TWI826873B - Pesticide spraying planning method and pesticide spraying system - Google Patents

Abstract

The present disclosure provides a pesticide spraying planning method including a target specie obtaining step, a random path generating step, a new path generating step, an old path and new path selecting step and a path deciding step. In the target specie obtaining step, a target specie dataset of a target specie in a target area is obtained, and the target specie dataset includes a plurality of points. In the random path generating step, a plurality of old paths are generated based on the points. In the new path generating step, a genetic algorithm is used to generate new paths. In the old path and new path selecting step, a simulate anneal arithmetic is used to select the old path or the corresponding new path. In the path deciding step, the path of the final updated path group, whose distance is shortest, is selected. Therefore, the suitable pesticide spraying path can be quickly found.

Description

Pesticide spraying planning method and pesticide spraying system

The present invention relates to a spray planning method and a spray system, and in particular, to a pesticide spray planning method and a pesticide spray system.

Taiwan's agricultural workforce is aging and the labor force is insufficient, and the sloping terrain requires a large number of laborers to spray pesticides. Therefore, some industry players have invented the use of drones to spray pesticides to save labor. Such drones can also be called plant protection drones.

Since the plant protection drone is not controlled by anyone, it needs to fly according to its set path. One known flight method is to calculate the area of the entire fruit tree and obtain the coordinates of the left edge of the fruit tree, so that the plant protection drone moves from the left edge to the direction in a square wave. Spray right until the entire spray is complete. However, this method does not take into account different tree species, and the flight path does not take into account the effectiveness of the path, and there is still room for improvement.

Therefore, when a target area contains multiple tree species, how to quickly set a better and more efficient pesticide spraying path based on the distribution size and location of the target tree species has always been a problem to be solved in related fields.

In order to solve the above problems, the present invention provides a pesticide spraying planning method and a pesticide spraying system, which quickly plans a more efficient pesticide spraying path through genetic algorithm and simulated annealing algorithm.

According to an embodiment of the present invention, a pesticide spraying planning method is provided, which includes a target tree species information acquisition step, a random path generation step, a new path generation step, a new and old path selection step, and a path determination step. In the target tree species information obtaining step, a target tree species information related to a target tree species in a target area is obtained, and the target tree species information includes plural points. In the random path generation step, the plurality of points are randomly sorted to generate a plurality of old paths, and an initial path group is obtained. In the new path generating step, a genetic algorithm is used to change the order of at least two points in each old path to generate a new path. In the old and new path selection step, a simulated annealing algorithm is used to select at least one old path and its corresponding new path to decide to retain the at least one old path or its corresponding new path, and obtain an updated path. group. In the path determination step, the new path generation step and the old and new path selection steps are repeated a preset number of times, and the shortest distance in the final updated path group is defined as a pesticide spraying path.

In this way, the genetic algorithm can be used to continuously find new and better paths, and the probability of the simulated annealing algorithm can be used to decide whether to replace the old path with the current better new path, so as to find out more quickly The most suitable pesticide spraying path.

The pesticide spraying planning method according to the aforementioned embodiment may further include a spraying time planning step and a pesticide spraying step. In the spraying time planning step, a residence time of a drone corresponding to each point is determined based on the density of multiple tree species of the target tree species information. In the pesticide spraying step, the drone is caused to spray the pesticide according to the order of the plurality of points in the pesticide spraying path and the residence time of each point.

According to the pesticide spraying planning method of the above embodiment, in the step of obtaining target tree species information, a drone equipped with a camera can be used to capture at least one image of the target area, and a processor can be used to identify and analyze the at least one image. , to obtain target tree species information.

According to the pesticide spraying planning method of the aforementioned embodiment, in the target tree species information acquisition step, an image analysis module of the processor can be used to identify the target tree species and perform color block processing to distinguish the target tree species, and obtain the plurality of longitudes of the target tree species distribution. and plural latitudes, and use a terrain component module of the processor to model the target area to obtain plural heights of the distribution of the target tree species, thereby obtaining the aforementioned plural points of the target tree species information.

According to the pesticide spraying planning method of the aforementioned embodiment, each longitude and each latitude can be converted into TWD97 coordinate values.

According to another embodiment of the present invention, a pesticide spraying system is provided, which includes a drone and a processor. The drone includes a camera, and the camera is used to photograph a target area to obtain at least one image. The processor signal is connected to the camera to receive the aforementioned at least one image and includes an image analysis module and a path planning module. The image analysis module identifies and analyzes the aforementioned at least one image to obtain a target tree species information related to a target tree species in the target area. The target tree species information includes a plurality of points. The path planning module receives the target tree species information to plan a pesticide spraying path. The path planning module randomly sorts the plurality of points to generate a plurality of old paths and obtains an initial path group, and then uses a genetic algorithm to Change the order of at least two points in each old path to generate a new path, and then use a simulated annealing algorithm to select at least one old path and its corresponding new path to decide whether to retain the aforementioned at least one old path or to retain its corresponding new path. The corresponding new path is obtained and the updated path group is obtained. The path group is continuously updated using the genetic algorithm and simulated annealing algorithm. The shortest distance in the final updated path group is defined as the pesticide spraying path. Among them, the drone sprays pesticides in the order of the plurality of points in the pesticide spraying path.

According to the pesticide spraying system of the aforementioned embodiment, the processor may further include a spraying time planning module that receives the target tree species information, and based on the plurality of tree species densities of the target tree species information, a stop of the drone corresponding to each point is given time.

According to the pesticide spraying system of the aforementioned embodiment, the processor may further include a terrain component module for modeling the target area to obtain plural heights of target tree species distribution, and an image analysis module for identifying the target tree species. And perform color block processing to distinguish the target tree species, obtain the plural longitudes and plural latitudes of the target tree species distribution, thereby obtaining the aforementioned plural points of the target tree species information.

Embodiments of the present invention will be described below with reference to the drawings. For the sake of clarity, many practical details will be explained together in the following narrative. The reader should understand, however, that these practical details should not be construed as limiting the invention. That is to say, in some embodiments of the present invention, these practical details are not necessary. In addition, in order to simplify the drawings, some commonly used structures and components will be illustrated in a simple schematic manner in the drawings; and repeated components may be represented by the same numbers or similar numbers.

In addition, the terms first, second, third, etc. in this article are only used to describe different components or components, and there is no limitation on the components/components themselves. Therefore, the first component/component can also be renamed as the second component/component. . Moreover, the combination of components/components/mechanisms/modules in this article is not a combination that is generally known, conventional or customary in this field. Whether the components/components/mechanisms/modules themselves are common knowledge cannot be used to determine whether their combination relationship is Easily accomplished by a person of ordinary skill in the technical field.

Please refer to Figures 1 and 2. Figure 1 illustrates a block diagram of a pesticide spraying system 100 according to an embodiment of the present invention, and Figure 2 illustrates the planning of the pesticide spraying path S1 of the embodiment in Figure 1. Schematic diagram. The pesticide spraying system 100 includes a drone 110 and a processor 120 . The drone 110 includes a camera 111. The camera 111 is used to photograph a target area to obtain at least one image. The processor 120 is connected to the camera 111 via signals to receive the at least one image and includes an image analysis module 121 and a path planning module 122 .

The image analysis module 121 identifies and analyzes the aforementioned at least one image to obtain a target tree species information related to a target tree species T1 in the target area. The target tree species information includes a plurality of points. The path planning module 122 receives the target tree species information to plan the pesticide spraying path S1. The path planning module 122 randomly sorts the plurality of points to generate a plurality of old paths, and obtains an initial path group, which is then inherited with a gene. The genetic algorithm changes the order of at least two points in each old path to generate a new path, and then uses a simulated anneal arithmetic algorithm to select at least one old path and its corresponding new path. To decide to retain at least one of the aforementioned old paths or to retain its corresponding new path, and obtain the updated path group, continue to update the path group using genetic algorithm and simulated annealing algorithm, and obtain the distance in the final updated path group. The shortest one is defined as the pesticide spraying path S1. Among them, the drone 110 sprays pesticides in the order of the plurality of points in the pesticide spraying path S1.

In this way, the genetic algorithm can be used to continuously find new and better paths, and the probability of the simulated annealing algorithm can be used to decide whether to replace the old path with the current better new path, so as to find out more quickly The most suitable pesticide spraying path S1. Details of the pesticide spray system 100 will be discussed later.

In this embodiment, the processor 120 is an independent back-end server and can be connected to the drone 110 in a wireless manner. The processor 120 receives the image of the camera 111 and generates a pesticide spraying path S1, and then converts the pesticide spraying path S1 After being transmitted to the controller of the drone 110, the drone 110 can spray pesticides according to the pesticide spraying path S1. In other embodiments, the processor may be a cloud server, or it may be installed on the drone and simultaneously serve as the control system of the drone, but is not limited thereto.

The image analysis module 121 can, for example, use deep learning neural network technology to identify the target tree species T1 and perform color block processing to distinguish the target tree species T1, and can obtain the plural longitudes and plural latitudes of the distribution of the target tree species T1. Specifically, the number of images can be plural, which allows the camera 111 to capture multiple images at different locations in the target area to form a global map. After identifying the target tree species T1, the image analysis module 121 can further use coloring technology to map the global map. The target tree species T1 is marked on the map, and then the global map is cut, for example, into a grid form, and the longitude and latitude of the target tree species T1 are recorded according to the distribution status, for example, the longitude and latitude of the center point of each grid are used to represent each The longitude and latitude of the point.

In addition, the processor 120 may further include a terrain component module 124, which is used to model the target area to obtain the complex heights of the distribution of the target tree species T1. The terrain component module 124 may further include a UAV post-processing software such as Pix4Dmapper, which can create a three-dimensional site map of the target area based on the images captured by the camera 111, and can read GPS records from the images, and finally export them. The data file includes longitude, latitude and corresponding height, from which the longitude, latitude and height of the T1 distribution of the corresponding target tree species can be obtained.

In addition, the processor 120 may further include a spraying time planning module 123, which receives the target tree species information and gives the UAV 110 a residence time corresponding to each point according to the plurality of tree species densities of the target tree species information. Specifically, by using coloring technology to mark the target tree species T1 and divide the grid, the density of the color blocks of the target tree species T1 in the grid can be confirmed as the tree species density, and this is used as the basis for the residence time of the UAV 110. The tree species If the density is high and the residence time is long, the amount of pesticides sprayed will be large; conversely, if the density of the tree species is small and the residence time is short, the amount of pesticides sprayed will be small, which is more conducive to spraying the right amount of pesticides.

Through the image analysis module 121, the spraying time planning module 123 and the terrain component module 124, the target tree species information as shown in Table 1 can be obtained, wherein the target tree species information can include the longitude, latitude, height and tree species density of 12 points. . Table 1. Target tree species information point longitude Latitude high tree species density 1 23.00507 120.471074 150.11 2 2 23.00496 120.471012 142.06 5 3 23.00495 120.470878 143.66 3 4 23.00485 120.470958 136.62 4 5 23.00477 120.470943 137.06 2 6 23.00466 120.470931 141.46 2 7 23.00475 120.470633 135.02 1 8 23.00488 120.470609 132.18 3 9 23.00486 120.470504 133.68 4 10 23.00498 120.470574 130.82 5 11 23.00505 120.470554 132.03 2 12 23.00508 120.470505 130.11 3

It should be noted here that the longitude and latitude of each point can be converted into TWD97 coordinate values, as shown in Table 2, where TWD97 X corresponds to longitude, TWD97 Y corresponds to latitude, because the unit is meters, and More suitable for calculating distances. Table 2. TWD97 coordinate values of target tree species information point TWD97X TWD97Y 1 195780.0762 2544942.793 2 195773.7298 2544931.199 3 195760.0065 2544929.244 4 195768.1893 2544918.496 5 195766.5897 2544909.875 6 195765.2618 2544897.001 7 195734.783 2544907.886 8 195732.4267 2544922.512 9 195721.6013 2544919.671 10 195728.7861 2544933.477 11 195726.7616 2544940.615 12 195721.7718 2544944.166

As shown in Tables 1 and 2, it contains a total of 12 points. If each arrangement order is taken into account, there will be a total of 479,001,600 paths. Although it can be directly calculated from these paths which distance is the shortest, it requires a lot of time. A lot of time for calculations and comparisons. Therefore, in the path planning module 122, for example, 100 old paths can be randomly generated based on the 12 points in Table 2 as the initial path group as shown in Table 3, and then continuously evolved to accelerate convergence. Table 3. Initial path group No. Point order old path 1 9à5à6à3à11à1à12à2à4à8à10à7 old path 2 11à8à7à3à9à12à6à2à4à4à5à10à1 old path 3 4à12à3à6à11à8à7à2à9à1à10à5 old path 4 3à7à4à9à12à5à1à2à6à8à10à11 old path 5 5à11à1à3à5à8à7à2à4à6à10à12 old path 6 7à2à5à3à11à1à9à2à8à12à10à6 … … old path 100 10à3à6à5à8à1à7à12à4à11à9à2

Afterwards, the path planning module 122 can use the genetic algorithm to change the new path. Among them, the gene genetic algorithm takes the middle section of the gene (the plurality of consecutive points in the middle) for replacement. For example, the order of points 1, 3, 6, 11, and 12 in the old path 1 is changed from 6à3à11à1à12 to 12à1à11à3à6 and so on. As a new path 1; or, take out several genes (points) and randomly insert them into the sequence, for example, replace the positions of points 3, 6, and 10 in the old path 2 to become a new path 2, etc. . No matter which method is used, the order of at least two points in the old path is changed to generate a new path, as shown in Table 4. Table 4. New path and old path No. Point order Distance(m) old path 1 9à5à6à3à11à1à12à2à4à8à10à7 235.6 new path 1 9à5à12à1à11à3à6à2à4à8à10à7 212.3 old path 2 11à8à7à3à9à12à6à2à4à4à5à10à1 213.7 new path 2 11à8à7à10à9à12à3à2à4à5à6à1 222.6 old path 3 4à12à3à6à11à8à7à2à9à1à10à5 225.7 new path 3 6à3à12à4à11à8à7à2à9à1à10à5 237.1 old path 4 3à7à4à9à12à5à1à2à6à8à10à11 198.5 new path 4 3à7à4à9à5à12à1à2à8à6à10à11 196.7 old path 5 5à11à1à3à5à8à7à2à4à6à10à12 203.7 new path 5 11à5à1à3à5à7à8à2à4à10à6à12 222.4 old path 6 7à2à5à3à11à1à9à2à8à12à10à6 233.8 new path 6 7à11à3à5à2à1à9à2à8à12à10à6 205.4 … … … old path 100 10à3à6à5à8à1à7à12à4à11à9à2 215.7 new path 100 10à2à6à5à7à1à3à12à4à11à9à8 210.8

It can be found from Table 4 that the distances of the new paths 2, 3, and 5 are longer than the distances of the old paths 2, 3, and 5, and the old paths 2, 3, and 5 are directly retained without making a selection; the new paths 1, 4, 6, and 100 are The distance is shorter than the old paths 1, 4, 6, and 100. However, in order to prevent the shortest new path from being selected every time to replace the original old path and falling into the local optimal solution, the path planning module 122 can further use simulation The probabilistic method of the annealing algorithm is used to decide whether to replace the current old path as the next path reference gene. The simulated annealing algorithm will increase the probability of selecting excellent genes each time it evolves the next generation, thereby quickly converging on a shortest path, but in the initial evolution process, it will try to retain the highest probability The original old path allows more changes in the genes to find the best way to change the path.

Specifically, the new path 1 is better than the old path 1, and through the probability of the simulated annealing algorithm, it is also decided to replace the old path 1 and use the new path 1 as the next generation gene. Although the new path 4 is better than the old path 4, it is determined by chance that the old path 4 should be retained as the next generation gene. After the probability determination of the simulated annealing algorithm, the new path 6 is selected, while the old path 100 is retained, thus generating a second generation path group for evolution, as shown in Table 5. Table 5. Second generation path group No. Point order Distance(m) new path 1 9à5à12à1à11à3à6à2à4à8à10à7 212.3 old path 2 11à8à7à3à9à12à6à2à4à4à5à10à1 213.7 old path 3 4à12à3à6à11à8à7à2à9à1à10à5 225.7 old path 4 3à7à4à9à12à5à1à2à6à8à10à11 198.5 old path 5 5à11à1à3à5à8à7à2à4à6à10à12 203.7 new path 6 7à11à3à5à2à1à9à2à8à12à10à6 205.4 … … … old path 100 10à3à6à5à8à1à7à12à4à11à9à2 215.7

The path group of the second generation can all be regarded as old paths, and new paths will be generated through repeated evolution. For example, the evolution can be set to 100 generations, and the final updated path group can be generated, as shown in Table 6. Please note that In Table 6, the old path or the new path is no longer marked, but only represented by the path. Table 6. Final updated path group No. Point order Distance(m) Path 1 10à2à6à5à7à1à3à12à4à11à9à8 195.3 Path 2 7à2à5à3à11à1à9à2à8à12à10à6 178.2 Path 3 4à12à3à6à11à8à7à2à9à1à10à5 133.4 Path 4 6à3à12à4à11à8à9à10à7à2à1à5 195.2 Path 5 5à11à1à3à5à8à7à2à10à4à6à12 169.1 Path 6 9à5à6à3à11à1à12à2à4à10à8à7 178.2 … … … path 100 10à3à6à5à8à1à7à12à4à11à9à2 177.2

As shown in Table 6, path 3 has the shortest distance, so path 3 can be defined as pesticide spraying path S1. In this way, the most suitable pesticide spraying path S1 can be quickly found, and the UAV 110 can spray pesticides according to the sequence of points 4à12à3à6à11à8à7à2à9à1à10à5 and their residence time.

Please refer to Figure 3, and refer to Figures 1 and 2 together. Figure 3 illustrates a block flow chart of a pesticide spraying planning method 200 according to another embodiment of the present invention. The pesticide spraying planning method 200 includes a target tree species information acquisition step 210, a random path generation step 220, a new path generation step 230, a new and old path selection step 240 and a path decision step 250. The details of the pesticide spraying planning method 200 will be described below with reference to the pesticide spraying system 100 in Figures 1 and 2 .

In the target tree species information obtaining step 210, a target tree species information related to a target tree species T1 in a target area is obtained, and the target tree species information includes plural points.

In the random path generation step 220, the plurality of points are randomly sorted to generate a plurality of old paths, and an initial path group is obtained.

In the new path generating step 230, a genetic algorithm is used to change the order of at least two points in each old path to generate a new path.

In the old and new path selection step 240, a simulated annealing algorithm is used to select each old path and its corresponding new path to decide whether to retain each old path or its corresponding new path, and obtain an updated path group.

In the path determination step 250, the new path generation step 230 and the old and new path selection step 240 are repeated for a preset number of times, and the shortest distance among the finally updated path groups is defined as a pesticide spraying path S1.

Specifically, in the target tree species information obtaining step 210, a drone 110 is equipped with a camera 111 to capture at least one image of the target area, and a processor 120 is used to identify and analyze the at least one image to obtain the target. Tree species information. Further, in the target tree species information acquisition step 210, an image analysis module 121 of the processor 120 is used to identify the target tree species T1 and perform color block processing to distinguish the target tree species T1, and obtain the plural longitudes and plural latitudes of the distribution of the target tree species T1. A terrain component module 124 of the processor 120 is then used to model the target area to obtain the plurality of heights of the distribution of the target tree species T1, thereby obtaining the aforementioned plurality of points of the target tree species information.

The image can be, for example, plural and constitute a global map, and the image analysis module 121 can, for example, use deep learning neural network technology to identify the target tree species T1, and use coloring technology to mark the target tree species T1 on the global map, and then use the global map to Cut, for example, into a grid form, and record its longitude and latitude according to the distribution of the target tree species T1. The terrain component module 124 can include a UAV post-processing software such as Pix4Dmapper, which can create a three-dimensional site map of the target area based on the images captured by the camera 111, and can export a data file to include longitude, latitude and their corresponding height. , based on which the longitude, latitude and height of the target tree species T1 distribution can be obtained, as shown in Table 1. Afterwards, the longitude and latitude can be converted into TWD97 coordinate values, as shown in Table 2.

In the random path generation step 220, the TWD97 coordinate values of each point can be used to randomly generate, for example, 100 old paths, and the initial path group in Table 3 can be obtained.

In the new path generation step 230, the path planning module 122 can use a genetic algorithm to generate a new path by selecting a plurality of consecutive points in the middle for replacement, or taking out several of the points and randomly inserting them again. into the sequence, so that the order of at least two points in the old path can be changed to generate a new path, as shown in Table 4.

In the old and new path selection step 240, the path planning module 122 can select through the simulated annealing algorithm to determine whether to replace the old path with a new path, and obtain the second generation path set as shown in Table 5. .

Through continuous repeated evolution, for example, 100 times, the final updated path group as shown in Table 6 can be obtained. Therefore, in the path determination step 250, the path of path 3 can be used as the pesticide spraying path S1.

In addition, the pesticide spraying planning method 200 further includes a spraying time planning step 260 and a pesticide spraying step 270. In the spraying time planning step 260, a residence time of the UAV 110 corresponding to each point is determined based on the plural tree species density of the target tree species information. In the pesticide spraying step 270, the drone 110 is caused to spray pesticides according to the order of the plurality of points in the pesticide spraying path S1 and the residence time of each point.

Among them, in the spraying time planning step 260, the tree species density can be confirmed through the image analysis module 121 of the processor 120. Although the spraying time planning step 260 is placed after the path determination step 250 in Figure 3, it can also be placed before the random path generation step 220, but is not limited to this.

As can be seen from the above embodiments, the present invention can use drones to take aerial images of the target area, identify tree species through deep learning neural network technology and perform color block processing to distinguish tree species categories, and find the two-dimensional distribution of the target tree species in the target area. , at the same time, the height of the target area can be obtained through terrain modeling, and further using genetic algorithms and annealing simulation algorithms, a better pesticide spraying path can be quickly obtained for pesticide spraying.

Although the present invention has been disclosed above through embodiments, they are not intended to limit the present invention. Anyone skilled in the art can make various modifications and modifications without departing from the spirit and scope of the present invention. Therefore, the protection of the present invention is The scope shall be determined by the appended patent application scope.

100:Pesticide spraying system 110: Drone 111:Camera 120: Processor 121:Image analysis module 122:Path planning module 123: Spraying time planning module 124:Terrain component module 200: Pesticide Spray Planning Methods 210: Steps to obtain target tree species information 220: Random path generation steps 230: New path generation steps 240: New and old path selection steps 250: Path determination steps 260: Spraying time planning steps 270: Pesticide spraying steps S1: Pesticide spraying path T1: Target tree species

Figure 1 illustrates a block diagram of a pesticide spraying system according to an embodiment of the present invention; Figure 2 shows a schematic planning diagram of the pesticide spraying path in the embodiment of Figure 1; and Figure 3 illustrates a block flow chart of a pesticide spraying planning method according to another embodiment of the present invention.

200: Pesticide Spray Planning Methods

210: Steps to obtain target tree species information

220: Random path generation steps

230: New path generation steps

240: New and old path selection steps

250: Path determination steps

260: Spraying time planning steps

270: Pesticide spraying steps

Claims (8)

A pesticide spraying planning method includes: a target tree species information acquisition step to obtain target tree species information about a target tree species in a target area, and the target tree species information includes a plurality of points; a random path generation step to randomly use the points Sorting to generate a plurality of old paths and obtaining an initial path group; a new path generation step, using a genetic algorithm to change the order of at least two points in each old path to generate a new path; a new and old path The path selection step uses a simulated annealing algorithm to select at least one old path and its corresponding new path to decide to retain the at least one old path or its corresponding new path, and in the early evolution process , the probability of retaining at least one old path with a longer distance is greater than the probability of retaining the corresponding new path with a shorter distance, and the updated path group is obtained; and a path determination step is repeated a preset number of times In the new path generation step and the old and new path selection step, the shortest distance in the finally updated path group is defined as a pesticide spraying path. The pesticide spraying planning method as described in claim 1 further includes: a spraying time planning step, which determines a residence time of a drone corresponding to each point according to the plurality of tree species densities of the target tree species information; and a pesticide spraying step , causing the UAV to spray pesticides according to the sequence of the points in the pesticide spraying path and the residence time at each point. The pesticide spraying planning method as described in claim 1, wherein in the step of obtaining the target tree species information, a drone is equipped with a camera to capture at least one image of the target area, and a processor is used to perform processing on the at least one image Identify and analyze to obtain information on the target tree species. The pesticide spraying planning method as described in claim 3, wherein in the step of obtaining the target tree species information, an image analysis module of the processor is used to identify the target tree species and perform color block processing to distinguish the target tree species, and obtain the target tree species. The plurality of longitudes and plurality of latitudes of the target tree species distribution, and using a terrain component module of the processor to model the target area to obtain the plurality of heights of the target tree species distribution, thereby obtaining the points of the target tree species information . The pesticide spraying planning method as described in claim 4, wherein each longitude and each latitude are converted into TWD97 coordinate values. A pesticide spraying system, including: a drone, including a camera, the camera is used to photograph a target area to obtain at least one image; and a processor, a signal is connected to the camera to receive the at least one image and includes: an image analysis A module that identifies and analyzes the at least one image to obtain a target tree species information related to a target tree species in the target area, where the target tree species information includes a plurality of points; and A path planning module receives the target tree species information to plan a pesticide spraying path, wherein the path planning module randomly sorts the points to generate a plurality of old paths, and obtains an initial path group, and then uses a The genetic algorithm changes the order of at least two points in each old path to generate a new path, and then uses a simulated annealing algorithm to select at least one old path and its corresponding new path to decide to retain The at least one old path or the corresponding new path is retained, and in the early evolution process, the probability of retaining the at least one old path with a longer distance is greater than the probability of retaining the corresponding new path with a shorter distance, and After obtaining the updated path group, continue to update the path group using the genetic algorithm and the simulated annealing algorithm, and the shortest distance in the finally updated path group is defined as the pesticide spraying path; where, The drone sprays pesticides in the order of the points in the pesticide spraying path. The pesticide spraying system as described in claim 6, wherein the processor further includes: a spraying time planning module that receives the target tree species information, and based on the plurality of tree species densities of the target tree species information, the drone corresponding to each tree species is given A dwell time at this point. The pesticide spraying system of claim 6, wherein the processor further includes a terrain component module for modeling the target area to obtain the plurality of heights of the target tree species distribution, and the image analysis module use To identify the target tree species and perform color block processing to distinguish the target tree species, obtain plural longitudes and plural latitudes of the distribution of the target tree species, thereby obtaining the points of the target tree species information.

Images