CN117907248B - A remote sensing monitoring method and system for root soil moisture content during the key growth period of winter wheat - Google Patents

Abstract

The invention provides a remote sensing monitoring method and a remote sensing monitoring system for the water content of root system soil in a key growth period of winter wheat, wherein the method comprises the steps of extracting spectral images in the whole growth period of winter wheat and preprocessing, and further comprises the following steps: acquiring irrigation area sample data; acquiring satellite remote sensing sample data; constructing winter wheat growth period classification characteristics; constructing a deep soil water content prediction model based on support vector machine regression and training; and (5) evaluating the accuracy of inversion results of the water content of the deep soil in the key growth period. According to the remote sensing monitoring method and system for the root system soil moisture content of the winter wheat in the key growth period, which are provided by the invention, the accurate identification of the growth period of the winter wheat is realized by setting the spectral threshold and calculating a plurality of normalization indexes, the accurate prediction of the root system soil moisture of the winter wheat is realized by a support vector machine regression model, the soil moisture content monitoring of the winter wheat in the key growth period is ensured, and effective technical support is provided for the construction of digital irrigation areas.

Description

A remote sensing monitoring method and system for root soil moisture content during the key growth period of winter wheat

Technical Field

The invention relates to the technical field of soil moisture remote sensing monitoring, in particular to a method and system for remote sensing monitoring of soil moisture content in the root system of winter wheat during a critical growth period.

Background technique

Existing studies have used the spectral curve thresholds of crops in different growth stages to distinguish the growth stages of crops, and finally determined that the growth stage of winter wheat can be identified through Sentinel-2 Level-2A data and multiple normalized indices.

At present, it is difficult to monitor deep soil moisture content with remote sensing technology. In order to solve the problem of deep soil moisture prediction, some studies have used mathematical methods such as multivariate regression fitting to connect shallow soil moisture content with deep soil moisture content in the root system. With the development of computer technology, more and more regression methods have been realized through machine learning, and it was finally determined that the prediction of deep soil moisture content from shallow soil moisture content can be achieved through support vector machine regression.

These studies only used a small amount of data to predict the relationship between surface soil moisture content and deep soil moisture content and to identify different crop growth stages. The errors in predicting deep soil moisture content were large, especially when there was rainfall or irrigation.

The invention patent application with publication number CN111721714A discloses a soil moisture estimation method based on multi-source optical remote sensing data, which combines the spectral characteristics of hyperspectral data and the high spatial resolution characteristics of high spatial resolution images, constructs high spatial resolution hyperspectral data through data compounding and splitting, and then uses the model to complete the soil moisture estimation. The disadvantage of this method is that the data used is optical remote sensing data, which is easily affected by weather and cloud thickness images, the data is easily disturbed, and the visible light band has a weak ability to penetrate the soil, making it difficult to monitor deep soil moisture data.

The invention patent application with publication number CN 107505265A discloses a remote sensing monitoring method for soil moisture content, including obtaining optical remote sensing image data of vegetation distribution in the monitoring area, extracting features from the optical remote sensing image data, dividing the monitoring area according to vegetation types and locations, determining vegetation coverage data of the monitoring area where different types of vegetation are located; introducing the growth cycle of different types of vegetation, establishing a data model of vegetation coverage degree in the monitoring area with respect to time and location; obtaining the surface temperature and evaporation of the monitoring area through satellite remote sensing, synchronously processing the data model of vegetation coverage degree in the monitoring area with respect to time and location, surface temperature and evaporation; determining the temperature vegetation drought index of the monitoring area; determining the surface soil moisture content with respect to the monitoring parameters; determining the surface soil moisture content with respect to the monitoring parameters; determining the surface soil moisture content with respect to the monitoring parameters. The disadvantages of this method are that only a single optical remote sensing data is used as a data source, the data is easily affected by the weather and cannot be used, and the optical remote sensing data has a long recurrence period and cannot be monitored in real time, and it does not consider monitoring the deep soil moisture content and can only monitor the surface moisture content.

Summary of the invention

In order to solve the above-mentioned technical problems, the present invention proposes a remote sensing monitoring method and system for the root soil moisture content in the critical growth period of winter wheat, which can accurately identify the growth period of winter wheat by setting spectral thresholds and calculating multiple normalized indexes, and accurately predict the root soil moisture of winter wheat by a support vector machine regression model, thereby ensuring the soil moisture monitoring in the critical growth period of winter wheat and providing effective technical support for the construction of digital irrigation areas.

The first object of the present invention is to provide a remote sensing monitoring method for root soil moisture content during the key growth period of winter wheat, comprising extracting spectral images of the whole growth period of winter wheat and performing preprocessing, characterized in that it also includes the following steps:

Step 1: Obtain irrigation area sample data;

Step 2: Obtain satellite remote sensing sample data;

Step 3: Construct classification characteristics of winter wheat growth period;

Step 4: Construct a deep soil moisture prediction model based on support vector machine regression and train it;

Step 5: Assess the accuracy of the inversion results of deep soil moisture content during the key growth period.

Preferably, the extracting and preprocessing of spectral images of winter wheat throughout its entire growth period includes obtaining a Sentinel-2 Level-2A surface reflectance product of winter wheat throughout its entire growth period in an irrigated area, and preprocessing the data to obtain an image set of the winter wheat growth period.

Preferably, in any of the above schemes, the method of obtaining the Sentinel-2 Level-2A surface reflectance product of winter wheat in the irrigation area throughout the entire growth period includes using visual interpretation and querying the irrigation area meteorological data to screen out remote sensing images where snowfall and heavy fog are located, selecting images without clouds or with less clouds, and cropping the remote sensing images within the irrigation area through the irrigation area vector map.

In any of the above schemes, preferably, the method for acquiring the winter wheat growth period image set includes band fusion and resampling the 12 spectral bands of the irrigation area image to obtain a 10-meter resolution spectral image set of the 12 spectral bands of winter wheat in the irrigation area.

In any of the above schemes, preferably, step 1 includes the following sub-steps:

Step 11: Collect soil moisture data at soil moisture stations in the irrigation area and obtain soil moisture data samples at different depths;

Step 12: Obtain rainfall data from rain gauges around the irrigation area to determine the rainfall date and amount;

Step 13: Obtain the soil texture data of the irrigation area and determine the distribution of soil texture in the irrigation area;

Step 14: Obtain irrigation area DEM data to calculate slope and aspect.

In any of the above schemes, preferably, step 2 includes obtaining SMAP-Derived 1-km surface soil moisture content, daily surface temperature data and sunlight reflectance data for the entire growth period of winter wheat in the irrigation area.

In any of the above schemes, preferably, step 3 includes using the winter wheat growing period image set to generate multiple vegetation indices and fusing them with the winter wheat growing period image set to form a synthetic image set, and determining the spectral characteristics of winter wheat in different growing periods based on the synthetic image set to form a spectral characteristic set of winter wheat growing period.

In any of the above schemes, preferably, step 3 includes the following sub-steps:

Step 31: Select corresponding spectral bands from the 12 spectral bands to perform band calculations to obtain 5 vegetation indices;

Step 32: The five vegetation indices are used as independent bands to perform band fusion with the 12 spectral bands to obtain a second fused image set of 17 bands for the entire growth period of winter wheat, and characteristic spectral curves of winter wheat in different growth periods are distinguished by analyzing the images of the second fused image set.

In any of the above schemes, preferably, the five vegetation indices include the normalized difference vegetation index NDVI, the difference vegetation index DVI, the normalized difference water index NDWI, the enhanced vegetation index EVI and the normalized building index NDBI, and the calculation formula is:

NDVI=(NIR- RED)/(NIR+ RED)

DVI=NIR-RED

NDWI= (Green-NIR)/(Green+NIR)

EVI = 2.5 * ((NIR – RED) / ((NIR) + (6 * RED) – (7.5 * BLUE) + 1))

NDBI = (SWIR - NIR) / (SWIR + NIR)

Among them, NIR is the reflection value of the near infrared band, RED is the reflection value of the red light band, Green is the reflection value of the green light band, BLUE is the reflection value of the blue light band, and SWIR is the reflection value of the shortwave infrared band.

In any of the above solutions, preferably, step 4 includes using the irrigation area sample data and the satellite remote sensing sample data as inputs of the support vector machine regression model to train the support vector machine regression model.

In any of the above schemes, preferably, step 4 includes the following sub-steps:

Step 41: Construct a regression model formula f(x)=w ^T x+b;

Step 42: Construct the input data matrix X=[x1, x2,…xn] and label matrix Y=[y1, y2,…yn] ^T ,

Step 43: Construct a Larange equation L and use the SMO method to solve it to obtain the a ^* and , the formula is

;

Step 44: Using the above Lagrange equation to obtain the partial derivative of w, we can obtain:

,

Step 45: Using the above Lagrange equation to obtain the partial derivative of b, we can obtain:

,

Step 46: Take the partial derivative of the above Lagrange equation and get:

,

Step 47: Substituting the above three equations to satisfy the KKT condition, we get:

,

,

Step 48: Calculate the minimum L value by SMO method to obtain a ^* and , we get the training equation f(x)=w ^T x+b;

Where w is the normal vector of the hyperplane, w ^T is the transpose of the normal vector of the hyperplane, x is the input matrix, b is the distance between the hyperplane and the origin, is a slack variable used to handle the inseparable situation, a is a Lagrange multiplier used to weight the constraints in the loss function, μ is a Lagrange multiplier used to weight the constraints of the slack variable, C is a penalty parameter used to balance the interval size and the weight of misclassification, n is the number of samples, a _i is the Lagrange multiplier of the i -th equation, x ⁱ is the i -th input variable, y ⁱ is the i -th output, a ^* and /> are two optimal Lagrange multipliers.

In any of the above schemes, preferably, step 5 includes obtaining the root soil moisture data of winter wheat in the irrigation area throughout the entire growth period through a deep soil moisture prediction model based on support vector machine regression, and classifying the root soil moisture data according to the spectral characteristics of the winter wheat growth period to obtain the root soil moisture data of the winter wheat in the key growth period for accuracy assessment.

In any of the above schemes, preferably, step 5 includes selecting root soil moisture data of winter wheat in the key growth period to perform accuracy assessment through root mean square error RMSE, and the calculation formula is:

,

The calculation formula of correlation coefficient R is:

,

Among them, SMP is the measured soil moisture content _at different depths during the critical growth period, _SMI is the soil moisture content during the corresponding critical growth period predicted by the regression model, and m is the effective number of samples.

The second object of the present invention is to provide a remote sensing monitoring system for root soil moisture content during the key growth period of winter wheat, including a remote sensing data processing module and the following modules:

The winter wheat growth period identification module is used to construct the classification characteristics of winter wheat growth period;

Root soil water content prediction module based on support vector machine regression: build a deep soil water content prediction model based on support vector machine regression and train it;

Accuracy detection module: evaluates the accuracy of the inversion results of deep soil moisture content during the key growth period;

The system adopts the method described in the first object to perform remote sensing monitoring of soil moisture content in the root system of winter wheat during its critical growth period.

Preferably, the remote sensing data processing module is used to extract spectral images of winter wheat throughout its growth period and perform preprocessing.

Preferably, in any of the above schemes, the extracting of spectral images of winter wheat throughout its entire growing period and preprocessing the images comprises obtaining the Sentinel-2 Level-2A surface reflectance product of winter wheat throughout its growing period in the irrigated area, and preprocessing the data to obtain an image set of the winter wheat growing period.

In any of the above schemes, preferably, the method for acquiring the winter wheat growth period image set includes band fusion and resampling the 12 spectral bands of the irrigation area image to obtain a 10-meter resolution spectral image set of the 12 spectral bands of winter wheat in the irrigation area.

In any of the above solutions, preferably, the remote sensing data processing module is also used to obtain irrigation area sample data and satellite remote sensing sample data.

In any of the above solutions, preferably, the method for obtaining irrigation area sample data includes the following sub-steps:

Step 11: Collect soil moisture data at soil moisture stations in the irrigation area and obtain soil moisture data samples at different depths;

Step 12: Obtain rainfall data from rain gauges around the irrigation area to determine the rainfall date and amount;

Step 13: Obtain the soil texture data of the irrigation area and determine the distribution of soil texture in the irrigation area;

Step 14: Obtain irrigation area DEM data to calculate slope and aspect.

In any of the above schemes, preferably, the satellite remote sensing sample data includes SMAP-Derived 1-km surface soil moisture content, daily surface temperature data and sunlight reflectance data of winter wheat in the irrigated area throughout the growth period.

In any of the above schemes, preferably, the method for constructing the classification characteristics of winter wheat growth period comprises the following sub-steps:

Step 31: Select corresponding spectral bands from the 12 spectral bands to perform band calculations to obtain 5 vegetation indices;

Step 32: The five vegetation indices are used as independent bands to perform band fusion with the 12 spectral bands to obtain a second fused image set of 17 bands for the entire growth period of winter wheat, and characteristic spectral curves of winter wheat in different growth periods are distinguished by analyzing the images of the second fused image set.

In any of the above schemes, preferably, the vegetation index includes the normalized difference vegetation index NDVI, the difference vegetation index DVI, the normalized difference water index NDWI, the enhanced vegetation index EVI and the normalized building index NDBI, and the calculation formula is:

NDVI=(NIR- RED)/(NIR+ RED)

DVI=NIR-RED

NDWI= (Green-NIR)/(Green+NIR)

EVI = 2.5 * ((NIR – RED) / ((NIR) + (6 * RED) – (7.5 * BLUE) + 1))

NDBI = (SWIR - NIR) / (SWIR + NIR)

Among them, NIR is the reflection value of the near infrared band, RED is the reflection value of the red light band, Green is the reflection value of the green light band, BLUE is the reflection value of the blue light band, and SWIR is the reflection value of the shortwave infrared band.

In any of the above schemes, preferably, the root soil water content prediction module based on support vector machine regression is used to apply irrigation area sample data and satellite remote sensing sample data as inputs of the support vector machine regression model to train the support vector machine regression model.

In any of the above solutions, preferably, the method for constructing the support vector machine regression model includes the following sub-steps:

Step 41: Construct a regression model formula f(x)=w ^T x+b;

Step 42: Construct the input data matrix X=[x1, x2,…xn] and label matrix Y=[y1, y2,…yn] ^T ,

Step 43: Construct a Larange equation L and use the SMO method to solve it to obtain the a ^* and , the formula is

;

Step 44: Using the above Lagrange equation to obtain the partial derivative of w, we can obtain:

,

Step 45: Using the above Lagrange equation to obtain the partial derivative of b, we can obtain:

,

Step 46: Take the partial derivative of the above Lagrange equation and get:

,

Step 47: Substituting the above three equations to satisfy the KKT condition, we get:

,

,

,

Step 48: Calculate the minimum L value by SMO method to obtain a ^* and , we get the training equation f(x)=w ^T x+b;

Where w is the normal vector of the hyperplane, w ^T is the transpose of the normal vector of the hyperplane, x is the input matrix, b is the distance between the hyperplane and the origin, is a slack variable used to handle the inseparable situation, a is a Lagrange multiplier used to weight the constraints in the loss function, μ is a Lagrange multiplier used to weight the constraints of the slack variable, C is a penalty parameter used to balance the interval size and the weight of misclassification, n is the number of samples, a _i is the Lagrange multiplier of the i -th equation, x ⁱ is the i -th input variable, y ⁱ is the i -th output, a ^* and /> are two optimal Lagrange multipliers.

In any of the above schemes, preferably, the accuracy detection module is used to obtain the root soil moisture data of winter wheat in the irrigation area throughout the growth period through a deep soil moisture prediction model based on support vector machine regression, and classify the root soil moisture data according to the spectral characteristics of the winter wheat growth period to obtain the root soil moisture data of the winter wheat in the key growth period for accuracy assessment.

In any of the above schemes, preferably, the accuracy detection module is also used to select the root soil moisture content data of winter wheat in the key growth period to perform accuracy assessment through the root mean square error RMSE, and the calculation formula is:

,

The calculation formula of correlation coefficient R is:

,

Among them, SMP is the measured soil moisture content _at different depths during the critical growth period, _SMI is the soil moisture content during the corresponding critical growth period predicted by the regression model, and m is the effective number of samples.

The present invention proposes a remote sensing monitoring method and system for the root soil moisture content during the critical growth period of winter wheat, which realizes the accurate identification of the growth period of winter wheat. The root soil moisture content during the critical growth period of winter wheat predicted based on support vector machine regression is highly accurate, thereby realizing the accurate prediction of the root soil moisture content during the critical growth period of winter wheat.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG. 1 is a flow chart of a preferred embodiment of a remote sensing monitoring method for root soil moisture content during a critical growth period of winter wheat according to the present invention.

FIG. 2 is a module diagram of a preferred embodiment of a remote sensing monitoring system for root soil moisture content during a critical growth period of winter wheat according to the present invention.

FIG3 is a flow chart of another preferred embodiment of the remote sensing monitoring method for root soil moisture content during the critical growth period of winter wheat according to the present invention.

FIG. 4 is a schematic diagram of the support vector machine regression principle of another preferred embodiment of the remote sensing monitoring method for root soil moisture content during the critical growth period of winter wheat according to the present invention.

Detailed ways

The present invention is further described below in conjunction with the accompanying drawings and specific embodiments.

Embodiment 1

As shown in FIGS. 1 and 2 , step 100 is executed, the remote sensing data processing module 200 extracts the spectral image of the whole growth period of winter wheat and performs preprocessing, obtains the Sentinel-2 Level-2A surface reflectance product of the whole growth period of winter wheat in the irrigation area, and preprocesses the data to obtain the image set of the growth period of winter wheat;

Acquiring Sentinel-2 Level-2A surface reflectance products for winter wheat in the irrigation area during the entire growth period involves using visual interpretation and querying the irrigation area's meteorological data to screen out remote sensing images where snow and fog are located, selecting images without clouds or with less clouds, and cropping remote sensing images within the irrigation area using irrigation area vector maps;

The method for acquiring the winter wheat growth period image set includes band fusion and resampling of the 12 spectral bands of the irrigation area image to obtain a 10-meter resolution spectral image set of the 12 spectral bands of winter wheat in the irrigation area.

Executing step 110, the remote sensing data processing module 200 obtains irrigation area sample data, including the following sub-steps:

Execute step 111 to collect soil moisture data from soil moisture stations in the irrigation area and obtain soil moisture data samples at different depths;

Execute step 112 to obtain rainfall data from rain gauges around the irrigation area and determine the rainfall date and rainfall amount;

Execute step 113 to obtain the soil texture data of the irrigation area and determine the distribution of the soil texture in the irrigation area;

Execute step 114 to obtain the irrigation area DEM data to calculate the slope and slope aspect.

Execute step 120, the remote sensing data processing module 200 obtains satellite remote sensing sample data, and obtains SMAP-Derived 1-km surface soil moisture content, daily surface temperature data, and sunlight reflectance data for the entire growth period of winter wheat in the irrigation area.

Execute step 130, the winter wheat growth period identification module 210 constructs the classification features of the winter wheat growth period, uses the winter wheat growth period image set to generate multiple vegetation indices and fuses them with the winter wheat growth period image set to form a synthetic image set, and determines the spectral features of winter wheat in different growth periods according to the synthetic image set to form a spectral feature set of the winter wheat growth period, including the following sub-steps:

Execute step 131, select the corresponding spectral band from the 12 spectral bands to perform band calculation to obtain 5 vegetation indices, the vegetation indices include the normalized difference vegetation index NDVI, the difference vegetation index DVI, the normalized difference water index NDWI, the enhanced vegetation index EVI and the normalized building index NDBI, and the calculation formula is:

NDVI=(NIR- RED)/(NIR+ RED)

DVI=NIR-RED

NDWI= (Green-NIR)/(Green+NIR)

EVI = 2.5 * ((NIR – RED) / ((NIR) + (6 * RED) – (7.5 * BLUE) + 1))

NDBI = (SWIR - NIR) / (SWIR + NIR)

Among them, NIR is the reflection value of the near infrared band, RED is the reflection value of the red light band, Green is the reflection value of the green light band, BLUE is the reflection value of the blue light band, and SWIR is the reflection value of the short-wave infrared band;

Execute step 132, take the 5 vegetation indices as independent bands and perform band fusion with the 12 spectral bands to obtain a second fused image set of 17 bands for the whole growth period of winter wheat, and distinguish the characteristic spectral curves of winter wheat in different growth periods by analyzing the images of the second fused image set.

Execute step 140, build a deep soil moisture prediction model based on support vector machine regression based on the support vector machine regression root soil moisture content prediction module 220 and train it, use the irrigation area sample data and the satellite remote sensing sample data as the input of the support vector machine regression model, and train the support vector machine regression model, including the following sub-steps:

Execute step 141 to construct a regression model formula f(x)=w ^T x+b;

Execute step 142 to construct the input data matrix X=[x1, x2,…xn] and the label matrix Y=[y1, y2,…yn] ^T ,

Execute step 143, construct a Lagrange equation L, and use the SMO method to solve the a ^* and , the formula is

;

Execute step 144, and use the above Lagrange equation to obtain the partial derivative of w:

,

Execute step 145, and use the above Lagrange equation to obtain the partial derivative of b:

,

Execute step 146, and obtain the partial derivative of the above Lagrange equation:

,

Execute step 147, and make the above three equations satisfy the KKT condition to obtain:

,

,

,

Execute step 148, calculate the minimum L value by SMO method to obtain a ^* and , we get the training equation f(x)=w ^T x+b;

Where w is the normal vector of the hyperplane, w ^T is the transpose of the normal vector of the hyperplane, x is the input matrix, b is the distance between the hyperplane and the origin, is a slack variable used to handle the inseparable situation, a is a Lagrange multiplier used to weight the constraints in the loss function, μ is a Lagrange multiplier used to weight the constraints of the slack variable, C is a penalty parameter used to balance the interval size and the weight of misclassification, n is the number of samples, a _i is the Lagrange multiplier of the i -th equation, x ⁱ is the i -th input variable, y ⁱ is the i -th output, a ^* and /> are two optimal Lagrange multipliers.

Execute step 150, the accuracy detection module 230 evaluates the accuracy of the inversion result of deep soil moisture content in the key growth period, and selects the root soil moisture data of the key growth period of winter wheat for accuracy evaluation through the root mean square error RMSE, and the calculation formula is:

,

The calculation formula of correlation coefficient R is:

,

Among them, SMP is the measured soil moisture content _at different depths during the critical growth period, _SMI is the soil moisture content during the corresponding critical growth period predicted by the regression model, and m is the effective number of samples.

Embodiment 2

The object of the present invention is to provide a technical solution of a remote sensing monitoring system for root soil moisture content in a key growth period of winter wheat based on support vector machine regression, and the solution is as follows:

Step S1, extraction and preprocessing of spectral images of winter wheat during the entire growth period: obtain the Sentinel-2 Level-2A surface reflectance product of winter wheat in the irrigation area during the entire growth period, and preprocess the data to obtain an image set of the winter wheat growth period.

In step S1, the specific method for obtaining the Sentinel-2 Level-2A surface reflectance product of winter wheat in the irrigation area during the entire growth period includes:

Visual interpretation and query of irrigation area meteorological data were used to screen out remote sensing images with snowfall and heavy fog, select images without clouds or with less clouds, and crop remote sensing images within the irrigation area through irrigation area vector maps;

The specific method of synthesizing the growth period images to obtain the growth period image set includes:

After band fusion and resampling of the 12 spectral bands of the irrigation area images, a 10-meter resolution spectral image set of 12 bands of winter wheat in the irrigation area was obtained.

Step S2, obtaining sample data of irrigation areas: Collect soil moisture data from soil moisture stations in irrigation areas and obtain samples of soil moisture data at different depths. Obtain rainfall data from rain gauges around the irrigation area to determine the date and amount of rainfall. Access the National Tibetan Plateau Data Center to obtain soil texture data in the irrigation area and determine the distribution of soil texture in the irrigation area. Access the geospatial data cloud to obtain DEM data of the irrigation area to calculate the slope and slope direction.

In step S2, the slope and aspect are calculated using the acquired DEM data to generate three raster data sets for subsequent calculations.

Step S3, satellite remote sensing sample data acquisition: SMAP-Derived 1-km surface soil moisture data of winter wheat in the irrigation area during the entire growth period is obtained for subsequent inversion of root soil moisture content, as well as daily surface temperature data, sunlight reflectance data, etc.

Step S4, constructing classification features of winter wheat growth period: using the above image set to generate multiple vegetation indices and merging them with the above image set to form a synthetic image set, and determining the spectral features of winter wheat in different growth periods based on the image set to form a spectral feature set of winter wheat growth period.

In step S4, the specific method of using the spectral image set of winter wheat in the irrigation area to generate multiple vegetation indices includes:

The corresponding bands of the 12 bands in the image set are selected for band calculation to obtain 5 vegetation index data: Normalized Difference Vegetation Index (NDVI), Difference Vegetation Index (DVI), Normalized Difference Water Index (NDWI), Enhanced Vegetation Index (EVI), Normalized Building Index (NDBI):

DVI=(NIR- RED)/(NIR+ RED);

DVI = NIR-RED;

NDWI= (Green-NIR)/(Green+NIR);

EVI = 2.5 * ((NIR – RED) / ((NIR) + (6 * RED) – (7.5 * BLUE) + 1)) ;

NDBI = (SWIR - NIR) / (SWIR + NIR);

The above five index features are used as independent bands and fused with the 12 bands of the above image set to obtain a second fused image set of 17 bands for the entire growth period of winter wheat. The characteristic spectral curves of winter wheat in different growth periods are distinguished by analyzing the images of the second fused image set.

Step S5, construction and training of a deep soil moisture prediction model based on support vector machine regression: using the irrigation area sample data and the satellite remote sensing sample data as inputs of the support vector machine regression model to train the support vector machine regression model.

Step S6, accuracy assessment of the inversion results of deep soil moisture content in the critical growth period: The root soil moisture data of winter wheat in the irrigation area throughout the entire growth period is obtained through a deep soil moisture prediction model based on support vector machine regression, and the root soil moisture data is classified according to the spectral characteristics of the winter wheat growth period to obtain the root soil moisture data of the winter wheat in the critical growth period for accuracy assessment.

In step S6, the specific method includes:

The root soil moisture data of winter wheat in the key growth period were selected for accuracy assessment using the root mean square error (RMSE), and the calculation formula was:

,

SMP : measured soil water content _at different depths during the critical growth period;

SM : regression model predicts soil water content during the corresponding key growth period;

m : effective number of samples;

The second aspect of the present invention discloses a remote sensing monitoring system for root soil moisture content during a key growth period of winter wheat based on support vector machine regression, the system comprising:

The remote sensing data processing module is configured to obtain data products from Sentinel-2 satellites, SMAP satellites, and MODIS satellites, and perform preprocessing such as atmospheric correction and radiation correction to obtain an image set, and then perform band fusion to obtain a first synthetic image set.

The winter wheat growth period identification module is configured to calculate each index feature using the above-mentioned synthetic image set, and fuse the calculated index features with the first synthetic image set, and then identify the growth period of the winter wheat based on the spectral characteristics and the differences in each index.

The root soil water content prediction module based on support vector machine regression is configured to divide sample data into a training set and a test set through stratified sampling, and use the training set and the remote sensing data product as inputs of the support vector machine regression model to train the support vector machine regression model.

The accuracy detection module is configured to use the root mean square error formula to detect the prediction accuracy of the root soil moisture content in the key growth period.

The third aspect of the present invention discloses an electronic device. The electronic device includes a memory and a processor, the memory stores a computer program, and when the processor executes the computer program, the steps in any one of the methods for remote sensing monitoring of soil moisture content of winter wheat root system during the key growth period based on support vector machine regression in the first aspect of the present disclosure are implemented.

The fourth aspect of the present invention discloses a storage medium. A computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps in any one of the methods for remote sensing monitoring of soil moisture content of winter wheat roots during a critical growth period based on support vector machine regression in the first aspect of the present disclosure are implemented.

Embodiment 3

As shown in FIG3 , a remote sensing monitoring method for root soil moisture content in a key growth period of winter wheat based on support vector machine regression of the present invention comprises:

Step S1, extraction and preprocessing of spectral images of winter wheat during the entire growth period: obtain the Sentinel-2 Level-2A surface reflectance product of winter wheat in the irrigation area during the entire growth period, and preprocess the data to obtain an image set of the winter wheat growth period.

In step S1, the spectral image extraction and preprocessing of winter wheat during the whole growth period includes:

(1) Perform atmospheric correction to eliminate the influence of clouds and fog.

(2) Perform radiation correction to eliminate or correct image distortion caused by radiation errors.

(3) Crop the spectrum according to the irrigation area and determine the spectrum range to be collected.

Step S2, obtaining sample data of irrigation areas: Collect soil moisture data from soil moisture stations in irrigation areas, and obtain samples of soil moisture data at different depths. Obtain rainfall data from rain gauges around the irrigation area to determine the date and amount of rainfall. Access the National Tibetan Plateau Data Center to obtain soil texture data in the irrigation area and determine the distribution of soil texture in the irrigation area. Access the geospatial data cloud to obtain DEM data of the irrigation area to calculate the slope and slope direction.

In step S2, the irrigation area sample data acquisition includes:

(1) Determine the scope of the irrigation area and download the soil texture distribution data within the irrigation area through the National Qinghai-Tibet Plateau Data Center

(2) Download the irrigation area DEM distribution map, calculate the slope and slope aspect using the downloaded irrigation area DEM data, and obtain three irrigation area information raster data maps.

Step S3, satellite remote sensing sample data acquisition: SMAP-Derived 1-km surface soil moisture data of winter wheat in the irrigation area during the entire growth period is obtained for subsequent inversion of root soil moisture content, as well as daily surface temperature data, sunlight reflectance data, etc.

Step S4, constructing classification features of winter wheat growth period: using the above image set to generate multiple vegetation indices and merging them with the above image set to form a synthetic image set, and determining the spectral features of winter wheat in different growth periods based on the image set to form a spectral feature set of winter wheat growth period.

In step S4, the winter wheat spectral feature set is constructed:

(1) The 12 bands of Sentinel-2 data collected by S1 were fused to obtain 12-band spectral images of winter wheat in the irrigation area.

(2) The normalized index was calculated using the bands in the winter wheat spectral image of the irrigation area to obtain a normalized index raster data map of the irrigation area, which was then combined with the winter wheat spectral image to identify the different growth stages of winter wheat.

Step S5, construction and training of a deep soil moisture prediction model based on support vector machine regression: using the irrigation area sample data and the satellite remote sensing sample data as inputs of the support vector machine regression model to train the support vector machine regression model.

In step S5, as shown in FIG4 , the deep soil moisture prediction model based on support vector machine regression is constructed and trained:

(1) The regression model formula is f(x)=w ^T x+b;

(2) Construct the input data matrix X = [x1, x2, ... xn], label matrix Y = [y1, y2, ... yn] ^T ;

(3) Construct a Larange equation L and solve it using the SMO method to obtain the a ^* and ;

,

By taking the partial derivative of w through the above Lagrange equation and setting it to 0, we get:

,

have to ;

By taking the partial derivative of b through the above Lagrange equation and setting it to 0, we get:

,

At the same time, the KKT conditions are met:

,

,

,

,

inferred ;

The minimum L value is calculated by the SMO method to obtain a ^* and , and then we get the expressions of b and w, and thus the training equation f(x)=w ^T x+b.

Step S6, accuracy assessment of the inversion results of deep soil moisture content in the critical growth period: The root soil moisture data of winter wheat in the irrigation area throughout the entire growth period is obtained through a deep soil moisture prediction model based on support vector machine regression, and the root soil moisture data is classified according to the spectral characteristics of the winter wheat growth period to obtain the root soil moisture data of the winter wheat in the critical growth period for accuracy assessment.

In step S6, the root soil moisture content prediction accuracy assessment during the key growth period of winter wheat includes the following:

(1) Root mean square error (RMSE), calculated as:

,

(2) Correlation coefficient (R), calculated as:

,

SMP : measured soil water content _at different depths during the critical growth period;

SM _i : regression model predicts soil water content in the corresponding key growth period;

m : effective number of samples;

The present invention realizes accurate identification of the growth period of winter wheat by setting spectral thresholds and calculating multiple normalized indexes, and realizes accurate prediction of soil moisture of winter wheat roots through a support vector machine regression model, thereby ensuring soil moisture monitoring during the critical growth period of winter wheat and providing effective technical support for the construction of digital irrigation areas.

In order to better understand the present invention, the above is described in detail in conjunction with the specific embodiments of the present invention, but it is not intended to limit the present invention. Any simple modifications made to the above embodiments based on the technical essence of the present invention still fall within the scope of the technical solution of the present invention. Each embodiment in this specification focuses on the differences from other embodiments, and the same or similar parts between the various embodiments can be referenced to each other. For the system embodiment, since it basically corresponds to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

Claims (2)

1. A remote sensing monitoring method for root soil moisture content during the key growth period of winter wheat, comprising extracting spectral images of the whole growth period of winter wheat and preprocessing them, characterized in that it also includes the following steps: The extracting and preprocessing of the spectral image of winter wheat during the whole growth period includes obtaining the Sentinel-2 Level-2A surface reflectance product of winter wheat during the whole growth period of the irrigation area, and preprocessing the data to obtain the image set of the winter wheat growth period. The Sentinel-2 Level-2A surface reflectance product for winter wheat in the irrigation area during the entire growth period is obtained by using visual interpretation and querying the meteorological data of the irrigation area to screen out the remote sensing images where snow and fog are located, select the images without clouds or with less clouds, and crop the remote sensing images within the irrigation area through the irrigation area vector map. The method for acquiring the winter wheat growth period image set comprises performing band fusion resampling on 12 spectral bands of the irrigation area image to obtain a 10-meter resolution spectral image set of 12 spectral bands of winter wheat in the irrigation area; Step 1: Obtain irrigation district sample data, including the following sub-steps: Step 11: Collect soil moisture data at soil moisture stations in the irrigation area and obtain soil moisture data samples at different depths; Step 12: Obtain rainfall data from rain gauges around the irrigation area to determine the rainfall date and amount; Step 13: Obtain the soil texture data of the irrigation area and determine the distribution of soil texture in the irrigation area; Step 14: Obtain irrigation area DEM data to calculate slope and aspect; Step 2: Obtain satellite remote sensing sample data and obtain SMAP-Derived 1-km surface soil moisture content, daily surface temperature data, and solar reflectance data for the entire growth period of winter wheat in the irrigation area; Step 3: Construct a deep soil moisture prediction model for winter wheat based on support vector machine regression and train the classification characteristics of the growth period. Use the above image set to generate multiple vegetation indices and fuse them with the above image set to form a synthetic image set. According to the image set, the spectral characteristics of winter wheat in different growth periods are determined to form a spectral feature set of winter wheat growth period, including the following sub-steps: Step 31: Select the corresponding spectral band from the 12 spectral bands to perform band calculation to obtain 5 vegetation indices, the 5 vegetation indices include the normalized difference vegetation index NDVI, the difference vegetation index DVI, the normalized difference water index NDWI, the enhanced vegetation index EVI and the normalized building index NDBI, and the calculation formula is: NDVI=(NIR- RED)/(NIR+ RED) DVI=NIR-RED NDWI= (Green-NIR)/(Green+NIR) EVI = 2.5 * ((NIR – RED) / ((NIR) + (6 * RED) – (7.5 * BLUE) + 1)) NDBI = (SWIR - NIR) / (SWIR + NIR) Among them, NIR is the reflection value of the near infrared band, RED is the reflection value of the red light band, Green is the reflection value of the green light band, BLUE is the reflection value of the blue light band, and SWIR is the reflection value of the short-wave infrared band; Step 32: using the five vegetation indices as independent bands to perform band fusion with the 12 spectral bands to obtain a second fused image set of 17 bands for the entire growth period of winter wheat, and distinguishing characteristic spectral curves of winter wheat in different growth periods by analyzing the images of the second fused image set; Step 4: Construct a deep soil moisture prediction model based on support vector machine regression and train it, including the following sub-steps: Step 41: Construct a regression model formula f(x)=w ^T x+b; Step 42: Construct the input data matrix X=[x1, x2,…xn] and label matrix Y=[y1, y2,…yn] ^T , Step 43: Construct the Lagrange equation L and use the SMO method to solve the a ^* and , the formula is , Step 44: Using the above Lagrange equation to obtain the partial derivative of w, we can obtain: , Step 45: Using the above Lagrange equation to obtain the partial derivative of b, we can obtain: , Step 46: Take the partial derivative of the above Lagrange equation and get: , Step 47: Substituting the above three equations to satisfy the KKT condition, we get: , , , Step 48: Calculate the minimum L value by SMO method to obtain a ^* and , we get the training equation f(x)=w ^T x+b; Where w is the normal vector of the hyperplane, w ^T is the transpose of the normal vector of the hyperplane, x is the input matrix, b is the distance between the hyperplane and the origin, is a slack variable used to handle the inseparable situation, a is a Lagrange multiplier used to weight the constraints in the loss function, μ is a Lagrange multiplier used to weight the constraints of the slack variable, C is a penalty parameter used to balance the interval size and the weight of misclassification, n is the number of samples, a _i is the Lagrange multiplier of the i -th equation, x ⁱ is the i -th input variable, y ⁱ is the i -th output, a ^* and /> are two optimal Lagrange multipliers; Step 5: Assess the accuracy of the inversion results of deep soil moisture content in the key growth period. The root soil moisture data of winter wheat in the irrigation area during the entire growth period is obtained through the deep soil moisture prediction model based on support vector machine regression. The root soil moisture data is classified according to the spectral characteristics of the winter wheat growth period to obtain the root soil moisture data of the key growth period of winter wheat for accuracy assessment. The root soil moisture data of the key growth period of winter wheat is selected for accuracy assessment through the root mean square error RMSE. The calculation formula is: , The calculation formula of correlation coefficient R is: , Among them, SMP is the measured soil moisture content _at different depths during the critical growth period, _SMI is the soil moisture content during the corresponding critical growth period predicted by the regression model, and m is the effective number of samples. 2. A remote sensing monitoring system for root soil moisture content during the key growth period of winter wheat, comprising a remote sensing data processing module, characterized in that it also includes the following modules: The winter wheat growth period identification module is used to construct the classification characteristics of winter wheat growth period; Root soil water content prediction module based on support vector machine regression: build a deep soil water content prediction model based on support vector machine regression and train it; Accuracy detection module: evaluates the accuracy of the inversion results of deep soil moisture content during the key growth period; The system uses the method as claimed in claim 1 to perform remote sensing monitoring of soil moisture content in the root system of winter wheat during its critical growth period.