Eucalyptus canopy unmanned aerial vehicle accurate topdressing method based on hyperspectral remote sensing prescription map

Abstract

The invention belongs to the technical field of unmanned aerial vehicle fertilization, and discloses a eucalyptus canopy unmanned aerial vehicle accurate topdressing method based on a hyperspectral remote sensing prescription map. Comprising the following steps: simultaneously acquiring a hyperspectral remote sensing image and a reference canopy image of a eucalyptus canopy in a target range, and segmenting the hyperspectral remote sensing image and the reference canopy image to obtain a hyperspectral remote sensing image and a reference canopy image corresponding to each eucalyptus; and unmixing the hyperspectral remote sensing image corresponding to each eucalyptus, establishing a eucalyptus canopy nitrogen inversion model, and establishing a hyperspectral remote sensing prescription map of each eucalyptus based on the eucalyptus canopy nitrogen inversion model. By setting the upper space, the flight space and the lower space in the mapping space, matching verification can be quickly performed on the canopy image and the corresponding reference canopy image, the situation that the current eucalyptus is wrongly applied or a non-corresponding topdressing amount is applied due to inaccurate positioning of the fertilization unmanned aerial vehicle is avoided, and thus accurate topdressing of the current eucalyptus is conveniently achieved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of unmanned aerial vehicle fertilization, more particularly, the present application relates to an eucalyptus canopy unmanned aerial vehicle precision topdressing method based on a hyperspectral remote sensing prescription map. BACKGROUND

[0002] In recent years, with the continuous development of hyperspectral technology, its application in the fields of agriculture and vegetation monitoring is becoming increasingly widespread. In the field of vegetation remote sensing, hyperspectral sensors have the ability to capture detailed spectral information. For example, with the help of hyperspectral data, key nutritional parameters such as the nitrogen content and chlorophyll concentration of vegetation can be accurately estimated. Moreover, unmanned aerial vehicles equipped with hyperspectral sensors can fly at a relatively low altitude to obtain high spatial resolution spectral data, which provides more detailed information support for vegetation monitoring. Eucalyptus, as an important economic tree, has the remarkable characteristics of fast growth and strong adaptability. However, the growth of eucalyptus has a high demand for nutrients, especially during the vigorous growth period, and reasonable fertilization is crucial to ensuring the healthy growth of eucalyptus and improving timber yield and quality.

[0003] The existing patent with the patent number CN111670668A: rice agricultural unmanned aerial vehicle precision topdressing method based on a hyperspectral remote sensing prescription map, establishes a rice tillering period topdressing prescription map by using unmanned aerial vehicle hyperspectral technology, on this basis, combines agricultural unmanned aerial vehicle operation parameters, divides the grid of the land to be topdressed, forms a spraying amount suitable for precision topdressing operation in the field, and finally carries out precision topdressing through an agricultural unmanned aerial vehicle, in order to provide data and model basis for unmanned aerial vehicle precision variable topdressing of cold region rice during the tillering period;

[0004] However, when the above-mentioned existing patent is applied to the precision topdressing of each eucalyptus canopy in the eucalyptus forest, due to the high planting density and complex terrain of eucalyptus in the eucalyptus forest, the positioning of the fertilization unmanned aerial vehicle when topdressing the eucalyptus is not accurate, and it is not convenient to carry out precision topdressing of the canopy of each eucalyptus.

[0005] In view of this, the present application proposes an eucalyptus canopy unmanned aerial vehicle precision topdressing method based on a hyperspectral remote sensing prescription map to solve the above-mentioned problems. SUMMARY

[0006] In order to overcome the above-mentioned defects of the prior art, in order to achieve the above-mentioned purpose, the present application provides the following technical scheme: an eucalyptus canopy unmanned aerial vehicle precision topdressing method based on a hyperspectral remote sensing prescription map, comprising the following steps:

[0007] Simultaneously collect the hyperspectral remote sensing image and the reference canopy image of the eucalyptus canopy in the target range, segment the hyperspectral remote sensing image and the reference canopy image to obtain the corresponding hyperspectral remote sensing image and reference canopy image of each eucalyptus;

[0008] The hyperspectral remote sensing image corresponding to each eucalyptus is unblended, and a eucalyptus canopy nitrogen inversion model is established; and a hyperspectral remote sensing prescription chart of each eucalyptus is established based on the eucalyptus canopy nitrogen inversion model;

[0009] The current eucalyptus canopy image is collected by using a fertilization unmanned aerial vehicle, and a virtual mapping space is established based on the corresponding target range, wherein the virtual mapping space comprises an upper space, a flight space and a lower space from top to bottom;

[0010] The canopy image and the reference canopy image are verified, and the hyperspectral remote sensing prescription chart corresponding to the reference canopy image meeting the preset verification condition is applied to the canopy of the current eucalyptus, so as to realize accurate topdressing of the current eucalyptus.

[0011] Further, the step of segmenting the hyperspectral remote sensing image and the reference canopy image to obtain the hyperspectral remote sensing image and the reference canopy image corresponding to each eucalyptus includes:

[0012] The geographical range of the eucalyptus forest is determined, and the geographical range of the eucalyptus forest is taken as the target range;

[0013] The hyperspectral remote sensing image and the reference canopy image of each eucalyptus in the eucalyptus forest in the target range are collected by using an unmanned aerial vehicle carrying a hyperspectral imager and an image collection device, wherein the hyperspectral remote sensing image and the reference canopy image of the eucalyptus collected at the same time are associated and bound with the corresponding GPS position information;

[0014] The hyperspectral remote sensing image and the reference canopy image are preprocessed, and an image segmentation algorithm is used to segment the hyperspectral remote sensing image and the reference canopy image to obtain the hyperspectral remote sensing image and the reference canopy image corresponding to each eucalyptus.

[0015] Further, the step of establishing the hyperspectral remote sensing prescription chart of each eucalyptus based on the eucalyptus canopy nitrogen inversion model includes:

[0016] The hyperspectral reflectance curve of the eucalyptus canopy is extracted, and a ground feature endmember wave library is constructed, and then the hyperspectral remote sensing image corresponding to each eucalyptus is unblended by using the ground feature endmember wave library and an orthogonal subspace projection method to obtain the hyperspectral information of each eucalyptus;

[0017] A nitrogen inversion model of eucalyptus leaves is established, and the topdressing amount of each eucalyptus is determined based on the nitrogen inversion model of eucalyptus leaves, and the hyperspectral remote sensing prescription chart of each eucalyptus is established based on the corresponding topdressing amount of each eucalyptus.

[0018] Further, the step of establishing the virtual mapping space corresponding to the target range includes:

[0019] corresponding association of hyperspectral remote sensing prescription image and distribution according to GPS position information to constitute the upper space;

[0020] The three-dimensional space in which the fertilizing unmanned plane flies in the target range is taken as the flight space; the crown layer image of the current eucalyptus is collected by the fertilizing unmanned plane, and the lower space storing the crown layer image is established corresponding to the fertilizing unmanned plane, that is, the virtual mapping space composed of the upper space, the flight space and the lower space is obtained.

[0021] Further, the step of verifying the crown layer image with the reference crown layer image comprises:

[0022] The center point of the eucalyptus crown layer in the reference crown layer image corresponding to the current GPS position information is taken as the center to establish a reference basic area, and the reference basic area is divided into: a plurality of sub-areas formed by the intersection between a plurality of concentric rings and eight azimuth sector blocks, the division mode of the reference basic area of the reference crown layer image is taken as the division mode of the fixed basic area in the lower space, so as to establish the fixed basic area in the lower space, the fixed basic area is unique, and the sub-area formed by the intersection of the fixed basic area and the reference basic area in the same concentric ring and the same azimuth sector block is taken as the homologous area;

[0023] A feature code library is established, the first feature information in the plurality of sub-areas of the reference crown layer image is extracted and coded based on the feature code library and filled into the sub-areas of the corresponding reference basic area, and the second feature information in the plurality of sub-areas of the crown layer image is extracted and coded based on the feature code library and filled into the sub-areas of the fixed basic area;

[0024] The reference port point and the detection port point are set corresponding to the fixed basic area in the lower space, and the crown layer image of the current eucalyptus collected by the fertilizing unmanned plane is matched and verified with the reference crown layer image in the upper space based on the reference port point and the detection port point.

[0025] Further, the step of matching and verifying the crown layer image of the current eucalyptus collected by the fertilizing unmanned plane with the reference crown layer image in the upper space based on the reference port point and the detection port point comprises:

[0026] The fixed basic area in the lower space is divided according to the division mode of the reference basic area of the reference crown layer image, the reference port point is configured one by one in the plurality of sub-areas of the fixed basic area, and the transmission verification channel is established between each reference port point and the fixedly associated detection port point;

[0027] The crown layer image currently collected by the fertilizing unmanned plane is adapted to the fixed basic area in the lower space, the reference port point acquires the code of the sub-area, and the code is provided to the detection port point for carrying;

[0028] According to the same region, the corresponding reference port point of the detection port point is projected to the sub-region of the current GPS position information corresponding to the reference crown layer image at the same time, and the detection port point detects the code of the sub-region of the reference crown layer image;

[0029] The code carried by the detection port point is matched and verified with the code of the corresponding sub-region.

[0030] Further, the step of applying the hyperspectral remote sensing prescription map corresponding to the reference crown layer image that meets the preset verification condition to the current eucalyptus crown layer, that is, realizing the precise topdressing of the current eucalyptus, comprises:

[0031] If the code carried by any of the detection port points is the same as the code of the corresponding sub-region of the reference crown layer image, the detection port point is withdrawn from the corresponding sub-region of the reference crown layer image;

[0032] If the code carried by any of the detection port points is different from the code of the corresponding sub-region of the reference crown layer image, all the detection port points are not withdrawn and are projected one by one to the sub-regions of other reference crown layer images within the deviation radius corresponding to the outside of the reference crown layer image and are matched until all the detection port points are withdrawn;

[0033] When all the detection port points are withdrawn, it is indicated that the reference crown layer image corresponding to all the detection port points meets the preset verification condition, and the topdressing amount of the hyperspectral remote sensing prescription map associated with the reference crown layer image is sprayed to the crown layer of the current eucalyptus, that is, the precise topdressing of the current eucalyptus is realized.

[0034] Further, in the feature code library:

[0035] The first feature information and the second feature information both include feature information of the thickness of branches and trunks of the eucalyptus crown layer and the density of leaves of the eucalyptus crown layer;

[0036] The feature degree of each feature information one-to-one corresponds to the code in the feature code library.

[0037] The eucalyptus crown layer unmanned aerial vehicle precise topdressing method based on the hyperspectral remote sensing prescription map has the following technical effects and advantages:

[0038] Through the setting of the upper space, the flight space and the lower space in the mapping space, the crown layer image and the corresponding reference crown layer image can be quickly matched and verified, the current eucalyptus is prevented from being wrongly fertilized or being fertilized with an incorrect topdressing amount due to inaccurate positioning of the unmanned aerial vehicle for fertilization, and thus the precise topdressing of the current eucalyptus is facilitated. BRIEF DESCRIPTION OF DRAWINGS

[0039] Fig. 1 It is a flowchart of the eucalyptus crown layer unmanned aerial vehicle precise topdressing method based on the hyperspectral remote sensing prescription map.

[0040] Fig. 2 Structure diagram for filling the first feature information of the reference canopy image in the present application into the sub-region of the reference base region. DETAILED DESCRIPTION

[0041] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work fall within the scope of protection of the present application.

[0042] Please refer to Figs. 1-2 The eucalyptus canopy unmanned aerial vehicle precision topdressing method based on the hyperspectral remote sensing prescription map described in the present embodiment includes the following steps:

[0043] S1, simultaneously collecting hyperspectral remote sensing images and reference canopy images of eucalyptus canopies in a target range, and segmenting the hyperspectral remote sensing images and the reference canopy images to obtain corresponding hyperspectral remote sensing images and reference canopy images of each eucalyptus tree;

[0044] S2, demixing the corresponding hyperspectral remote sensing images of each eucalyptus tree and establishing a eucalyptus canopy nitrogen inversion model, and based on the eucalyptus canopy nitrogen inversion model, establishing a hyperspectral remote sensing prescription map of each eucalyptus tree;

[0045] S3, collecting a canopy image of the current eucalyptus tree by using a topdressing unmanned aerial vehicle, and based on the corresponding virtual mapping space established for the target range, the virtual mapping space includes an upper space, a flight space and a lower space in sequence from top to bottom;

[0046] S4, verifying the canopy image and the reference canopy image, and applying the hyperspectral remote sensing prescription map corresponding to the reference canopy image meeting the preset verification condition to the canopy of the current eucalyptus tree, so as to realize precision topdressing of the current eucalyptus tree;

[0047] In the above embodiment, by setting the upper space, the flight space and the lower space in the mapping space, the canopy image can be quickly matched and verified with the corresponding reference canopy image, so as to avoid incorrect topdressing or topdressing of an incorrect amount of topdressing for the current eucalyptus tree due to inaccurate positioning of the topdressing unmanned aerial vehicle, thereby facilitating precision topdressing of the current eucalyptus tree. The problem that it is inconvenient to perform precision topdressing on the canopy of each eucalyptus tree due to high planting density of eucalyptus trees in eucalyptus forests and complex terrain, and the topdressing unmanned aerial vehicle is prone to inaccurate positioning when topdressing the eucalyptus trees is solved.

[0048] In one embodiment, the step of segmenting the hyperspectral remote sensing image and the reference canopy image to obtain the hyperspectral remote sensing image and the reference canopy image corresponding to each eucalyptus tree comprises:

[0049] S11, determining the geographical range of the eucalyptus forest, and taking the geographical range of the eucalyptus forest as a target range;

[0050] S12, simultaneously collecting the hyperspectral remote sensing image and the reference canopy image of each eucalyptus tree in the eucalyptus forest in the target range by using a drone carrying a hyperspectral imager and an image collection device, wherein the hyperspectral remote sensing image and the reference canopy image of the eucalyptus tree collected at the same time are associated and bound to corresponding GPS position information;

[0051] S13, pre-processing the hyperspectral remote sensing image and the reference canopy image, and segmenting the hyperspectral remote sensing image and the reference canopy image by using an image segmentation algorithm to obtain the hyperspectral remote sensing image and the reference canopy image corresponding to each eucalyptus tree;

[0052] Specifically, in determining the target range, the geographical range of the eucalyptus forest is collected, the determined boundary data is accurately drawn on the electronic map through field investigation, a clear eucalyptus forest geographical range map is formed, and the accurate coordinate information of the range is recorded, including longitude, latitude and altitude, etc., so as to provide accurate basis for subsequent flight planning of unmanned aerial vehicles carrying hyperspectral imagers and image acquisition equipment and unmanned aerial vehicles for fertilization; in collecting hyperspectral remote sensing images and reference canopy images of the eucalyptus forest: a hyperspectral imager with a wavelength range of 400-1000 nm and a resolution of 3.5 nm is selected, and a high-definition digital camera is equipped as an image acquisition device, the hyperspectral imager and the image acquisition device are installed on the unmanned aerial vehicle, the power supply and data line are connected and fully debugged until the pre-set standard is reached; the flight route of the unmanned aerial vehicle is designed according to the geographical range and topography of the eucalyptus forest, to ensure that the route completely covers the target range, the flight parameters are set according to the eucalyptus height and equipment performance, such as flight speed control at 5-10 m / s, and the take-off point, landing point and safety height are set; the hyperspectral remote sensing images and the reference canopy images of the same eucalyptus tree are collected at the same time point, and the collected hyperspectral remote sensing images and the reference canopy images of the same eucalyptus tree are associated and bound with the GPS position information, so as to determine the corresponding reference canopy image and the corresponding hyperspectral remote sensing prescription map of the current eucalyptus tree according to the canopy image; the hyperspectral remote sensing images and the reference canopy images are pretreated: the hyperspectral remote sensing images are subjected to radiation correction and atmospheric correction, the radiation correction can eliminate the influence of the sensor's own characteristics on the images, so that the spectral reflectance of the images is more accurate, the atmospheric correction can remove the interference of atmospheric scattering and other factors on the spectral information of the images, and restore the true spectral characteristics of the ground objects, the reference canopy images are subjected to image enhancement such as brightness and contrast adjustment, to improve the clarity and readability of the images, for subsequent analysis and processing; when the hyperspectral remote sensing images and the reference canopy images are segmented by using an image segmentation algorithm: the Sobel, Canny and other edge detection operators are used to identify the boundary between the eucalyptus canopy and the background, so as to segment the hyperspectral remote sensing images and the reference canopy images corresponding to each eucalyptus tree, to obtain the hyperspectral remote sensing images and the reference canopy images corresponding to each eucalyptus tree without background (only the eucalyptus canopy exists), and then the segmented hyperspectral remote sensing images facilitate subsequent determination of the corresponding fertilizer amount of each eucalyptus tree, and the segmented reference canopy images facilitate subsequent matching and verification of the canopy images and the reference canopy images.

[0053] In one embodiment, the step of establishing the hyperspectral remote sensing prescription map of each eucalyptus tree based on the eucalyptus canopy nitrogen inversion model comprises:

[0054] S21, extract the hyperspectral reflectance curve of the eucalyptus canopy and construct a ground object endmember library, then use the ground object endmember library and the orthogonal subspace projection method to unmix the hyperspectral remote sensing image corresponding to each eucalyptus tree to obtain the hyperspectral information of each eucalyptus tree;

[0055] S22, establish a nitrogen inversion model of eucalyptus leaves, and determine a topdressing amount of each eucalyptus based on the nitrogen inversion model of eucalyptus leaves, and establish a hyperspectral remote sensing prescription map of each eucalyptus based on the topdressing amount of each eucalyptus;

[0056] Specifically, the reflectivity curves of the segmented hyperspectral remote sensing image are analyzed by using a spectral analysis software such as ENVI, the spectral characteristics of eucalyptus canopy leaves, branches, soil, and shadow are obtained by analyzing the reflectivity curves, and are classified and stored in the ground object endmember spectrum library as a basis for subsequent unmixing analysis; based on a linear mixing model, the ground object endmember spectrum library and the hyperspectral remote sensing image are input into a software such as MATLAB by means of matrix operation and projection transformation, and the hyperspectral information of each eucalyptus is separated out; eucalyptus leaf samples are obtained by field sampling, and the nitrogen content thereof is measured, and the spectral reflectivity curves of the corresponding leaves are extracted from the hyperspectral image, and a quantitative relationship model between the spectral reflectivity and the nitrogen content of the leaves is established by using a statistical analysis method such as multiple linear regression or partial least squares regression, so that the spectral information in the hyperspectral image can be converted into an estimated value of the nitrogen content, the nitrogen content of the eucalyptus canopy is inversed, and a nitrogen inversion model of the eucalyptus canopy leaves is obtained; by comparing the inversed nitrogen content with the ideal nitrogen content required for the growth of the eucalyptus, the nitrogen deficiency amount of each eucalyptus is calculated, and the topdressing amount thereof is determined, the determination of the topdressing amount can comprehensively consider the influence of factors such as the growth stage of the eucalyptus and soil fertility, so as to ensure the scientificity and rationality of the topdressing amount, and then the corresponding hyperspectral remote sensing prescription map of each eucalyptus is established according to the topdressing amount, so as to provide guidance for precise fertilization of the eucalyptus; further, the nitrogen information of the corresponding canopy of each eucalyptus can be extracted from the hyperspectral image, and a precise hyperspectral remote sensing prescription map is formulated for each eucalyptus accordingly.

[0057] In one embodiment, the step of establishing a virtual mapping space based on the target range comprises:

[0058] S31, one-to-one corresponding associate the hyperspectral remote sensing prescription map with all reference canopy images in the target range and distribute them according to the GPS position information to constitute the upper space;

[0059] S32, the three-dimensional airspace in which the fertilization unmanned aerial vehicle flies in the target range is taken as the flight space; the canopy image of the current eucalyptus is collected by using the fertilization unmanned aerial vehicle, and the lower space in which the canopy image is stored is established and stored for the fertilization unmanned aerial vehicle, so as to obtain the virtual mapping space constituted by the upper space, the flight space, and the lower space;

[0060] Specifically, the upper space and the lower space can be cloud servers, the upper space contains reference canopy images corresponding to all eucalyptus trees in the eucalyptus forest and their associated hyperspectral remote sensing prescription maps and GPS position information; the flight space is a three-dimensional airspace for physical flight of the unmanned aerial vehicle, mainly used for navigation; the lower space is replaced by the new canopy image corresponding to the new eucalyptus tree to realize updating; when the unmanned aerial vehicle for fertilization fertilizes, the unmanned aerial vehicle for fertilization is positioned by GPS in the flight space, and the canopy image of the current eucalyptus tree is collected (the lower space is updated at the same time), the corresponding reference canopy image in the upper space is determined according to the current GPS position information of the current eucalyptus tree collected by the unmanned aerial vehicle for fertilization, and then each sub-region of the canopy image and the reference canopy image is matched and verified through the transmission verification channel, if the preset verification condition is met, the unmanned aerial vehicle for fertilization fertilizes the canopy of the current eucalyptus tree according to the fertilizer amount of the reference canopy image associated with the hyperspectral remote sensing prescription map, so as to avoid the error of fertilizing the current eucalyptus tree or applying the corresponding fertilizer amount due to inaccurate positioning of the unmanned aerial vehicle for fertilization, thereby facilitating precise fertilization of the current eucalyptus tree.

[0061] In one embodiment, the step of verifying the canopy image with the reference canopy image comprises:

[0062] S41, establishing a reference basic region with the center point of the eucalyptus canopy in the reference canopy image corresponding to the current GPS position information as the center, dividing the reference basic region into: a plurality of sub-regions formed by the intersection between a plurality of concentric rings and eight azimuth sector blocks, taking the division mode of the reference basic region of the reference canopy image as the division mode of the fixed basic region in the lower space, thereby establishing the fixed basic region in the lower space, the fixed basic region is unique, and the sub-region formed by the fixed basic region and the reference basic region in the same concentric ring and the same azimuth sector block is regarded as a homologous region;

[0063] S42, establishing a feature code library, extracting first feature information in a plurality of sub-regions of the reference canopy image based on the feature code library for coding and filling into the sub-regions of the corresponding reference basic region, and extracting second feature information in a plurality of sub-regions of the canopy image based on the feature code library for coding and filling into the sub-regions of the fixed basic region;

[0064] S43, setting a reference port point and a detection port point corresponding to the fixed basic region in the lower space, and matching and verifying the canopy image of the current eucalyptus tree collected by the unmanned aerial vehicle for fertilization with the reference canopy image in the upper space based on the reference port point and the detection port point;

[0065] Specifically, the feature code library is used to store various feature information in the first feature information and the second feature information, and the corresponding extent range and the corresponding code of each feature information, for example, when the feature information is the branch thickness, the code corresponding to the branch thickness less than 3cm is b1; in the reference basic area: a plurality of concentric rings take the center point of the eucalyptus crown layer in the reference crown layer image as the center, and the radius difference of the plurality of concentric rings is the same, and the radius difference is diffused from the center to the edge of the eucalyptus crown layer; eight azimuth sectors correspond to east, south, west, north, southeast, northeast, southwest and northwest respectively; through the establishment of the fixed basic area and the reference basic area, the reference crown layer image can be matched with each sub-area divided in the crown layer image, through the setting of the homologous area, the reference port point set by the fixed basic area can project the detection port point to the corresponding sub-area of the reference basic area, through the matching verification of the code carried by the detection port point and the code of the corresponding sub-area of the reference crown layer image, it can be determined whether the code carried by the detection port point and the code of the corresponding sub-area of the reference crown layer image are the same, through the matching verification of the codes carried by a plurality of detection port points and the codes of the corresponding sub-areas of the reference crown layer image, it can be determined whether the reference crown layer image meets the preset verification condition, so as to determine the reference crown layer image corresponding to the crown layer image, so as to determine the current eucalyptus fertilizer amount according to the corresponding reference crown layer image.

[0066] In one embodiment, the step of matching the crown layer image of the current eucalyptus collected by the fertilizer unmanned aerial vehicle with the reference crown layer image in the upper space based on the reference port point and the detection port point, comprises:

[0067] S431, the fixed basic area in the lower space is divided according to the reference basic area division mode of the reference crown layer image, and the reference port point is configured one by one in a plurality of sub-areas in the fixed basic area, and a transmission verification channel is established between each reference port point and the detection port point associated with the fixed;

[0068] S432, the crown layer image currently collected by the fertilizer unmanned aerial vehicle is adapted to the fixed basic area in the lower space, the reference port point obtains the code of the sub-area, and the code is provided to the detection port point for carrying;

[0069] S433, according to the homologous area, the detection port point corresponding to each reference port point is projected to the sub-area of the reference crown layer image corresponding to the current GPS position information at the same time, and the detection port point detects the code of the sub-area of the reference crown layer image;

[0070] S434, the code carried by the detection port point is matched with the code of the corresponding sub-area.

[0071] In one embodiment, the step of applying the hyperspectral remote sensing prescription image corresponding to the reference canopy image meeting the preset verification condition to the canopy of the current eucalyptus to achieve precise topdressing of the current eucalyptus comprises:

[0072] If the code carried by any of the detection port points is the same as the code of the corresponding sub-region of the reference canopy image, the detection port point is withdrawn from the corresponding sub-region of the reference canopy image;

[0073] If the code carried by any of the detection port points is different from the code of the corresponding sub-region of the reference canopy image, all the detection port points are not withdrawn and are projected one by one to the sub-regions of other reference canopy images within the offset radius corresponding to the outside of the reference canopy image and are matched until all the detection port points are withdrawn;

[0074] When all the detection port points are withdrawn, it means that the reference canopy images corresponding to all the detection port points meet the preset verification condition, and the fertilization unmanned aerial vehicle sprays the topdressing amount of the hyperspectral remote sensing prescription image associated with the reference canopy image to the canopy of the current eucalyptus, that is, precise topdressing of the current eucalyptus is achieved;

[0075] Specifically, in the process of adapting the canopy image to the fixed base area in the lower space, the adaptation method adopted is based on the orientation of the canopy image collected by the unmanned aerial vehicle for fertilization of the current eucalyptus, so that the canopy image can be adapted and filled to the fixed base area in the same orientation as the reference canopy image, and registration between the two corresponding sub-regions during matching verification can be avoided; the deviation radius, i.e. the distance of historical GPS positioning deviation, is generally between 10-40 meters; the same division method is used for the fixed base area as the reference base area, and this same division method makes the sub-regions of the reference canopy image and the canopy image have the same structure, facilitating the matching of each sub-region in the canopy image and the reference canopy image; through the setting of the same position area, the reference port point can be quickly and accurately projected to the sub-region in the reference base area through the fixed associated probe port point; when all the probe port points carry the same code as the corresponding sub-region of the reference canopy image, they are all collected, so as to determine that the reference canopy image meets the preset verification condition; when the codes are different, the probe port points are not collected, and all the probe port points are one-to-one projected to the sub-regions of other reference canopy images within the deviation radius outside the reference canopy image and are matched and verified, so that the probe port points corresponding to the fixed base area can flexibly match and verify the sub-regions of other reference canopy images within the deviation radius, so as to facilitate the foliar fertilization of the current eucalyptus through the trace fertilization amount corresponding to the hyperspectral remote sensing prescription of the associated other reference canopy image, so as to realize the precise trace fertilization of the current eucalyptus by the unmanned aerial vehicle for fertilization. In addition, by having only one fixed base area in the lower space, the data storage amount in the lower space can be effectively reduced. At the same time, since the reference base area sub-region of each reference canopy image is not configured with a communication port, the data amount and communication resources of the corresponding sub-region of each reference canopy image in the upper space are reduced, thereby saving the storage space of the upper space. When matching and verifying the canopy image and the reference canopy image in the upper space, by means of the reference port points of the fixed base area, the probe port points fixedly associated and carrying the corresponding sub-region code can reduce the data transmission amount and reduce the occupation of communication resources. Fig. 2 After the canopy image is collected through the current GPS position information of the unmanned aerial vehicle for fertilization, the reference canopy image corresponding to the current GPS position information is determined according to the current GPS position information, and the code of the feature information in the first feature information is filled into the sub-region of the corresponding base area of the reference canopy image (such as a2, b1, such as a1).

[0076] In one embodiment, in the feature code library:

[0077] The first feature information and the second feature information both include feature information of the thickness of the branches and trunks of the eucalyptus canopy and the density of the leaves of the eucalyptus canopy;

[0078] The feature degree of each feature information corresponds to a code in the code library in one-to-one correspondence;

[0079] Specifically, the branch edge is extracted using a Canny edge detection algorithm, the branch connectivity is enhanced through morphological operations of dilation and erosion, the findContours function of OpenCV is used to extract the branch contour, and the area and bounding box of the contour are calculated to determine the thickness of the branch, for example, the code b1 corresponds to a branch with a thickness of less than 3 cm, the code b2 corresponds to a branch with a thickness of 3-6 cm, and so on; the sub-region growing algorithm is used to separate the eucalyptus leaf region from the background, the gray level co-occurrence matrix (GLCM) is used to extract the leaf texture feature, and the leaf density is obtained; each feature information corresponds to multiple feature degrees, and in the eucalyptus canopy leaf density: the code a1 corresponds to a leaf coverage rate of less than 20% in unit area (extremely sparse), the code a2 corresponds to a leaf coverage rate of 20%-50% in unit area (sparse), the code a3 corresponds to a leaf coverage rate of 50%-80% in unit area (dense), and the code a4 corresponds to a leaf coverage rate of more than 80% in unit area (extremely dense), wherein the range of the degree of each feature information corresponds to a code in the feature code library.

[0080] Those skilled in the art can appreciate that the units and algorithm steps of the examples described in combination with the embodiments disclosed in the present application can be realized in electronic hardware or a combination of computer software and electronic hardware. Whether the functions are realized in hardware or software depends on the specific application and design constraints of the technical solution. A person skilled in the art can use different methods to realize the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0081] In several embodiments provided by the present application, it should be understood that the disclosed system, device and method can be implemented in other ways. For example, the device embodiments described above are only schematic, for example, the division of the units is only one, and actual implementation can have another division manner, for example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the units shown or discussed can be indirect coupling or communication connection through some interface, device or unit, and can be electrical, mechanical or other forms.

[0082] The above description is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto, and any person skilled in the art can easily think of changes or replacements within the technical range disclosed by the present application, which should be covered within the protection scope of the present application.

[0083] Finally: the above only for the preferred embodiments of the present application, and not for limiting the present application, any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application, should be included in the scope of protection of the present application.

Claims

1. An eucalyptus canopy unmanned aerial vehicle precision topdressing method based on a hyperspectral remote sensing prescription map, characterized in that, The method comprises the following steps: Meanwhile, the hyperspectral remote sensing image and the reference crown layer image of the target range are collected, and the hyperspectral remote sensing image and the reference crown layer image are segmented to obtain the hyperspectral remote sensing image and the reference crown layer image corresponding to each eucalyptus tree; The hyperspectral remote sensing image corresponding to each eucalyptus tree is unblended, and a eucalyptus crown layer nitrogen inversion model is established, and a hyperspectral remote sensing prescription map of each eucalyptus tree is established based on the eucalyptus crown layer nitrogen inversion model; The crown layer image of the current eucalyptus tree is collected by using the unmanned aerial vehicle for fertilization, and a virtual mapping space is established corresponding to the target range, wherein the virtual mapping space comprises an upper space, a flight space and a lower space from top to bottom; The crown layer image and the reference crown layer image are verified, and the hyperspectral remote sensing prescription map corresponding to the reference crown layer image meeting the preset verification condition is applied to the crown layer of the current eucalyptus tree, so as to realize accurate topdressing of the current eucalyptus tree.

2. The eucalyptus canopy unmanned aerial vehicle precision topdressing method based on a hyperspectral remote sensing prescription map according to claim 1, characterized in that, The step of segmenting the hyperspectral remote sensing image and the reference crown layer image to obtain the hyperspectral remote sensing image and the reference crown layer image corresponding to each eucalyptus tree comprises: The geographical range of the eucalyptus forest is determined, and the geographical range of the eucalyptus forest is taken as the target range; The hyperspectral remote sensing image and the reference crown layer image of each eucalyptus tree in the eucalyptus forest in the target range are collected simultaneously by using the unmanned aerial vehicle carrying the hyperspectral imager and the image collection device, wherein the hyperspectral remote sensing image and the reference crown layer image of the eucalyptus tree collected at the same time are associated and bound with the corresponding GPS position information; The hyperspectral remote sensing image and the reference crown layer image are preprocessed, and the hyperspectral remote sensing image and the reference crown layer image are segmented by using an image segmentation algorithm to obtain the hyperspectral remote sensing image and the reference crown layer image corresponding to each eucalyptus tree.

3. The eucalyptus canopy unmanned aerial vehicle precision topdressing method based on a hyperspectral remote sensing prescription map according to claim 2, characterized in that, The step of establishing the hyperspectral remote sensing prescription map of each eucalyptus tree based on the eucalyptus crown layer nitrogen inversion model comprises: The hyperspectral reflectance curve of the eucalyptus crown layer is extracted, and a ground feature endmember wave library is constructed, then the hyperspectral remote sensing image corresponding to each eucalyptus tree is unblended by using the ground feature endmember wave library and an orthogonal subspace projection method to obtain the hyperspectral information of each eucalyptus tree; A nitrogen inversion model of eucalyptus leaves is established, and the topdressing amount of each eucalyptus tree is determined based on the nitrogen inversion model of eucalyptus leaves, and the hyperspectral remote sensing prescription map of each eucalyptus tree is established corresponding to the topdressing amount of each eucalyptus tree.

4. The eucalyptus canopy unmanned aerial vehicle precision topdressing method based on a hyperspectral remote sensing prescription map according to claim 3, characterized in that, The step of establishing the virtual mapping space corresponding to the target range comprises: All the reference crown layer images in the target range are one-to-one associated with the hyperspectral remote sensing prescription map and distributed according to the GPS position information to form the upper space; The three-dimensional space where the unmanned aerial vehicle for fertilization flies in the target range is taken as the flight space, and the lower space storing the crown layer image of the current eucalyptus tree is established corresponding to the unmanned aerial vehicle for fertilization, so as to obtain the virtual mapping space composed of the upper space, the flight space and the lower space.

5. The precision topdressing method for eucalypt crown layer based on hyperspectral remote sensing prescription map according to claim 4, characterized in that, The step of verifying the crown layer image and the reference crown layer image comprises: A reference basic region is established with the center point of the eucalyptus canopy in the reference canopy image corresponding to the current GPS position information as the center, the reference basic region is divided into a plurality of sub-regions formed by the intersection of a plurality of concentric rings and eight azimuth sector blocks, the division mode of the reference basic region of the reference canopy image is used as the division mode of the fixed basic region in the lower space, and the fixed basic region in the lower space is established, the fixed basic region is unique, and the sub-region formed by the intersection of the fixed basic region and the reference basic region in the same concentric ring and the same azimuth sector block is regarded as a homologous region; A feature code library is established, first feature information in a plurality of sub-regions in the reference canopy image is extracted based on the feature code library, encoded and filled into the sub-regions of the corresponding reference basic region, and second feature information in a plurality of sub-regions in the canopy image is extracted based on the feature code library, encoded and filled into the sub-regions of the fixed basic region; A reference basic region is established with the center point of the eucalyptus canopy in the reference canopy image corresponding to the current GPS position information as the center, the reference basic region is divided into a plurality of sub-regions formed by the intersection of a plurality of concentric rings and eight azimuth sector blocks, the division mode of the reference basic region of the reference canopy image is used as the division mode of the fixed basic region in the lower space, and the fixed basic region in the lower space is established, the fixed basic region is unique, and the sub-region formed by the intersection of the fixed basic region and the reference basic region in the same concentric ring and the same azimuth sector block is regarded as a homologous region; 6. The eucalyptus canopy unmanned aerial vehicle precision topdressing method based on a hyperspectral remote sensing prescription map according to claim 5, characterized in that, The step of matching and verifying the canopy image of the current eucalyptus collected by the unmanned aerial vehicle for fertilization with the reference canopy image in the upper space based on the reference port point and the detection port point comprises: The fixed basic region in the lower space is divided according to the reference basic region division mode of the reference canopy image, the reference port point is configured in the plurality of sub-regions of the fixed basic region one by one, and a transmission verification channel is established between each reference port point and the detection port point associated with the fixed basic region; The canopy image currently collected by the unmanned aerial vehicle for fertilization is adapted to the fixed basic region in the lower space, the reference port point obtains the code of the sub-region, and the code is provided to the detection port point for carrying; According to the homologous region, the detection port point corresponding to each reference port point is simultaneously projected to the sub-region of the reference canopy image corresponding to the current GPS position information, and the detection port point detects the code of the sub-region of the reference canopy image; The code carried by the detection port point is matched and verified with the code of the corresponding sub-region.

7. The eucalyptus canopy unmanned aerial vehicle precision topdressing method based on a hyperspectral remote sensing prescription map according to claim 6, characterized in that, The step of applying the hyperspectral remote sensing prescription image corresponding to the reference canopy image meeting the preset verification condition to the canopy of the current eucalyptus to realize precise topdressing of the current eucalyptus comprises: If the code carried by any detection port point is the same as the code of the corresponding sub-region of the reference canopy image, the detection port point is withdrawn from the corresponding sub-region of the reference canopy image; If the code carried by any detection port point is different from the code of the corresponding sub-region of the reference canopy image, all detection port points are not withdrawn and are projected one by one to the sub-regions of other reference canopy images within the offset radius outside the reference canopy image and are matched until all detection port points are withdrawn; When all detection port points are withdrawn, it is indicated that the reference canopy images corresponding to all detection port points meet the preset verification condition, and the unmanned aerial vehicle for fertilization sprays the topdressing amount of the hyperspectral remote sensing prescription image associated with the reference canopy image to the canopy of the current eucalyptus, thereby realizing precise topdressing of the current eucalyptus.

8. The eucalyptus canopy unmanned aerial vehicle precision topdressing method based on a hyperspectral remote sensing prescription map according to claim 7, characterized in that, In the feature code library: The first feature information and the second feature information both include feature information of thickness of branches and trunks of the eucalyptus canopy and density of leaves of the eucalyptus canopy; The feature degree of each feature information one-to-one corresponds to a code in the feature code library.

Images