CN118230200A - Tea garden high-temperature drought composite disaster investigation method based on unmanned aerial vehicle - Google Patents

Abstract

The invention discloses a tea garden high-temperature drought composite disaster investigation method based on an unmanned aerial vehicle. And then extracting the image element of the damaged tea tree under the high-temperature drought composite condition based on the unmanned aerial vehicle image. And then constructing a tea garden damage severity index model under the high-temperature drought composite condition according to the damaged tea tree pixels, and calculating the tea garden damage severity index under different spatial resolutions. Finally, different spatial resolution tea garden damage severity indexes based on unmanned aerial vehicle image high-temperature drought composite conditions are obtained, a pixel severity index chart is drawn, and drought composite disaster investigation is completed. The invention realizes quantitative evaluation of the high-temperature drought composite disasters in the tea garden, and improves the scientificity, repeatability and comparability of investigation results.

Description

Tea garden high-temperature drought composite disaster investigation method based on unmanned aerial vehicle

Technical Field

The invention belongs to the field of unmanned aerial vehicle image acquisition and processing, machine learning and agronomic application intersection, and particularly relates to a tea garden high-temperature drought composite disaster investigation method based on an unmanned aerial vehicle.

Background

The traditional monitoring of the high-temperature drought event of the tea tree mainly depends on the human power census or the analysis of meteorological monitoring data, the human power census consumes time and labor, the meteorological monitoring data depend on meteorological sites, and the difference of each area cannot be comprehensively reflected. The unmanned aerial vehicle has the characteristics of accuracy, flexibility and high efficiency, and can carry on different sensors to acquire various high-spatial-resolution remote sensing data. With the progress of science and technology, unmanned aerial vehicle remote sensing systems, whether hardware devices or information processing methods, develop rapidly, and are widely applied to industries such as mapping, glacier, agriculture, forestry, vegetation drawing, electric power, atmosphere detection, geological disasters and the like. However, the research on the extraction of the disaster information of the tea garden under the high-temperature drought stress by utilizing the unmanned aerial vehicle remote sensing technology is not reported, the development of the disaster investigation technology of the tea garden under the high-temperature drought composite disaster based on the unmanned aerial vehicle is urgently needed, the disaster information under the high-temperature drought composite stress is timely, objectively and accurately obtained, and the basis is provided for the management, the reconstruction after disaster and the insurance claim of the tea garden.

The method specifically comprises the steps of acquiring and preprocessing data of the unmanned aerial vehicle in the tea garden under high-temperature drought composite stress, manufacturing a label based on SEGMENT ANYTHING unmanned aerial vehicle images, extracting and checking precision of a tea tree victim pixel based on the unmanned aerial vehicle images under high-temperature drought composite stress, constructing a tea garden victim severity index model under high-temperature drought composite stress, and drawing tea garden victim severity indexes (Compound drought-HEATWAVE DAMAGED SEVERITY index, CDH_DSI) under different spatial scales of unmanned aerial vehicle images under high-temperature drought composite stress to form a complete unmanned aerial vehicle-based tea garden high-temperature drought composite disaster investigation technical system, so that unmanned aerial vehicle can be used alone for tea garden unmanned aerial vehicle tea garden high-temperature drought composite disaster monitoring, and satellite remote sensing monitoring can be verified by utilizing unmanned aerial vehicle monitoring results.

Disclosure of Invention

The invention aims at solving the problems of small area and small area of the tea garden land, and because most tea gardens are positioned in hilly mountain areas and are easily subjected to high-temperature drought compound stress, a rapid, accurate, even and efficient disaster investigation method is urgently needed to develop, so that the problems of time and labor waste, high labor intensity, low efficiency and low precision of the traditional method are solved, and the unmanned-plane-based tea garden high-temperature drought compound disaster investigation method is provided.

The invention comprises the following steps: acquiring and preprocessing unmanned aerial vehicle data of a tea garden under high-temperature drought compound stress; labeling based on SEGMENT ANYTHING (dividing everything) unmanned aerial vehicle images; extracting a tea tree damage pixel and checking precision under high-temperature drought compound stress based on unmanned aerial vehicle images; constructing a tea garden damage severity index model under high-temperature drought compound stress; the tea garden CDH_DSI drawing of the unmanned aerial vehicle images with different spatial scales forms a set of complete unmanned aerial vehicle-based tea garden high-temperature drought composite disaster monitoring technical system, so that the unmanned aerial vehicle can be used for carrying out tea garden unmanned aerial vehicle high-temperature drought composite disaster monitoring, and satellite remote sensing monitoring can be verified by utilizing unmanned aerial vehicle monitoring results. The extraction result can be used as a basis for recovery after the disaster of tea farmers and accurate management of water and fertilizer in tea gardens, and a basis is provided for disaster loss assessment and damage assessment of insurance companies.

A tea garden high-temperature drought composite disaster investigation method based on unmanned aerial vehicle comprises the following steps:

and step 1, acquiring image data of the unmanned aerial vehicle under the high-temperature drought compound condition of the tea garden and preprocessing the image data.

And 2, label making based on SEGMENT ANYTHING unmanned aerial vehicle images, which is used for model training and accuracy verification.

And 3, extracting a tea tree damage pixel based on the high-temperature drought composite condition of the unmanned aerial vehicle image, and calculating severity indexes of the tea garden damage under different spatial resolutions.

And 4, constructing a tea garden damage severity index model under the high-temperature drought composite condition according to the damaged tea tree pixels, and calculating the tea garden damage severity index under different spatial resolutions.

And 5, acquiring severity indexes of damage of tea gardens with different spatial resolutions based on unmanned aerial vehicle image high-temperature drought composite conditions, drawing a pixel severity index chart, and completing drought composite disaster investigation.

The following is a preferred technical solution of the present invention:

In step 1, the method comprises the following steps: the method comprises the steps of obtaining meteorological element conditions such as local temperature, precipitation and soil moisture content through a local meteorological department, obtaining management measures such as main cultivation varieties, fertilization pruning and harvesting and tea garden distribution through a local agricultural department, obtaining damage conditions of a tea garden subjected to high-temperature drought composite disasters through tea enterprises and tea farmers, determining a representative area, designing unmanned aerial vehicle flight range and route, and obtaining unmanned aerial vehicle photos of healthy and damaged tea gardens. And splicing by using the acquired unmanned aerial vehicle photo and professional software to obtain the regional unmanned aerial vehicle image.

In the step 2, the unmanned aerial vehicle image finished in the step 1 is utilized to be segmented into sub-images with different spatial resolutions, and the sub-images are selected to manufacture labels of healthy tea trees, damaged tea trees, bare soil, water, buildings and the like. Because of huge workload, the human-computer interaction interpretation is performed by adopting SEGMENT ANYTHING algorithm, and finally the healthy tea tree pixels, the victim tea tree pixels, the bare soil pixels, the water body pixels and the building pixels are obtained.

And 3, dividing training data and verification data by using the healthy tea tree, the damaged tea tree, bare soil, the water body, the building label and the corresponding unmanned aerial vehicle sub-image which are finished in the step 2.

And (3) extracting healthy tea trees, damaged tea trees, bare soil, water bodies and buildings by adopting a plurality of machine learning algorithms, selecting the algorithm with the highest precision as a damaged tea tree pixel extraction model according to an evaluation result, and extracting the damaged tea tree pixels of all unmanned aerial vehicle images.

In step 4, the method comprises the steps of calculating the number of the victim tea tree pixels in the sub-images with different spatial resolutions by using the victim tea tree pixels extracted in step 3, dividing the number of the victim tea tree pixels in the corresponding sub-images by the number of all the pixels in the corresponding sub-images, and calculating:

wherein, DTP is victim tea tree pixel quantity, and R represents spatial resolution, and the unit is meter, and Ru is unmanned aerial vehicle image spatial resolution, and the unit is meter.

In the step 5, the image of the unmanned aerial vehicle segmented corresponding to satellite data with different spatial resolutions is calculated by utilizing the image element of the victim tea tree extracted in the step 3 and the severity index of the victim in the tea garden defined in the step 4, and CDH_DSI is calculated, so that CDH_DSI drawing based on the image of the unmanned aerial vehicle is realized.

Compared with the prior art, the invention has the advantages that:

1. From point to face: at present, no report related to the investigation of the high-temperature drought composite disasters of the tea garden is developed by using an unmanned aerial vehicle, and the traditional investigation of the high-temperature drought composite disasters usually adopts artificial ground investigation, so that a plurality of lines run at most, and the disaster condition of the area cannot be obtained.

2. The quantitative method can be used for: the traditional artificial ground investigation is strong in subjectivity and lacks quantitative indexes, and the technology realizes quantitative evaluation of the high-temperature drought composite disasters of the tea garden by defining the severity index of the damage of the tea garden under the high-temperature drought composite stress.

3. Objective repeatable: the traditional artificial ground investigation is easy to influence professional background, experience knowledge and even benefit of investigation staff because the judgment of disaster situation is different, different investigation staff investigation results are quite different, and the unmanned aerial vehicle images are utilized to have true graphs. The extraction result can be used as a basis for recovery after the disaster of tea farmers and accurate management of water and fertilizer in tea gardens, and a basis is provided for disaster loss assessment and damage assessment of insurance companies.

Drawings

FIG. 1 is a photograph of an unmanned aerial vehicle under the high-temperature drought compound stress obtained by Zhejiang, and a pretreatment and splicing flow thereof;

FIG. 2 is a representation of a health and victim tea label made using SEGMENT ANYTHING using a unmanned aerial vehicle sub-image, providing data for model training and verification in accordance with the present invention;

fig. 3 shows a healthy (light) tea garden and a damaged (dark) tea garden extracted by a method XGBoost finally determined based on unmanned aerial vehicle images by comparing different methods, and the extraction result is utilized to calculate the severity index of the disaster under the high-temperature drought compound stress of the tea garden.

Detailed Description

The invention will be further described with reference to the following specific drawings and examples. The invention relates to a tea garden high-temperature drought composite disaster investigation technology based on an unmanned aerial vehicle, which comprises the following steps of:

Step 1, investigating and determining a representative tea garden subjected to high-temperature drought composite disasters

Through collecting local meteorological element condition, tea tree main cultivation variety and management measure and tea garden distribution condition, understand the tea garden and receive the compound calamity victim condition of high temperature drought through tea enterprise and tea farmer, confirm representative region, formulate unmanned aerial vehicle flight range, acquire healthy and the unmanned aerial vehicle photo of victim tea garden.

The height, angle and transverse direction of the flight path of the unmanned aerial vehicle also need to meet the data quantity of control points required by splicing. By using the obtained unmanned aerial vehicle photo, the professional software is adopted for splicing, and the main process comprises the steps of photo establishment, dense point cloud establishment, grid generation, DEM generation and orthographic image generation, and finally, the regional unmanned aerial vehicle image is obtained, as shown in fig. 1.

Step 2, label making based on SEGMENT ANYTHING unmanned aerial vehicle images

And (3) setting sub-images according to the spatial resolutions of the four satellite data by using the unmanned aerial vehicle images finished in the step (1), wherein the sub-images are 5 meters, 10 meters, 16 meters and 30 meters respectively. And the unmanned aerial vehicle image is re-divided into sub-images with different spatial resolutions.

And selecting the sub-images to manufacture labels of healthy tea trees, damaged tea trees, bare soil, water, buildings and the like, wherein green parts of the tea trees are marked as healthy tea tree labels, and brown parts are marked as damaged tea tree labels.

Because of the huge workload, the human-computer interaction interpretation is performed by adopting an image segmentation basic model-SEGMENT ANYTHING algorithm, and a result example is shown in fig. 2.

And 3, extracting pixels of the damage of the tea tree under the high-temperature drought compound stress based on the unmanned aerial vehicle image and checking the precision.

And (3) using the healthy tea trees, the damaged tea trees, bare soil, the water body, the building labels and the corresponding unmanned aerial vehicle sub-images which are finished in the step (2) to take 70% of the labels as model training data and 30% of the labels as model verification data.

And extracting healthy tea trees, damaged tea trees, bare soil, water bodies and buildings by adopting a plurality of machine learning methods, and selecting an optimal extraction model, wherein the accuracy verification indexes comprise a determination coefficient, a root mean square error and an average absolute error.

Taking logistic regression, naive bayes regression algorithm, random forest sum XGBoost as examples, logistic regression and naive bayes regression algorithm all belong to the supervised parameter classifier. If the data are unimodal and normally distributed they produce good results. While applications have limitations in extracting target categories based on multimodal datasets. Nonparametric supervised classifiers, such as XGBoost and random forest classifiers, perform classification without any data frequency distribution assumptions and outperform the supervised parametric classifier. The model established by the four algorithms is verified with 30% of label data, and the result shows that XGBoost algorithm has the highest precision;

Therefore, the XGBoost algorithm will be used to extract all the victim tea tree pixels of the drone image.

And 4, constructing a tea garden damage severity index model under the high-temperature drought composite condition, and calculating the tea garden damage severity index under different spatial resolutions.

The method comprises the steps of calculating the number of the victim tea tree pixels in the sub-images with different spatial resolutions by using the victim tea tree pixels extracted in the step 3, and dividing the number by the number of all the pixels of the corresponding sub-image (the spatial resolution of the unmanned aerial vehicle image is 0.025 m). The calculation formula of each spatial resolution sub-image CDH_DSI is as follows:

(1) 5 meters:

The number of pixels per sub-picture is 40,000, and cdh_dsi is defined as:

(2) 10 meters:

The number of pixels per sub-picture is 160000, and cdh_dsi is defined as:

(3) 16 meters:

the number of pixels per sub-picture is 409600, and cdh_dsi is defined as:

(4) 30 meters:

The number of pixels per sub-picture is 144000, and CDH_DSI is defined as:

And 5, drawing the pixel severity index, and obtaining CDH_DSI with different spatial resolutions based on the high-temperature drought composite condition of the unmanned aerial vehicle image.

And (3) calculating unmanned aerial vehicle sub-images corresponding to different spatial resolution satellite data (5 meters, 10 meters, 16 meters and 30 meters) for segmentation by using the image elements of the damaged tea tree extracted in the step (3) and the severity index of the damage of the tea garden defined in the step (4), and calculating CDH_DSI, thereby realizing CDH_DSI drawing based on the unmanned aerial vehicle images, as shown in figure 3.

Claims (5)

1. A tea garden high-temperature drought composite disaster investigation method based on unmanned aerial vehicle is characterized by comprising the following steps:

Step 1, acquiring image data of an unmanned aerial vehicle under the high-temperature drought compound condition of a tea garden and preprocessing the image data;

Step 2, labeling based on SEGMENT ANYTHING unmanned aerial vehicle images;

Step 3, extracting a damaged tea tree pixel based on the high-temperature drought composite condition of the unmanned aerial vehicle image;

step 4, constructing a tea garden damage severity index model under the high-temperature drought composite condition according to the damaged tea tree pixels, and calculating the tea garden damage severity index under different spatial resolutions;

and 5, acquiring severity indexes of damage of tea gardens with different spatial resolutions based on unmanned aerial vehicle image high-temperature drought composite conditions, drawing a pixel severity index chart, and completing drought composite disaster investigation.

2. The unmanned aerial vehicle-based tea garden high-temperature drought composite disaster investigation method according to claim 1, wherein the step 1 specifically comprises: the method comprises the steps of acquiring meteorological element conditions through a meteorological department, acquiring management measures and tea garden distribution through a local agricultural department, acquiring the damage condition of a high-temperature drought composite disaster on a tea garden through a tea enterprise and a tea farmer, designing the flight range and the route of an unmanned aerial vehicle, acquiring unmanned aerial vehicle photos of a healthy and damaged tea garden, and splicing by utilizing the acquired unmanned aerial vehicle photos to obtain regional unmanned aerial vehicle images.

3. The unmanned aerial vehicle-based tea garden high-temperature drought composite disaster investigation method according to claim 2, wherein the specific operation of the step 2 is as follows: dividing the unmanned aerial vehicle image into sub-images with different spatial resolutions, and selecting the sub-images to manufacture healthy tea trees, damaged tea trees, bare soil, water bodies and building labels; and performing man-machine interaction interpretation by adopting SEGMENT ANYTHING algorithm to obtain healthy tea tree pixels, victim tea tree pixels, bare soil pixels, water body pixels and building pixels.

4. The unmanned aerial vehicle-based tea garden high-temperature drought composite disaster investigation method according to claim 3, wherein the specific process of the step3 is as follows:

Dividing training data and verification data by using healthy tea trees, damaged tea trees, bare soil, water bodies, building labels and corresponding unmanned aerial vehicle sub-images;

and (3) extracting healthy tea trees, damaged tea trees, bare soil, water bodies and buildings by adopting a plurality of machine learning algorithms, selecting the algorithm with the highest precision as a damaged tea tree pixel extraction model according to an evaluation result, and extracting the damaged tea tree pixels of all unmanned aerial vehicle images.

5. The unmanned aerial vehicle-based tea garden high-temperature drought composite disaster investigation method according to claim 4, wherein the tea garden damage severity index in step 4 is specifically calculated as: calculating the number of the victim tea tree pixels in the sub-images with different spatial resolutions by utilizing the victim tea tree pixels extracted in the step 3, dividing the number by the number of all the pixels corresponding to the sub-images, and calculating the severity index CDH_DSI of the victim tea tree in the tea garden:

wherein, DTP is victim tea tree pixel quantity, and R represents spatial resolution, and the unit is meter, and Ru is unmanned aerial vehicle image spatial resolution, and the unit is meter.