Photovoltaic power station unmanned aerial vehicle three-dimensional route planning method and system based on field of view angle self-adaption and terrain modeling

Abstract

The application discloses a photovoltaic station unmanned aerial vehicle three-dimensional route planning method and system based on field of view angle self-adaption and terrain modeling. The method comprises the following steps: acquiring photovoltaic panel geographic position information; determining a vertical field of view angle according to an unmanned aerial vehicle image device parameter; generating a reference point based on the geographic position information, performing horizontal clustering by using a density clustering algorithm to form an initial row grouping, and then completing row and column grouping by linear fitting completion and correction of row structure; dynamically determining a flight height for each flight point by a geometric model calculation based on the grouping result, a vertical field of view angle and a pitch angle, and generating a three-dimensional flight point; determining a flight point connection sequence based on a distance optimization principle, detecting an optimized acute angle turning, and generating a smooth three-dimensional flight path. The application realizes full-automatic three-dimensional planning of a photovoltaic inspection route, can adapt to complex terrains, guarantees shooting quality and improves inspection efficiency.

Description

Technical Field

[0001] This invention belongs to the field of autonomous flight path planning technology for unmanned aerial vehicles (UAVs), and specifically relates to a three-dimensional flight path planning method and system for photovoltaic power station UAVs based on adaptive field of view and terrain modeling. Background Technology

[0002] With the rapid development of the photovoltaic power generation industry, the scale of photovoltaic power plants continues to increase, and their site selection is becoming increasingly complex, with many power plants built in mountainous, hilly, and other terrain-undulating areas. Traditional manual inspection methods suffer from low efficiency, high safety risks, and strong subjectivity when dealing with large-scale power plants in complex terrain, and can no longer meet the needs of efficient operation and maintenance of modern power plants. Therefore, using drones for automated inspection has become an inevitable trend in the industry, and the core component—automatic planning technology for drone inspection routes—directly determines the efficiency and effectiveness of the inspection.

[0003] Currently, the methods for drone flight path planning in photovoltaic power plants can be mainly divided into the following categories:

[0004] 1. Manual Planning. This method relies entirely on the field experience and flight knowledge of experts, with waypoints, flight altitudes, and shooting parameters set manually on electronic maps one by one. Its advantage lies in its high accuracy and controllability for planning small-scale, regular, and flat photovoltaic power plants. However, for large power plants covering thousands or even millions of square meters, manual planning is extremely labor-intensive, time-consuming, and costly. More importantly, in complex terrain environments, the human eye struggles to accurately determine safe distances and optimal shooting angles in three-dimensional space, easily leading to problems such as flight path conflicts with terrain, incomplete shooting coverage, or poor image quality, resulting in poor adaptability.

[0005] 2. Automated Two-Dimensional Path Planning. To overcome the efficiency bottleneck of manual planning, automated two-dimensional path planning methods have emerged. These methods typically treat the photovoltaic panels as a set of two-dimensional points on the ground, abstracting the path planning problem into a Multi-Traveling Salesman Problem (MTSP) or a covering path planning problem, and employing metaheuristic algorithms such as genetic algorithms, particle swarm optimization, and ant colony optimization to solve it. A common strategy is to generate a bow-shaped round-trip path covering the entire site, supplemented by region partitioning to reduce computational complexity.

[0006] Existing limitations: Ignoring terrain elevation: Two-dimensional planning cannot perceive and adapt to the three-dimensional undulations of the terrain. In mountainous areas, the planned planar path may collide with the hillside during actual flight, or the constant flight altitude may result in an unsatisfactory shooting angle, leading to distance from the photovoltaic panel surface. Low path efficiency: The standard bow-shaped path generates a large number of invalid flights and redundant turns when facing complex layouts with varying photovoltaic panel row lengths and staggered arrangements, increasing flight time and energy consumption. Coverage blind spots: From a two-dimensional perspective, the algorithm struggles to handle shooting blind spots caused by terrain shading or the three-dimensional arrangement of photovoltaic panels (such as dual-wave modules), potentially leading to some modules not being effectively inspected.

[0007] 3. Automated 3D Path Planning. 3D path planning technology is being introduced for complex terrain environments. This type of method constructs a 3D model of the environment using a digital elevation model (DEM) or laser point cloud, and considers the performance constraints and environmental threats of the UAV in 3D space, using metaheuristic algorithms or sampling-based algorithms (such as RRT, PRM) for path searching.

[0008] Existing drawbacks: High computational cost: The search dimension in 3D space is much higher than in 2D, resulting in slow algorithm convergence and high computational resource consumption, making it difficult to meet the rapid planning requirements of ultra-large-scale photovoltaic power plants. Prone to getting trapped in local optima: Many metaheuristic algorithms (such as genetic algorithms) tend to converge prematurely in complex 3D scenes, failing to find the globally optimal path. Inaccurate environmental modeling: Existing 3D modeling often only provides a single elevation value, failing to fully consider sensor errors, model uncertainties, and the fine structure of the photovoltaic modules themselves (such as supports and shading relationships), leading to safety risks in actual execution of the planned flight path. Disconnected from inspection tasks: Many general-purpose 3D path planning algorithms are not tightly coupled with the specific task requirements of photovoltaic inspection (such as ensuring the clarity of each string of modules and optimizing shooting angles). The planned path may be safe to fly, but the inspection quality will be low.

[0009] In summary, existing technologies suffer from a core contradiction: efficient 2D planning cannot adapt to complex 3D terrain, while refined 3D planning faces challenges such as computational complexity, low efficiency, and low coupling with the task. Furthermore, both 2D and 3D methods have shortcomings in the intelligence of path generation. For example, they lack intelligent clustering analysis of the spatial distribution of photovoltaic panels, cannot dynamically adjust flight altitude and field of view based on terrain and component attitude, and lack smoothness optimization at path connections and turns, leading to reduced flight efficiency.

[0010] Therefore, there is an urgent need in this field for a comprehensive solution that can deeply integrate high-precision three-dimensional environmental perception, intelligent identification of photovoltaic modules, adaptive adjustment of field of view, and efficient three-dimensional path planning, so as to achieve fully automatic, safe, efficient and high-coverage route planning for UAV photovoltaic inspection in various complex terrains and large-scale scenarios. Summary of the Invention

[0011] The purpose of this invention is to overcome the shortcomings of existing manual planning methods, such as low efficiency, inability of two-dimensional automatic planning to adapt to complex terrain, and computational complexity and low coupling with inspection tasks in three-dimensional automatic planning. This invention provides a method and system for three-dimensional flight path planning for photovoltaic power plant UAVs based on adaptive field of view and terrain modeling. This method aims to achieve complete automation of photovoltaic inspection flight path generation, dynamically optimizing flight paths, altitudes, and shooting angles based on actual terrain and photovoltaic panel layout. This significantly improves inspection efficiency while ensuring flight safety and image quality.

[0012] In a first aspect, embodiments of this application provide a three-dimensional flight path planning method for photovoltaic power station UAVs based on adaptive field of view and terrain modeling, the method comprising:

[0013] Obtain the geographical location information of each photovoltaic panel within the photovoltaic power station;

[0014] Determine the vertical field of view based on the parameters of the UAV image acquisition equipment;

[0015] Based on the geographic location information, path planning reference points for each photovoltaic panel are generated. Density clustering algorithm is used to cluster the reference points horizontally to form initial row groups. Then, linear fitting method is used to fit the reference points in each initial row group to complete and correct the row structure, thus completing the row and column grouping of the photovoltaic panels.

[0016] Based on the completed grouping of photovoltaic panels, and combined with the vertical field of view and preset pitch angle of the UAV, the flight altitude of each waypoint is dynamically determined through geometric model calculation, thereby generating three-dimensional waypoints that ensure complete coverage of the photovoltaic panels;

[0017] Based on the principle of optimal distance, the connection order of three-dimensional waypoints between different rows of photovoltaic panels is dynamically determined, and acute turns in the waypoint path are detected and optimized to generate a smooth three-dimensional flight path.

[0018] The generated three-dimensional flight path is sent to the UAV flight control system, which drives the UAV to perform the inspection task according to the planned path.

[0019] Secondly, embodiments of this application provide a 3D flight path planning system for photovoltaic power station UAVs based on adaptive field of view and terrain modeling, applied to the 3D flight path planning method for photovoltaic power station UAVs based on adaptive field of view and terrain modeling as described in the first aspect. The system includes:

[0020] The data acquisition module is used to acquire the geographical location information of each photovoltaic panel within the photovoltaic power station;

[0021] The field of view determination module is used to determine the vertical field of view based on the parameters of the UAV image acquisition device.

[0022] The photovoltaic panel grouping module is used to generate path planning reference points for each photovoltaic panel based on the geographical location information, and to use a density clustering algorithm to perform horizontal clustering of the reference points to form initial row groups. Then, a linear fitting method is used to fit the reference points in each initial row group to complete and correct the row structure, thereby completing the row and column grouping of the photovoltaic panels.

[0023] The 3D waypoint generation module is used to dynamically determine the flight altitude for each waypoint based on the completed grouping of photovoltaic panels, combined with the vertical field of view of the UAV and the preset pitch angle, through geometric model calculation, thereby generating 3D waypoints that ensure complete coverage of the photovoltaic panels;

[0024] The path optimization module is used to dynamically determine the connection order of three-dimensional waypoints between different rows of photovoltaic panels based on the principle of optimal distance, and to detect sharp turns in the optimized waypoint path to generate a smooth three-dimensional flight path.

[0025] The route execution module is used to send the generated three-dimensional flight path to the UAV flight control system, driving the UAV to perform inspection tasks according to the planned path.

[0026] Thirdly, embodiments of this application provide an electronic device, including:

[0027] processor;

[0028] Memory used to store processor-executable instructions;

[0029] The processor is configured to implement the photovoltaic power station UAV three-dimensional flight path planning method based on field of view adaptation and terrain modeling as described in the first aspect when executing the instructions.

[0030] Fourthly, embodiments of this application provide a computer-readable storage medium storing a program that instructs a device to execute the photovoltaic power station UAV three-dimensional flight path planning method based on field-of-view adaptation and terrain modeling as described in the first aspect.

[0031] Beneficial effects:

[0032] 1. Fully automated and highly efficient: It achieves fully automated generation from photovoltaic panel data to the final flight path, completely eliminating the reliance on expert experience and greatly improving planning efficiency, especially suitable for large-scale photovoltaic sites.

[0033] 2. Three-dimensional adaptation and high precision: Through precise field-of-view model and trigonometric calculation, the optimal flight altitude is dynamically calculated for each waypoint, ensuring that the best shooting distance and angle are maintained between the UAV and the photovoltaic panel in complex terrain. This achieves true three-dimensional adaptive planning and ensures the clarity and coverage integrity of the inspection images.

[0034] 3. Intelligent Path Optimization: Through DBSCAN clustering and linear regression, it can intelligently identify and complete irregular and misaligned rows of photovoltaic panels, adapting to the actual power plant layout. The dynamic path connection strategy effectively reduces the overall flight distance. The acute-angle turn optimization mechanism improves flight smoothness and safety, reducing energy consumption and mechanical wear.

[0035] 4. Strong versatility and reliability: The core of this method lies in the intelligent understanding and mathematical modeling of the spatial distribution of photovoltaic panels. It does not rely on specific drone platforms or expensive sensors. The algorithm is robust and can be widely applied to various types of photovoltaic power plants, such as flat and open sites and complex mountainous areas. Attached Figure Description

[0036] Figure 1 This is a schematic diagram of a three-dimensional flight path planning method for photovoltaic power station UAVs based on adaptive field of view and terrain modeling, provided as an embodiment of this application.

[0037] Figure 2 The system architecture diagram of the photovoltaic power station UAV three-dimensional flight path planning system based on field of view adaptation and terrain modeling provided in this application.

[0038] Figure 3 A schematic diagram of an electronic device provided in an embodiment of this application. Detailed Implementation

[0039] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them.

[0040] It should be noted that in the embodiments of this application, "at least one" refers to one or more, and "more than one" refers to two or more. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used in the specification of this application is for the purpose of describing particular embodiments only and is not intended to be limiting of this application.

[0041] Based on the embodiments described in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0042] Example 1

[0043] Figure 1 This is a schematic flowchart illustrating a method for 3D flight path planning of a photovoltaic power station UAV based on adaptive field of view and terrain modeling, provided as an embodiment of this application. Figure 1 As shown, the method includes:

[0044] S110. Obtain the geographical location information of each photovoltaic panel within the photovoltaic power station. Photovoltaic panels refer to the solar panels being inspected, which are the target objects captured by the drone. Geographical location information refers to the precise location data of each photovoltaic panel in the real world, usually existing in the form of a series of coordinate points. This is the data foundation for the entire flight path planning. Without precise location information, all subsequent planning is castles in the air. Its function is to construct a digital, computer-processable map of the photovoltaic power station.

[0045] Specifically, obtaining the geographical location information of each photovoltaic panel within the photovoltaic power station includes:

[0046] Acquire laser point cloud data or oblique photogrammetry 3D models of the photovoltaic power station. This step aims to collect high-precision raw 3D spatial data of the photovoltaic power station. Laser point cloud data is obtained through lidar scanning and consists of a massive number of 3D coordinate points. Its advantages lie in directly and accurately acquiring the 3D geometric information of the ground surface, unaffected by lighting conditions, penetrating vegetation gaps, and exhibiting excellent representation of complex terrain and structures. Oblique photogrammetry 3D models are generated by calculating and processing photographs taken from different angles to create 3D models with realistic textures. Its advantages include simultaneously acquiring geometric information and realistic visual (texture) information, relatively low cost, and a more intuitive model. A drone equipped with lidar scans the entire photovoltaic power station, obtaining a collection (point cloud) containing all surface points of objects such as terrain, photovoltaic panels, supports, and utility poles, with each point having X, Y, and Z coordinates.

[0047] The spatial contours and vertex coordinates of each photovoltaic panel are automatically identified and extracted from the point cloud data or 3D model. This step intelligently locates and separates the target object—the photovoltaic panel—from the mixed raw data and accurately obtains its boundary information. This is a crucial step in transforming raw data into usable information. Extracting the spatial contours and vertex coordinates quantifies the geometry and position of the photovoltaic panel. After identifying it as a photovoltaic panel, it is necessary to accurately know its boundaries, especially the positions of several corner points (vertices), as this is the basis for subsequent waypoint calculations. For example, when a computer algorithm processes point cloud data, it finds an area with rectangular planar features, determines it to be a photovoltaic panel, and then uses a corner detection algorithm to accurately output the 3D coordinates of the four corners of the panel.

[0048] The vertex coordinates are transformed to a geodetic coordinate system through georeferencing to obtain the geographic location information. This step unifies the coordinate reference, converting the object's local, relative coordinates into global, absolute geographic coordinates. This ensures that the generated flight path can seamlessly interface with the UAV's GPS navigation system and be accurately executed in the real world. Georeferencing establishes coordinate transformation relationships. Point clouds or models may be generated in an independent coordinate system and need to be transformed to a geodetic coordinate system such as WGS-84 or CGCS2000 using known ground control points. Transforming to a geodetic coordinate system gives the coordinates actual geographic meaning. After transformation, a coordinate point (e.g., longitude 115.5°, latitude 38.8°, altitude 100 meters) corresponds to a precise location on Earth.

[0049] S120. Determine the vertical field of view (VOP) of the drone based on the parameters of its image acquisition equipment. The image acquisition equipment refers to the camera mounted on the drone. The parameters mainly refer to the camera's sensor size and lens focal length. The vertical field of view refers to the range of angles the camera can capture in the vertical direction. Determine the drone's vertical coverage capability in a single shot. This is a crucial optical parameter for subsequent calculations of flight altitude, ensuring that the photovoltaic panels are fully captured in the image.

[0050] Specifically, determining the vertical field of view (VOP) based on the parameters of the UAV image acquisition device includes: calculating the initial vertical field of view based on the sensor size and lens focal length of the image acquisition device; dynamically adjusting the vertical field of view based on the reflectivity of the photovoltaic panel surface and the ambient light intensity; establishing a coupling relationship model between the field of view and flight altitude, and jointly optimizing the flight altitude based on the environmental adaptation coefficient; verifying the rationality of the field of view setting through the effective coverage of the actual captured images, and automatically recalculating the field of view parameters when the coverage is lower than a set threshold; wherein, the environmental adaptation coefficient is dynamically calculated and determined based on the cleanliness of the photovoltaic panel surface and the ambient visibility.

[0051] The specific implementation method is as follows: the vertical field of view is determined based on the parameters of the UAV image acquisition device, which is achieved through the following formula:

[0052] ,

[0053] Where FOV is the vertical field of view, D is the diagonal size of the sensor of the image acquisition device, and f is the focal length of the lens.

[0054] It also includes a dynamic field-of-view optimization step: adjusting the vertical field of view in real time based on the reflectivity of the photovoltaic panel surface; dynamically optimizing the focal length parameter f according to the ambient light intensity; and establishing a coupling relationship model between the field of view and flight altitude, performing joint optimization through the following formula:

[0055] ,

[0056] in, To achieve the optimal flight altitude, For environmental adaptability coefficient, This is the adjusted field of view. The function of this model is to strongly correlate and synergistically optimize optical parameters (field of view) and flight control parameters (altitude). It no longer calculates altitude in isolation, but recognizes that the field of view and flight altitude are a pair of variables that need to be decided together to achieve global optimum.

[0057] The environmental adaptability coefficient Dynamic calculations based on photovoltaic panel surface cleanliness and environmental visibility:

[0058] ,

[0059] in, Based on the coefficient, This is the surface cleanliness factor. Dirt on the panel will reduce image contrast, and parameter adjustments are needed to compensate for it. The environmental visibility factor refers to factors such as fog and haze, which reduce image clarity and affect the effective recognition distance. Using this factor, the system can intelligently fine-tune its flight strategy in complex and variable real-world environments, such as automatically flying closer in low visibility conditions. These are weighting parameters; the function of this step is to quantify the impact of environmental factors on flight safety and shooting quality, and to use them as core decision variables.

[0060] It also includes a field-of-view verification mechanism: the rationality of the field-of-view setting is verified by the effective coverage of the actual captured images. When the coverage is lower than a set threshold, the field-of-view parameters are automatically recalculated. The function of this mechanism is to form a closed-loop control system of planning-execution-verification-feedback-replanning. It does not rely on the absolute correctness of the theoretical model, but rather uses actual results to test and correct the theoretical planning. For example, after the UAV flies and captures images according to the planned parameters, the transmitted images are analyzed in real time. If it is found that a large number of photovoltaic panels at the edge of the photo are not fully captured (low coverage), a recalculation is immediately triggered, adjusting the field-of-view or flight altitude, and the new parameters are used in subsequent flight segments, thereby ensuring the quality closed loop of the entire inspection mission.

[0061] The above steps construct a multi-layered, adaptive, and feedback-enabled intelligent perception and decision-making system: the foundational layer provides theoretical calculations; the optimization layer dynamically adjusts core parameters based on environmental perception; and the validation layer provides feedback and corrections based on practical results. This significantly enhances the robustness, adaptability, and reliability of the entire route planning method in real-world complex environments, which is one of its core advantages over traditional static planning methods.

[0062] S130. Based on the geographical location information, generate path planning reference points for each photovoltaic panel, and use density clustering algorithm to perform horizontal clustering of the reference points to form initial row groups. Then, use linear fitting method to fit the reference points in each initial row group to complete and correct the row structure, thus completing the row and column grouping of the photovoltaic panels.

[0063] Path planning reference points are points generated near each photovoltaic panel to represent the locations the drone should traverse. Their function is to transform panel inspection into point-based path planning. Density clustering algorithms (such as DBSCAN) are algorithms that automatically group points based on their density distribution. Their function is to intelligently identify reference points for photovoltaic panels that are spatially close as belonging to the same row, even if these rows are not perfectly straight. Linear fitting refers to fitting a set of scattered points with a straight line. Its function is to correct for minor irregularities in the arrangement of photovoltaic panels and infer the complete direction and length of the row, laying the foundation for generating a straight flight path. This step is the core intelligent processing stage, aiming to organize the disorganized individual photovoltaic panels into a regular, logical row and column structure through clustering and fitting, thus simplifying the problem from covering countless points to covering several lines. For example, photovoltaic panels on a hillside may be slightly curved due to the terrain. DBSCAN will cluster the reference points of these panels into several groups (rows). Then, a linear fit is performed on each group to obtain a straight line representing the optimal direction of the row, and the row is completed with the correct length based on this line.

[0064] Specifically, the process of grouping the photovoltaic panels into rows and columns includes: calculating geometric features based on the vertex coordinates of the photovoltaic panels to generate path planning reference points; using a density clustering algorithm to cluster the reference points horizontally to form initial row groups; performing linear fitting on the reference points within each initial row group to obtain a standard straight line equation; and performing length completion and vertical grouping on the photovoltaic panel rows based on the standard straight line equation to complete the row and column grouping.

[0065] The specific implementation method is as follows: the horizontal clustering grouping and the vertical linear grouping and completion include: calculating the length L and width W of the photovoltaic panel based on the geographical coordinates of the four vertices of the photovoltaic panel.

[0066] , ,

[0067] in This step assigns the coordinates of the four vertices of the photovoltaic panel; its function is to quantify the physical dimensions and spatial orientation of each photovoltaic panel. It's not just a simple size calculation, but more importantly, it involves selecting specific pairs of vertices for calculation (e.g., using...). arrive To calculate the length, use arrive The formula (calculating width) implicitly identifies the long side direction (i.e., row direction) and short side direction of the photovoltaic panel. This forms the basis for subsequent meaningful row grouping. For example, if a panel is a rectangle placed in a north-south direction, this calculation will yield a larger L value (length) and a smaller W value (width), thus clarifying its dominant direction.

[0068] The number of reference points for each photovoltaic (PV) panel is determined based on its aspect ratio; this step enables adaptive path planning density. It intelligently decides how many waypoints to generate to cover the panel based on its shape, thus optimizing path efficiency while ensuring coverage integrity. For long panels, more reference points (waypoints) may be needed to ensure clear imaging along their entire length during flight. For short panels, fewer reference points may be required for coverage. For example, the system might set a rule to generate one reference point per meter of panel length. A 10-meter-long panel would then generate 10 reference points, while a 5-meter-long panel would only generate 5.

[0069] The DBSCAN clustering algorithm is used to cluster the reference points horizontally, and the Euclidean distance formula is used as the distance metric in the clustering process.

[0070] .

[0071] Where d is the distance between reference point i and reference point j; this step intelligently divides all reference points into different rows based on spatial proximity. The core function of the DBSCAN algorithm is to automatically discover regions with sufficient point density and identify them as a cluster (i.e., a row). It can effectively handle noisy points (such as isolated or misplaced points) and does not require pre-specifying the number of rows. The Euclidean distance metric defines the standard of proximity, that is, the straight-line distance between two points on a horizontal plane. This is the most intuitive and physically consistent way of defining distance. For example, all reference points that are roughly on the same horizontal line and are close to each other will be automatically identified by DBSCAN and grouped into the first row; another group of points above them will be grouped into the second row, and so on.

[0072] For each initial row group, a linear regression algorithm is used to fit a straight line to the reference points, resulting in the equation of the straight line. Where m is the slope and b is the intercept. This step corrects for irregular alignment and extracts the idealized center axis for each row. Due to installation errors or terrain undulations, the reference points for the same row of photovoltaic panels may not be strictly aligned. Linear regression finds the best-fitting straight line that best represents the overall orientation of the row by minimizing the error. This straight line equation (slope m and intercept b) defines the precise direction and position of the row, providing the mathematical basis for generating straight, parallel flight paths.

[0073] Based on the aforementioned linear equation, the length of the photovoltaic panel rows is completed to form a complete photovoltaic panel row structure. This step addresses potential gaps or inconsistencies at the beginning and end of the photovoltaic panel rows, constructing complete and continuous row units. Using the fitted straight line, the system can infer the theoretical starting and ending points of the row. It can fill in row incompleteness caused by missing data or damaged edge panels. It can unify the y-coordinates (or x-coordinates) of all reference points in the row, ensuring that the generated flight path is a perfect straight line, completely parallel to the direction of the photovoltaic panel rows, thereby obtaining the optimal shooting angle.

[0074] These five steps constitute a refined and automated processing pipeline from raw coordinates to orderly rows and columns: Understanding individuals (calculating panel size and orientation); Planning density (determining the number of reference points); Identifying groups (clustering into rows); Abstracting the model (fitting the centerline); Refining the structure (completing row lengths). Its overall function is to transform the potentially uneven and irregularly arranged photovoltaic panel arrays in reality into an orderly set of rows that computers can process efficiently and accurately, laying the most crucial foundation for generating high-quality 3D flight paths.

[0075] The method also includes an intelligent grouping optimization step based on the spatial distribution of photovoltaic panels:

[0076] After grouping the photovoltaic panels into rows and columns, the vertical coordinate difference between adjacent photovoltaic panels within a group is calculated. "Within a group" refers to an initial row group obtained through DBSCAN clustering. Adjacent photovoltaic panels are two photovoltaic panels that are adjacent to each other within the same row group. The vertical coordinate difference typically refers to the difference in Y-coordinate (or latitude) between the center points or specific reference points of these two adjacent photovoltaic panels. This value reflects their alignment along the row direction. This step quantitatively assesses the internal consistency of the current grouping. It checks whether the members (photovoltaic panels) in each initially grouped row are truly arranged neatly enough. For example, suppose 5 panels are initially grouped into the first row. Calculations show that the second and third panels differ by 2 meters in the Y-direction, while other adjacent panels differ by only 0.1 meters. This 2-meter difference is an anomaly that requires attention.

[0077] A vertical threshold is set. When the coordinate difference exceeds the threshold, the current group is horizontally subdivided. The vertical threshold is a pre-defined maximum distance that allows photovoltaic panels within the same row to be misaligned vertically. For example, it can be set to 1.5 times the width of a photovoltaic panel. Horizontal subdivision refers to splitting the current excessively large and irregular row group into two or more smaller, more internally aligned subgroups in the horizontal direction (X direction). This step automatically identifies and corrects erroneous groupings. Its core purpose is to ensure that the internal members of each final group are spatially compact and aligned, thereby guaranteeing that the flight path generated for each subgroup is efficient and safe. For example, continuing the previous example, the threshold is set to 1 meter. If 2 meters > 1 meter, the system will automatically cut the first row between the second and third panels, subdividing it into two sub-rows: [panel 1, panel 2] and [panel 3, panel 4, panel 5].

[0078] After standardizing all row units according to their length, they are grouped vertically. The specific steps are as follows:

[0079] A progressive grouping strategy based on a distance threshold. By iteratively calculating the row spacing and using the distance threshold as a hard boundary, row units that meet the set criteria are aggregated into the same row cluster.

[0080] Assuming a threshold of 25 meters, using sub-row 1 as the reference anchor point, the distances to subsequent rows are calculated sequentially. For example, the distance between sub-row 1 and sub-row 2 is 13, and the distance between sub-row 1 and sub-row 3 is 26. When the distance first reaches or exceeds the threshold (25), all previously consecutive row units that have not exceeded the threshold (sub-row 1 and sub-row 2) are grouped together, their original index [0,1] is recorded, and they are removed from the candidate set. Then, using sub-row 3 as the new reference anchor point, the above calculation and grouping process is repeated until all row units are aggregated to form different row clusters. In addition, the threshold can be dynamically set according to the target cluster density, with the core basis being the row interval and the expected number of rows per cluster.

[0081] Based on the subdivided grouping results, a topological connection relationship is established for the photovoltaic panel rows. This topological connection relationship describes the spatial adjacency and connection order between different row clusters. It defines which row should be flown next after completing one row. This step builds a logical flight sequence on top of the physical grouping. Due to the further subdivision, the number and structure of rows change, and their connections must be re-established; otherwise, path planning cannot proceed. For example, after subdivision, the system records that the end of sub-row 1 is close to the beginning of sub-row 2, so after completing sub-row 1, the next row should be sub-row 2; and the end of sub-row 2 is close to the beginning of the second row, and so on. This forms an ordered flight list.

[0082] Based on the actual spatial distribution characteristics of photovoltaic panel rows, the neighborhood radius parameter of the DBSCAN clustering algorithm is dynamically adjusted. The vertical threshold is set based on the actual physical dimensions of the photovoltaic panels and the standard row spacing. Horizontal subdivision includes subgrouping adjacent photovoltaic panels exceeding the threshold while maintaining linearity within each group. The neighborhood radius parameter (ε) is a core parameter in the DBSCAN algorithm, used to determine whether two points are adjacent. This step enables self-tuning of the clustering algorithm's parameters, improving its robustness in different scenarios. The panel spacing may vary in different areas of a photovoltaic power station, and using a fixed neighborhood radius may lead to grouping errors. This step allows the system to learn and adjust itself. For example, in areas with dense panel spacing, the system automatically uses a smaller ε value (e.g., 0.5 meters) to prevent misgrouping panels from different rows; in areas with sparse panel spacing, a larger ε value (e.g., 2 meters) is used to prevent separating panels that should be in the same row.

[0083] The aforementioned intelligent grouping optimization steps constitute a closed-loop process of quality control and self-improvement. Through a process of inspection (step 1) -> correction (step 2) -> reconstruction (step 3) -> learning (step 4), it ensures that the final photovoltaic panel grouping result not only reflects spatial distribution but also meets the high-order requirements of flight path planning for row regularity, continuity, and efficiency. This enables the entire solution to flexibly adapt to complex and irregular real-world photovoltaic site layouts.

[0084] Specifically, the execution process of the DBSCAN clustering algorithm also includes a bi-cluster center distance optimization mechanism, specifically including:

[0085] Step 1: Establish a dual-feature clustering model based on the spatial distribution characteristics of photovoltaic panels and the terrain elevation characteristics. Spatial distribution characteristics refer to the positional relationship of the photovoltaic panels on the horizontal plane (X, Y axes). Terrain elevation characteristics refer to the altitude (Z axis) of the photovoltaic panel locations. This step upgrades clustering analysis from a two-dimensional plane to three-dimensional space. It recognizes that in complex terrain, photovoltaic panels belonging to the same row should not only be close in the horizontal direction but also have elevation continuity. This is key to overcoming the limitations of traditional two-dimensional clustering.

[0086] Step 2: Calculate the cluster centers based on spatial distribution characteristics. Cluster centers of topographic elevation features The function of this step is to quantify the two core dimensions that characterize each potential group. This represents the center point of the photovoltaic panel group in planar position. This represents the center value of the photovoltaic panel group at the average elevation. This provides a specific, calculable data object for subsequent optimization.

[0087] Step 3: Employ a bi-cluster center distance optimization algorithm to determine the optimal cluster partition by minimizing the weighted sum of the distances between the two cluster centers.

[0088] ,

[0089] in, This represents the spatial distance between the centers of two clusters. Its function is to force the algorithm to find a group of points that are clustered together both horizontally and vertically. Minimizing this term implies an ideal cluster partition where the horizontal and vertical centers should almost coincide, indicating that the group is a compact population in three-dimensional space. These represent the intra-class compactness of spatial distribution and topographic elevation classes, respectively, measuring the dispersion of points within each group in terms of plane and elevation. Their function is to ensure that members within each group are as similar as possible. This refers to the weighting coefficients; their function is to balance the importance of the two optimization objectives: class center overlap and intra-class compactness. The weighting coefficients... The algorithm dynamically adjusts based on the density of photovoltaic panel distribution and the degree of terrain undulation; the intra-cluster density is measured by calculating the average distance from each point within a cluster to the cluster center. Unlike the standard DBSCAN, which only considers the distance between points, this algorithm uses a more advanced and comprehensive objective function to guide the clustering process, seeking a grouping scheme that is optimal in three-dimensional space.

[0090] Step 4: Based on the optimized clustering results, dynamically adjust the neighborhood radius parameter ε and the minimum number of points minPts in the DBSCAN algorithm. This step enables adaptive feedback adjustment of the clustering parameters. Traditional DBSCAN uses fixed parameters, but the terrain complexity varies across different regions. In areas with dramatic terrain undulations, the grouping should be more refined; therefore, the neighborhood radius ε is automatically reduced, and minPts is decreased, making the algorithm more sensitive to small, localized point clusters. In flat areas, ε can be increased to form larger, more coherent groupings. This significantly improves the robustness and adaptability of the algorithm across different areas of the entire site.

[0091] The dual-cluster center distance optimization mechanism is an advanced intelligent grouping scheme for complex 3D terrain. Its overall functionality can be summarized as follows: Multi-dimensional perception: simultaneously considering planar position and elevation information; Global optimization: finding the most reasonable grouping in 3D space by minimizing a comprehensive objective function; Parameter adaptation: dynamically adjusting the sensitivity of the underlying clustering algorithm based on the optimization results and terrain features. Ultimately, it ensures that the grouping results of photovoltaic panels not only conform to their planar layout but also to the 3D terrain features they are situated in, providing the most fundamental and accurate guarantee for generating safe and efficient 3D flight paths in complex sites such as mountainous areas.

[0092] S140. Based on the grouped rows of photovoltaic panels, and combining the UAV's vertical field of view and preset pitch angle, the flight altitude for each waypoint is dynamically determined through geometric model calculation, thereby generating 3D waypoints that ensure complete coverage of the photovoltaic panels. This is the core step in achieving 3D adaptation and complete coverage. It uses geometric relationships to dynamically calculate how high to fly at each location so that the camera's field of view precisely covers the target row of photovoltaic panels.

[0093] Specifically, the process of dynamically determining the flight altitude for each waypoint using a geometric model includes the following steps:

[0094] Establishing a waypoint geometric calculation model based on pitch angle α and vertical field of view β: This step defines the theoretical framework and key input parameters for solving the problem. It clarifies that the calculation of flight altitude is not an isolated guess, but is governed by a rigorous geometric model whose core variables are pitch angle (α) and vertical field of view (β). This model abstracts the UAV, camera, and solar panel rows into a geometric problem, where the UAV is the vertex, the camera's field of view constitutes a viewpoint, and the solar panel rows are the base edges that need to be covered.

[0095] Define the spacing between photovoltaic rows (g), the outward expansion distance (d), and the upper and lower factors a1 and a2 for the outward expansion distance. This step quantifies the safety and coverage requirements in actual inspection tasks and transforms these requirements into calculable parameters for the model. Calculate the UAV's field of view (L): The spacing *g* between photovoltaic panels defines the width of the core target that needs to be covered. The outward extension distance *d* and factors *a1* and *a2* add safety and coverage margins on both sides of the core target. This ensures that the camera's field of view covers not only the photovoltaic panels themselves but also a portion of the area beyond their edges, thus avoiding incomplete coverage due to positioning errors or slight flight drift. *a1* and *a2* allow for asymmetrical adjustments to the outward extension ratio in front of and behind the UAV to adapt to different flight directions or terrains. Here, *d*×*a1* is the outward extension distance in front of the waypoint (flight direction), and *d*×*a2* is the outward extension distance behind. Together with the spacing *g* between the panels, they constitute the total ground width *L* that needs to be covered by the camera's field of view.

[0096] Solve for the horizontal distance D from the waypoint to the photovoltaic panel and the flight altitude H using the following system of trigonometric equations:

[0097]

[0098] ;

[0099] ;

[0100] This step is the core computational engine for the entire 3D waypoint generation process. Its function is to transform optical constraints (field of view, pitch angle) and mission constraints (coverage range L) into precise, executable flight control parameters (horizontal distance D and flight altitude H). The first formula defines the total ground coverage requirement (L). The latter two trigonometric equations establish a mathematical model with two unknowns (D and H). By solving this system of equations, it is possible to uniquely determine how high the UAV should fly (H) and how far the waypoint should be located from the solar panel (D) so that the upper and lower edges of the camera's field of view precisely cover the preset boundaries. Essentially, this projects the camera's field of view (an isosceles trapezoid) onto the ground and adjusts the altitude H so that this trapezoid precisely covers the desired ground area [D, D+L].

[0101] Based on the calculated flight altitude H and horizontal distance D, combined with the geographical coordinates of the photovoltaic panel reference point, the three-dimensional coordinates of each waypoint are calculated. The pitch angle α ranges from -30° to 30°, and the outward extension distance factors a1 and a2 are dynamically adjusted according to the specific layout characteristics of the photovoltaic panel. This step transforms the calculated abstract geometric parameters into concrete three-dimensional spatial coordinates that can be executed by the UAV flight control system. Using the coordinates of the photovoltaic panel reference point and the calculated horizontal distance D, the longitude and latitude (planar position) of the waypoint can be determined. Using the calculated flight altitude H, combined with the ground elevation of that point (obtainable from a digital terrain model), the absolute altitude of the waypoint can be determined. The final output of this step is a three-dimensional waypoint as (longitude, latitude, altitude), rather than a two-dimensional point.

[0102] These four steps constitute a precise and automated conversion chain from task requirements to spatial coordinates: Modeling: Defining the problem framework. Setting parameters: Quantifying task requirements (coverage and safety). Core calculation: Converting requirements into flight parameters through a geometric model. Coordinate generation: Mapping flight parameters to real geographic space. Its overall function is to fully realize three-dimensional adaptive flight path planning, ensuring that the altitude of each generated waypoint is not a preset fixed value, but an optimal value that has been rigorously calculated and can dynamically adapt to different photovoltaic panel row positions and terrain undulations, thereby achieving 100% shooting coverage and optimal shooting angles while ensuring safety.

[0103] S150: Based on the principle of optimal distance, the system dynamically determines the connection order of 3D waypoints between different rows of photovoltaic panels and detects and optimizes sharp turns in the waypoint path, generating a smooth 3D flight path. The principle of optimal distance is a path planning strategy that always selects the connection method with the shortest total flight distance. Sharp turns refer to small, sharp turns that can cause the UAV to decelerate, sway, consume more power, or even fail to execute. A smooth 3D flight path refers to a continuous, smooth flight route without sharp turns. This step, after generating individual waypoints, is responsible for optimizing the efficiency and safety of the entire flight path. Optimal distance ensures inspection efficiency and reduces flight time and energy consumption. Optimizing sharp turns ensures flight safety and stability, while also improving the stability of the captured footage. For example, after flying the first row, should it fly the second row on the left or the second row on the right? The system calculates the distance from the current endpoint to the starting point of these two rows and selects the closer one. If the system detects that a turn is needed to fly from this endpoint to the starting point of the next line, forming a sharp V-shaped turn, it will insert an additional waypoint in the middle, forming a U-shaped trapezoidal path to make the turn smoother.

[0104] Specifically, the process of dynamically determining the connection order of three-dimensional waypoints between different rows of photovoltaic panels based on the principle of optimal distance, and detecting and optimizing acute-angle turns in the waypoint path, includes the following steps:

[0105] Module 1: Path Connection Optimization Steps. This module addresses the question of where to start the next flight after completing the current one, aiming to minimize the total distance of the global flight path.

[0106] Step 1: Calculate the Euclidean distance between the last waypoint of the current line and the first and last waypoints of the next line:

[0107] ;

[0108] ;

[0109] in, The coordinates of the last waypoint in the current row. , These are the coordinates of the first and last waypoints of the next row, respectively. This step quantifies the cost of transitioning from the current state to the next row for two possible options. It calculates the spatial distance between the two choices: turning around to the start of the next row and continuing to the end of the next row.

[0110] Step 2: Based on the principle of optimal distance, select... and The waypoint corresponding to the smaller value in the equation is used as the starting point to determine the flight direction; the function of this step is to make a globally optimal decision. This is achieved through comparison... and The system dynamically determines the flight direction for the next line. If If smaller, the next line flies from left to right. If If the distance is smaller, the next line flies from right to left. This avoids the long empty flights that can result from a fixed bow-shaped path. As shown in the diagram below, it automatically selects an S-shaped path with a shorter total distance, instead of a bow-shaped path, significantly reducing wasted flights.

[0111] Module Two: Sharp Angle Turn Optimization Steps. This module identifies and eliminates sharp turns in the flight path, improving the flight path from flyable to smooth, safe, and efficient.

[0112] Step 1: Calculate the direction vectors of adjacent path segments:

[0113] ;

[0114] ;

[0115] in, : The starting coordinates of the current path segment. : The coordinates of the endpoint of the current path segment (also the last waypoint of the current line). Vector It refers to the direction from the end of the path segment to the starting point. It defines the orientation of the drone as it completes this flight segment. This is the coordinate of the first waypoint of the next path segment. It is also the end point of the current path segment, i.e., the turning point. Vector It refers to the direction from the turning point towards the starting point of the next flight segment. It defines the expected orientation of the drone at the beginning of the next flight segment.

[0116] Step 2: Calculate the angle θ between the direction vectors: ;

[0117] Direction vector , These represent the directions of the current flight path segment and the next flight path segment, respectively. The function of this step is to accurately diagnose health issues within the path—such as acute-angle turns. By calculating the included angle θ, the degree of abruptness of the turn can be objectively determined.

[0118] When the included angle θ is less than a preset threshold, the last waypoint of the current line is copied and its coordinates are translated to form a trapezoidal transition path. The preset threshold is 90°, and the translation distance is dynamically determined based on the UAV's turning radius and flight speed. This step functions as a treatment for a diagnosed emergency. Its core is path smoothing. Copying and translating a waypoint is equivalent to adding a buffer point at the angle of a sharp turn. Forming a trapezoidal transition path transforms a sharp V-shaped turn into a smooth U-shaped turn. This is similar to straightening curves or adding transition curves in road design.

[0119] The overall function of this formula is to transform the geometry of the flight path into a computable mathematical problem. Through vector operations, it accurately diagnoses sharp-angle turns (θ < 90°) that are detrimental to flight, thus providing a quantitative and reliable decision-making basis for subsequent path smoothing optimization (forming a trapezoidal transition path).

[0120] The two modules optimize the path from both a macro-strategy and micro-local perspective: path connectivity optimization ensures the shortest global flight distance, improving mission efficiency; and acute-angle turn optimization ensures a smooth local flight trajectory, improving flight safety, stability, and economy. Together, they generate a usable, efficient, safe, and professional three-dimensional flight path.

[0121] S160. The generated three-dimensional flight path is sent to the UAV flight control system to drive the UAV to perform the inspection task according to the planned path.

[0122] Specifically, in this embodiment, step S160 further includes a real-time closed-loop control process: establishing a flight status monitoring channel to receive GPS / RTK positioning data, IMU attitude data, and battery status information of the UAV in real time; constructing a trajectory tracking error model and calculating the position deviation using the following formula: ,in These are the theoretical waypoint coordinates. For the actual positioning coordinates, design an adaptive PID controller to dynamically adjust the flight control parameters based on the position deviation δ. ,in The gain coefficient varies with δ; an abnormal state handling mechanism is established, which automatically triggers emergency strategies when the following situations occur: when the continuous positioning deviation exceeds the safety threshold, the flight path is replanned; when the image transmission quality is lower than the set standard, the current flight segment is automatically restarted; when the battery power is lower than the warning value, the return path is automatically generated; real-time transmission and analysis of inspection data are realized, and the captured images are transmitted to the ground station through the 5G / 4G network. Online preliminary identification of photovoltaic panel defects is performed based on deep learning algorithms; a structured inspection report is generated, which automatically records the flight path, shooting parameters, abnormal events and defect identification results.

[0123] Example 2

[0124] like Figure 2 As shown, this application provides a system architecture diagram of a photovoltaic power station UAV three-dimensional flight path planning system based on adaptive field of view and terrain modeling. It is applied to the photovoltaic power station UAV three-dimensional flight path planning system based on adaptive field of view and terrain modeling as described in Embodiment 1, including: a data acquisition module 210, a field of view determination module 220, a photovoltaic panel grouping module 230, a three-dimensional waypoint generation module 240, a path optimization module 250, and a flight path execution module 260.

[0125] The data acquisition module 210 is used to acquire the geographical location information of each photovoltaic panel in the photovoltaic power station.

[0126] The field of view determination module 220 is used to determine the vertical field of view based on the parameters of the UAV image acquisition device.

[0127] The photovoltaic panel grouping module 230 is used to generate path planning reference points for each photovoltaic panel based on the geographical location information, and to use a density clustering algorithm to perform horizontal clustering of the reference points to form initial row groups. Then, a linear fitting method is used to fit the reference points in each initial row group to complete and correct the row structure, thereby completing the row and column grouping of the photovoltaic panels.

[0128] The 3D waypoint generation module 240 is used to dynamically determine the flight altitude for each waypoint based on the completed grouping of photovoltaic panel rows and columns, combined with the UAV's vertical field of view and preset pitch angle, through geometric model calculation, thereby generating 3D waypoints that ensure complete coverage of the photovoltaic panels.

[0129] The path optimization module 250 is used to dynamically determine the connection order of three-dimensional waypoints between different rows of photovoltaic panels based on the principle of optimal distance, and to detect sharp turns in the optimized waypoint path to generate a smooth three-dimensional flight path.

[0130] The route execution module 260 is used to send the generated three-dimensional flight path to the UAV flight control system, driving the UAV to perform inspection tasks according to the planned path.

[0131] Figure 3 This is an electronic device provided in one embodiment of this application. For example... Figure 3 As shown, the electronic device includes at least the following components: processor 301 and memory 300, communication interface 303, and bus 302.

[0132] In this embodiment of the application, memory 300 is used to store executable instructions of processor 301, which, when configured to execute instructions, implements the method as described in the first aspect.

[0133] In embodiments of this application, a computer-readable storage medium includes instructions that instruct a device to perform the method as described in the first aspect. For example, the instructions instruct the device to perform... Figure 1 The method is shown in the process steps.

[0134] In one embodiment of this application, the program operating in the electronic device may be a program that controls a central processing unit (CPU) or similar device to achieve the functions of the above-described embodiments of the present invention (a program that enables the computer to function). Information processed by these systems is then temporarily stored in random access memory (RAM) during processing, and subsequently stored in various ROMs such as read-only memory (FlashROM) and hard disk drives (HDDs), and read, corrected, and written by the CPU as needed.

[0135] It should be noted that a portion of the electronic device described above can also be implemented using a computer. In this case, the program for implementing the control function can be recorded on a computer-readable recording medium, and the program recorded on the recording medium can be read into the computer and executed.

[0136] It should be noted that the computer mentioned here refers to a computer built into an electronic device, employing hardware including an operating system and peripheral devices. Furthermore, computer-readable recording media refers to removable media such as floppy disks, magneto-optical disks, ROMs, and CD-ROMs, as well as storage systems such as hard drives built into the computer.

[0137] Furthermore, computer-readable recording media can include: media that dynamically stores programs for short periods of time, such as communication lines used when transmitting programs via networks like the Internet or communication lines like telephone lines; and media that store programs for fixed periods of time, such as volatile memory inside a computer that serves as a server or client in this case. In addition, the aforementioned program can be a program used to implement the above-mentioned functions, or it can be a program that can implement the above-mentioned functions by combining them with programs already recorded in the computer.

[0138] Furthermore, the electronic device in the above embodiments can also be implemented as an assembly (system group) composed of multiple systems. Each system constituting the system group can possess some or all of the functions or functional blocks of the electronic device in the above embodiments. As a system group, it is sufficient to have all the functions or functional blocks of the electronic device.

[0139] Those skilled in the art should recognize that the above embodiments are only used to illustrate this application and are not intended to limit this application. Any appropriate changes and variations made to the above embodiments within the essential spirit and scope of this application fall within the scope of protection claimed in this application.

Claims

1. A three-dimensional flight path planning method for photovoltaic power station UAVs based on adaptive field of view and terrain modeling, characterized in that, Includes the following steps: Obtain the geographical location information of each photovoltaic panel within the photovoltaic power station; The vertical field of view is determined based on the parameters of the UAV image acquisition equipment, including: The initial vertical field of view is calculated based on the sensor size and lens focal length of the image acquisition device. The vertical field of view is dynamically adjusted based on the reflectivity of the photovoltaic panel surface and the ambient light intensity. A coupled relationship model between field of view and flight altitude is established, and the flight altitude is jointly optimized based on the environmental adaptability coefficient; The rationality of the field of view setting is verified by the effective coverage of actual captured images. When the coverage is lower than the set threshold, the field of view parameters are automatically recalculated. The environmental adaptability coefficient is determined based on dynamic calculations of the photovoltaic panel surface cleanliness and environmental visibility. Based on the geographic location information, path planning reference points for each photovoltaic panel are generated. Density clustering algorithm is used to cluster the reference points horizontally to form initial row groups. Then, linear fitting method is used to fit the reference points in each initial row group to complete and correct the row structure, thus completing the row and column grouping of the photovoltaic panels. Based on the completed grouping of photovoltaic panels, and combined with the vertical field of view and preset pitch angle of the UAV, the flight altitude of each waypoint is dynamically determined through geometric model calculation, thereby generating three-dimensional waypoints that ensure complete coverage of the photovoltaic panels; Based on the principle of optimal distance, the connection order of three-dimensional waypoints between different rows of photovoltaic panels is dynamically determined, and acute turns in the waypoint path are detected and optimized to generate a smooth three-dimensional flight path. The generated three-dimensional flight path is sent to the UAV flight control system, which drives the UAV to perform the inspection task according to the planned path.

2. The method for three-dimensional flight path planning of photovoltaic power station UAVs based on adaptive field of view and terrain modeling as described in claim 1, characterized in that, The acquisition of the geographical location information of each photovoltaic panel within the photovoltaic power station includes: Acquire laser point cloud data or oblique photogrammetry 3D models of photovoltaic power plants; Automatically identify and extract the spatial contours and vertex coordinates of each photovoltaic panel from the point cloud data or 3D model; The vertex coordinates are converted to a geodetic coordinate system through georeferencing to obtain the geographic location information.

3. The method for three-dimensional flight path planning of photovoltaic power station UAVs based on adaptive field of view and terrain modeling as described in claim 1, characterized in that, The completion of the row and column grouping of the photovoltaic panels includes: Based on the geometric features calculated from the vertex coordinates of the photovoltaic panel, a path planning reference point is generated. Density clustering algorithm is used to cluster reference points horizontally to form initial row groups; A standard straight line equation is obtained by performing a linear fit on the reference points within each initial row group; Based on the standard straight line equation, the photovoltaic panel rows are length-completed and vertically grouped to complete the row and column grouping.

4. The method for three-dimensional flight path planning of photovoltaic power station UAVs based on adaptive field of view and terrain modeling as described in claim 3, characterized in that, The horizontal clustering, forming initial row groups and the length padding and vertical grouping, includes: The calculation of geometric features based on the vertex coordinates of the photovoltaic panel includes calculating the length and width of the photovoltaic panel; The number of reference points for each photovoltaic panel is determined based on its aspect ratio. The DBSCAN clustering algorithm is used to cluster the reference points horizontally, and Euclidean distance is used as the metric during the clustering process. A linear regression algorithm is used to fit a straight line to the reference points within each initial row group. The length of the photovoltaic panel rows is supplemented based on the fitted linear equation to form a complete photovoltaic panel row structure.

5. The method for three-dimensional flight path planning of photovoltaic power station UAVs based on adaptive field of view and terrain modeling according to claim 3, characterized in that, It also includes intelligent grouping optimization steps: After grouping the photovoltaic panels into rows and columns, calculate the vertical coordinate difference between adjacent photovoltaic panels within the group; When the coordinate difference exceeds a preset vertical threshold, the current group is further subdivided horizontally. Based on the subdivided grouping results, the topological connection relationship of the photovoltaic panel rows is established; The neighborhood radius parameter of the clustering algorithm is dynamically adjusted based on the actual spatial distribution characteristics of the photovoltaic panel rows. The vertical threshold is set based on the actual physical size of the photovoltaic panel and the standard spacing between rows.

6. The method for three-dimensional flight path planning of photovoltaic power station UAVs based on adaptive field of view and terrain modeling according to claim 3, characterized in that, The execution process of the density clustering algorithm also includes a bicluster center distance optimization mechanism: Establish a dual-feature clustering model based on the spatial distribution characteristics of photovoltaic panels and the terrain elevation characteristics; Calculate the cluster centers based on spatial distribution characteristics and topographic elevation characteristics, respectively; The optimal cluster partition is determined by minimizing the weighted sum of the distances between the two cluster centers using a bi-cluster center distance optimization algorithm. Based on the optimized clustering results, the neighborhood radius parameter and minimum number of points of the clustering algorithm are dynamically adjusted. The weighting coefficients are dynamically adjusted based on the density of photovoltaic panel distribution and the degree of terrain undulation.

7. The method for three-dimensional flight path planning of photovoltaic power station UAVs based on adaptive field of view and terrain modeling as described in claim 1, characterized in that, The process of dynamically determining the flight altitude for each waypoint using a geometric model includes: Establish a waypoint geometry calculation model based on pitch angle and vertical field of view; Set the spacing between photovoltaic rows, the outer extension distance, and the outer extension distance factor; Calculate the UAV's field of view and solve for the horizontal distance from the waypoint to the photovoltaic panel and the flight altitude using trigonometric functions; Based on the obtained flight altitude and horizontal distance, combined with the geographical coordinates of the photovoltaic panel reference point, the three-dimensional coordinates of each waypoint are calculated; The outward expansion distance factor is dynamically adjusted according to the specific layout characteristics of the photovoltaic panel.

8. A method for three-dimensional flight path planning of photovoltaic power station UAVs based on adaptive field of view and terrain modeling as described in claim 5, characterized in that, The dynamic determination of waypoint connection order and optimization of acute-angle turns based on the principle of optimal distance includes: Calculate the Euclidean distance between the last waypoint of the current row and the first and last waypoints of the next row; Based on the principle of optimal distance, waypoints with shorter distances are selected as the starting points for connection to determine the flight direction; Calculate the angle between the direction vectors of adjacent path segments, and identify acute-angle turns when the angle is less than a preset threshold. For identified acute-angle turns, a trapezoidal transition path is formed by copying and translating waypoints; The distance of coordinate translation is dynamically determined based on the UAV's turning radius and flight speed.

9. A 3D flight path planning system for photovoltaic power station UAVs based on adaptive field of view and terrain modeling, applied to the 3D flight path planning method for photovoltaic power station UAVs based on adaptive field of view and terrain modeling as described in any one of claims 1 to 8, characterized in that, The system includes: The data acquisition module is used to acquire the geographical location information of each photovoltaic panel within the photovoltaic power station; The field of view determination module is used to determine the vertical field of view based on the parameters of the UAV image acquisition device. The photovoltaic panel grouping module is used to generate path planning reference points for each photovoltaic panel based on the geographical location information, and to use a density clustering algorithm to perform horizontal clustering of the reference points to form initial row groups. Then, a linear fitting method is used to fit the reference points in each initial row group to complete and correct the row structure, thereby completing the row and column grouping of the photovoltaic panels. The 3D waypoint generation module is used to dynamically determine the flight altitude for each waypoint based on the completed grouping of photovoltaic panels, combined with the vertical field of view of the UAV and the preset pitch angle, through geometric model calculation, thereby generating 3D waypoints that ensure complete coverage of the photovoltaic panels; The path optimization module is used to dynamically determine the connection order of three-dimensional waypoints between different rows of photovoltaic panels based on the principle of optimal distance, and to detect sharp turns in the optimized waypoint path to generate a smooth three-dimensional flight path. The route execution module is used to send the generated three-dimensional flight path to the UAV flight control system, driving the UAV to perform inspection tasks according to the planned path.

Images