Unmanned aerial vehicle multi-target image acquisition path planning method and device and storage medium

Abstract

The invention belongs to the technical field of navigation, and relates to an unmanned aerial vehicle multi-target image acquisition path planning method, equipment and a storage medium. The method comprises the following steps: acquiring a city live-action three-dimensional model; rasterizing the building model to create a three-dimensional grid map; acquiring distribution information of obstacles through the live-action three-dimensional model, judging whether the corresponding grid units are occupied by the obstacles or not, and generating an obstacle map according to a judgment result; dividing and storing space nodes in the grid map based on the obstacle map; performing initial path search by calculating a cost function of an initial node; and taking the obtained initial path as the basis of initial pheromone distribution, and obtaining an optimized path by adopting an ant colony optimization algorithm. According to the method, the influence of building shielding on image quality is avoided, time and energy waste of the unmanned aerial vehicle on an unnecessary path is reduced, the target image is obtained in the shortest time and the optimal path, and a flight task is completed.

Description

Technical Field

[0001] The present invention belongs to the technical field of navigation, and relates to a method, device and storage medium for multi-target image acquisition path planning of an unmanned aerial vehicle (UAV). Background Technique

[0002] With the progress of sensor, control system, communication technology and battery technology, the application scope of modern UAVs has been continuously expanded, covering multiple fields such as agricultural monitoring, environmental protection, disaster rescue, precision logistics, and geographical mapping. The core advantage of UAVs lies in their ability to perform aerial operations, breaking through the limitations of traditional ground operations and possessing the capabilities of efficient data acquisition and real-time monitoring. At the same time, the combination of artificial intelligence and machine learning has promoted the breakthroughs of UAVs in aspects such as autonomous flight and image recognition. However, whether it is used for geographical mapping, agricultural crop monitoring or environmental investigation, how to scientifically, accurately and efficiently plan the UAV path to ensure the integrity of the mission objectives and the high-quality acquisition of image data is still a key problem faced by the current application of UAVs.

[0003] UAV path planning refers to designing a flight route for the UAV to execute a mission from the starting point to the target area and finally return safely on the basis of meeting the mission requirements. Path planning is not just a simple design of the flight path. It involves the precise positioning of mission objectives, the optimization of flight time, the guarantee of flight safety, and how to avoid obstacles and risks in a complex environment. Reasonable path planning can significantly improve the mission execution efficiency, save time and energy, ensure the safety of UAV flight, and obtain the target images within the shortest time.

[0004] Although the current UAV path planning algorithms have made significant progress in technology, traditional algorithms such as Dijkstra's algorithm are inefficient and lack real-time performance when dealing with complex and dynamic environments; other meta-heuristic algorithms such as heuristic search and genetic algorithms can better handle uncertainties and find approximate optimal solutions, but they have high computational costs and are not suitable for all application scenarios. With the booming rise of high-rise buildings in cities, the actual operation scenarios of UAVs have undergone profound changes. High-rise buildings in cities not only change the air flow distribution on the ground surface, making UAVs face risks such as airframe shaking and flight trajectory deviation during flight; the narrow channels between high-rise buildings and the surrounding electromagnetic interference environment also add obstacles to the safe and stable flight of UAVs. Such a complex urban high-rise building environment undoubtedly brings varying degrees of challenges to the intelligent path planning of UAVs. Through various means such as UAV oblique photography, real-scene three-dimensional technology comprehensively and meticulously collects objects in the real scene, and then generates a three-dimensional model that highly restores the real scene. Through real-scene three-dimensional technology, the comprehensive environmental information of the UAV operation area can be accurately depicted, so that the UAV can timely respond to environmental changes, enhance the safety of path planning, and contribute to coordinating multi-path tasks. In addition, based on the real-time updated three-dimensional map data, effective coverage of multiple target points can be achieved, ensuring that the UAV efficiently completes the acquisition of image data for all targets with the optimal path. Summary of the Invention

[0005] To solve the problems brought by the complex urban high-rise building environment to UAV path planning, ensure that UAVs avoid affecting the image quality due to factors such as building occlusion, and at the same time take into account the image acquisition efficiency, avoid wasting time and resources on unnecessary paths, so that UAVs in complex environments can balance the relationship between image quality and acquisition efficiency, and obtain target images that meet the requirements with the shortest time and the optimal path. In addition, by leveraging the rich environmental information provided by real-scene three-dimensional, it is possible to overcome problems such as traditional algorithms being prone to falling into local optimal solutions, large computational amounts, non-smooth paths, and large node overheads, and efficiently plan the optimal path that meets the flight performance and task requirements of UAVs.

[0006] One of the technical solutions provided by the present invention is: a method for UAV multi-target image acquisition path planning, including the following steps: (1) Obtain the urban real-scene three-dimensional model; (2) Perform rasterization processing on the building model to create a three-dimensional raster map; through the urban real-scene three-dimensional model, obtain the distribution information of obstacles, judge whether the corresponding raster cells are occupied by obstacles, and generate an obstacle map according to the judgment results; (3) Divide and store the spatial nodes in the grid map based on the obstacle map obtained in step (2), and define the starting node and multiple waypoint target nodes of the UAV flight path according to the task requirements; perform an initial path search by calculating the cost function of the starting node; (4) Use the initial path obtained in step (3) as the basis for the initial pheromone distribution, and adopt the ant colony optimization algorithm to obtain the optimized path; (5) Smooth the optimized path obtained in step (4) to obtain the final UAV flight path.

[0007] Preferably, in step (1), first call the existing high-precision real-scene 3D model of the city. For the parts lacking data, obtain the texture information of ground objects through UAV photography; use photo control points for aerial triangulation encryption, and generate a high-precision real-scene 3D model of the city through software processing.

[0008] Preferably, in step (2), use an array or matrix to represent grid cells, and each grid includes the height information of buildings and obstacle attributes; whether each grid cell is passable is determined by the height threshold corresponding to the grid cell: ; Wherein, is the height of the grid and is the UAV flight height threshold; 0 represents a passable area, and 1 represents an obstacle area.

[0009] Preferably, in step (3), the cost function is defined as where is the actual cost from the starting node to the current node ; is the estimated cost from the current node to the target node.

[0010] Preferably, in step (3), the process of the initial path search is as follows: Calculate the cost function of the starting node, compare the function values of each node, and find the node with the smallest value as the target node; and backtrack through the parent node of each node until returning to the starting node, finally obtaining a complete initial path from the starting node to the target node.

[0011] Preferably, to smooth the optimized path in step (4), specifically: interpolate the optimized path by linear interpolation to increase the number of nodes in the path; use a Bezier curve to smooth the interpolated path to obtain the final path.

[0012] Preferably, the latest obstacle map is updated by using the environmental data collected and analyzed in real time by sensors, and steps (3)-(5) are executed to dynamically adjust the flight path of the drone in real time.

[0013] Another technical solution provided by the present invention is a multi-target image acquisition path planning device for a drone. The device includes a processor and a memory coupled to the processor for storing instructions. When the instructions are executed by the processor, the multi-target image acquisition path planning method for the drone is implemented.

[0014] The present invention also provides a storage medium storing a computer program, which implements the multi-target image acquisition path planning method for the drone when executed by a processor.

[0015] The beneficial effects of the present invention are as follows: By using the rich environmental information provided by the urban real-scene three-dimensional model, the present invention avoids the influence of building occlusion on the image quality, reduces the time and energy waste of the drone on unnecessary paths, obtains target images with the shortest time and the optimal path, and completes the flight mission. Compared with the conventional drone path planning method, the present invention can not only achieve the task of collecting multiple target data in one flight, but also avoid the problems of local optimal solutions and high computational complexity, generate a smoother path, reduce node overhead, and improve the overall task execution efficiency while ensuring the quality of image targets. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 is a flowchart of the multi-target image acquisition path planning method for the drone in the implementation of the present invention; Figure 2 is a schematic diagram of rasterization of a building model; Figure 3 is a schematic diagram of the path for the drone to complete the multi-target image acquisition task. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0017] To facilitate the understanding of the present invention, the present invention will be described in more detail below with reference to the accompanying drawings and specific embodiments. However, the present invention can be implemented in many different forms and is not limited to the embodiments described in this specification. On the contrary, these embodiments are provided to make the understanding of the disclosed content of the present invention more thorough and comprehensive.

[0018] Embodiment 1 The multi-target image acquisition path planning method for the drone provided in this embodiment has a process as Figure 1 shown, and the specific steps are as follows: Step 1, obtaining a city building model.

[0019] First, call the existing high-precision real-scene 3D model of the city. For the parts lacking data, use a multi-lens camera system carried on a drone platform to photograph the target area to obtain the texture information of the ground features; use photo control points for image matching and geometric calculations, and generate a high-precision real-scene 3D model of the city through software processing.

[0020] Step 2: Rasterize the building model.

[0021] Determine the 3D spatial range of rasterization according to the boundary of the building 3D model, and create a 3D raster map according to the determined raster parameters. Use an array or matrix to represent raster cells, and each raster includes the height information of the building and the obstacle attribute. According to the building 3D model data, identify and label obstacles such as buildings and trees, judge whether each raster cell is occupied by an obstacle, and mark each raster cell according to the obstacle judgment result. Usually, 0 represents the passable area and 1 represents the obstacle area, so as to generate an obstacle map, as Figure 2 shown.

[0022] The resolution of the raster map is determined by the formula where represents the map size, represents the number of pixels; the state (whether passable) of each raster cell is determined by the height threshold corresponding to the cell, and the passability is: ; where is the height of raster , is the drone flight height threshold, and its value is specifically determined by the drone device parameters and the urban airspace management regulations, etc.

[0023] Step 3: Search for the initial path.

[0024] This step guides the path search by defining and evaluating the cost function of each node to find a complete path that meets the requirements. The cost function is defined as where is the actual cost from the starting node to the current node , usually the cumulative cost calculated along the path, is the estimated cost from the current node to the target node . Considering the actual operation situation of the drone flight, select the Euclidean distance formula applicable to any moving direction for specific calculation. The Euclidean distance formula is: ; where is the node coordinates of are the coordinates of the target node.

[0025] Define the starting point and multiple waypoint targets of the UAV flight path according to the task requirements, and map them to the corresponding grid cells of the grid map. Divide the grid map into multiple nodes. According to the obstacle map obtained in step 2, clarify which nodes are passable and which nodes are where the obstacles are located. Initialize the open list and the closed list . The open list is used to store the nodes to be explored. When initialized, only the starting node is included in the open list; the closed list is used to store the explored nodes. When initialized, the list is empty. Calculate the cost function of the starting node . For the starting node, . Repeat the following steps (1)-(7) in sequence until a complete path that meets the requirements is obtained: (1) Traverse the open list, compare the values of each node in the list, and find the node with the smallest value as the target node , and the search ends; (2) Starting from the target node , backtrack the parent node of each node to obtain the complete path from the starting node to the target node ; (3) Remove the current node from the open list and add it to the closed list , indicating that the node has been explored and will not be searched in the next loop; (4) Obtain all adjacent nodes of the current target node , and process all adjacent nodes in sequence; (5) If the adjacent node is an obstacle node or has been stored in the closed list, ignore the node and continue to expand the next adjacent node; otherwise, calculate the actual cost of reaching the adjacent node from the starting node through the current target node . The calculation formula is , where is the distance from the target node to the adjacent node , which can be calculated according to the grid map resolution; (6) Determine whether the adjacent node is in the open list . If not, calculate the estimated cost of the adjacent node using the Euclidean distance formula , and obtain the cost of this node , and at the same time add this adjacent node to the open list, and set the target node as the parent node of the adjacent node ; (7) If the adjacent node is already in the open list , then define the one obtained in step (5) as , and compare it with the original actual cost ; if , it means that a new path with a smaller cost than the original path has been found. Therefore, update the value of the cost function and set the target node as the parent node of the adjacent node ; otherwise, do nothing.

[0026] By performing the above steps, find the target node that meets the requirements, and backtrack through the parent node of each node until returning to the starting node, and finally obtain the complete initial path from the starting node to the target node.

[0027] Step 4, Global path optimization.

[0028] This step uses the Ant Colony Optimization (ACO) algorithm, which simulates the behavior of ants leaving pheromones during the process of finding food. The path with a higher pheromone concentration has a greater probability of being selected. As the number of iterations increases, the pheromone is continuously updated, and the pheromone concentration on the path with a smaller path cost gradually increases, and finally converges to a better path.

[0029] Use the initial path obtained in step 3 as the basis for the initial pheromone distribution. For the node pairs on the path , set its initial pheromone concentration to a smaller value . Repeat the following steps until the maximum number of iterations is reached: (1) Randomly initialize a certain number (p) of ants. Each ant starts from the starting node and selects the next node according to the pheromone concentration and heuristic information to construct its own path. The probability formula for an ant to select the next node is: ; where is the pheromone concentration from node to node ; is the heuristic information ( is the node ​Distance to the node ); is a parameter for controlling the importance of pheromone concentration; is a parameter for controlling the importance of heuristic information; is the ant a set of nodes that the ant is allowed to choose from (i.e., nodes that have not been visited and are not obstacles), is a node in representing all possible next nodes that the ant can choose when it is at the current node ; is the sum of the weighted sums of the pheromone concentrations and heuristic information of all allowed nodes, ensuring that the sum of probabilities is 1; represents the weighted product of the pheromone concentration and heuristic information from node to node ; (2) Calculate the path cost of each ant , and the path cost can be defined as the sum of the distances between all node pairs on the path; (3) Update the pheromone distribution. The pheromone update formula is: ; where is the pheromone evaporation coefficient, ; is the pheromone increment left by the -th ant between node and node , is a constant; the smaller the path cost of the ant, the greater the pheromone increment left, thus attracting more ants to choose this path.

[0030] The path obtained through the above steps is the optimized path.

[0031] Step 5: Path smoothing.

[0032] Smooth the optimized path obtained in step 4, so that the UAV flight trajectory is more natural, reduce sharp turns, reduce flight energy consumption and improve flight stability at the same time.

[0033] First, interpolate the optimized path by linear interpolation to increase the number of nodes on the path. For example, for two adjacent nodes and , if intermediate nodes are inserted, the coordinate calculation formula for the intermediate point is , , , where , indicating the relative position ratio of the midpoint.

[0034] Bézier curves define the curve shape through control points. For control points , , , , the interpolation path is smoothed using the Bézier curve formula: .

[0035] This method obtains the points on the curve by multiplying each control point by the corresponding combination number and performing weighted summation. When varies from 0 to 1, it depicts the Bézier curve. The path smoothed by the Bézier curve is the final flight path of the drone.

[0036] Step 6, Result output.

[0037] Set an interface as the bridge between the path planning system and the drone control system. Through this interface, flight instructions are transmitted to the drone control system in real time, and the smoothed path is converted into a specific flight instruction set, including but not limited to key parameters such as speed adjustment (e.g., cruise speed), direction change (e.g., steering angle), altitude control (e.g., ascending or descending to a specific altitude), as Figure 3 shown, to ensure that it can fly accurately along the predetermined path.

[0038] 7. Real-time dynamic adjustment.

[0039] Through the highly sensitive integrated sensors carried by the drone, continuously collect and analyze the surrounding environmental data in real time, including but not limited to building distribution, temporary obstacle positions, etc. When the drone detects potential environmental conflicts on the traveling path, it immediately pauses the current flight route and quickly triggers the dynamic adjustment module of the system. The system marks the current conflict area as a temporary no-fly zone and transmits the relevant data to the control system to update and generate the latest obstacle map. Subsequently, starting from the current drone position, with the remaining uncompleted target points as the new task objectives, based on the updated environmental data, steps 3 to 6 are cyclically executed to quickly plan and generate a new flight path to ensure that the drone continues to perform the flight task safely and efficiently.

[0040] Embodiment 2 This embodiment provides a drone multi-target image acquisition path planning device, which includes a processor and a memory coupled to the processor for storing instructions. When the instructions are executed by the processor, the drone multi-target image acquisition path planning method of Embodiment 1 is implemented.

[0041] Embodiment 3. This embodiment provides a storage medium that stores a computer program. When the computer program is executed by a processor, it implements the method for planning the multi-target image acquisition path of the drone in Embodiment 1.

Claims

1. A method for path planning of multi-target image acquisition by an unmanned aerial vehicle, characterized in that It includes the following steps: (1) Obtain a three-dimensional model of the urban real scene; (2) Perform rasterization processing on the building model to create a three-dimensional raster map; obtain the distribution information of obstacles through the three-dimensional model of the urban real scene, determine whether the corresponding raster cell is occupied by an obstacle, and generate an obstacle map according to the judgment result; (3) Divide and store the spatial nodes in the raster map based on the obstacle map obtained in step (2), define the starting node and multiple target nodes along the flight path of the UAV according to the task requirements; perform an initial path search by calculating the cost function of the starting node; (4) Use the initial path obtained in step (3) as the basis for the initial pheromone distribution, and adopt the ant colony optimization algorithm to obtain an optimized path; (5) Smooth the optimized path obtained in step (4) to obtain the final flight path of the UAV.

2. The method for planning the multi-target image acquisition path of a drone according to claim 1, wherein, In step (1), first call the existing high-precision three-dimensional model of the city. For the parts lacking data, obtain the texture information of ground objects through UAV photography; perform aerial triangulation encryption using photo control points, and generate a high-precision three-dimensional model of the city through software processing.

3. The method for planning the multi-target image acquisition path of a drone according to claim 1, wherein In step (2), use an array or matrix to represent the raster cell, and each raster includes the height information of the building and the obstacle attribute; whether each raster cell is passable is determined by the height threshold corresponding to the raster cell: ； Among them, is the height of the grid and is the UAV flight altitude threshold; 0 represents a passable area, and 1 represents an obstacle area.

4. The method for planning the multi-target image acquisition path of a drone according to claim 1, wherein, In the said step (3), the cost function is defined as , where is the actual cost from the starting node to the current node ; is the estimated cost from the current node to the target node.

5. The method for planning the multi-target image acquisition path of a drone according to claim 1, wherein, In step (3), the process of the initial path search is as follows: calculate the cost function of the starting node, compare the function values of each node, and find the node with the smallest value as the target node; and backtrack through the parent node of each node until returning to the starting node, finally obtaining a complete initial path from the starting node to the target node.

6. The method for planning the multi-target image acquisition path of the unmanned aerial vehicle according to claim 1, characterized in that For the smoothing process of the optimized path in step (5), specifically: interpolate the optimized path by linear interpolation to increase the number of nodes in the path; use a Bezier curve to smooth the interpolated path to obtain the final path.

7. The method for planning the multi-target image acquisition path of an unmanned aerial vehicle according to claim 1, wherein, With the help of the environmental data collected and analyzed in real time by the sensor, update and generate the latest obstacle map, and execute steps (3)-(5) to perform real-time dynamic adjustment of the UAV flight path.

8. Drone multi-target image acquisition path planning device, characterized in that It includes a processor and a memory coupled to the processor for storing instructions, and when the instructions are executed by the processor, the method described in any one of claims 1-7 is implemented.

9. A storage medium, characterized in that, The storage medium stores a computer program, and when the computer program is executed by a processor, the method described in any one of claims 1-7 is implemented.

Images