateau environment. The plateau adaptability of the proposed method is verified from both elevation and forest stand dimensions.

[0071] Based on the same inventive concept, an embodiment of the present invention provides a drone inspection path planning system.

[0072] The UAV inspection path planning system of this invention can be installed in an electronic device. Depending on the functions implemented, the UAV inspection path planning system includes: The acquisition module is used to acquire digital elevation models, digital surface models, and laser point cloud data of the plateau forest area. The processing module is used to determine the three-dimensional voxel space of the plateau forest area based on the digital elevation model, digital surface model and laser point cloud data, and to determine the reachability of the three-dimensional voxel space based on the vertical safety margin to determine the reachable area of ​​the UAV. A pathfinding module is used to optimize the pathfinding process within the drone's reach area, utilizing improved... The algorithm performs a three-dimensional path search from the starting point to the ending point; The trajectory generation module is used to smooth the searched 3D path and apply maximum turning radius constraints and maximum climb rate constraints to generate the final 3D trajectory.

[0073] The module described in this invention can also be called a unit, which refers to a series of computer program segments that can be executed by the processor of an electronic device and can perform a fixed function, and are stored in the memory of the electronic device.

[0074] The various variations and specific examples of the UAV inspection path planning method provided in the above embodiments are also applicable to the UAV inspection path planning system of this embodiment. Through the foregoing detailed description of the UAV inspection path planning method, those skilled in the art can clearly understand the implementation method of the UAV inspection path planning system in this embodiment. For the sake of brevity, they will not be described in detail here.

[0075] This application also discloses an electronic device, such as Figure 9The diagram shown is a schematic representation of the electronic device used in a drone inspection path planning method according to an embodiment of the present invention. The electronic device may include at least one processor 10, a memory 11 communicatively connected to the at least one processor, a communication bus 12, and a communication interface 13. It may also include a computer program, such as a drone inspection path planning method program, stored in the memory 11 and executable on the processor 10.

[0076] In some embodiments, the processor 10 may be composed of integrated circuits, such as a single packaged integrated circuit or multiple integrated circuits with the same or different functions, including combinations of one or more central processing units (CPUs), microprocessors, digital processing chips, graphics processors, and various control chips. The processor 10 is the control unit of the electronic device, connecting various components of the entire electronic device through various interfaces and lines. It executes programs or modules stored in the memory 11 (e.g., executing a UAV inspection path planning method) and calls data stored in the memory 11 to perform various functions of the electronic device and process data.

[0077] The memory 11 includes at least one type of readable storage medium, including flash memory, portable hard drive, multimedia card, card-type memory (e.g., SD or DX memory), magnetic memory, magnetic disk, optical disk, etc. In some embodiments, the memory 11 can be an internal storage unit of an electronic device, such as a portable hard drive. In other embodiments, the memory 11 can be an external storage device of the electronic device, such as a plug-in portable hard drive, Smart Media Card (SMC), Secure Digital (SD) card, Flash Card, etc. Furthermore, the memory 11 can include both internal and external storage units of the electronic device. The memory 11 can be used not only to store application software and various types of data installed on the electronic device, such as the code of a method program for drone inspection path planning, but also to temporarily store data that has been output or will be output.

[0078] The communication bus 12 can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. This bus can be divided into an address bus, a data bus, a control bus, etc. The bus is configured to enable communication between the memory 11 and at least one processor 10, etc.

[0079] Communication interface 13 is used for communication between the aforementioned electronic device and other devices, including a network interface and a user interface. Optionally, the network interface may include a wired interface and / or a wireless interface (such as a Wi-Fi interface, Bluetooth interface, etc.), typically used to establish communication connections between the electronic device and other electronic devices. The user interface may be a display, an input unit (such as a keyboard), or optionally, a standard wired or wireless interface. Optionally, in some embodiments, the display may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, or an OLED (Organic Light-Emitting Diode) touchscreen, etc. The display may also be appropriately referred to as a screen or display unit, used to display information processed in the electronic device and to display a visual user interface.

[0080] Figure 9 Only electronic devices with components are shown; it will be understood by those skilled in the art that... Figure 9 The structure shown does not constitute a limitation on the electronic device and may include fewer or more components than shown, or combine certain components, or have different component arrangements.

[0081] For example, although not shown, the electronic device may also include a power supply (such as a battery) to power various components. Preferably, the power supply can be logically connected to at least one processor 10 via a power management device, thereby enabling functions such as charging management, discharging management, and power consumption management. The power supply may also include one or more DC or AC power supplies, recharging devices, power fault detection circuits, power converters or inverters, power status indicators, and other arbitrary components. The electronic device may also include various sensors, Bluetooth modules, Wi-Fi modules, etc., which will not be described in detail here.

[0082] It should be understood that the embodiments are for illustrative purposes only and are not limited to this structure in the scope of the patent application.

[0083] Furthermore, if the modules / units integrated into the electronic device are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. The computer-readable storage medium can be volatile or non-volatile.

[0084] This application provides a computer-readable storage medium, including, for example, any entity or device capable of carrying the computer program code, a recording medium, a USB flash drive, a portable hard drive, a magnetic disk, an optical disk, a computer memory, or a read-only memory (ROM). The computer-readable storage medium stores a computer program that can be loaded by a processor and execute the UAV inspection path planning method of the above embodiments.

[0085] In the description of this specification, the references to terms such as "an embodiment," "some embodiments," "example," "specific example," "a implementation," "a preferred implementation," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0086] Although embodiments of the invention have been shown and described, those skilled in the art will understand that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the claims and their equivalents.

Claims

1. A method for unmanned aerial vehicle (UAV) inspection path planning, characterized in that, The method includes: Acquire digital elevation models, digital surface models, and laser point cloud data of the plateau forest area; The three-dimensional voxel space of the plateau forest area is determined based on digital elevation model, digital surface model and laser point cloud data, and the accessibility of the three-dimensional voxel space is determined based on vertical safety margin to determine the area accessible by UAV. In the reachable area of the UAV, a heuristic search algorithm based on is used to search a three-dimensional path from the start point to the end point; the heuristic search algorithm based on generates a target flight node by using a terrain perception heuristic function when searching the three-dimensional path, and calculates the inter-node cost between the current position node and the target flight node by using a multi-factor cost function containing a corner penalty term and an obstacle proximity penalty term, and obtains the three-dimensional path of the UAV according to the target flight node and the inter-node cost. The obtained 3D path is smoothed and constraints on maximum turning radius and maximum climb rate are applied to generate the final 3D track. The expression for the terrain-aware heuristic function is: ; in, This represents the terrain-aware heuristic function. This represents the three-dimensional Euclidean distance from the current position to the target point. Indicates a high penalty weight. This indicates the operation of taking the maximum value of the two parameters in parentheses. This represents the height value of the current position. This represents the height value of the target point; The expression for the multi-factor cost function is: , in, The adjacency distance, The weighting coefficient for the corner penalty term. Penalty for turning corners; The weighting coefficient for the obstacle proximity penalty. Penalty for proximity to obstacles; This represents the index of a 3D mesh node.

2. The UAV inspection path planning method as described in claim 1, characterized in that, Vertical safety margin is a parameter representing the flight safety status of a drone. It is determined based on the drone's basic safe altitude, wind-induced sway amplitude, and drone positioning error.

3. The UAV inspection path planning method as described in claim 2, characterized in that, The process involves determining the three-dimensional voxel space of the plateau forest area based on digital elevation models, digital surface models, and laser point cloud data, and then determining the reachability of the three-dimensional voxel space based on vertical safety margins to identify the accessible area for the UAV, including: The canopy thickness of the plateau forest area was calculated based on the digital elevation model and digital surface model, and the canopy thickness of the plateau forest area was locally corrected using laser point cloud data to obtain the corrected canopy thickness information. The three-dimensional voxel space of the plateau forest area was determined based on laser point cloud data and corrected canopy thickness information. Accessibility is determined in three-dimensional voxel space based on vertical safety margin to identify the reachable area of ​​the UAV.

4. The UAV inspection path planning method as described in any one of claims 1 to 3, characterized in that, The obtained 3D path is smoothed using cubic B-splines.

5. The UAV inspection path planning method as described in claim 1, characterized in that, The three-dimensional path search employs jump point search and / or bidirectional expansion strategy and / or adaptive step size expansion strategy.

6. A drone inspection path planning system, used to implement the drone inspection path planning method according to any one of claims 1 to 5, characterized in that, include: The acquisition module is used to acquire digital elevation models, digital surface models, and laser point cloud data of the plateau forest area. The processing module is used to determine the three-dimensional voxel space of the plateau forest area based on the digital elevation model, digital surface model and laser point cloud data, and to determine the reachability of the three-dimensional voxel space based on the vertical safety margin to determine the reachable area of ​​the UAV. A pathfinding module is used to optimize the pathfinding process within the drone's reach area, utilizing improved... The algorithm performs a three-dimensional path search from the starting point to the ending point; the algorithm based on The heuristic search algorithm uses a terrain-aware heuristic function to generate target flight nodes when performing 3D path search, and uses a multi-factor cost function that includes corner penalty and obstacle proximity penalty to calculate the inter-node cost between the current position node and the target flight node. The 3D path of the UAV is obtained based on the target flight node and the inter-node cost. The trajectory generation module is used to smooth the searched 3D path and apply maximum turning radius constraints and maximum climb rate constraints to generate the final 3D trajectory.

7. An electronic device, characterized in that, The electronic device includes: At least one processor (10); and, A memory (11) communicatively connected to the at least one processor (10); The memory (11) stores a computer program that can be executed by the at least one processor (10) to enable the at least one processor (10) to perform the UAV inspection path planning method as described in any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program; when the computer program is executed by a processor, it implements the UAV inspection path planning method as described in any one of claims 1 to 5.

Images