CN-122024050-A - Multi-level investigation method for desert vegetation of single-voyage unmanned aerial vehicle

CN122024050ACN 122024050 ACN122024050 ACN 122024050ACN-122024050-A

Abstract

The invention discloses a multilayer investigation method for desert vegetation of a single-voyage unmanned aerial vehicle, and belongs to the technical field of ecological environment monitoring and remote sensing. The invention provides a method for synchronously collecting visible light and multispectral images through single-voyage flight, which ensures that data are highly consistent in time, space and attitude parameters, reduces multisource data registration errors from sources, provides a basis for realizing high-precision fusion, simultaneously provides a digital elevation model generation method based on spectrum index mask constraint, can automatically and accurately separate vegetation points and ground points in a sparse vegetation area, thereby eliminating false terrain lifting and obtaining high-fidelity terrain data, and further provides an ecological index inversion method for individual-plaque-landscape progressive aggregation, which realizes hierarchical association and collaborative expression of different scale indexes, thereby solving the problem of index dispersion and improving the structuring degree of a sample investigation data set.

Inventors

DING ZIFAN
WANG XUSHENG
HAN PENGFEI
BAI TAIYA

Assignees

中国地质大学（北京）

Dates

Publication Date: 20260512
Application Date: 20260126

Claims (10)

1. A multilayer investigation method for desert vegetation of a single-voyage unmanned aerial vehicle is characterized by comprising the following steps: S1, determining a sample range of a region to be measured, arranging ground control points on the ground surface in a target region, and measuring accurate coordinates of the ground control points; S2, controlling an unmanned aerial vehicle aerial survey platform carrying a high-precision positioning device, a visible light imaging device and a multispectral imaging device, executing single aerial survey flight, and synchronously collecting multi-view RGB images of a target area and multispectral images containing red light and near infrared bands; S3, based on the image information acquired in the S2, completing three-dimensional reconstruction by adopting a motion recovery structure to generate dense point cloud, and obtaining a visible light grid, a red light wave band reflectivity grid, a near infrared wave band reflectivity grid and a high resolution digital surface model grid through orthographic correction; S4, calculating a vegetation index grid based on the red-light wave band reflectivity grid and the near-infrared wave band reflectivity grid; S5, calculating a threshold value of the bare soil pixel by adopting an image segmentation method, comparing the vegetation index grid with the obtained threshold value to determine a bare soil area, and further obtaining a vegetation coverage mask for distinguishing vegetation and the bare soil area; S6, extracting pixel elevation values marked as bare soil areas from the high-resolution digital surface model grid by using a vegetation cover mask to serve as ground surface height Cheng Yangben points, performing spatial interpolation based on the ground surface height Cheng Yangben points, and reconstructing to obtain a digital elevation model grid of the target area; s7, performing pixel-by-pixel differential calculation on the high-resolution digital surface model grid and the digital elevation model grid to obtain a vegetation canopy height model grid; s8, distinguishing the preparation types of the target area based on a supervision classification method by taking the vegetation index grid as an input characteristic to obtain a vegetation type classification grid; s9, calculating a landscape pattern index based on the vegetation type classification grid.
2. The method of claim 1, wherein the single-voyage flight course in S2 is designed as a "m" type intersecting course, including a "+" type orthographic course along the edge of the sample plot and an "x" type oblique course disposed at an angle to the orthographic course.
3. The method for multi-level investigation of desert vegetation by a single-voyage unmanned aerial vehicle according to claim 1, wherein the vegetation index in S4 is a soil-adjusted vegetation index or a normalized vegetation index; the calculation formula of the soil adjustment vegetation index is as follows: wherein NIR represents the reflectivity of the near infrared band, RED represents the reflectivity of the RED light band, L represents the soil adjustment parameter; the calculation formula of the normalized vegetation index is as follows: 。
4. the method for multi-level investigation of desert vegetation by a single-voyage unmanned aerial vehicle according to claim 1, wherein the image segmentation method in S5 is adaptive threshold segmentation based on the oxford method, a classification model based on supervised learning or a classification method based on unsupervised clustering.
5. The method for multi-level investigation of desert vegetation by a single-voyage unmanned aerial vehicle according to claim 1, wherein the method for spatial interpolation in S6 is inverse distance weighted interpolation, common kriging interpolation, spline interpolation or triangular net linear interpolation.
6. The method for multi-level investigation of desert vegetation by a single-voyage unmanned aerial vehicle according to claim 1, wherein in S7, after calculating the vegetation canopy height model grid, the pixel value in which the elevation value is less than 0 is set to 0.
7. The method for multi-level investigation of desert vegetation of a single-voyage unmanned aerial vehicle according to claim 1, wherein the supervision and classification method in S8 is a random forest, a support vector machine or a gradient lifting decision tree model, and the input features further comprise a visible light grid, a red light wave band reflectivity grid and/or a near infrared wave band reflectivity grid.
8. The method of claim 1, wherein the landscape pattern index in S9 comprises at least one of shannon diversity index, aggregation index and landscape shape index.
9. A computer device comprising a processor and a memory, wherein the memory stores at least one instruction, at least one program, code set, or instruction set, the instruction, program, code set, or instruction set being loaded and executed by the processor to implement a method of multi-level investigation of desert vegetation for a single voyage unmanned aerial vehicle according to any of claims 1 to 8.
10. A computer readable storage medium having stored therein at least one instruction, at least one program, code set or instruction set loaded and executed by a processor to implement a single-voyage unmanned aerial vehicle desert vegetation multi-level investigation method according to any of claims 1-8.

Description

Multi-level investigation method for desert vegetation of single-voyage unmanned aerial vehicle Technical Field The invention belongs to the technical field of ecological environment monitoring and remote sensing, and particularly relates to a multilayer investigation method for desert vegetation of a single-voyage unmanned aerial vehicle, which is suitable for multi-scale vegetation structure and landscape pattern investigation of a desert grassland and other land vegetation ecosystems with low vegetation coverage arid regions. Background The vegetation ecosystem in the desert grasslands and arid regions has the characteristics of low vegetation coverage, sparse species composition, staggered distribution of vegetation patches and bare lands and the like, and the traditional vegetation sample investigation mainly relies on manual measurement and point scale observation, so that the problems of low efficiency, high working strength, insufficient space representativeness and the like exist, and the requirements of regional scale and multi-scale ecological monitoring are difficult to meet. With the development of unmanned aerial vehicle remote sensing technology, high-resolution visible light images and multispectral images based on unmanned aerial vehicles are widely applied to vegetation investigation. In the prior art, one type of method calculates a vegetation index, such as a normalized vegetation index (NDVI), based on multispectral orthographic images for estimating plaque scale indexes such as vegetation coverage, biomass and the like, and the other type of method utilizes a multi-view visible light photo to reconstruct a three-dimensional point cloud and a digital surface model for estimating individual scale structure information such as canopy height and the like. There are also some studies to analyze vegetation diversity on a landscape scale by identifying vegetation types. In connection with the above discussion, the prior art has mainly the following drawbacks: (1) The multi-source data are acquired respectively through multiple flights, and the visible light and the multi-spectrum image are inconsistent in time, space and attitude parameters, so that the subsequent registration error is larger, and high-precision multi-source fusion is difficult to realize; (2) In a sparse vegetation area, the traditional photogrammetry method cannot effectively distinguish vegetation points from ground points, so that a generated digital elevation model has systematically raised false terrains at vegetation positions. (3) The index system is fragmented and lacks systematic fusion, the existing method is mainly focused on acquiring single or few parameters (such as vegetation classification or coverage calculation), and each parameter is independently produced and lacks synergy. This results in the end result being a series of "static layers" of tiles, rather than a dynamically associated, logically consistent ecological index system. In order to solve the problems, the invention provides a multilayer investigation method for desert vegetation of a single-voyage unmanned aerial vehicle. Disclosure of Invention The invention aims to provide a multi-level investigation method for desert vegetation of a single-voyage unmanned aerial vehicle, which aims to solve the technical problems that in the prior art, data registration errors are large due to multi-voyage or multi-source data acquisition, multi-scale information of the desert vegetation is difficult to acquire in the same time phase, and a Digital Elevation Model (DEM) is easily interfered by vegetation lifting under a sparse vegetation condition, so that vegetation height inversion and index system construction are influenced. In order to achieve the above purpose, the invention adopts the following technical scheme: A multilayer investigation method for desert vegetation of single-voyage unmanned aerial vehicle comprises the following steps: S1, determining a sample range of a region to be measured, arranging ground control points on the ground surface in a target region, and measuring accurate coordinates of the ground control points; S2, controlling an unmanned aerial vehicle aerial survey platform carrying a high-precision positioning device, a visible light imaging device and a multispectral imaging device, executing single aerial survey flight, and synchronously collecting multi-view RGB images of a target area and multispectral images containing red light and near infrared bands; S3, based on the image information acquired in the S2, adopting a motion recovery structure (Structure from Motion, sfM) to complete three-dimensional reconstruction and generate dense point cloud, and obtaining a visible light (RGB) grid, a RED light wave band Reflectivity (RED) grid, a near infrared wave band reflectivity (NIR) grid and a high-resolution Digital Surface Model (DSM) grid through orthographic correction; S4, calculating a vegetation index grid based on a RED