A method and system for spraying areas by using a drone for forest pest control

Abstract

The present invention relates to the technical field of image segmentation, and proposes a method and system for spraying areas of a drone for forest pest control, including: collecting forest images and multispectral images, dividing suspected forest areas and obtaining edge pixels and non-edge pixels; marking target pixels, respectively determining the edge continuity score of each row in the local window of the target pixels, determining the continuity of the target pixels and the tree confidence, and dividing the tree area; determining the disease impact of the tree area according to the peaks of the red light band and the near-infrared light band in the multispectral image corresponding to the position area of ​​the tree area; determining the amount of medicine sprayed in the tree area according to the disease impact of the tree area and the tree confidence of all pixels contained in the tree area, so as to realize drone spraying for forest pest control. The present invention solves the problem that the inaccurate vegetation area division makes the drone spraying and application effect poor.

Description

Technical Field

[0001] The present invention relates to the technical field of image segmentation, and in particular to a method and system for spraying an area of ​​a forest pest control unmanned aerial vehicle. Background Art

[0002] Drones equipped with cameras and other equipment can conduct all-round monitoring of forest structure, vegetation coverage, vegetation health, etc., which can help forest managers keep abreast of the dynamics of pests and diseases and develop scientific prevention and control methods. At the same time, drones equipped with cameras and other equipment can also achieve precise pesticide application, reducing the waste of sprayed pesticides and pollution and damage to the environment. In order to achieve precise pesticide application, it is necessary to analyze the vegetation photos taken by the drone's camera, divide the vegetation areas in the vegetation photos, and then use the drone to apply pesticides to the vegetation areas.

[0003] Generally, the vegetation areas in vegetation photos are divided out using their color features. However, the color of vegetation varies in different seasons, and the colors of different types of vegetation also differ. The vegetation areas divided only based on their color features are not accurate, which often leads to poor spraying effects by drones. Summary of the invention

[0004] The present invention provides a method and system for spraying areas of a forest pest control drone to solve the problem of inaccurate vegetation area division, which results in poor spraying effect of drones. The technical solutions adopted are as follows:

[0005] In a first aspect, an embodiment of the present invention provides a method for spraying a forest pest control drone spraying area, the method comprising the following steps:

[0006] Collect forest images and multispectral images at the same location in the area where forest pest control is required, divide the suspected forest area according to the grayscale value of the pixel points in the forest image, and obtain the edge pixel points and non-edge pixel points;

[0007] Any pixel in any suspected forest area is recorded as a target pixel, a local window of the target pixel is set, and according to the position distribution of edge pixels and non-edge pixels contained in the local window of the target pixel in the same row, the edge continuity score of each row in the local window of the target pixel is determined respectively, and the suspected tree row of the target pixel is divided, and the continuity of the target pixel is determined according to the edge continuity scores of all suspected tree rows in the local window of the target pixel and the number of suspected tree rows of the target pixel, as well as the positional relationship between the row where the target pixel is located and the suspected tree row of the target pixel in the local window of the target pixel, and the tree confidence of all pixels of the target pixel is determined according to the continuity of the target pixel and the number of edge pixels contained in the local window of the target pixel to divide the tree area;

[0008] Determine the disease impact of the tree area according to the peaks of the red light band and the near infrared light band in the multispectral image corresponding to the tree area;

[0009] According to the disease impact of the tree area and the tree confidence of all pixels contained in the tree area, the amount of medicine spraying in the tree area is determined to realize drone spraying for forest pest control.

[0010] Furthermore, the specific method of determining the edge continuity score of each row in the local window of the target pixel point includes:

[0011] For each row of pixels in the local window of the target pixel, the difference between the number of edge pixels and twice the number of non-edge pixels in the same row is recorded as the edge number distribution difference of the same row;

[0012] Two adjacent pixels in the same row of pixels that are arranged in the order of edge pixels and non-edge pixels are recorded as a first-class pixel pair, and two adjacent pixels in the same row of pixels that are arranged in the order of non-edge pixels and edge pixels are recorded as a second-class pixel pair; the sum of the number of first-class pixel pairs and second-class pixel pairs that appear in the same row of pixels in the local window is recorded as the number of neighbors in the same row;

[0013] The difference between the edge quantity distribution difference of the same row and the adjacent quantity within the local window of the target pixel is recorded as the edge continuity score of the row.

[0014] Furthermore, the continuity of the target pixel is determined based on the edge continuity scores of all suspected tree rows in the local window of the target pixel and the number of suspected tree rows of the target pixel, as well as the positional relationship between the row where the target pixel is located and the suspected tree row of the target pixel in the local window of the target pixel, including the specific method of:

[0015] For the suspected tree rows above the target pixel in the local window, the product of the mean of the edge continuity scores of all the suspected tree rows of the target pixel and the number of the suspected tree rows of the target pixel is recorded as the upper row distribution continuity of the target pixel;

[0016] For the suspected tree rows in the local window of the target pixel and below the target pixel, the product of the mean of the edge continuity scores of all the suspected tree rows of the target pixel and the number of the suspected tree rows of the target pixel is recorded as the downward distribution continuity of the target pixel;

[0017] The continuity of the target pixel is obtained according to the positional relationship between the row where the target pixel is located and the suspected tree row of the target pixel in the local window of the target pixel, as well as the continuity of the upper and lower distributions of the target pixel.

[0018] Furthermore, the continuity of the target pixel is obtained according to the positional relationship between the row where the target pixel is located and the suspected tree row of the target pixel in the local window of the target pixel, as well as the continuity of the upper and lower distributions of the target pixel, including the specific method of:

[0019] The minimum number of rows between the row of suspected trees on the upper side of the target pixel and the row where the target pixel is located in the local window of the target pixel is recorded as the upper distance of the target pixel, and the minimum number of rows between the row of suspected trees on the lower side of the target pixel and the row where the target pixel is located in the local window of the target pixel is recorded as the lower distance of the target pixel;

[0020] The ratio of the upper distribution continuity of the target pixel to the upper distance is recorded as the upper continuity of the target pixel, and the ratio of the lower distribution continuity of the target pixel to the lower distance is recorded as the lower continuity of the target pixel;

[0021] The continuity of the target pixel point is determined according to the upper continuity and the lower continuity of the target pixel point.

[0022] Further, the determination of the continuity of the target pixel point according to the upper continuity and the lower continuity of the target pixel point includes the following specific methods:

[0023] The maximum value of the upper continuity and the lower continuity of the target pixel is recorded as the continuity of the target pixel.

[0024] Furthermore, the tree confidence of the target pixel is determined according to the continuity of the target pixel and the number of edge pixels contained in the local window of the target pixel, and the tree area is divided according to the tree confidence of all pixels in the forest image, including the specific method of:

[0025] The normalized value of the product of the number of edge pixels contained in the local window of the target pixel and the continuity of the target pixel is recorded as the tree confidence of the target pixel;

[0026] Pixels whose tree confidence is greater than a third preset threshold are recorded as tree pixels, and an area composed of tree pixels in the suspected forest area is recorded as a tree area.

[0027] Furthermore, the disease impact of the tree area is determined according to the peaks of the red light band and the near infrared light band in the multispectral image at the corresponding position of the tree area, and the specific method includes:

[0028] Collect multispectral images of healthy forests and obtain corresponding tree areas, record the average of all peaks of the red light band of the tree area corresponding to the healthy trees in the multispectral as the red light peak benchmark, and record the average of all peaks of the near-infrared light band of the tree area corresponding to the healthy trees in the multispectral as the near-infrared light peak benchmark;

[0029] The disease impact of the tree area is determined based on the difference between the peak of the red light band in the multispectral image and the red light peak benchmark, as well as the difference between the peak of the near-infrared light band and the near-infrared light peak benchmark at the corresponding position of the tree area in the multispectral image.

[0030] Further, the disease impact of the tree area is determined according to the difference between the peak of the red light band in the multispectral image and the red light peak benchmark, and the difference between the peak of the near infrared light band and the near infrared light peak benchmark, including the specific method of:

[0031] The average of all the peaks of the red light band of the corresponding position area of ​​the tree area in the multispectral image is recorded as the red light peak average of the tree area, and the absolute value of the difference between the red light peak average of the corresponding position area of ​​the tree area in the multispectral image and the red light peak benchmark is recorded as the red light peak difference of the tree area;

[0032] The mean of all the peaks of the near-infrared light band of the corresponding position area of ​​the tree area in the multispectral image is recorded as the near-infrared light peak mean of the tree area, and the absolute value of the difference between the near-infrared light peak mean of the tree area and the near-infrared light peak benchmark is recorded as the near-infrared light peak difference of the tree area;

[0033] The product of the red light peak difference and the near-infrared light peak difference in the tree area is recorded as the disease impact degree of the tree area.

[0034] Furthermore, the method of determining the amount of medicine sprayed in the tree area according to the disease impact of the tree area and the tree confidence of all pixels contained in the tree area, and realizing drone spraying for forest pest control includes the following specific methods:

[0035] The normalized value of the product of the mean of the tree confidence of all pixels in the tree area and the disease impact of the tree area is recorded as the adjustment weight of the tree area, the sum of the adjustment weight of the tree area and the number 1 is recorded as the adjustment amount of the tree area, and the product of the adjustment amount of the tree area and the standard amount of drug spraying is recorded as the drug spraying amount of the tree area;

[0036] When the UAV flies to the collection location of the forest image, the drug is sprayed according to the corresponding drug spraying amount of all tree areas in the forest image, thereby realizing UAV spraying for forest pest control.

[0037] In the second aspect, an embodiment of the present invention also provides a forest pest control drone spraying area spraying system, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor implements the steps of any one of the above-mentioned methods when executing the computer program.

[0038] The beneficial effects of the present invention are:

[0039] The present application first divides the suspected forest area according to the grayscale value of the pixel points in the forest image, and then analyzes each pixel point in the suspected forest area. First, according to the characteristics of the large number of tree leaves, small area, and clustered distribution, which correspond to a large number of unevenly distributed edge pixels in the image, the continuity of edge pixels and non-edge pixels in the local range of the pixel points is analyzed to determine the suspected tree row of the target pixel point. The suspected tree row is the data row in the local range of the target pixel point that shows obvious tree leaf characteristics. Furthermore, from the upper, lower, and edge of the target pixel point in the local range, the The continuity of pixel points and non-edge pixel points is further analyzed to determine the continuity of target pixel points, and the confidence of the target pixel points corresponding to the tree is determined based on the continuity of the target pixel points, and the tree area is divided; then, the degree of influence of trees on plants and insect pests can be reflected based on chlorophyll-related information, and the degree of disease impact of the tree area can be determined; finally, the amount of medicine sprayed in the tree area is determined based on the disease impact of the tree area and the tree confidence of all pixels contained in the tree area, so as to realize drone spraying for forest pest control and solve the problem of inaccurate vegetation area division, which makes drone spraying and application effect poor. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0041] Figure 1 A schematic diagram of a flow chart of a method for spraying an area using a forest pest control drone provided by an embodiment of the present invention;

[0042] Figure 2 A continuity acquisition flow chart provided for one embodiment of the present invention. DETAILED DESCRIPTION

[0043] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0044] See also Figure 1 , which shows a flow chart of a method for spraying a forest pest control drone spraying area provided by an embodiment of the present invention, the method comprising the following steps:

[0045] Step S001, collecting forest images and multispectral images at the same location in the area where forest pest control is required, dividing suspected forest areas according to the grayscale values ​​of pixels in the forest image, and obtaining edge pixels and non-edge pixels.

[0046] Use the camera carried by the drone to sequentially capture images of areas that require forest pest control at the same height to obtain forest images. Use the multispectral camera carried by the drone to collect multispectral images at the same location where the forest images were obtained.

[0047] It is understandable that forest images contain not only various trees, but also streams, roads and other objects that do not need to be sprayed. In order to prevent and control forest pests, the areas corresponding to the trees need to be sprayed. There is an obvious grayscale difference between the trees that need to be sprayed and the objects that do not need to be sprayed; each forest image corresponds to a multispectral image.

[0048] Perform threshold segmentation on the forest image to divide the suspected forest area in the forest image.

[0049] Preferably, as an embodiment of the present application, the forest image is threshold segmented using the maximum inter-class variance method to obtain a segmentation threshold, and pixels with grayscale values ​​greater than the segmentation threshold are recorded as suspected forest pixels. Connected domain analysis is used for the suspected forest pixels, and each connected domain obtained is recorded as a suspected forest area.

[0050] At this point, the suspected forest area in the forest image is obtained.

[0051] Different leaves have different shades of color. When drones are photographed in the air, the leaves corresponding to each pixel of the corresponding tree are smaller and in different positions. There are also differences between the grayscale values, and there are also differences between the pixel values ​​of adjacent pixels, resulting in a large number of unevenly distributed edge pixels. Roads are man-made facilities with uniform textures, and the differences between the pixel values ​​of adjacent pixels are small, resulting in a large number of continuously distributed non-edge pixels.

[0052] Perform edge detection on the forest image to obtain the forest edge image, and obtain edge pixel points and non-edge pixel points in the forest edge image.

[0053] At this point, edge pixels and non-edge pixels in the forest edge image are obtained.

[0054] Step S002, record any pixel in any suspected forest area as a target pixel, set a local window of the target pixel, determine the edge continuity score of each row in the local window of the target pixel according to the position distribution of edge pixels and non-edge pixels in the same row contained in the local window of the target pixel, and divide the suspected tree rows of the target pixel, determine the continuity of the target pixel according to the edge continuity scores of all suspected tree rows in the local window of the target pixel and the number of suspected tree rows of the target pixel, and the positional relationship between the row where the target pixel is located in the local window of the target pixel and the suspected tree row of the target pixel, and determine the tree confidence of all pixels of the target pixel according to the continuity of the target pixel and the number of edge pixels contained in the local window of the target pixel to divide the tree area.

[0055] Any pixel in any suspected forest area is recorded as a target pixel, a local window with a side length of a first preset threshold is established with the target pixel as the center, and the number of edge pixels contained in the local window of the target pixel is calculated.

[0056] Preferably, as an embodiment of the present application, the value of the first preset threshold is 25.

[0057] Furthermore, the continuity of edge pixels contained in the local window of the target pixel is analyzed.

[0058] According to the position distribution of edge pixels and non-edge pixels contained in the local window of the target pixel in the same row, the edge continuity score of each row in the local window of the target pixel is determined respectively.

[0059] Preferably, as an embodiment of the present application, for each row of pixels in the local window of the target pixel, the difference between the number of edge pixels in the same row and twice the number of non-edge pixels is recorded as the edge number distribution difference of the same row. Two adjacent pixels in the same row of pixels arranged in the order of edge pixels and non-edge pixels are recorded as a type I pixel pair, and two adjacent pixels in the same row of pixels arranged in the order of non-edge pixels and edge pixels are recorded as a type II pixel pair. The sum of the number of type I pixel pairs and type II pixel pairs appearing in the same row of pixels in the local window is recorded as the adjacent number of the same row, and the difference between the edge number distribution difference and the adjacent number of the same row in the local window of the target pixel is recorded as the edge continuity score of the row.

[0060] The edge continuity score of each row in the local window of the target pixel is compared with the second preset threshold, and the row corresponding to the edge continuity score greater than the second preset threshold is recorded as the suspected tree row of the target pixel.

[0061] Preferably, as an embodiment of the present application, the value of the second preset threshold is 6.

[0062] According to the edge continuity scores of all suspected tree rows in the local window of the target pixel and the number of suspected tree rows of the target pixel, the upward distribution continuity and downward distribution continuity of the target pixel are determined respectively.

[0063] Preferably, as an embodiment of the present application, for the suspected tree rows within the local window of the target pixel and on the upper side of the target pixel, the product of the mean of the edge continuity scores of all suspected tree rows of the target pixel and the number of suspected tree rows of the target pixel is recorded as the upper distribution continuity of the target pixel.

[0064] Preferably, as an embodiment of the present application, for the suspected tree rows within the local window of the target pixel and below the target pixel, the product of the mean of the edge continuity scores of all suspected tree rows of the target pixel and the number of suspected tree rows of the target pixel is recorded as the downward distribution continuity of the target pixel.

[0065] In the local window of the target pixel, on the upper side of the target pixel, the greater the edge continuity score of all suspected tree rows of the target pixel and the greater the number of suspected tree rows of the target pixel, the denser and more continuous the distribution of edge pixels on the upper side of the local window of the target pixel, and at this time, the greater the continuity of the upper row distribution of the target pixel. The same is true for the lower side of the target pixel in the local window of the target pixel.

[0066] Furthermore, according to the positional relationship between the row where the target pixel is located in the local window of the target pixel and the suspected tree row of the target pixel, as well as the upper distribution continuity and the lower distribution continuity of the target pixel, the upper continuity and the lower continuity of the target pixel are obtained.

[0067] Preferably, as an embodiment of the present application, the minimum number of rows between the row of suspected trees on the upper side of the target pixel and the row where the target pixel is located within the local window of the target pixel is recorded as the upper distance of the target pixel, the minimum number of rows between the row of suspected trees on the lower side of the target pixel and the row where the target pixel is located within the local window of the target pixel is recorded as the lower distance of the target pixel, the ratio of the upper row distribution continuity of the target pixel to the upper distance is recorded as the upper continuity of the target pixel, and the ratio of the lower row distribution continuity of the target pixel to the lower distance is recorded as the lower continuity of the target pixel.

[0068] When the row where the target pixel is located in the local window of the target pixel is closer to the suspected tree row of the target pixel, the upper distribution continuity of the target pixel is greater, and the lower distribution continuity of the target pixel is greater, the upper continuity and lower continuity of the target pixel are greater. At this time, the continuity of the edge pixels contained in the local window of the target pixel is more significant, and the possibility that the target pixel corresponds to a forest tree is greater.

[0069] The tree confidence of the target pixel is determined according to the upper continuity, lower continuity and the number of edge pixels contained in the local window of the target pixel.

[0070] Preferably, as an embodiment of the present application, the maximum value of the upper continuity and the lower continuity of the target pixel is recorded as the continuity of the target pixel. The normalized value of the product of the number of edge pixels contained in the local window of the target pixel and the continuity of the target pixel is recorded as the tree confidence of the target pixel.

[0071] Among them, the continuous acquisition flow chart is as follows Figure 2 shown.

[0072] It should be noted that this embodiment uses the Z-Score standard normalization method to calculate the normalized value. In actual application, the implementer may use other methods in the prior art such as the maximum and minimum value normalization method to calculate the normalized value, which is not limited here.

[0073] Pixels whose tree confidence is greater than the third preset threshold are recorded as tree pixels, and pixels whose tree confidence is less than or equal to the third preset threshold are recorded as road pixels. Connected domain analysis is performed on the tree pixels in the suspected forest area to obtain the tree area.

[0074] It can be understood that the tree area is an area composed of pixel points corresponding to trees in the suspected forest area, and the tree area is the location in the forest image where drone spraying is required.

[0075] Preferably, as an embodiment of the present application, the value of the third preset threshold is 6.

[0076] At this point, the tree area in each suspected forest area in the forest image is obtained, that is, all the tree areas in the forest image are obtained.

[0077] Step S003, determining the disease impact of the tree area according to the peaks of the red light band and the near infrared light band in the multispectral image at the corresponding position area of ​​the tree area.

[0078] Since each forest image corresponds to a multispectral image, the tree areas corresponding to the position areas in the multispectral image are divided according to all the tree areas in the forest image, and the divided areas are recorded as forest areas.

[0079] When spraying pesticides on tree areas in forest images that need to be sprayed by drones, the amount of pesticide spraying needs to be determined according to the degree to which the trees are affected by pests and diseases in order to achieve the purpose of pest control. Information such as the chlorophyll content of trees in the tree area in the multispectral image can reflect the degree to which the trees are affected by pests and diseases to a certain extent. Chlorophyll is the main pigment responsible for photosynthesis in plant leaves and has significant spectral characteristics for red light and near-infrared light. Among them, the red light band is the 600-700 nanometer band in the multispectral spectrum, and the near-infrared light band is the 700-1100 nanometer band in the multispectral spectrum.

[0080] According to the same method of obtaining the corresponding tree area in the multispectral image, multispectral images of the healthy forest are collected and the corresponding tree areas are obtained. The average of all peaks of the tree area corresponding to the healthy trees in the 600-700 nanometer band in the multispectrum is recorded as the red light peak benchmark, and the average of all peaks of the tree area corresponding to the healthy trees in the 700-1100 nanometer band in the multispectrum is recorded as the near-infrared light peak benchmark.

[0081] The disease impact of the forest area is determined based on the difference between the peak of the red light band in the multi-spectrum and the red light peak benchmark, as well as the difference between the near-infrared light band and the near-infrared light peak benchmark.

[0082] Preferably, as an embodiment of the present application, the mean of all peaks of the forest area in the 600-700 nanometer band in the multi-spectrum is recorded as the red light peak mean of the forest area, and the absolute value of the difference between the red light peak mean of the forest area and the red light peak benchmark is recorded as the red light peak difference of the forest area; the mean of all peaks of the forest area in the 700-1100 nanometer band in the multi-spectrum is recorded as the near-infrared light peak mean of the forest area, and the absolute value of the difference between the near-infrared light peak mean of the forest area and the near-infrared light peak benchmark is recorded as the near-infrared light peak difference of the forest area; the product of the red light peak difference and the near-infrared light peak difference of the forest area is recorded as the disease impact degree of the forest area.

[0083] The greater the difference between the peak of the red light band in the multi-spectrum and the red light peak benchmark, and the greater the difference between the near-infrared light band and the near-infrared light peak benchmark, the greater the impact of the disease in the forest area. At this time, the trees in the forest area are more affected by diseases and pests.

[0084] It can be understood that the location area corresponding to the tree area in the multispectral image is the forest area, each tree area corresponds to a forest area in the multispectral image, and the disease impact degree of the tree area is the disease impact degree of the location area corresponding to the tree area in the multispectral image.

[0085] At this point, the disease impact of the tree area is obtained.

[0086] Step S004, determining the amount of medicine sprayed in the tree area according to the disease impact of the tree area and the tree confidence of all pixels contained in the tree area, and realizing drone spraying for forest pest control.

[0087] The amount of medicine sprayed in the tree area is determined according to the disease impact of the tree area and the tree confidence of all pixels contained in the tree area.

[0088] Preferably, as an embodiment of the present application, the normalized value of the product of the mean of the tree confidence of all pixels contained in the tree area and the disease impact of the tree area is recorded as the adjustment weight of the tree area, the sum of the adjustment weight of the tree area and the number 1 is recorded as the adjustment amount of the tree area, and the product of the adjustment amount of the tree area and the standard amount of drug spraying is recorded as the drug spraying amount of the tree area.

[0089] When the disease impact of the tree area is greater and the tree confidence of all pixels contained in the tree area is greater, the trees in the tree area are more affected by diseases and pests. At this time, the greater the amount of medicine sprayed in the tree area, the more medicine should be sprayed to reduce the impact of diseases and pests on the trees.

[0090] When the UAV flies to the collection location of the forest image, the drug is sprayed according to the drug spraying amount corresponding to each tree area in the forest image, thereby realizing UAV spraying for forest pest control.

[0091] Based on the same inventive concept as the above method, an embodiment of the present invention also provides a forest pest control drone spraying area spraying system, including a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it implements the steps of any one of the above-mentioned forest pest control drone spraying area spraying methods.

[0092] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for spraying an area using a drone for forest pest control, characterized in that: The method comprises the following steps: Collect forest images and multispectral images at the same location in the area where forest pest control is required, divide the suspected forest area according to the grayscale value of the pixel points in the forest image, and obtain the edge pixel points and non-edge pixel points; Any pixel in any suspected forest area is recorded as the target pixel, and a local window of the target pixel is set. According to the position distribution of edge pixels and non-edge pixels in the local window of the target pixel in the same row, the edge continuity score of each row in the local window of the target pixel is determined respectively, and the suspected tree rows of the target pixel are divided. For the suspected tree rows in the local window of the target pixel and above the target pixel, the product of the mean of the edge continuity scores of all the suspected tree rows of the target pixel and the number of the suspected tree rows of the target pixel is recorded as the upper row distribution continuity of the target pixel. For the suspected tree rows in the local window of the target pixel and below the target pixel, the product of the mean of the edge continuity scores of all the suspected tree rows of the target pixel and the number of the suspected tree rows of the target pixel is recorded as the upper row distribution continuity of the target pixel. The product of the mean of the edge continuity scores of all suspected tree rows of the target pixel and the number of suspected tree rows of the target pixel is recorded as the downstream distribution continuity of the target pixel; the continuity of the target pixel is obtained according to the positional relationship between the row where the target pixel is located and the suspected tree row of the target pixel in the local window of the target pixel, as well as the upstream distribution continuity and downstream distribution continuity of the target pixel, and the normalized value of the product of the number of edge pixels contained in the local window of the target pixel and the continuity of the target pixel is recorded as the tree confidence of the target pixel; the pixel whose tree confidence is greater than the third preset threshold is recorded as the tree pixel, and the area composed of tree pixels in the suspected forest area is recorded as the tree area; Determine the disease impact of the tree area according to the peaks of the red light band and the near infrared light band in the multispectral image corresponding to the tree area; According to the disease impact of the tree area and the tree confidence of all pixels contained in the tree area, the amount of medicine spraying in the tree area is determined to realize drone spraying for forest pest control.

2. A method for spraying an area using a drone for forest pest control according to claim 1, characterized in that: The specific method of determining the edge continuity score of each row in the local window of the target pixel point includes: For each row of pixels in the local window of the target pixel, the difference between the number of edge pixels and twice the number of non-edge pixels in the same row is recorded as the edge number distribution difference of the same row; Two adjacent pixels in the same row of pixels that are arranged in the order of edge pixels and non-edge pixels are recorded as a first-class pixel pair, and two adjacent pixels in the same row of pixels that are arranged in the order of non-edge pixels and edge pixels are recorded as a second-class pixel pair; the sum of the number of first-class pixel pairs and second-class pixel pairs that appear in the same row of pixels in the local window is recorded as the number of neighbors in the same row; The difference between the edge quantity distribution difference of the same row and the adjacent quantity within the local window of the target pixel is recorded as the edge continuity score of the row.

3. A method for spraying an area using a drone for forest pest control according to claim 1, characterized in that: The specific method of obtaining the continuity of the target pixel point according to the positional relationship between the row where the target pixel point is located and the suspected tree row of the target pixel point in the local window of the target pixel point, as well as the continuity of the upper and lower distributions of the target pixel point, includes: The minimum number of rows between the row of suspected trees on the upper side of the target pixel and the row where the target pixel is located in the local window of the target pixel is recorded as the upper distance of the target pixel, and the minimum number of rows between the row of suspected trees on the lower side of the target pixel and the row where the target pixel is located in the local window of the target pixel is recorded as the lower distance of the target pixel; The ratio of the upper distribution continuity of the target pixel to the upper distance is recorded as the upper continuity of the target pixel, and the ratio of the lower distribution continuity of the target pixel to the lower distance is recorded as the lower continuity of the target pixel; The continuity of the target pixel point is determined according to the upper continuity and the lower continuity of the target pixel point.

4. A method for spraying an area using a drone for forest pest control according to claim 3, characterized in that: The specific method of determining the continuity of the target pixel point according to the upper continuity and the lower continuity of the target pixel point is as follows: The maximum value of the upper continuity and the lower continuity of the target pixel is recorded as the continuity of the target pixel.

5. The method for spraying an area using a drone for forest pest control according to claim 1, characterized in that: The specific method of determining the disease impact of the tree area according to the peaks of the red light band and the near infrared light band in the multispectral image corresponding to the position area of ​​the tree area includes: Collect multispectral images of healthy forests and obtain corresponding tree areas, record the average of all peaks of the red light band of the tree area corresponding to the healthy trees in the multispectral as the red light peak benchmark, and record the average of all peaks of the near-infrared light band of the tree area corresponding to the healthy trees in the multispectral as the near-infrared light peak benchmark; The disease impact of the tree area is determined based on the difference between the peak of the red light band in the multispectral image and the red light peak benchmark, as well as the difference between the peak of the near-infrared light band and the near-infrared light peak benchmark at the corresponding position of the tree area in the multispectral image.

6. A method for spraying an area using a drone for forest pest control according to claim 5, characterized in that: The method of determining the disease impact of the tree area according to the difference between the peak of the red light band and the red light peak benchmark in the multispectral image of the corresponding position area of ​​the tree area, and the difference between the peak of the near infrared light band and the near infrared light peak benchmark, includes the following specific methods: The average of all the peaks of the red light band of the corresponding position area of ​​the tree area in the multispectral image is recorded as the red light peak average of the tree area, and the absolute value of the difference between the red light peak average of the corresponding position area of ​​the tree area in the multispectral image and the red light peak benchmark is recorded as the red light peak difference of the tree area; The mean of all the peaks of the near-infrared light band of the corresponding position area of ​​the tree area in the multispectral image is recorded as the near-infrared light peak mean of the tree area, and the absolute value of the difference between the near-infrared light peak mean of the tree area and the near-infrared light peak benchmark is recorded as the near-infrared light peak difference of the tree area; The product of the red light peak difference and the near-infrared light peak difference in the tree area is recorded as the disease impact degree of the tree area.

7. A method for spraying an area using a drone for forest pest control according to claim 1, characterized in that: The specific method of determining the amount of medicine sprayed in the tree area according to the disease impact of the tree area and the tree confidence of all pixels contained in the tree area to realize the drone spraying for forest pest control includes: The normalized value of the product of the mean of the tree confidence of all pixels in the tree area and the disease impact of the tree area is recorded as the adjustment weight of the tree area, the sum of the adjustment weight of the tree area and the number 1 is recorded as the adjustment amount of the tree area, and the product of the adjustment amount of the tree area and the standard amount of drug spraying is recorded as the drug spraying amount of the tree area; When the UAV flies to the collection location of the forest image, the drug is sprayed according to the corresponding drug spraying amount of all tree areas in the forest image, thereby realizing UAV spraying for forest pest control.

8. A forest pest control drone spraying area spraying system, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

Images