Unmanned aerial vehicle flight path optimization algorithm for land consolidation monitoring
By introducing a genetic algorithm of greedy mechanism in UAV route planning, the problems of incomplete coverage and low efficiency of traditional route planning in hybrid morphological pattern monitoring are solved, efficient flight route planning is achieved, and monitoring efficiency is significantly improved and cover quality.
Patent Information
- Application Number
- CN202210385512.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-04-13
- Publication Date
- 2024-11-26
- Estimated Expiration
- 2042-04-13
AI Technical Summary
When facing mixed-form natural resource survey maps, traditional drone route planning often has problems such as missed inspection, redundancy, incomplete coverage, and unreasonable track settings, making it difficult to efficiently monitor land reclamation with scattered distributions and different shapes. Picture spot.
Design a genetic algorithm that takes into account greed mechanisms, and extract flight points through drone hardware parameters, optimize flight route planning, realize optimal flight point coding, and improve monitoring efficiency and coverage quality.
It significantly improves the efficiency of drone monitoring, reduces the collection time and storage space requirements, ensures full coverage of map spots and reasonable routes, and saves the number of images and storage space.
Smart Images

Figure CN114740890B_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of unmanned aerial vehicle (UAV) route planning, and in particular relates to an unmanned aerial vehicle (UAV) flight path optimization algorithm for land consolidation pattern monitoring. Background Art
[0002] As a new type of instrument that integrates high-altitude photography, remote control, telemetry technology, video image microwave transmission and computer information processing, drones can quickly obtain high-resolution images and are widely used in many fields such as aerial photography, agriculture, plant protection, emergency rescue, and surveying and mapping. Applying drones to natural resource monitoring can effectively make up for the shortcomings of insufficient temporal and spatial resolution of satellite monitoring, limited coverage of tower-based monitoring, and high reliance on manual inspections and monitoring on the ground, and can achieve "fast, clear, and accurate" vision, thus improving the efficiency of natural resource monitoring.
[0003] Traditional route planning is an "S"-shaped route. Its planning process is to clarify the scope of aerial survey - confirm the heading and lateral overlap - confirm the altitude - plan the route. It is suitable for situations that need to be observed within the scope of aerial survey. However, it is oriented to mixed forms of natural resource survey patches, such as: narrow and long industrial and mining waste land reclamation project patches along the road, scattered village settlement reclamation land reclamation project patches, scattered suspected illegal construction land patches, large-scale industrial land transfer construction project patches, etc., and often there are missed inspections, redundancy, incomplete coverage, and unreasonable track settings.
[0004] To this end, domestic and foreign scholars have carried out a lot of research on the UAV trajectory planning problem. For example, they use the K-means algorithm and simulated annealing algorithm to perform trajectory planning for multi-task and multi-UAV under multiple conditions, increase the coverage of sub-target areas in the effective cruise area, and use hierarchical directed dynamic programming to solve the applicability and algorithm complexity of dynamic programming in special terrain environments, improve computational efficiency, and design a greedy MB-RRT* algorithm to improve the speed of UAV trajectory planning and save flight time by sacrificing a certain amount of trajectory quality. Summary of the invention
[0005] Purpose of the invention: The present invention aims to simulate the distribution of images and extract flight points based on the hardware parameters of the UAV, design a genetic algorithm that takes into account the greedy mechanism, efficiently solve the optimal sorting of the flight point codes, and realize the UAV flight route planning, so as to optimize the solution to the problem of image acquisition of land consolidation spots with scattered distribution and various shapes, improve monitoring efficiency, reduce acquisition time, and save storage space.
[0006] Technical solution: The UAV flight path optimization algorithm for land consolidation pattern monitoring of the present invention includes the following steps:
[0007] Step 1: Extract flight points: Input the monitored patch, population size, genetic probability ga_choose_ratio, mutation probability ga_mutate_ratio, and iteration number t. Design a flight point extraction method for mixed-form land consolidation patches. Select the skeleton points with the least number, uniform distribution, and reasonableness within the patch monitoring area as the flight points for the UAV. The number of flight points is recorded as n.
[0008] Step 2 Population initialization: Calculate the distance between each flight point according to the flight point coordinates to form a distance matrix D. For any flight point SP i (i=1,2,…,n), with SP i As the current route L i Starting point, using the greedy mechanism, all the routes L1, L2, ..., L size The set of , the initial population of the genetic algorithm.
[0009] Step 3: Calculate the population fitness: For any individual L in the population i (i=1,2,…,n), L i The reciprocal of the flight route distance is taken as the evaluation value of its fitness, denoted as A i . A i The larger the value, the higher the fitness, and vice versa.
[0010] Step 4: Get the parent population: The individuals in the original population are divided according to the fitness evaluation value A. i Arrange them in order from large to small, and select the first p_size individuals as the parent population of the next generation genetic algorithm.
[0011] Step 5: Genetic operation: Based on the parent population obtained in step 5, after size times of random selection, crossover, mutation, and inheritance, size offspring individuals with excellent genes are generated and enter the next generation population.
[0012] Step 6 Iterate and output the result: Repeat the operations from step 3 to step 5 until the preset number of genetic iterations t is reached. After calculating and comparing the fitness evaluation value, the optimal flight point sequence, i.e., the optimal flight route of the UAV, is obtained and output.
[0013] Furthermore, in step 1, the method for extracting flight points from mixed-form natural resource monitoring maps is adopted, and the specific steps are as follows:
[0014] Step 1.1: Calculate the ground size b corresponding to a picture taken by the drone based on the drone chip width a, focal length f and flight altitude h, see formula (1).
[0015]
[0016] Step 1.2: According to parameter b, simulate the pictures taken by the drone during flight in the natural resource monitoring area and organize the pictures into a surface layer.
[0017] Step 1.3: Use the natural resource monitoring map to perform spatial screening on the surface layer obtained in step 1.2, and select surfaces that can be fully covered and have a smaller shooting area.
[0018] Step 1.4: Extract the centroid of the surface obtained in step 1.3, which is the flight point.
[0019] Furthermore, in step 2, the greedy mechanism process is as follows: first establish two sets S and VS, add the starting point and the end point to the S set, store the flight points to be traversed in the VS, select the point closest to the starting point from the VS and add it to the set S, and remove the point from the VS, and continue to loop until all flight points are traversed to obtain the final set S.
[0020] Furthermore, in step 3, for any population individual Li, the essence of Li is the permutation set S of n flight points, and the evaluation value Fi of its fitness is calculated, see formula (2).
[0021]
[0022] Where d is the route L i The distance between adjacent flight points.
[0023] Furthermore, in step 4, the number of individuals p_zize of the parent population is calculated, see formula (3).
[0024] p_size = n × ga_choose_rate Formula (3)
[0025] Furthermore, in step 5, based on the parent population obtained in step 4, size numbers of offspring individuals with good genes are obtained through size random selection, crossover, and mutation processes, and enter the next generation population. The specific operation steps are as follows:
[0026] Step 5.1 Selection: Calculate the probability P of each individual in the original population being selected i , using the roulette wheel selection method, randomly select two individuals, denoted as gene_x and gene_y. Among them, the probability of any individual being selected is P i The calculation of can be found in formula (4).
[0027]
[0028] Step 5.2 Crossover: Based on individuals gene_x and gene_y, use the two-point crossover method to exchange part of the chromosomes (flight point arrangement fragments) between the two, and exchange some of the conflicting genes to ensure that there are no conflicting genes, thereby generating candidate offspring population individuals, which are recorded as gene_x_new and gene_y_new respectively.
[0029] Step 5.3 mutation: Based on the preset mutation probability ga_mutate_ratio, select gene_x_new and gene_y_new for mutation at selected gene positions (flight points), that is, reverse some fragments of the mutant chromosome.
[0030] Step 5.4 generates offspring: re-evaluate the fitness of gene_x_new and gene_y_new, and after comparison, add individuals with high fitness evaluation values to the next generation population.
[0031] Step 5.5 Repeat steps 5.1 to 5.4 until the number of individuals in the next generation population reaches the preset size.
[0032] Beneficial effects: Compared with the prior art, the present invention has the following advantages:
[0033] (1) The present invention introduces a greedy mechanism into the traditional genetic algorithm, which provides a good initial population for the route planning process. Compared with the traditional genetic algorithm, it not only reduces the route distance but also ensures full monitoring coverage and improves the search efficiency.
[0034] (2) The present invention is aimed at the problem of collecting images of land consolidation spots with scattered distribution and various shapes. Based on the hardware parameters of the UAV, the image distribution is simulated and the flight points are extracted. A genetic algorithm with a greedy mechanism is designed to efficiently solve the optimal sorting of the flight point codes and realize the UAV flight route planning. It can significantly improve the UAV monitoring efficiency, save the UAV collection time, the number of images, and the storage space. It has good application value in UAV monitoring scenarios in the fields of natural resources. BRIEF DESCRIPTION OF THE DRAWINGS
[0035] Figure 1 Flow chart of the present invention
[0036] Figure 2 Extracting schematic diagram for flight point
[0037] Figure 3 A cross diagram
[0038] Figure 4 The route planning results of the genetic algorithm and the traditional genetic algorithm taking into account the greedy mechanism
[0039] Figure 5 This is the algorithm convergence curve
[0040] Figure 6 Orthophoto map of Zhangqiao Village land consolidation area DETAILED DESCRIPTION
[0041] The technical solution of the present invention is described in detail below, but the protection scope of the present invention is not limited to the embodiments.
[0042] This embodiment selects the land consolidation map monitoring of Zhangqiao Village in the west of Zhangdian Town, Jiangyan City, Jiangsu Province as the experimental data, and uses the DJI UAV M300 RTK with the Zenmuse P1 camera to collect the orthophoto data of the map.
[0043] like Figure 1 As shown, the UAV flight route planning method for mixed-form natural resource monitoring images of this embodiment adopts a greedy mechanism to generate an initialization population, reduce the blind search of the subsequent genetic algorithm, and obtain the optimal solution of the UAV route through multiple iterative calculations of the genetic algorithm while improving efficiency. Specifically, it includes the following steps:
[0044] Step 1: Extract flight points: Figure 2 As shown in (a), input the Zhangqiao Village land consolidation map to be monitored, set the population size size = 335, genetic probability ga_choose_ratio = 0.2, mutation probability ga_mutate_ratio = 0.05, and iteration number t = 300. Design a method for extracting flight points for mixed-form land consolidation maps. Within the experimental area, select the skeleton points with the least number, uniform location distribution, and reasonableness as the flight points for UAV flight. The number of flight points is recorded as 35. The specific steps are as follows:
[0045] Step 1.1: Figure 2 As shown in (b), based on the Zenmuse P1 camera chip width d = 24mm, focal length f = 2.8mm, and flight altitude of 270m, the corresponding size of a single image is calculated to be b =, and the ground size corresponding to a picture taken by the drone is approximately 2314m.
[0046] Step 1.2: Figure 2 As shown in (c), the experimental area is simulated by taking pictures taken by the UAV during flight according to parameter b, and the pictures are combined into a surface layer.
[0047] Step 1.3: If Figure 2 As shown in (d), the natural resource monitoring patches are used to perform spatial screening on the surface layer obtained in step 1.2 to screen out surfaces that can be fully covered and have a smaller shooting area.
[0048] Step 1.4: Figure 2 As shown in (e), the centroid points of the surface obtained in step 1.3 are extracted, which are the flight points, totaling 35.
[0049] Step 2 Population initialization: Calculate the distance between each flight point according to the flight point coordinates to form a distance matrix D. For any flight point SP i (i=1,2,…,n), with SP i As the current route L i Starting point, using the greedy mechanism, all the routes L1, L2, ..., L 355 The set of , the initial population of the genetic algorithm.
[0050] In particular, the introduced greedy mechanism process is as follows: first establish two sets S and VS, add the starting point and the end point to the S set, store the flight points to be traversed in VS, select the point closest to the starting point from VS and add it to the set S, and then remove the point from VS, and continue to cycle until all flight points are traversed to obtain the final set S, which is the route L.
[0051] Step 3: Calculate the population fitness: Refer to formula (2). For any individual L in a population, i , calculate the evaluation value of its fitness, recorded as A i . A i The larger the value, the higher the fitness, and vice versa.
[0052] Step 4: Obtain the parent population: Refer to formula (3) to calculate the number of individuals in the parent population as 71. i Arrange them in order from largest to smallest, and select the first 71 individuals as the parent population of the next generation genetic algorithm.
[0053] Step 5: Genetic operation: Based on the parent population obtained in step 5, 355 offspring individuals with excellent genes are generated through 355 random selections, crossovers, and mutations, and enter the next generation population. The specific steps are as follows:
[0054] Step 5.1 Selection: Calculate the probability P of each individual in the parent population being selected according to formula (4) i , using the roulette wheel selection method, randomly select 2 individuals, denoted as gene_x and gene_y respectively.
[0055] Step 5.2 Crossover: Figure 3 As shown, based on individuals gene_x and gene_y, the two-point crossover method is used to exchange part of the chromosomes (flight point arrangement fragments) between the two, and gene exchange is performed on some of the conflicting genes to ensure that there are no conflicting genes, thereby generating candidate offspring population individuals, which are recorded as gene_x_new and gene_y_new respectively.
[0056] Step 5.3 mutation: Based on the preset mutation probability ga_mutate_ratio, select gene_x_new and gene_y_new for mutation at selected gene positions (flight points), that is, reverse some fragments of the mutant chromosome.
[0057] Step 5.4 generates offspring: re-evaluate the fitness of gene_x_new and gene_y_new, and after comparison, add individuals with high fitness evaluation values to the next generation population.
[0058] Step 5.5: Repeat steps 5.1 to 5.4 until the number of individuals in the next generation population reaches the preset size of 355.
[0059] Step 6 Iterate and output the result: Repeat the operations from step 3 to step 5. After 300 iterations of genetics, for the final population, calculate and compare the fitness evaluation value, obtain and output the flight point sequence with the largest fitness evaluation value as the optimal flight route for the UAV. Figure 4 (a) with Figure 4 (b) are the route planning results obtained by the genetic algorithm taking into account the greedy mechanism of the present invention and the traditional genetic algorithm after 300 iterations. It is obvious that the route planned by the algorithm proposed in this paper is more orderly and reasonable than the traditional genetic algorithm path. Figure 5 The convergence curves of the two algorithms show that when the number of iterations approaches 280, the objective function of the traditional genetic algorithm converges to a certain value, which is the planned route distance of 18657.988. The genetic algorithm taking into account the greedy mechanism begins to converge to a fixed value of 11264.053 when the number of iterations is 30. Compared with the traditional genetic algorithm, the algorithm of the present invention not only reduces the route distance but also ensures full monitoring coverage and improves the search efficiency.
[0060] In order to more intuitively demonstrate the advantages of the present invention, the traditional route planning and the drone flight route planning for mixed-form land consolidation spots of the present invention are used to collect orthophoto data of the Zhangqiao Village land consolidation spots. As shown in Table 1, under the same drone flight parameters, the route planning of the present invention takes 12 minutes and collects 126 pictures to achieve full coverage monitoring of the spots, and the image storage space is 2.68G, while the traditional route planning takes 30 minutes and collects 400 pictures to achieve full coverage monitoring of the spots, and the image storage space is 4.55G. Obviously, the route planning of the present invention greatly improves the efficiency of drones in monitoring mixed-form spots, saving nearly 60% of the collection time, reducing the number of pictures by 68.5%, and saving nearly 41% of the image storage space.
[0061] Table 1 Experimental results of orthophoto acquisition of Zhangqiao Village land consolidation map
[0062]
[0063] like Figure 6 (a1) is the orthophoto of the land reclamation map of Zhangqiao Village collected by the route planning of the present invention, such as Figure 6 (b1) is the orthophoto of the land consolidation map of Zhangqiao Village collected for the route planning of the present invention. Figure 6 (a2) and Figure 6 (b2), Figure 6 (a3) and Figure 6 (b3) are images of the same image patch collected under two flight routes. It can be seen that the quality of the orthophoto images collected using the route planning path of the present invention is no different from that of the traditional route, which meets the business needs of natural resource land consolidation.
[0064] In summary, the present invention introduces a greedy mechanism into the traditional genetic algorithm, which provides a good initial population for the route planning process. Compared with the traditional genetic algorithm, it not only reduces the route distance but also ensures full monitoring coverage and improves the search efficiency.) The present invention is aimed at the problem of land consolidation image acquisition with scattered distribution and various shapes. Based on the hardware parameters of the drone, the distribution of images is simulated and the flight points are extracted. A genetic algorithm that takes into account the greedy mechanism is designed to efficiently solve the optimal sorting of the flight point coding, realize the drone flight route planning, and significantly improve the drone monitoring efficiency, save nearly 60% of the acquisition time, reduce the number of images by 68.5%, and save nearly 41% of the image storage space. It has good application value in drone monitoring scenarios in natural resources and other fields.
Claims
1. An optimization algorithm for UAV flight paths for land consolidation monitoring, characterized by: The algorithm targets scattered and morphological patches and uses a greedy algorithm to generate an initial population to reduce the blind search of the subsequent genetic algorithm. Through multiple iterative calculations of the genetic algorithm, the optimal solution for the UAV route is obtained while improving efficiency. The algorithm includes the following steps: Step 1: Extraction of flight points: Input the image patch to be monitored, population size size, genetic probability ga_choose_ratio, mutation probability ga_mutate_ratio, and iteration number t. By designing a flight point extraction method for mixed-form land consolidation patches, select the skeleton points with the least number, uniform position distribution, and reasonableness within the patch monitoring area as the flight points for the UAV flight. The number of flight points is recorded as n. Flight point calculation process: Calculate the ground size b corresponding to a picture taken by the UAV according to the width a, focal length f, and flight height h of the UAV chip, see formula (1). According to the parameter b, simulate the pictures taken by the UAV during flight in the land consolidation patch area, and form a surface layer with the pictures. Use the land consolidation patch to perform spatial screening on the obtained surface layer, and select the surface that can be completely covered and has a small shooting area. Extract the centroid point of the obtained surface, which is the flight point. Step 2 Population initialization: Calculate the distance between each flight point according to the flight point coordinates to form a distance matrix D. For any flight point SP i , i = 1, 2, ..., n, with SP i As the current route L i Starting point, using the greedy algorithm, all the routes L1, L2, ..., L size The set of is used as the initial population of the genetic algorithm; Step 3: Calculate the population fitness: For any individual L in the population i ,i=1,2,…,sizeL i The reciprocal of the flight route distance is taken as the evaluation value of its fitness, denoted as A i , A i The larger the value, the higher the fitness, and vice versa; Step 4: Get the parent population: The individuals in the original population are divided according to the fitness evaluation value A. i Arrange them in order from large to small, select the first p_size individuals, and use them as the parent population of the next generation genetic algorithm; Step 5 Genetic operation: Based on the parent population obtained in step 4, after size times of random selection, crossover, and mutation, size offspring individuals with excellent genes are generated and enter the next generation population; Step 6 Iterate and output the result: Repeat the operations from step 3 to step 5 until the preset number of genetic iterations t is reached. After calculating and comparing the fitness evaluation values, the best flight point sequence, i.e. the best flight route of the UAV, is obtained and output.
2. The UAV flight path optimization algorithm for land consolidation pattern monitoring according to claim 1 is characterized by: In step 2, the greedy algorithm introduced has the following process: first establish two sets S and VS, add the starting point and the end point to the S set, store the flight points to be traversed in the VS, select the point closest to the starting point from the VS and add it to the set S, and remove the point from the VS, and continue to loop until all flight points are traversed to obtain the final set S.
3. The UAV flight path optimization algorithm for land consolidation pattern monitoring according to claim 1 is characterized by: In step 3, for any population individual L i , L i The essence of is the permutation set S of n flight points, and the fitness evaluation value F is calculated i , see formula (2): Where d is the route L i The distance between adjacent flight points.
4. The UAV flight path optimization algorithm for land consolidation pattern monitoring according to claim 1 is characterized by: In step 4, the number of individuals p_size of the parent population is calculated, see formula (3): p_size=n×ga_choose_rattio formula (3).
5. The UAV flight path optimization algorithm for land consolidation pattern monitoring according to claim 1 is characterized by: In step 5, based on the parent population obtained in step 4, size numbers of offspring individuals with good genes are obtained through size random selection, crossover, and mutation processes, and enter the next generation population. The specific operation steps are as follows: Step 5.1 Selection: Calculate the probability P of each individual in the original population being selected i , using the roulette wheel selection method, randomly select two individuals, denoted as gene_x and gene_y, where the probability of any individual being selected is P i The calculation of can be seen in formula (4): Step 5.2 Crossover: Based on individuals gene_x and gene_y, use the two-point crossover method to exchange part of the chromosomes between them, that is, the flight point arrangement fragments, and exchange some of the conflicting genes to ensure that there are no conflicting genes, so as to generate candidate offspring population individuals, which are recorded as gene_x_new and gene_y_new respectively; Step 5.3 Mutation: Based on the preset mutation probability ga_mutate_ratio, select gene positions for gene_x_new and gene_y_new to mutate, that is, reverse some segments of the mutant chromosome; Step 5.4 Generate offspring: Re-evaluate the fitness of gene_x_new and gene_y_new, and after comparison, add individuals with high fitness evaluation values to the next generation population; Step 5.5 Repeat steps 5.1 to 5.4 until the number of individuals in the next generation population reaches the preset size.
