Unmanned aerial vehicle aerial photography reservoir route planning method and system based on digital twinning

Abstract

The application relates to the technical field of route planning, in particular to a UAV aerial photography reservoir route planning method and system based on digital twinning, which comprises the following steps: acquiring terrain fluctuation data, performing noise reduction filtering processing on the terrain fluctuation data, generating a three-dimensional terrain model, and determining obstacle position data; performing block statistics on the obstacle position data to obtain obstacle distribution data, identifying an abnormal density area based on the obstacle distribution data; comparing the abnormal density area with a preset risk threshold to screen out a potential risk area; performing safety path planning on the potential risk area and screening to obtain a candidate path range; performing smoothing processing on the candidate path range, and screening according to a preset distance constraint to obtain a standard path sequence; and performing iterative optimization on the standard path sequence to obtain a UAV flight route. The method realizes efficient planning of a UAV aerial flight path.

Description

Technical Field

[0001] This invention relates to the field of route planning technology, and in particular to a method and system for planning routes for unmanned aerial vehicles (UAVs) to take aerial photographs of reservoirs based on digital twins. Background Technology

[0002] In modern water management and resource monitoring, drone aerial photography technology has become an indispensable tool due to its high efficiency, flexibility, and wide coverage. Especially in inspection and data collection tasks in complex terrain environments such as reservoirs, drones can significantly improve work efficiency and reduce the risks of manual operation, thus promoting the intelligent development of efficient reservoir management.

[0003] In one existing technology, sensors on a drone are used to collect reservoir terrain data. Based on this data, high-density obstacle areas are analyzed, and more detailed route planning or detours are performed in these areas to obtain a pre-set drone flight route for aerial photography of the reservoir. However, in the complex and ever-changing obstacle environment of a reservoir, systems based on traditional processor chips and data processing logic struggle to fully leverage the advantages of high-performance chips such as NPUs and FPGAs in digital twins. They cannot accurately assess the impact of obstacle spatial characteristics and height differences. When the drone passes through high-density obstacle areas, it may deviate from its path, waste energy, or even face the risk of collision.

[0004] In summary, existing technologies lack a comprehensive consideration of the deep-seated characteristics of the environment, making it difficult to efficiently plan the flight path of drone aerial photography, thus failing to meet the diverse needs of actual tasks. Summary of the Invention

[0005] This invention provides a method and system for planning routes for drone aerial photography of reservoirs based on digital twins, so as to achieve efficient planning of drone aerial photography flight paths.

[0006] Firstly, to address the aforementioned technical problems, this invention provides a method for planning routes for unmanned aerial vehicle (UAV) aerial photography of reservoirs based on digital twins, comprising:

[0007] Acquire terrain undulation data, perform noise reduction and filtering on the terrain undulation data to generate a three-dimensional terrain model, and determine obstacle location data based on the three-dimensional terrain model;

[0008] The obstacle location data is divided into blocks and statistically analyzed according to a preset classification standard to obtain obstacle distribution data. Based on the obstacle distribution data, regions with density higher than a preset density threshold are identified to obtain abnormal density regions.

[0009] The abnormal density areas are compared with a preset risk threshold to screen out potential risk areas;

[0010] Safe path planning is performed on the potential risk areas to obtain a set of avoidance paths. The set of avoidance paths is then filtered to obtain a range of candidate paths.

[0011] The candidate path range is smoothed, and risky paths are selected based on preset distance constraints. The risky paths are then replanned to obtain a standard path sequence.

[0012] Based on the standard path sequence, the path correction vector obtained from real-time analysis is fused, and the standard path sequence is iteratively optimized to obtain the UAV flight path.

[0013] In one optional implementation, the steps of acquiring terrain undulation data, performing noise reduction filtering on the terrain undulation data to generate a three-dimensional terrain model, and determining obstacle location data based on the three-dimensional terrain model include:

[0014] Topographic relief data is obtained by layer-by-layer scanning and terrain scanning. Noise filtering is then applied to the topographic relief data to obtain a denoised dataset.

[0015] Based on the denoised dataset, a 3D environment is constructed, and a 3D terrain model is generated;

[0016] Feature points are extracted from the three-dimensional terrain model to obtain obstacle location data.

[0017] In one optional implementation, the step of dividing the obstacle location data into blocks according to a preset classification standard to obtain obstacle distribution data, and identifying areas with density exceeding a preset density threshold based on the obstacle distribution data to obtain abnormal density areas, includes:

[0018] The obstacle location data is statistically analyzed according to a preset block segmentation standard. If the obstacle density value in a block is higher than a preset abnormal threshold, it is marked to obtain an abnormal block set.

[0019] The abnormal block set is labeled to obtain abnormal region boundary data, and the abnormal region boundary data is integrated to obtain abnormal density regions.

[0020] In one optional implementation, the step of comparing the abnormal density regions with a preset risk threshold to filter out potential risk areas includes:

[0021] Real-time environmental data within the abnormal density area is acquired, and the real-time environmental data is preliminarily cleaned to obtain dynamic environmental information.

[0022] Based on the environmental dynamic information, the density value of the abnormal density area is compared with historical data to determine whether a change has occurred, thereby obtaining the density change area.

[0023] If the density value of the density change region exceeds a preset risk threshold, the density change region is delineated to obtain a potential risk area.

[0024] In one optional implementation, the step of performing safe path planning on the potential risk area to obtain a set of avoidance paths, and then filtering the set of avoidance paths to obtain a range of candidate paths, includes:

[0025] Based on the potential risk areas, a path is generated in conjunction with the path offset magnitude to construct a set of avoidance paths;

[0026] The set of avoidance paths is compared, and if it does not meet the preset area boundary limit, the non-compliant path is removed to obtain a subset of filtered paths;

[0027] Based on the selected path subset and combined with dynamic adjustment conditions, the feasibility of the selected path subset is ranked to obtain a range of candidate paths; wherein, the dynamic adjustment conditions include dynamic data such as real-time wind speed and terrain changes.

[0028] In one optional implementation, the step of smoothing the candidate path range, filtering out risky paths based on preset distance constraints, and replanning the risky paths to obtain a standard path sequence includes:

[0029] The path sequences within the candidate path range are initially smoothed to obtain a set of path sequences;

[0030] If the path points in the path sequence set do not meet the preset safety standard, the interval between the path points and the obstacles is compared to obtain the path interval to be adjusted.

[0031] For the path interval to be adjusted, combined with the dynamic adjustment conditions, a dynamic offset range is generated to obtain the adjusted path sequence;

[0032] The adjusted path sequence is subjected to regional boundary detection to obtain the detection result. If the detection result does not conform to the preset regional boundary limit, the path point position is adjusted again to obtain the standard path sequence.

[0033] In one optional implementation, the step of iteratively optimizing the path sequence based on the standard path sequence, fusing the path correction vector obtained from real-time analysis, to obtain the UAV flight route includes:

[0034] Real-time environmental data is acquired and integrated to obtain obstacle density data;

[0035] Based on the obstacle density data, areas with obstacle density higher than a preset density threshold are marked and compared with historical obstacle data to obtain newly added obstacle areas;

[0036] Based on the newly added obstacle area and combined with the preset distance constraints, a set of adjustment directions is obtained;

[0037] Boundary evaluation is performed on the set of adjustment directions. If the set of adjustment directions does not conform to the preset boundary range constraint, it is re-selected to obtain the path correction vector.

[0038] The path correction vector is verified. If the path correction vector does not meet the safe distance constraint, it is re-filtered to obtain the UAV flight path.

[0039] Secondly, the present invention provides a drone aerial photography reservoir route planning system based on digital twins, comprising:

[0040] The data preprocessing module is used to acquire terrain undulation data, perform noise reduction and filtering on the terrain undulation data, generate a three-dimensional terrain model, and determine obstacle location data based on the three-dimensional terrain model.

[0041] The region classification module is used to perform block statistics on the obstacle location data according to a preset classification standard to obtain obstacle distribution data. Based on the obstacle distribution data, it identifies regions with a density higher than a preset density threshold to obtain abnormal density regions.

[0042] The risk detection module is used to compare the abnormal density areas with a preset risk threshold and filter out potential risk areas.

[0043] The path planning module is used to plan safe paths for the potential risk areas, obtain a set of avoidance paths, and filter the set of avoidance paths to obtain a range of candidate paths.

[0044] The path filtering module is used to smooth the range of candidate paths, filter out risky paths according to preset distance constraints, and replan the risky paths to obtain a standard path sequence.

[0045] The dynamic iteration module is used to iteratively optimize the standard path sequence based on the standard path sequence and integrate the path correction vector obtained from real-time analysis to obtain the UAV flight path.

[0046] Compared with the prior art, the present invention has the following beneficial effects:

[0047] This invention acquires terrain undulation data through layer-by-layer scanning and terrain scanning, then performs noise filtering to obtain a denoised dataset, and then constructs a three-dimensional environment to generate a three-dimensional terrain model, extracts feature point data, performs spatial clustering analysis on the location distribution of obstacles, and obtains obstacle location data by comparing the data with a preset density threshold.

[0048] (1) This invention obtains terrain undulation data through layer-by-layer scanning and terrain scanning, generates a three-dimensional terrain model through noise reduction and filtering, and determines obstacle location data by combining spatial clustering analysis. This solution makes full use of the advantages of digital twin technology to accurately restore the reservoir terrain and obstacle distribution, and solves the problem that existing technologies cannot meet the diverse needs of actual tasks.

[0049] (2) This invention identifies abnormal density areas and screens potential risk areas by performing block statistics on obstacle location data, and plans a set of avoidance paths in a targeted manner. It focuses on risk areas to carry out differentiated path design, avoiding the defects of insufficient consideration of risk areas in traditional planning. This helps to reduce path redundancy and energy waste, improve flight efficiency, and make path planning more in line with the actual needs of the complex environment of the reservoir.

[0050] (3) This invention performs smoothing and risk screening on alternative paths, optimizes the path sequence by combining dynamic adjustment conditions, and performs multiple rounds of screening and adjustment to eliminate unreasonable turning points in the path, effectively improving the reliability of the planned path. At the same time, this solution adapts to environmental changes in real time, reduces the difficulty of operating the UAV, and improves the stability and accuracy of data collection.

[0051] (4) This invention obtains the UAV flight route by iteratively optimizing the standard path sequence through the fusion of real-time path correction vectors. This scheme provides a dynamic response to environmental changes, making up for the shortcomings of traditional fixed paths in adapting to real-time scenarios. It solves the problem that traditional methods are difficult to accurately capture environmental features, helps improve obstacle recognition accuracy, provides reliable data support for subsequent path planning, and ensures flight safety. Attached Figure Description

[0052] Figure 1 This is a schematic diagram of the method for planning a reservoir route by drone aerial photography based on digital twin provided in an embodiment of the present invention;

[0053] Figure 2 This is a schematic diagram of the avoidance path set provided in the embodiments of the present invention;

[0054] Figure 3 This is a schematic diagram of the structure of a drone aerial photography reservoir route planning system based on digital twin provided in an embodiment of the present invention. Detailed Implementation

[0055] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0056] Reference Figure 1 This invention provides a method for planning routes for drone aerial photography of reservoirs based on digital twins, including:

[0057] S11, acquire terrain undulation data, perform noise reduction filtering on the terrain undulation data, generate a three-dimensional terrain model, and determine obstacle location data based on the three-dimensional terrain model;

[0058] S12, the obstacle location data is divided into blocks and statistically analyzed according to a preset classification standard to obtain obstacle distribution data. Based on the obstacle distribution data, areas with density higher than a preset density threshold are identified to obtain abnormal density areas.

[0059] S13, compare the abnormal density areas with a preset risk threshold to screen out potential risk areas;

[0060] S14, perform safe path planning on the potential risk area to obtain a set of avoidance paths, and filter the set of avoidance paths to obtain a range of candidate paths;

[0061] S15, smooth the range of candidate paths, filter out risky paths according to preset distance constraints, replan the risky paths, and obtain a standard path sequence.

[0062] S16. Based on the standard path sequence, the path correction vector obtained from real-time analysis is fused, and the standard path sequence is iteratively optimized to obtain the UAV flight route.

[0063] In step S11, terrain undulation data is acquired, the terrain undulation data is subjected to noise reduction filtering to generate a three-dimensional terrain model, and obstacle location data is determined based on the three-dimensional terrain model.

[0064] It should be noted that the first step is to acquire terrain undulation data. This data is collected by a UAV equipped with high-precision sensors using a preset layer-by-layer scanning method to survey the terrain surrounding the reservoir. The sampling frequency is adjusted for different terrain areas to obtain raw terrain undulation data, resulting in a preliminary terrain point cloud dataset. The preset layer-by-layer scanning method divides the scanning space into several continuous parallel layers of fixed thickness (e.g., 1m) based on the elevation range of the reservoir area. The UAV performs a full-coverage scan within each layer according to a preset path. The sampling frequency is dynamically adjusted for different terrain types, such as plains and mountains (mountains are defined as having a slope greater than 15°); for example, plains are scanned once per square meter, while mountainous areas are scanned three times per square meter. The raw data acquired through layered scanning is then integrated into a preliminary terrain point cloud dataset after coordinate alignment.

[0065] In one implementation, the step of performing noise reduction filtering on the terrain undulation data to generate a three-dimensional terrain model, and determining obstacle location data based on the three-dimensional terrain model, includes:

[0066] Topographic relief data is obtained by layer-by-layer scanning and terrain scanning. Noise filtering is then applied to the topographic relief data to obtain a denoised dataset.

[0067] Based on the denoised dataset, a 3D environment is constructed, and a 3D terrain model is generated;

[0068] Feature points are extracted from the three-dimensional terrain model to obtain obstacle location data.

[0069] Layer-by-layer scanning and terrain scanning refer to dividing the target area vertically into several continuous and uniformly thick parallel layers according to preset spatial layering rules, and then collecting data from each layer using high-precision sensing equipment mounted on the UAV. The preset spatial layering rules are used to define the vertical spatial layers of the target area when the UAV collects terrain undulation data, and include two parts: layer thickness parameters and sampling frequency parameters. These rules can be set in three modes—normal, efficiency, and precision—based on terrain complexity and aerial photography requirements, achieving a dynamic balance between data acquisition accuracy and scanning efficiency. The layer thickness determines the height interval of each vertical scan by the UAV, and the sampling frequency determines the horizontal sampling density within each layer. The system can automatically select appropriate parameter combinations based on terrain characteristics to adapt to the scanning needs of different terrain scenarios.

[0070] For example, in a data collection project targeting the topographic relief of a reservoir area, the scanning range included both plains and mountainous areas. The plains area had gentle topography and sparse vegetation, so an efficiency-oriented setting was used, with a layer thickness of 2m and a sampling frequency of 1 time per square meter to ensure rapid coverage. In contrast, the mountainous area had complex topography, so a precision-oriented setting was used, reducing the layer thickness to 1m and increasing the sampling frequency to 3 times per square meter to obtain detailed topographic relief data.

[0071] Noise filtering involves two stages of denoising. First, median filtering is used to initially remove discrete noise points in the original terrain undulation data caused by sensor interference or abnormal reflections. Median filtering uses a sliding window to statistically analyze the height values ​​of adjacent sampling points. When the window size is set to 3×3, it effectively removes isolated abnormal high or low values ​​while preserving the edge features of the data. Subsequently, Gaussian filtering is applied to the median-filtered data for smoothing, reducing local fluctuations and improving overall continuity. The standard deviation parameter σ of the Gaussian filter can be set according to the complexity of the terrain: σ is 0.5 in gently sloping areas and 1 in areas with significant terrain undulations; the filter window radius can be set to 5 sampling points.

[0072] The 3D environment construction includes three steps: point cloud coordinate fusion, spatial meshing, and surface fitting. First, multi-layer terrain data acquired by the UAV during layer-by-layer scanning is registered and fused according to geographic coordinates and altitude information to ensure alignment of each layer's scanning results within the same spatial coordinate system. Second, the fused point cloud data undergoes spatial meshing, dividing the scanned area into uniform 0.5m × 0.5m × 0.5m 3D mesh units. Each mesh unit stores the average elevation value and surface feature point information within its corresponding spatial range to reflect the local undulations of the terrain. Finally, a triangular mesh interpolation algorithm is used to fit the surface based on the meshed data, generating a continuous 3D terrain surface model. The triangular mesh interpolation algorithm is a spatial interpolation method for constructing continuous terrain surfaces based on discrete point sets. It uses filtered terrain sampling points as vertices and generates an irregular triangular network according to the spatial distance relationship between adjacent points, covering the entire terrain area with a series of interconnected triangular patches. The elevation value within each triangular patch is calculated from the vertex coordinates through linear interpolation, thus forming a continuous terrain surface.

[0073] It should be noted that the feature point data is extracted from the 3D environment model. Specifically, firstly, elevation gradient analysis is performed on the 3D terrain model, the average elevation value of each 3D grid cell is recorded, and the average elevation difference between that cell and its adjacent cells on the same horizontal plane is calculated. The average elevation difference of all adjacent faces is taken as the average elevation difference value of the grid. If the absolute value of this average elevation difference is greater than 0.3m, it is determined that there is a significant terrain abrupt change in the grid, and its geometric center point is extracted as the preliminary feature point.

[0074] Furthermore, spatial clustering is performed on the obtained preliminary feature point set. The aggregation criteria are: the shortest straight-line distance does not exceed 1m and the absolute value of the elevation difference between feature points does not exceed 0.1m (0.1m is set at 10% of the elevation difference threshold of 0.3m to ensure consistency). Points meeting these criteria are grouped into the same feature cluster. Finally, the geometric center point of each feature cluster is calculated, and this center point is used as the final feature point data. Each feature point represents a three-dimensional spatial coordinate with a terrain abrupt change, indicating the possible presence of obstacles at that location. For example, if the range of grid cell X is [10.0 m, 10.5 m], the range of Y is [20.0 m, 20.5 m], and the range of Z is [30.0 m, 30.5 m], then the center of X is (10.0 + 10.5) / 2 = 10.25 m, the center of Y is (20.0 + 20.5) / 2 = 20.25 m, and the center of Z is (30.0 + 30.5) / 2 = 30.25 m. Therefore, the geometric center point is (10.25, 20.25, 30.25).

[0075] In step S12, the obstacle location data is segmented and statistically analyzed according to a preset segmentation standard to obtain obstacle distribution data. Based on the obstacle distribution data, regions with density exceeding a preset density threshold are identified to obtain abnormal density regions, including:

[0076] The obstacle location data is statistically analyzed according to a preset block segmentation standard. If the obstacle density value in a block is higher than a preset abnormal threshold, it is marked to obtain an abnormal block set.

[0077] The abnormal block set is labeled to obtain abnormal region boundary data, and the abnormal region boundary data is integrated to obtain abnormal density regions.

[0078] The preset segmentation standard divides the 3D environment model into several non-overlapping, continuously covering square blocks on a horizontal plane. Each block has a side length of 10m, representing a 10m × 10m horizontal spatial area, facilitating the segmentation and analysis of obstacle location data. Each block includes attributes such as its corresponding spatial coordinate range and the number of feature points. The preset anomaly threshold is a pre-set quantitative standard value used to measure obstacle density within a block. The anomaly density threshold for each block is set by the mean and standard deviation of the number of obstacles in that block over the past 30 days. The threshold can be set as the sum of the mean of the historical obstacle count over 30 days and k times the standard deviation of the historical data over 30 days, where k is an empirical constant, typically set to 2, to cover approximately 95% of the normal data range. Subsequently, the data from all anomaly blocks are integrated to obtain an anomaly block set.

[0079] After obtaining the set of anomalous blocks, adjacency is determined using Euclidean distance. Specifically, if the Euclidean distance between the center points of two anomalous blocks is equal to the block side length (10m), the two blocks are considered adjacent. All adjacent anomalous blocks are connected and integrated to form anomalous regions. The convex hull algorithm is used to identify the boundary of each anomalous region. First, the center points of the blocks are sorted by their horizontal (and vertical) coordinates. When constructing the lower convex hull from left to right, the cross product of the vector formed by the new point and the last two points of the convex hull is checked. If the cross product is positive, the vertex is retained (left-turned); otherwise, the vertex is removed. The upper convex hull is constructed similarly and then merged to obtain the boundary of the smallest convex polygon. The anomalous region boundary data is the set of coordinate points at the intersection of all anomalous and non-annomous regions, providing a reliable basis for subsequent path planning and risk assessment. Subsequently, the anomalous region boundary data is organized. Overlapping anomalous region boundaries can be merged to form a complete anomalous region boundary. All grids within the anomalous region boundary are marked to obtain the anomalous density region.

[0080] In step S13, the abnormal density regions are compared against a preset risk threshold to filter out potential risk areas, including:

[0081] Real-time environmental data within the abnormal density area is acquired, and the real-time environmental data is preliminarily cleaned to obtain dynamic environmental information.

[0082] Based on the environmental dynamic information, the density value of the abnormal density area is compared with historical data to determine whether a change has occurred, thereby obtaining the density change area.

[0083] If the density value of the density change region exceeds a preset risk threshold, the density change region is delineated to obtain a potential risk area.

[0084] It should be noted that the real-time environmental data is obstacle distribution data continuously monitored by multiple sensor devices. These sensors are deployed in areas of abnormal density and include obstacle monitoring devices mounted on drones or on the ground. They can monitor the three-dimensional spatial coordinates, size, and category of obstacles within their area and transmit the data stream to the central processing system in real time.

[0085] It is worth noting that the acquisition of data streams needs to ensure real-time performance and completeness. Therefore, data can be collected every 30 minutes to ensure timely updates and provide a dynamic and continuous environmental information foundation for subsequent analysis. The initial cleaning process involves denoising and formatting the collected raw data. In reservoir perimeter monitoring, sensors may collect abnormal values ​​due to external interference, such as sudden signal interruptions or numerical jumps. One possible approach is to mark data points that significantly deviate from a preset average threshold as invalid and insert the historical 10-day average value at that position into the invalid data bits. The preset average threshold is set by the mean and standard deviation of data points collected in the past 10 days for that block. The threshold is set as the sum of the mean of historical data points within the past 10 days and k times the standard deviation of historical data within the past 10 days, where k is an empirical constant, typically set to 2, to cover approximately 95% of the normal data range. Subsequently, the remaining data is processed into a time series format to form structured environmental dynamic information. This cleaning process helps improve the accuracy of subsequent analysis.

[0086] Furthermore, the environmental dynamic information is analyzed to distinguish between current and historical data. Based on the current obstacle distribution data, the corresponding historical obstacle distribution data and fluctuation range within the region are extracted. The standard deviation of the average distribution density over the past 10 days within the current region is set as the density deviation threshold for the current region. For example, in a certain area around a reservoir, the historical average distribution density is 5 per cubic meter, and the standard deviation is 2 per cubic meter. Therefore, a reasonable obstacle deviation threshold for the current region is set to a change of 2 per cubic meter. If the current data stream shows a sudden increase in distribution density to 9 per cubic meter, it can be preliminarily determined that there is a density change in the area, and the area is identified as a density change area. Conversely, if the deviation is within the density deviation threshold, it can be attributed to measurement error, and there is no density anomaly in the area.

[0087] The preset risk threshold is an upper limit threshold for obstacle density used for early warning, obtained based on historical density data analysis. For example, if 12 points per cubic meter are set as the risk threshold, when analyzing areas of density change, if a density value of 15 points per cubic meter is detected in an area, it is determined to be a risk area. After detection, the boundary lines of all risk areas are marked, and the boundary line data is integrated to obtain potential risk areas. These potential risk areas are then drawn on the 3D environment model or map, with their boundaries highlighted using different colors.

[0088] In step S14, safe path planning is performed on the potential risk area to obtain a set of avoidance paths. The set of avoidance paths is then filtered to obtain a range of candidate paths, including:

[0089] Based on the potential risk areas, a path is generated in conjunction with the path offset magnitude to construct a set of avoidance paths;

[0090] The set of avoidance paths is compared, and if it does not meet the preset area boundary limit, the non-compliant path is removed to obtain a subset of filtered paths;

[0091] Based on the selected path subset and combined with dynamic adjustment conditions, the feasibility of the selected path subset is ranked to obtain the range of candidate paths.

[0092] It should be noted that during path generation, a preliminary path is first calculated based on the terrain, starting point, and ending point. This preliminary path is typically the shortest path and may traverse some potential risk areas. The path offset represents the extent to which the flight path deviates from the original path when the path needs to cross or avoid a potential risk area. The path offset is calculated from the obstacle location data and the preset risk distance constraint between the preliminary path to ensure that the UAV can safely bypass potential risk areas. The preset risk distance constraint is obtained by multiplying the maximum cruising speed of the aerial drone by the response delay time. It limits the minimum straight-line distance between the UAV's trajectory and the potential risk area. For example, if the maximum cruising speed of the aerial drone is 12 m / s and the system response delay is 0.5 s, then the preset risk distance constraint is 12 × 0.5 = 6 m, ensuring that the UAV will not collide or cause other dangers due to getting too close to obstacles during flight.

[0093] It should be noted that the path offset calculation steps are as follows: First, for path points overlapping with potential risk areas, denoted as overlapping path points, the path offset direction is determined based on the positive or negative direction of the line connecting the path point to the center of the potential risk area. The longest distance from this path point to the boundary of the potential risk area in the offset direction is extracted, and this longest distance is added to a preset risk distance constraint to obtain the path offset magnitude of the path point. Second, for path points not overlapping with potential risk areas, to ensure path smoothness, their offset direction is consistent with the offset direction of the nearest overlapping path point. The shortest distance from this path point to the boundary of the potential risk area in the offset direction is extracted, and the preset risk distance constraint is subtracted from this shortest distance to obtain the path offset magnitude of the path point. Through the above calculations, it can be ensured that the path maintains a safe distance from the risk boundary while avoiding potential risk areas, and also maintains the overall continuity and smoothness of the path.

[0094] Reference Figure 2This invention provides an example of an avoidance path set. The dashed line from point O to point P represents the initial path, which sequentially traverses potential risk zones A and B, thus requiring path correction. A preset risk distance constraint of 3.6m is set. Based on this constraint, the path offset magnitude is calculated for each path point on the standard path: for path points overlapping with potential risk zones A or B, the offset direction is southwest or northeast, and the corresponding offset magnitude is calculated; for path points not overlapping with potential risk zones, their offset direction is consistent with the offset direction of the nearest overlapping path point, and the corresponding path offset magnitude is calculated, resulting in four path offset magnitude schemes. Based on these schemes, the initial path is corrected, resulting in four different avoidance paths represented by solid lines, denoted as path ①, path ②, and path ③. These three avoidance paths are then integrated to obtain an avoidance path set.

[0095] Furthermore, boundary constraints are applied to the set of avoidance paths. If an avoidance path is detected to exceed the preset area boundary limit, it is discarded, resulting in a filtered subset of paths. The preset area boundary limit is a flight boundary limit based on factors such as the drone's endurance, flight start and end points, used to reduce the possibility of unforeseen situations such as signal interference or insufficient battery life during drone flight. For example, in a flight path planning scenario, the drone's endurance is 20km. The midpoint of the shortest straight line between the flight start and end points is found, and a circle with a radius of 10km centered at this point is set as the area boundary limit. The avoidance path set obtained in this path planning includes the north path, the south path, and the far perimeter path. It is detected that the north path exceeds the area boundary limit near the reservoir edge, potentially preventing the drone from safely completing its mission; in this case, the path is marked as non-compliant and discarded. The south path and the far perimeter path are retained because they do not exceed the boundary limit, forming the filtered path subset. This method ensures that path planning always takes place within a safe area.

[0096] Furthermore, based on the selected path subset and combined with dynamic adjustment conditions, the feasibility of the selected path subset is ranked to obtain a range of candidate paths. The dynamic adjustment conditions include dynamic data such as real-time wind speed and terrain changes. During the ranking process, the system assigns equal weight to four indicators: safety, smoothness, path length, and environmental stability, and selects the optimal path by comprehensively considering all four indicators. For example, when ranking feasibility, the southern path, being located in a valley and close to potential risk areas, is nearly straight and suitable for rapid passage, but its location in a valley necessitates consideration of the impact of wind speed on path stability. The more distant outer paths, although detouring, are far from all potential risk areas and avoid complex environments such as valleys, resulting in a more stable environment. Based on the four indicators of safety, smoothness, path length, and environmental stability, the scores for the far-end outer path were 9, 6, 5, and 8, respectively, for a total of 28 points. The scores for the south-side path were 6, 7, 8, and 5, for a total of 26 points. Ultimately, the far-end outer path ranked higher due to its higher total score, while the south-side path was considered an alternative. This ranking method helps in selecting the most suitable path in dynamic environments.

[0097] In step S15, the range of candidate paths is smoothed, and risky paths are selected based on preset distance constraints. These risky paths are then replanned to obtain a standard path sequence, including:

[0098] The path sequences within the candidate path range are initially smoothed to obtain a set of path sequences;

[0099] If the path points in the path sequence set do not meet the preset safety standard, the interval between the path points and the obstacles is compared to obtain the path interval to be adjusted.

[0100] For the path interval to be adjusted, combined with the dynamic adjustment conditions, a dynamic offset range is generated to obtain the adjusted path sequence;

[0101] The adjusted path sequence is subjected to regional boundary detection to obtain the detection result. If the detection result does not conform to the preset regional boundary limit, the path point position is adjusted again to obtain the standard path sequence.

[0102] It should be noted that the preliminary smoothing operation includes extracting the coordinates of path points sequentially from the path sequence within the candidate path range, and performing smoothing calculations using a moving average method on the coordinate values ​​of adjacent path points. Specifically, taking any path point on the path as the center, a sliding window is formed by selecting five path points before and after it, calculating the average value of the path point coordinates within the window, and replacing the original path point coordinates with this average value to generate a smooth path sequence. The preset safety standard is to ensure the safe flight of the UAV. The preset safe distance between the UAV and obstacles refers to the minimum safe distance that the UAV must maintain from obstacles, risk areas, or terrain changes on the flight path. This standard can be determined based on two basic parameters: the UAV's body size and cruise speed. For example, in a path planning scenario, using a UAV with an arm span of 0.6m and a cruise speed of 5m / s, the preset safety standard is set to the sum of three times the body width and the dynamic safe distance (cruise speed multiplied by the system response time), which is 3 × 0.6m + 5 m / s × 0.5 s = 1.8m + 2.5m = 4.3m. Furthermore, it is determined whether the distance between the obstacle and the path point in the path meets the preset safety standard. If it does not meet the standard, it is marked as a path point to be adjusted. The path interval to be adjusted is a set of path points to be adjusted.

[0103] When generating the dynamic offset amplitude, it is necessary to consider not only static adjustment conditions such as the relative position of the obstacle to the current path interval and the path direction, but also dynamic adjustment conditions such as terrain changes and wind speed. To ensure the safety of the UAV, the relative position of the obstacle to the current path interval is the primary consideration, followed by fine adjustments based on terrain changes, wind speed, and other adjustment conditions. The vector pointing from the path point to the obstacle is used as the reference direction, and its opposite direction is taken as the basis for avoidance. Combined with the path tangent vector at the path point (calculated by the coordinate difference of adjacent points), the offset lateral direction is determined through the vector cross product, ensuring that the offset direction of all points is consistent. Subsequently, in the determined offset direction, according to preset safety standards, the shortest and longest possible offset distances of the path points are calculated to obtain the basic offset amount that meets the safety interval. Finally, the basic offset amount is corrected based on dynamic factors. When environmental disturbances are strong, the median value of the offset interval is taken to leave sufficient safety distance; conversely, the lower limit of the interval can be appropriately selected to shorten the path distance. The path interval to be adjusted is modified according to the dynamic offset amplitude to obtain the adjusted path sequence.

[0104] For example, there is a section of the path north of the reservoir that needs adjustment. Based on the relative positions of obstacles and the current path section, a suitable offset range is calculated to be 3m to 7m eastward. Analysis of dynamic environmental data shows that the wind speed is higher on the east side of the path section to be adjusted. To ensure high safety of the modified path, the median of the offset range is taken, resulting in a final dynamic offset of 5m eastward. This dynamic adjustment method can adapt to environmental changes and ensure the practicality of the path.

[0105] Furthermore, regional boundary detection is performed on the adjusted path sequence to ensure that the path conforms to preset regional boundary constraints. This regional boundary detection checks the trajectory of each path to ensure that the trajectory of each path does not deviate from the preset regional boundary constraints. For paths exceeding the regional boundary constraints, the positions of the path points on that path are readjusted according to dynamic adjustment conditions. After adjustment and passing the direction consistency check, a standard path sequence is obtained. The preset regional boundary constraints are flight boundary constraints inferred based on factors such as the UAV's endurance, flight start and end points, used to reduce the possibility of unforeseen situations such as signal interference and insufficient battery life during UAV flight. In one flight path planning, the UAV's endurance is 20km. The midpoint of the shortest straight line between the flight start and end points is found, and a circle with a radius of 10km centered at this point is set as the regional boundary constraint. For paths exceeding the regional boundary constraints, the positions of the path points on that path are readjusted in conjunction with dynamic adjustment conditions.

[0106] In step S16, the step of iteratively optimizing the path sequence based on the standard path sequence and fusing the path correction vector obtained from real-time analysis to obtain the UAV flight route includes:

[0107] Real-time environmental data is acquired and integrated to obtain obstacle density data;

[0108] Based on the obstacle density data, areas with obstacle density higher than a preset density threshold are marked and compared with historical obstacle data to obtain newly added obstacle areas;

[0109] Based on the newly added obstacle area and combined with the preset distance constraints, a set of adjustment directions is obtained;

[0110] Boundary evaluation is performed on the set of adjustment directions. If the set of adjustment directions does not conform to the preset boundary range constraint, it is re-selected to obtain the path correction vector.

[0111] The path correction vector is verified. If the path correction vector does not meet the preset distance constraints, it is re-filtered to obtain the UAV flight route.

[0112] The preset density threshold for each block is set by the mean and standard deviation of the number of obstacles in that block over the past 30 days. The threshold is set as the sum of the mean of the number of historical obstacles over the past 30 days and k times the standard deviation of the historical data for the same period (k=2). The obstacle data is compared with the preset density threshold, and areas where the obstacle density exceeds the preset abnormality threshold are marked as abnormal density areas. Furthermore, to accurately assess changes in these abnormal density areas, real-time abnormal density areas are compared with historical abnormal density areas. Areas not previously present in historical abnormal density areas are statistically analyzed to obtain newly added obstacle areas. For example, if an area was previously a non-abnormal density area but recently experienced strong winds causing tree branches to accumulate, forming an abnormal density area, this area will be identified and marked as a newly added obstacle area. This method can promptly identify potential obstacles, clarify the scope of path segments requiring adjustment, and provide clear guidance for subsequent path planning.

[0113] The preset distance constraint is a distance standard set during path planning to ensure sufficient spatial distance between the path and newly added obstacle areas. This constraint is obtained by multiplying the maximum cruising speed by the system response delay time, and represents the minimum Euclidean distance between the flight path and potential risk areas. For example, if the maximum cruising speed of the aerial drone is 12 m / s and the system response delay is 0.5 s, then the preset risk distance constraint is 12 × 0.5 = 6 m. This value serves as the preset distance constraint.

[0114] For newly added obstacle areas on the standard path sequence, a set of adjustment directions needs to be generated to correct the path. Based on preset distance constraints, multiple possible adjustment directions are generated from the standard path sequence and organized into a set to avoid the newly added obstacle areas. Using the newly added obstacle area as a reference, rays are generated at preset angle intervals (e.g., 15°) on both sides of the tangent direction of the standard path. Ray directions that satisfy the preset constraint on the distance to the obstacle boundary are extracted as candidate adjustment directions. This set of adjustment directions will provide candidate solutions for subsequent path optimization.

[0115] Furthermore, the set of adjustment directions is compared with the preset boundary range limit. Adjustment directions that do not conform to the preset boundary range limit are re-selected and organized to obtain the path correction vector. The preset area boundary limit, mentioned in step S15, is a flight boundary limit inferred based on factors such as the UAV's endurance, flight start and end points, and maximum flight altitude, combined with environmental factors such as reservoir terrain. It is used to reduce the possibility of unforeseen situations such as signal interference and insufficient battery life during UAV flight.

[0116] Finally, the path correction vector is subjected to multiple safety distance constraint checks. The safety distance constraint is the larger of a preset boundary range constraint and a preset distance constraint, ensuring that both boundary restrictions and obstacle avoidance safety requirements are met simultaneously. This is used for global verification of the path correction vector. If the path correction vector does not meet the preset distance constraints, it is re-selected and verified again. Only after passing the verification is the UAV flight path generated by combining the path correction vector and the standard path sequence. This UAV flight path will ensure that all identified obstacle areas are avoided, while maintaining path smoothness and effectiveness, meeting all safety and efficiency requirements, and ensuring that the UAV can safely and stably complete its mission during flight.

[0117] In summary, this invention provides a method for planning drone aerial photography routes in reservoirs based on digital twins. This method can comprehensively consider the deep-level characteristics of the reservoir environment, accurately identify obstacles within the reservoir, and thus efficiently plan the drone's aerial photography flight path.

[0118] Reference Figure 3 This invention provides a drone aerial photography reservoir route planning system based on digital twins, comprising:

[0119] The data preprocessing module is used to acquire terrain undulation data, perform noise reduction and filtering on the terrain undulation data, generate a three-dimensional terrain model, and determine obstacle location data based on the three-dimensional terrain model.

[0120] The region classification module is used to perform block statistics on the obstacle location data according to a preset classification standard to obtain obstacle distribution data. Based on the obstacle distribution data, it identifies regions with a density higher than a preset density threshold to obtain abnormal density regions.

[0121] The risk detection module is used to compare the abnormal density areas with a preset risk threshold and filter out potential risk areas.

[0122] The path planning module is used to plan safe paths for the potential risk areas, obtain a set of avoidance paths, and filter the set of avoidance paths to obtain a range of candidate paths.

[0123] The path filtering module is used to smooth the range of candidate paths, filter out risky paths according to preset distance constraints, and replan the risky paths to obtain a standard path sequence.

[0124] The dynamic iteration module is used to iteratively optimize the standard path sequence based on the standard path sequence and integrate the path correction vector obtained from real-time analysis to obtain the UAV flight path.

[0125] It should be noted that the UAV aerial photography reservoir route planning system based on digital twin provided in this embodiment of the invention is used to execute all the process steps of the UAV aerial photography reservoir route planning method based on digital twin in the above embodiment. The working principles and beneficial effects of the two are one-to-one, so they will not be described again.

[0126] This invention also provides an electronic device. The electronic device includes a processor, a memory, and a computer program stored in the memory and executable on the processor, such as a drone-based aerial photography route planning program for reservoirs based on digital twins. When the processor executes the computer program, it implements the steps described in the various embodiments of the drone-based aerial photography route planning method for reservoirs based on digital twins, for example... Figure 1 The step S11 shown. Alternatively, when the processor executes the computer program, it implements the functions of each module / unit in the above-described device embodiments, such as the data preprocessing module.

[0127] For example, the computer program may be divided into one or more modules / units, which are stored in the memory and executed by the processor to complete the present invention. The one or more modules / units may be a series of computer program instruction segments capable of performing a specific function, which describe the execution process of the computer program in the electronic device.

[0128] The electronic device may be a desktop computer, laptop, handheld computer, or smart tablet, etc. The electronic device may include, but is not limited to, a processor and memory. Those skilled in the art will understand that the above components are merely examples of electronic devices and do not constitute a limitation on the electronic device. It may include more or fewer components than described above, or combine certain components, or different components. For example, the electronic device may also include input / output devices, network access devices, buses, etc.

[0129] The processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor. The processor is the control center of the electronic device, connecting all parts of the electronic device via various interfaces and lines.

[0130] The memory can be used to store the computer programs and / or modules. The processor implements various functions of the electronic device by running or executing the computer programs and / or modules stored in the memory and by calling data stored in the memory. The memory may mainly include a program storage area and a data storage area. The program storage area may store the operating system, at least one application program required for a function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created according to the use of the mobile phone (such as audio data, phonebook, etc.). In addition, the memory may include high-speed random access memory, and may also include non-volatile memory, such as hard disk, memory, plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0131] Wherein, if the modules / units integrated in 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. Based on this understanding, all or part of the processes in the methods of the above embodiments of the present invention can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the content included in the computer-readable medium can be appropriately added or removed according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electrical carrier signals and telecommunication signals.

[0132] It should be noted that the device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Furthermore, in the accompanying drawings of the device embodiments provided by this invention, the connection relationships between modules indicate that they have communication connections, which can be specifically implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement this without any creative effort.

[0133] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention for those skilled in the art.

Claims

1. A method for planning routes for drone aerial photography of reservoirs based on digital twins, characterized in that, include: Acquire terrain undulation data, perform noise reduction and filtering on the terrain undulation data to generate a three-dimensional terrain model, and determine obstacle location data based on the three-dimensional terrain model; The obstacle location data is divided into blocks and statistically analyzed according to a preset classification standard to obtain obstacle distribution data. Based on the obstacle distribution data, regions with density higher than a preset density threshold are identified to obtain abnormal density regions. The abnormal density areas are compared with a preset risk threshold to screen out potential risk areas; Safe path planning is performed on the potential risk areas to obtain a set of avoidance paths. The set of avoidance paths is then filtered to obtain a range of candidate paths. The candidate path range is smoothed, and risky paths are selected based on preset distance constraints. The risky paths are then replanned to obtain a standard path sequence. Based on the standard path sequence, the path correction vector obtained from real-time analysis is fused, and the standard path sequence is iteratively optimized to obtain the UAV flight path; Specifically, the step of performing safe path planning on the potential risk area to obtain a set of avoidance paths, and then filtering the set of avoidance paths to obtain a range of candidate paths, includes: Based on the potential risk areas, a path is generated in conjunction with the path offset magnitude to construct a set of avoidance paths; The initial directions of the avoidance path set are compared. If they do not meet the preset area boundary restrictions, the non-compliant paths are eliminated to obtain a subset of filtered paths. Based on the selected path subset and combined with dynamic adjustment conditions, the feasibility of the selected path subset is ranked to obtain a range of candidate paths; wherein, the dynamic adjustment conditions include real-time wind speed and dynamic data on terrain changes. The process of smoothing the candidate path range, filtering out risky paths based on preset distance constraints, and replanning the risky paths to obtain a standard path sequence includes: The path sequences within the candidate path range are initially smoothed to obtain a set of path sequences; If the path points in the path sequence set do not meet the preset safety standard, the interval between the path points and the obstacles is compared to obtain the path interval to be adjusted. For the path interval to be adjusted, combined with the dynamic adjustment conditions, a dynamic offset range is generated to obtain the adjusted path sequence; The adjusted path sequence is subjected to regional boundary detection to obtain the detection result. If the detection result does not conform to the preset regional boundary limit, the path point position is adjusted again to obtain the standard path sequence.

2. The method for planning drone aerial photography routes for reservoirs based on digital twins according to claim 1, characterized in that, The process of acquiring terrain undulation data, performing noise reduction and filtering on the terrain undulation data to generate a three-dimensional terrain model, and determining obstacle location data based on the three-dimensional terrain model includes: Topographic relief data is obtained by layer-by-layer scanning and terrain scanning. Noise filtering is then applied to the topographic relief data to obtain a denoised dataset. Based on the denoised dataset, a 3D environment is constructed, and a 3D terrain model is generated; Feature points are extracted from the three-dimensional terrain model to obtain obstacle location data.

3. The method for planning drone aerial photography routes for reservoirs based on digital twins according to claim 1, characterized in that, The obstacle location data is segmented and statistically analyzed according to a preset classification standard to obtain obstacle distribution data. Based on the obstacle distribution data, regions with density exceeding a preset density threshold are identified to obtain abnormal density regions, including: The obstacle location data is statistically analyzed according to a preset block segmentation standard. If the obstacle density value in a block is higher than a preset abnormal threshold, it is marked to obtain an abnormal block set. The abnormal block set is labeled to obtain abnormal region boundary data, and the abnormal region boundary data is integrated to obtain abnormal density regions.

4. The method for planning drone aerial photography routes for reservoirs based on digital twins according to claim 1, characterized in that, The step of comparing the abnormal density regions with a preset risk threshold to filter out potential risk areas includes: Real-time environmental data within the abnormal density area is acquired, and the real-time environmental data is preliminarily cleaned to obtain dynamic environmental information. Based on the environmental dynamic information, the density value of the abnormal density area is compared with historical data to determine whether a change has occurred, thereby obtaining the density change area. If the density value of the density change region exceeds a preset risk threshold, the density change region is delineated to obtain a potential risk area.

5. The method for planning drone aerial photography routes for reservoirs based on digital twins according to claim 1, characterized in that, The method involves iteratively optimizing the path sequence based on the standard path sequence, integrating the path correction vector obtained from real-time analysis, to obtain the UAV flight route, including: Real-time environmental data is acquired and integrated to obtain obstacle density data; Based on the obstacle density data, areas with obstacle density higher than a preset density threshold are marked and compared with historical obstacle data to obtain newly added obstacle areas; Based on the newly added obstacle area and combined with the preset distance constraints, a set of adjustment directions is obtained; Distance evaluation is performed on the set of adjustment directions. If the set of adjustment directions does not conform to the preset boundary range constraint, it is re-filtered to obtain the path correction vector. The path correction vector is verified. If the path correction vector does not meet the safe distance constraint, it is re-filtered to obtain the UAV flight path.

6. A drone-based aerial photography route planning system for reservoirs based on digital twins, characterized in that, The method for planning drone aerial photography routes for reservoirs based on digital twins as described in any one of claims 1 to 5 includes: The data preprocessing module is used to acquire terrain undulation data, perform noise reduction and filtering on the terrain undulation data, generate a three-dimensional terrain model, and determine obstacle location data based on the three-dimensional terrain model. The region classification module is used to perform block statistics on the obstacle location data according to a preset classification standard to obtain obstacle distribution data. Based on the obstacle distribution data, it identifies regions with a density higher than a preset density threshold to obtain abnormal density regions. The risk detection module is used to compare the abnormal density areas with a preset risk threshold and filter out potential risk areas. The path planning module is used to plan safe paths for the potential risk areas, obtain a set of avoidance paths, and filter the set of avoidance paths to obtain a range of candidate paths. The path filtering module is used to smooth the range of candidate paths, filter out risky paths according to preset distance constraints, and replan the risky paths to obtain a standard path sequence. The dynamic iteration module is used to iteratively optimize the standard path sequence based on the standard path sequence and integrate the path correction vector obtained from real-time analysis to obtain the UAV flight path.

Images