UAV Full Coverage Path Planning Method and System Based on Adaptive Viewpoint Sampling

Abstract

The present invention discloses a method and system for unmanned aerial vehicle full-coverage path planning based on adaptive viewpoint sampling, including: adaptive viewpoint generation: discretizing the position and direction in space to generate initial viewpoints, using adaptive sampling comparison to identify the areas not covered by the initial viewpoints, and iterating within the unrecognized areas until a qualified set of viewpoints is generated; full-coverage path generation: clustering the final viewpoints and connecting each cluster internally, connecting each cluster with its surrounding neighbor clusters to generate a search graph, and using the concepts of entropy and neighbor distance as heuristic functions to generate a full-coverage path; path coverage rate evaluation: calculating the volume of the structure to be detected that has been covered by accumulating along the path generated in the previous step, so as to calculate the path coverage percentage; which can greatly reduce the time and path length consumed by multi-rotor unmanned aerial vehicles when inspecting complex large structures, and greatly improve the energy utilization rate of the unmanned aerial vehicles.

Description

Technical Field

[0001] The present invention belongs to the field of multi-rotor UAV path planning, and particularly relates to a UAV full-coverage path planning method and system based on adaptive viewpoint sampling. Background Art

[0002] In traditional large-structure surface detection tasks, workers need to visually inspect the surface, or combine a lift table and a crane to achieve the surface detection of the object to be inspected. However, such a detection method consumes a lot of time and requires professional tools and equipment for cooperation, resulting in a high cost. In recent years, multi-rotor UAVs have been widely used in various fields due to their advantages of flexibility, lightness, and low cost. The present invention further introduces multi-rotor UAVs in the field of surface detection tasks of large and complex structures such as buildings and large airliners, reducing labor, effectively reducing costs, greatly improving the detection efficiency, and enhancing the detection accuracy.

[0003] Generally, when using a multi-rotor UAV for large-object surface detection, it is necessary to consider that the flight path of the UAV fully covers the three-dimensional object surface to prevent missed detection by the UAV, and at the same time ensure the safe distance between the UAV and the three-dimensional object surface to ensure the safety of the UAV and the object to be detected. However, using a handheld remote control to operate the UAV flight makes it impossible to evaluate whether the UAV flight path achieves full coverage, and it is difficult to keep the distance between the UAV and the object to be detected constant, increasing the detection time and danger. Therefore, there is an urgent need for an efficient full-coverage path planning method for large three-dimensional structures.

[0004] Existing UAV full-coverage path planning methods for three-dimensional structures mostly divide the three-dimensional space into layers and convert the three-dimensional full-coverage problem into a two-dimensional full-coverage problem for processing. Such a processing method will lead to problems such as high path repetition rate and unreasonable paths, reducing the efficiency of large three-dimensional structure detection. Summary of the Invention

[0005] The main purpose of the present invention is to overcome the deficiencies in the existing three-dimensional full-coverage path planning algorithm technology, and provide a UAV full-coverage path planning method based on adaptive viewpoint sampling, which not only has a high detection coverage rate, high efficiency, and low cost, but also is not affected by the model structure, has robustness, can greatly reduce the time and path length consumed by the multi-rotor UAV when inspecting complex large structures, and greatly improves the energy utilization rate of the UAV.

[0006] To achieve the above object, the technical solution adopted by the present invention is: a UAV full-coverage path planning method based on adaptive viewpoint sampling, including the following steps:

[0007] Adaptive Viewpoint Generation

[0008] First, discretize the space in terms of position and direction to generate initial viewpoints; then, use adaptive sampling comparison to identify the areas not covered by the initial viewpoints; finally, iterate within the unrecognized areas until a qualified set of viewpoints is generated;

[0009] Full Coverage Path Generation

[0010] Cluster the final set of viewpoints generated in the previous step and connect each cluster internally; then, connect each cluster to its surrounding neighbor clusters to generate a search graph; finally, use the concepts of entropy and neighbor distance as heuristic functions to generate a full coverage path;

[0011] Path Coverage Evaluation

[0012] Accumulate and calculate the volume of the to-be-detected structure that has been covered along the full coverage path to calculate the path coverage percentage. During this process, occlusion culling of the reference model is required.

[0013] The specific steps to generate the initial viewpoints are as follows: First, perform position-based discretization using the set grid resolution to generate a set of sample UAV waypoints that define the x, y, z positions in the 3D space;

[0014] Generate a set of yaw angle directions by performing direction-based discretization on each position sample. The generated set of sample waypoints S is represented by the x, y, z coordinates and the ψ yaw angle:

[0015]

[0016] where: x i , y i , z i are the coordinates of the UAV of the i-th sample waypoint in the world coordinate system respectively; ψ i is the yaw angle of the UAV of the i-th sample waypoint in the world coordinate system;

[0017] Then, generate each sensor viewpoint by applying a 4×4 transformation matrix, which defines the position of the sensor relative to the UAV body frame; the transformation matrix is defined as follows:

[0018]

[0019] where: is the pose transformation matrix between the body coordinate system A and the sensor coordinate system B, which transforms the waypoint described in the body coordinate system A to be described in the sensor coordinate system B; is the rotation transformation matrix between the body coordinate system A and the sensor coordinate system B, which describes the attitude relationship between these two coordinate systems; A P Bis the translation transformation matrix between two coordinate systems A and B, describing the positional relationship between the two coordinate systems, and the pose transformation matrix is determined after the drone is assembled;

[0020] Finally, the initial waypoints are filtered using methods based on collision and maximum and minimum distances; The collision-based filtering method eliminates sample viewpoints that are inside the model or collide with the model; The distance-based filtering method makes the sample viewpoints available within a defined distance from the structural model, which is determined based on the minimum and maximum effective ranges of the sensor, as follows:

[0021]

[0022] where: intersection() is a function that calculates the number of intersection points between a ray and the model; Ray[(x i ,y i ,z i ),(a,b,c)] is a ray pointing from point i(x i ,y i ,z i ) to point (a,b,c); model is the model of the three-dimensional structure to be inspected; d min is the minimum distance from the viewpoint i to the model; d max is the maximum distance from the viewpoint i to the model; d i is the actual distance from point i to the model;

[0023] Through the above method, a set of initial sensor viewpoints is generated:

[0024]

[0025] When using adaptive sampling comparison to identify the areas not covered by the initial viewpoints, the uniform sampling process in the previous step is iteratively repeated at different discretization resolutions, and at each discretization level, the uncovered voxels are identified by comparing the voxels generated by the model of the three-dimensional structure to be inspected and the voxels visible to the viewpoints;

[0026] Using the nearest neighbor method and the calculated volume occupancy rate, the identified uncovered voxels are regionally divided by using Euclidean clustering to finally find the uncovered regions;

[0027] The regions with lower accuracy are identified by calculating the standard deviation of the depth errors of each point in the covered part.

[0028] When iterating within the unrecognized regions until a qualified set of viewpoints is generated, it is iteratively repeated within the uncovered regions at a discretization resolution different from that during adaptive viewpoint generation;

[0029] At each discretization level when using adaptive sampling comparison to identify areas not covered by the initial viewpoints, areas with no coverage and low precision are targeted and resampled at a finer resolution until a sufficient number of viewpoint samples are generated to ensure model coverage with the required precision.

[0030] When generating a full-coverage path, a heuristic function R is used to generate the path. The heuristic function R used is:

[0031] R = Ee -λδd (1.5)

[0032] Where: R is the heuristic function used for full-coverage path planning; E represents the total entropy of the path where the current waypoint is located; λ is the distance contribution factor; δd is the distance between the current waypoint and its parent waypoint;

[0033] During the process of full-coverage path planning, waypoints with larger R values are selected to minimize δd and maximize the entropy E. Minimizing δd means a shorter travel distance, and maximizing the entropy E is maximizing the information gain.

[0034] When evaluating the path coverage rate, the integrity of the coverage path planning is evaluated by quantifying the percentage of the covered volume of the structure and the predicted 3D structure volume on the generated path, which is specifically expressed as follows:

[0035]

[0036] Where Coverage represents the coverage rate of the path; CV Voxel represents the surface volume reconstructed by collecting data along the full-coverage path generated by the algorithm; OV Voxel represents the surface volume of the model of the three-dimensional structure to be inspected.

[0037] Meanwhile, a multi-rotor UAV full-coverage path planning system based on adaptive viewpoint sampling is provided, including an adaptive viewpoint generation module, a full-coverage path generation module, and a path coverage rate evaluation module; the adaptive viewpoint generation module is used to discretize the position and direction of space to generate initial viewpoints; then use adaptive sampling comparison to identify areas not covered by the initial viewpoints; finally, iterate within the unrecognized areas until a qualified set of viewpoints is generated;

[0038] The full-coverage path generation module is used to cluster the final set of viewpoints generated by the adaptive viewpoint generation module and connect each cluster internally; then connect each cluster with its surrounding neighbor clusters to generate a search graph; finally, use the concepts of entropy and neighbor distance as heuristic functions to generate a full-coverage path;

[0039] The path coverage evaluation module is used to accumulate and calculate the volume of the structure to be detected that has been covered along the full-coverage path, so as to calculate the path coverage percentage. During this process, occlusion culling needs to be performed on the reference model.

[0040] In addition, the present invention also provides a computer device, including a processor and a memory. The memory is used to store computer-executable programs. The processor reads the computer-executable programs from the memory and executes them. When the processor executes the computer-executable programs, it can implement the unmanned aerial vehicle full-coverage path planning method based on adaptive viewpoint sampling described herein.

[0041] At the same time, a computer-readable storage medium can be provided. A computer program is stored in the computer-readable storage medium. When the computer program is executed by a processor, it can implement the unmanned aerial vehicle full-coverage path planning method based on adaptive viewpoint sampling described herein.

[0042] Compared with the prior art, the present invention has at least the following beneficial effects:

[0043] Based on the method described in the present invention, different discretizations of space can be performed during the process of generating viewpoints. Areas with low coverage rate and low coverage accuracy are identified, and re-sampling is performed at a finer resolution in these areas until enough viewpoint samples are generated to ensure the coverage of the model with the required accuracy. This ensures the coverage rate and coverage accuracy, while reducing computing resources and computing time; the viewpoints are connected into clusters, and each cluster is connected internally, and then each cluster is connected to the nearest cluster neighbor points externally to generate a search graph, and then a path is generated by using the graph search method; the path generation speed is improved, and the processing ability for large point sets is improved compared with the same type of methods, ensuring the feasibility of the path; the coverage volume is used to estimate the path coverage rate, and the coverage rate calculation is more accurate, and the evaluation of the path is more objective.

[0044] Furthermore, a heuristic function R is used to generate a path, minimizing δd and maximizing E, ensuring that the path has a lower path length cost, and the detection process has lower noise, lower uncertainty, and better model accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 It is a flowchart of an implementable method of the present invention;

[0046] Figure 2 It is the implementation process of the present invention for a fighter model;

[0047] Figure 3 It is the implementation result of the present invention for a fighter model;

[0048] Figure 4Schematic diagram of the adaptive viewing point generated for the fighter model according to the present invention;

[0049] Figure 5 Clustering schematic diagram generated for the fighter model according to the present invention;

[0050] Figure 6 Schematic diagram of the fighter model. Specific implementation manner

[0051] The technical solutions in the embodiments of the present invention will be further described in detail below:

[0052] The flow of the method proposed by the present invention is as Figure 1 shown below:

[0053] Step 1: Adaptive viewing point generation. First, discretize the position and direction in space to generate an initial viewing point for the fighter model; then use adaptive sampling to compare and identify the areas not covered by the initial viewing point; finally, iterate within the unrecognized areas until a qualified set of viewing points is generated;

[0054] Step 2: Full-coverage path generation. First, cluster the final viewing points generated in the previous step and connect each cluster internally; then connect each cluster with its surrounding neighbor clusters to generate a search graph; finally, use the concepts of entropy and neighbor distance as heuristic functions to generate a full-coverage path for the fighter model;

[0055] Step 3: Path coverage rate evaluation. Accumulate and calculate the volume of the fighter model that has been covered along the path generated in the previous step to calculate the path coverage percentage. During this process, occlusion culling of the fighter model is required. The UAV poses obtained after completing the above three steps are as Figure 6 shown.

[0056] Furthermore, the generation of the adaptive sampling viewing point described in Step 1 mainly includes the following steps: 1) Initial viewing point generation; 2) Identifying the uncovered areas using adaptive sampling; 3) Repeating the iteration within the uncovered areas at different discretization resolutions.

[0057] 1) Initial viewing point generation:

[0058] First, perform position-based discretization using a specific grid resolution to generate a set of sample UAV waypoints that define the x, y, z positions in 3D space;

[0059] Generate a set of yaw angle directions by performing direction-based discretization on each position sample. The generated set of sample waypoints S is represented by x, y, z coordinates and ψ yaw angle:

[0060]

[0061] where: x i , y i , z i are the coordinates of the i-th sample waypoint UAV in the world coordinate system respectively; ψ i is the yaw angle of the i-th sample waypoint UAV in the world coordinate system.

[0062] Then, each sensor view point is generated by applying a 4×4 transformation matrix, which defines the position of the sensor relative to the UAV body frame. The transformation matrix is generally defined as follows:

[0063]

[0064] where: is the pose transformation matrix between the body coordinate system A and the sensor coordinate system B, which can transform the waypoint described in the body coordinate system A to the sensor coordinate system B for description; is the rotation transformation matrix between the body coordinate system A and the sensor coordinate system B, which describes the attitude relationship between these two coordinate systems; A P B is the translation transformation matrix between the two coordinate systems A and B, which describes the position relationship between these two coordinate systems. The pose transformation matrix is generally determined after the UAV is assembled.

[0065] Finally, a method based on collision and maximum and minimum distances is used to filter the initial waypoints. The collision-based filtering method eliminates the sample view points inside the model or colliding with the model; the distance-based filtering method makes the sample view points available within the defined distance from the fighter model, which is determined based on the minimum and maximum effective ranges of the sensor. These two constraints are defined as follows:

[0066]

[0067] where: intersection() is a function to calculate the number of intersection points between a ray and the model; Ray[(x i , y i , z i ), (a, b, c)] is a ray pointing from point i(x i , y i , z i ) to point (a, b, c); model is the fighter model; d min is the minimum distance from the view point i to the model; d max is the maximum distance from the view point i to the model; d i is the actual distance from point i to the model.

[0068] A set of initial viewpoints of the sensors can be generated through the above method:

[0069]

[0070] 2) Use adaptive sampling to identify uncovered areas

[0071] In order to improve the viewpoints, in this step, the uniform sampling process in the previous step is iteratively repeated at different discretization resolutions, and at each discretization level, the uncovered voxels are identified by comparing the voxels generated by the known fighter model and the voxels visible from the viewpoints.

[0072] Using the nearest neighbor method and the calculated volume occupancy rate, the identified uncovered voxels are partitioned into regions by using Euclidean clustering, and finally the uncovered regions are found. In addition, the present invention will also identify the regions with lower accuracy by calculating the standard deviation of the depth errors of each point in the covered part.

[0073] 3) Iteratively repeat at different discretization resolutions within the uncovered regions

[0074] At each discretization level in the previous step, the uncovered and low-accuracy regions are taken as targets for resampling at a finer resolution until enough viewpoint samples are generated to ensure the coverage of the model with the required accuracy.

[0075] The adaptive viewpoints generated for the fighter model are as Figure 4 shown.

[0076] Furthermore, in step 2, the viewpoints are connected into clusters, each cluster is connected internally, and then each cluster is connected to the nearest adjacent cluster externally to generate a search graph, and then a search path is generated by using graph search. The clusters generated for the fighter model are as Figure Five shown.

[0077] Furthermore, the heuristic function R used in step 2:

[0078] R = Ee -λδd (1.5)

[0079] where: R is the heuristic function used for full-coverage path planning; E represents the total entropy of the path where the current waypoint is located; λ is the distance contribution factor; δd is the distance between the current waypoint and its parent waypoint.

[0080] During the process of full - coverage path planning, waypoints with larger R values are selected to minimize δd and maximize the entropy E. Minimizing δd means a shorter travel distance, which can effectively save the energy consumption of the UAV and increase its endurance. Maximizing the entropy E is actually maximizing the Information Gain (IG), which means that the entire full - coverage path has lower noise, lower uncertainty, and better model accuracy.

[0081] The full - coverage path generated for the fighter model is as Figure 3 shown.

[0082] Furthermore, in step 3, the integrity of the coverage path planning is evaluated by quantifying the percentage of the covered volume of the fighter model and the predicted 3D structure volume on the generated path. The specific description is as follows:

[0083]

[0084] Where: Coverage represents the coverage rate of the path; CV Voxel represents the surface volume reconstructed by collecting data along the full - coverage path generated by the algorithm; OV Voxel represents the surface volume of the model of the three - dimensional structure to be inspected.

[0085] The present invention also provides a multi - rotor UAV full - coverage path planning system based on adaptive view - point sampling, including an adaptive view - point generation module, a full - coverage path generation module, and a path coverage rate evaluation module; the adaptive view - point generation module is used to discretize the position and direction of the space to generate initial view - points; then it uses adaptive sampling to compare and identify the areas not covered by the initial view - points; finally, it iterates within the unrecognized areas until a qualified set of view - points is generated;

[0086] The full - coverage path generation module is used to cluster the final set of view - points generated by the adaptive view - point generation module and connect each cluster internally; then it connects each cluster with its surrounding neighbor clusters to generate a search graph; finally, it uses the concepts of entropy and neighbor distance as heuristic functions to generate a full - coverage path;

[0087] The path coverage rate evaluation module is used to accumulate and calculate the volume of the structure to be detected that has been covered along the full - coverage path, so as to calculate the path coverage percentage. During this process, occlusion culling of the reference model is required.

[0088] In addition, the present invention can also provide a computer device, including a processor and a memory. The memory is used to store computer-executable programs. The processor reads part or all of the computer-executable programs from the memory and executes them. When the processor executes part or all of the computer-executable programs, it can implement the method for full-coverage path planning of an unmanned aerial vehicle based on adaptive viewpoint sampling according to the present invention.

[0089] On the other hand, the present invention provides a computer-readable storage medium. A computer program is stored in the computer-readable storage medium. When the computer program is executed by a processor, it can implement the method for full-coverage path planning of an unmanned aerial vehicle based on adaptive viewpoint sampling according to the present invention.

[0090] The computer device can be a laptop computer, a desktop computer or a workstation.

[0091] The processor can be a central processing unit (CPU), a digital signal processor (DSP), an application-specific integrated circuit (ASIC) or a field-programmable gate array (FPGA).

[0092] For the memory according to the present invention, it can be an internal storage unit of a laptop computer, a desktop computer or a workstation, such as a memory or a hard disk; it can also use an external storage unit, such as a mobile hard disk or a flash card.

[0093] The computer-readable storage medium can include a computer storage medium and a communication medium. The computer storage medium includes volatile and non-volatile, removable and non-removable media implemented by any method or technology for storing information such as computer-readable instructions, data structures, program modules or other data. The computer-readable storage medium can include: read-only memory (ROM), random access memory (RAM), solid state drives (SSD) or optical discs, etc. Among them, the random access memory can include resistive random access memory (ReRAM) and dynamic random access memory (DRAM).

Claims

1. An unmanned aerial vehicle full-coverage path planning method based on adaptive viewpoint sampling, characterized in that, It includes the following steps: Adaptive viewpoint generation First, discretize the position and orientation of the space to generate initial viewpoints; Then, use adaptive sampling comparison to identify the areas not covered by the initial viewpoints; finally, iterate within the unrecognized areas until a qualified set of viewpoints is generated; Full-coverage path generation Cluster the set of final viewpoints generated in the previous step and connect each cluster internally; then connect each cluster to its neighboring clusters around it to generate a search graph; finally, use the concepts of entropy and neighbor distance as heuristic functions to generate a full-coverage path; when generating the full-coverage path, use the heuristic function to generate the path, and the heuristic function used : (1.5) Wherein: is the heuristic function used for full-coverage path planning; represents the total entropy of the path where the current waypoint is located; is the distance contribution factor; is the distance between the current waypoint and its parent waypoint; During the process of full-coverage path planning, select waypoints with larger values to minimize , and maximize entropy , minimizing means a shorter travel distance, and maximizing entropy is to maximize the information gain; Path coverage rate evaluation Accumulatively calculate the volume of the to-be-detected structure that has been covered along the full-coverage path to calculate the path coverage percentage. During this process, occlusion culling of the reference model is required.

2. The method for full-coverage path planning of an unmanned aerial vehicle based on adaptive view-point sampling according to claim 1, wherein Generate an initial viewpoint as follows: First, perform position-based discretization using the set grid resolution to generate a set of sample UAV waypoints that define positions in 3D space; A set of yaw angle directions is generated by performing direction-based discretization on each position sample, and the generated set of sample waypoints consisting of coordinates and yaw angles is represented as: (1.1) Wherein: are respectively the coordinates of the UAV at the th sample waypoint in the world coordinate system; is the yaw angle of the UAV at the th sample waypoint in the world coordinate system; Generate each sensor viewpoint through a 4×4 transformation matrix, and the transformation matrix defines the position of the sensor relative to the UAV airframe frame; the transformation matrix is defined as follows: (1.2) Wherein: is the body coordinate system and the sensor coordinate system The pose transformation matrix between the two coordinate systems is used to transform the waypoint described in the body coordinate system to the sensor coordinate system for description; is the body coordinate system and the sensor coordinate system The rotation transformation matrix between the two coordinate systems describes the attitude relationship between the two coordinate systems; is the translation transformation matrix between the two coordinate systems, which describes the position relationship between the two coordinate systems. The pose transformation matrix is determined after the drone is assembled; Finally, use the methods based on collision and maximum-minimum distance to filter the initial waypoints; the collision-based filtering method eliminates the sample viewpoints inside the model or in collision with the model; the distance-based filtering method makes the sample viewpoints available within the defined distance from the structure model, and this distance is determined based on the minimum and maximum effective ranges of the sensor, as follows: (1.3) Wherein: is a function for calculating the number of intersection points between a ray and a model; is a ray pointing from point to point ; is a model of the three-dimensional structure to be inspected; is the viewing point the minimum distance from the model; is the viewing point the maximum distance from the model; is the actual distance of point from the model; Generate a set of initial sensor viewpoints through the above method: (1.4)。 3. The method for full-coverage path planning of an unmanned aerial vehicle based on adaptive viewpoint sampling according to claim 1, wherein When using adaptive sampling comparison to identify the areas not covered by the initial viewpoints, iteratively repeat the uniform sampling process at different discretization resolutions, and at each discretization level, identify the uncovered voxels by comparing the voxels generated by the model of the three-dimensional structure to be inspected and the voxels visible to the viewpoints; Use the nearest neighbor method and the calculated volume occupancy rate to partition the identified uncovered voxels into regions through Euclidean clustering, and finally find the uncovered regions; Identify the regions with lower accuracy by calculating the standard deviation of the depth errors of each point in the covered part.

4. The method for full-coverage path planning of an unmanned aerial vehicle based on adaptive view-point sampling according to claim 1, wherein When iterating within the unrecognized areas until a qualified set of viewpoints is generated, iteratively repeat at a discretization resolution different from that in adaptive viewpoint generation within the uncovered areas; At each discretization level when using adaptive sampling comparison to identify the areas not covered by the initial viewpoints, target the areas that are not covered and have low accuracy, and resample at a finer resolution until enough viewpoint samples are generated to ensure the coverage of the model with the required accuracy.

5. The UAV full-coverage path planning method based on adaptive viewpoint sampling according to claim 1, wherein When evaluating the path coverage rate, evaluate the integrity of the coverage path planning by quantifying the percentage of the covered volume of the structure and the predicted 3D structure volume on the generated path, and the specific description is as follows: (1.6) wherein, represents the coverage rate of the path; represents the surface volume reconstructed by collecting data along the full-coverage path generated by the algorithm; represents the surface volume of the model of the three-dimensional structure to be inspected.

6. Multi-rotor UAV full-coverage path planning system based on adaptive viewpoint sampling, characterized in that, It includes an adaptive viewpoint generation module, a full-coverage path generation module, and a path coverage rate evaluation module; The adaptive viewpoint generation module is used to discretize the position and orientation of the space to generate initial viewpoints; Then, use adaptive sampling comparison to identify the areas not covered by the initial viewpoints; finally, iterate within the unrecognized areas until a qualified set of viewpoints is generated; The full coverage path generation module is used to cluster the final viewpoint set generated by the adaptive viewpoint generation module and connect each cluster internally; then connect each cluster with its surrounding neighbor clusters to generate a search graph; finally, use the concepts of entropy and neighbor distance as heuristic functions to generate a full coverage path; when generating the full coverage path, use the heuristic function to generate the path, and the heuristic function used : (1.5) Wherein: is the heuristic function used for full-coverage path planning; represents the total entropy of the path where the current waypoint is located; is the distance contribution factor; is the distance between the current waypoint and its parent waypoint; During the process of full-coverage path planning, select waypoints with larger values to minimize , and maximize entropy . Minimizing means a shorter travel distance, and maximizing entropy is to maximize the information gain; The path coverage rate evaluation module is used to accumulatively calculate the volume of the to-be-detected structure that has been covered along the full-coverage path to calculate the path coverage percentage. During this process, occlusion culling of the reference model is required.

7. A computer device, characterized in that, It includes a processor and a memory. The memory is used to store computer-executable programs. The processor reads and executes the computer-executable programs from the memory. When the processor executes the computational executable programs, it can implement the method for full-coverage path planning of an unmanned aerial vehicle based on adaptive viewpoint sampling described in any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that, A computer program is stored in a computer-readable storage medium. When the computer program is executed by a processor, it can implement the method for full-coverage path planning of an unmanned aerial vehicle based on adaptive viewpoint sampling described in any one of claims 1 to 5.

Images