Method for rapidly measuring soil conductivity by using hyperspectral remote sensing technology

Abstract

The invention provides a method for rapidly measuring soil conductivity by using a hyperspectral remote sensing technology, and belongs to the field of remote sensing inversion of soil parameters. The method comprises the following steps: distributing points in a to-be-measured area according to a system grid, and collecting hyperspectral data and conductivity data of a soil sample area; extracting spectral characteristic parameters of different dimensions, including a one-dimensional spectral index set, a two-dimensional spectral index set and a three-dimensional spectral index set; selecting important spectral indexes by using SHAP analysis, and establishing a multiple linear regression model; after a hyperspectral remote sensing image is processed, a corresponding spectral index wave band pixel value is extracted and substituted into the model, so that accurate, rapid and large-area synchronous measurement of the field soil conductivity is realized. The method solves the problems of time consuming, labor consuming, high cost, result lagging and the like of a traditional measurement method, and provides an effective means for soil salinization monitoring.

Description

Technical Field

[0001] The present invention belongs to the field of remote sensing inversion of soil parameters, and specifically relates to a method for rapid measurement of soil electrical conductivity using hyperspectral remote sensing technology. Background Art

[0002] Soil salinization is a significant form of soil degradation, severely impacting soil structure and significantly deteriorating its physical and chemical properties. Furthermore, soil salinization can lead to a significant decrease in soil fertility, declining soil productivity, and an imbalance in soil acidity and alkalinity, ultimately reducing crop yields. Electrical conductivity is widely recognized as a reliable indicator of soil salinity and an internationally recognized parameter for characterizing soil salinity. Soil conductivity is primarily measured using conductivity, electromagnetic induction, and remote sensing. Due to its high accuracy, the conductivity method is widely used in laboratories to prepare soil extracts at varying soil-water ratios for measurement. However, this traditional method is limited in measurement area, labor-intensive, and expensive. Furthermore, the conductivity method requires sample collection and laboratory processing, which often results in delayed soil conductivity acquisition and prevents real-time information on soil salt content. Although electromagnetic induction can also achieve non-contact measurement of saline-alkali soil conductivity, this method has several drawbacks, primarily the large and expensive instrumentation and its high sensitivity to soil moisture content, atmospheric temperature, and other environmental conditions. Therefore, developing a non-destructive, rapid, and accurate method for measuring the electrical conductivity of saline soils is imperative. Remote sensing technology provides comprehensive and effective data and is commonly used for large-scale and multi-temporal soil salinization monitoring. However, the accuracy of remote sensing inversion results is often affected by factors such as atmospheric conditions, electromagnetic radiation transmission processes, and spectral aliasing, resulting in inaccurate results. Therefore, this technology is primarily used to determine the spatial distribution of soil salinity and the classification of soil salinization. Summary of the Invention

[0003] Based on the above shortcomings, the present invention provides a rapid soil conductivity measurement method using hyperspectral remote sensing technology. It fuses multispectral simulated remote sensing images with hyperspectral measured remote sensing images, and uses spectral parameter extraction results of different dimensions to establish a soil conductivity prediction model, ultimately achieving accurate, rapid and large-area synchronous measurement of soil conductivity, which is used to solve a series of problems of traditional soil conductivity measurement methods such as time-consuming, labor-intensive, high cost, and time lag in measurement results.

[0004] The technology adopted by the present invention is as follows: a method for rapidly measuring soil electrical conductivity using hyperspectral remote sensing technology, comprising the following steps:

[0005] Step 1: Set up a regular square grid sampling area in the area to be measured according to the systematic grid point method. Use a 350-2500nm ground object spectrometer to perform 10 repeated spectral measurements at the center point of each sampling grid area at a measurement angle of 25° and a height of 50cm vertically above the ground. Calculate the mean of the 10 measured spectral data by band and resample to 10nm.

[0006] Step 2: In each grid sample area where spectral measurements have been completed, determine an equilateral triangle with a side length of 2m in the north, east, and west directions, with the center of the spectrometer surface as the triangle center. Collect 0-20 cm surface soil samples at the three vertices of the triangle. Bring them back to the laboratory for mixing, drying, grinding, and passing through a 2mm sieve. Prepare a standard soil suspension with a water-soil mass ratio of 5:1, and measure the soil electrical conductivity value using the electrode method;

[0007] Step 3: Based on the Pearson correlation coefficient between the reflectance of each band and the electrical conductivity at the sampling point, the spectral reflectance of the band with the highest correlation coefficient is extracted as the one-dimensional spectral index set; the reflectance of any two bands is used to calculate six two-dimensional soil salinity spectral indices, and a two-dimensional correlation coefficient diagram of the indices with electrical conductivity is plotted. The optimal spectral band combination corresponding to the highest correlation coefficient between each type of two-dimensional spectral index and electrical conductivity is determined, and the six two-dimensional spectral index sets for all sampling points under this combination are calculated; the reflectance of any three bands is used to calculate seven three-dimensional soil salinity spectral indices, and a three-dimensional correlation coefficient diagram of the indices with electrical conductivity is plotted. The optimal spectral band combination corresponding to the highest correlation coefficient between each type of three-dimensional spectral index and electrical conductivity is determined, and the seven three-dimensional spectral index sets for all sampling points under this combination are calculated;

[0008] Step 4: Using the SHAP analysis method, the two most important two-dimensional spectral indices and three three-dimensional spectral indices were selected. A multivariate linear regression model was established with the one-dimensional spectral index, the selected two-dimensional spectral index, and the three-dimensional spectral index as independent variables, and the conductivity measurement value as the dependent variable. The hyperspectral remote sensing image of the area to be measured was downloaded, and radiation correction, geometric correction, and spectral resampling were performed. The band with the highest correlation with conductivity was extracted as the one-dimensional spectral index band. The two most important two-dimensional spectral index bands and three three-dimensional spectral index bands were extracted and calculated. The pixel values ​​of each spectral index band were introduced into the conductivity prediction model, and the conductivity prediction results of all pixels were traversed and calculated to achieve synchronous measurement of soil conductivity in the field.

[0009] Furthermore, the ground object spectrometer in step 1 includes a VNIR detector and a SWIR detector. The VNIR detector has a measurement range of 350-1000 nm, a sampling interval of 1.4 nm, and a spectral resolution of 3 nm; the SWIR detector has a measurement range of 1000-2500 nm, a sampling interval of 2 nm, and a spectral resolution of 10 nm. 3. The method for rapid soil conductivity measurement using hyperspectral remote sensing technology according to claim 1 is characterized in that the six two-dimensional soil salinity spectral indices in step 3 include:

[0010] DI(R λ1 ,R λ2 )=R λ1 -R λ2

[0011] NDI(R λ1 ,R λ2 )=(R λ1 -R λ2 ) / (R λ1 +R λ2 )

[0012] RI(R λ1 ,R λ2 )=R λ1 / R λ2

[0013] PSI(R λ1 ,R λ2 )=(R λ1 +M*R λ2 ) / (M 2 +1) 0.5

[0014] PI(R λ1 ,R λ2 )=R λ1 *R λ2

[0015] SRI(R λ1 ,R λ2 )=(R λ1 2 +R λ2 2 ) 0.5

[0016] Among them, R λ1 and R λ2 Respectively represent the reflectance data of any two bands between 350nm and 2500nm, M represents the slope of the soil line, M 1.0563;

[0017] For each type of two-dimensional soil salinity spectral index, a two-dimensional correlation coefficient diagram between its calculation results and electrical conductivity under different band combinations was drawn, and the optimal spectral band combination corresponding to the highest correlation coefficient between the two-dimensional spectral index and electrical conductivity was determined. According to the optimal spectral band combination of each type of two-dimensional spectral index, six different types of two-dimensional spectral index sets corresponding to all sampling points were calculated, including DI two-dimensional spectral index set D2-1, NDI two-dimensional spectral index set D2-2, RI two-dimensional spectral index set D2-3, PSI two-dimensional spectral index set D2-4, PI two-dimensional spectral index set D2-5, and SRI two-dimensional spectral index set D2-6.

[0018] Furthermore, the seven three-dimensional soil salinity spectral indices in step 3 include: TI1(R λ1 ,R λ2 ,R λ3 )=R λ1 / (R λ2 *R λ3 )

[0019] TI2(R λ1 ,R λ2 ,R λ3 )=R λ1 / (R λ2 +R λ3 )

[0020] TI3(R λ1 ,R λ2 ,R λ3 )=(R λ1 -R λ2 ) / (R λ1 +R λ3 )

[0021] TI4(R λ1 ,R λ2 ,R λ3 )=(R λ1 -R λ2 ) / (R λ1 -R λ3 )

[0022] TI5(R λ1 ,R λ2 ,R λ3 )=(R λ1 +R λ2 ) / R λ3

[0023] TI6(R λ1 ,R λ2 ,R λ3 )=(R λ1 -R λ2 ) / [(Rλ1 -R λ2 )-(R λ2 -R λ3 )]

[0024] TI7(R λ1 ,R λ2 ,R λ3 )=(R λ1 -R λ2 )-(R λ2 -R λ3 )

[0025] Among them, Rλ1, Rλ2 and Rλ3 represent the reflectance data of any three bands between 350nm and 2500nm. For each type of three-dimensional soil salinity spectral index, the three-dimensional correlation coefficient diagram between its calculation results and conductivity under different band combinations is plotted, and the optimal spectral band combination corresponding to the highest correlation coefficient between the three-dimensional spectral index and conductivity is determined. According to the optimal spectral band combination of each type of three-dimensional spectral index, seven different types of three-dimensional spectral index sets corresponding to all sampling points are calculated, including TI1 three-dimensional spectral index set D3-1, TI2 three-dimensional spectral index set D3-2, TI3 three-dimensional spectral index set D3-3, TI4 three-dimensional spectral index set D3-4, TI5 three-dimensional spectral index set D3-5, TI6 three-dimensional spectral index set D3-6, and TI7 three-dimensional spectral index set D3-7.

[0026] Furthermore, in step 4, the SHAP analysis method is used to determine the importance of the six two-dimensional spectral index sets D2-1, D2-2, D2-3, D2-4, D2-5, and D2-6, as well as the importance of the seven three-dimensional spectral indices D3-1, D3-2, D3-3, D3-4, D3-5, D3-6, and D3-7; for the two-dimensional spectral index set, the two indices with the highest SHAP analysis importance are recorded as X1 and X2; for the three-dimensional spectral index set, the three indices with the highest SHAP analysis importance are recorded as X3, X4, and X5; the one-dimensional spectral index set D1-1 is recorded as X6; all sample conductivity data E The data set Ci is denoted as Y. The six spectral index sets of different dimensions X1i (i = 1, 2, 3, ..., 200), X2i (i = 1, 2, 3, ..., 200), X3i (i = 1, 2, 3, ..., 200), X4i (i = 1, 2, 3, ..., 200), X5i (i = 1, 2, 3, ..., 200), and X6i (i = 1, 2, 3, ..., 200) of the soil samples at all sampling points are used as independent variables, and the conductivity data set Yi (i = 1, 2, 3, ..., 200) of the soil samples at all sampling points is used as the dependent variable to establish a multiple linear regression model. The formula is:

[0027] Yi=a1*X1+a2*X2+a3*X3+a4*X4+a5*X5+a6*X6+a0

[0028] The partial least squares method is used to determine the model coefficients a0, a1, a2, a3, a4, a5, and a6 to complete the modeling of the soil conductivity hyperspectral prediction model; the hyperspectral remote sensing image HSI in the area to be measured is downloaded, and after preprocessing operations such as radiation correction and geometric correction, the hyperspectral remote sensing image HSI is resampled to the same band setting as the field measured spectral data, and the resampled hyperspectral remote sensing image is saved as HSI-1(I,J), where I is the total number of pixel rows of the remote sensing image, and J is the total number of pixel columns of the remote sensing image; The band with the highest correlation with conductivity in the hyperspectral image HSI-1(I,J) is taken as the one-dimensional spectral index band. The data results extracted from the two two-dimensional spectral index bands are recorded as X1(I,J) and X2(I,J), the data results extracted from the three three-dimensional spectral index bands are recorded as X3(I,J), X4(I,J) and X5(I,J), and the data result extracted from the one-dimensional spectral index band is recorded as X6(I,J). The value of each pixel in the spectral index band is brought into the conductivity prediction model to predict the conductivity of the pixel. The formula is:

[0029] Y(I,J)=a1*X1(I,J)+a2*X2(I,J)+a3*X3(I,J)+a4*X4(I,J)+a5*X5(I,J)+a6*X6(I,J)+a0

[0030] The conductivity prediction results of all pixels in the area to be measured are traversed and calculated to achieve synchronous measurement of field soil conductivity in the area to be measured.

[0031] Furthermore, the spectral width of the hyperspectral remote sensing image after resampling in step 4 is 10 nm.

[0032] The present invention has the following advantages and beneficial effects:

[0033] 1. Fast and efficient: Traditional measurement methods have limited measurement areas and large workloads. However, this invention, by setting up sampling grid areas and utilizing hyperspectral remote sensing technology and established prediction models, can achieve large-area synchronous measurement of soil conductivity, greatly shortening measurement time and improving measurement efficiency. It can quickly obtain large-area soil conductivity data, providing timely data support for soil research and management.

[0034] 2. Lower cost: The conductivity method requires sample collection and laboratory processing, which is expensive; the electromagnetic induction method also requires expensive instrumentation. This method eliminates the need for complex and expensive equipment and extensive sample collection and processing, reducing measurement costs and making soil conductivity measurement more economical and feasible, making it suitable for large-scale application.

[0035] 3. High Accuracy: Traditional remote sensing technology suffers from unstable inversion results due to atmospheric and other factors. This method uses systematic spectral measurements, calculates a variety of spectral indices, and uses the SHAP analysis method to screen important spectral indices to establish a multivariate linear regression model. This reduces interference factors and improves model accuracy. It can accurately measure soil electrical conductivity and provide reliable data for soil salinization monitoring and control.

[0036] 4. Non-destructive measurement: Compared with the traditional conductivity method that requires collecting soil samples, the present invention uses hyperspectral remote sensing technology, which does not require destroying the soil. It is a non-destructive measurement that can maintain the original state of the soil without affecting the soil ecological environment, and meets the environmental protection requirements of modern soil monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 Schematic diagram of hyperspectral field measurement in the soil sample area;

[0038] Figure 2 Schematic diagram of field conductivity measurement in the soil sample area. DETAILED DESCRIPTION

[0039] Example 1

[0040] Step 1: The software used includes MATLAB, and the hardware used includes a laptop, ruler, and ground object spectrometer. Spectral measurements were performed using the ASD FieldSpec 3, a portable field hyperspectral spectrometer produced by Analytical Spectral Devices (ASD). The spectrometer has a measurement range of 350 nm to 2500 nm and contains two detectors: a VNIR detector measures the spectral characteristics of the target object from 350 nm to 1000 nm, with a sampling interval of 1.4 nm and a spectral resolution of 3 nm; and a SWIR detector measures the spectral characteristics of the target object from 1000 nm to 2500 nm, with a sampling interval of 2 nm and a spectral resolution of 10 nm. After the spectrometer measures the ground object, the instrument resamples the spectral measurement results of different bands to 1 nm. The specific surface spectral measurement process is as follows: after determining the area to be measured for conductivity, 200 square regular grid sampling areas are evenly set up at intervals of 100 m in the north-south and east-west directions within the measurement area according to the systematic grid point distribution method. Each sampling point in the grid sampling area is recorded as Pi (i = 1, 2, 3, ..., 200), and the longitude information XPi (i = 1, 2, 3, ..., 200) and latitude position information YPi (i = 1, 2, 3, ..., 200) of each sampling point are recorded using a handheld GPS. For each sampling point Pi in the grid sampling area, a fiber optic lens with a measuring angle of 25° was used. A ruler was used to measure and ensure that the center of the fiber optic lens was 50 cm vertical to the ground surface. The spectral measurement of the sampling point Pi in the grid sample area was repeated 10 times and the spectral measurement results were saved. The 10 repeated measurement results of the sampling point Pi were read using MATLAB software, and the average of the 10 spectral reflectance data was calculated according to the spectral band. The 10 average spectral reflectance data of the full band of 350-2500nm measured at the sampling point Pi were further spectrally resampled with a step size of 10nm, and the 10nm spectral resampled reflectance data of the sampling point Pi were recorded as RPij (i=1,2,3,...,200,j=350,360,370,...,2500).

[0041] Step 2: The hardware used includes a tape measure, a protractor, a cutting machine, a steel plate with a thickness of 2mm and a width of 1cm, an electric drill, a 3.2mm high-hardness drill bit, 3mm screws and nuts, a thin wire, a compass, a ring cutter, a shovel, a plastic ziplock bag, an electronic scale, a dryer, a measuring cup, a glass rod, a conductivity meter, etc. Use a cutting machine to cut three 1m long steel plates, install a 3.2mm high-hardness drill bit on the electric drill and drill holes at the center position 1cm away from the edges of both sides of each cut steel plate, fix the three cut and drilled steel plates with 3mm screws and nuts and combine them into an equilateral triangle steel sample frame with a side length of 1m, use one section of three thin wires to fix the three vertices of the equilateral triangle steel sample frame, use a tape measure to measure and determine the center position point of each vertex of the equilateral triangle steel sample frame opposite to the side length, fix the other end of the thin wire after each vertex is fixed at the center position point of each side length, and use a protractor to ensure that the angle between the thin wire and the side length is perpendicular. The intersection of the three thin wires is the midpoint of the equilateral triangle steel sample frame, and finally use a thin wire to fix the midpoint of the sampling frame.

[0042] Within the area where soil conductivity is to be measured, after performing a spectral measurement at sampling point Pi, mark the projection of the fiber optic probe on the ground. Then, use a compass to adjust the equilateral triangle steel sample frame, ensuring that the center of the sample frame coincides with the projection of the fiber optic probe on the ground and that one side of the sample frame is parallel to the east-west direction. Then, mark the positions of the three vertices of the sample frame on the ground at this time, Pi1, Pi2, and Pi3. Use a circular cutter to collect soil samples from Pi1, Pi2, and Pi3, 0-20 cm apart, mix them, and place them in a ziplock bag. This serves as the soil sample Qi (i = 1, 2, 3, ..., 200) for sampling point Pi. The soil sample Qi was mixed, dried, ground, and passed through a standard 2 mm sieve. 20 g of the sieved soil sample was weighed at each sampling point Pi using an electronic scale. 100 ml of distilled water was weighed using a measuring cup. The 20 g of sieved soil sample was placed in a 100 ml distilled water measuring cup. The sample was thoroughly stirred with a glass rod and allowed to stand for 20 minutes. The soil electrical conductivity (ECi) at each sampling point Pi was then measured using a conductivity meter (i = 1, 2, 3, ..., 200).

[0043] Step 3: Based on the Pearson correlation coefficient between the reflectance data of all sampling points Pi (i = 1, 2, 3, ..., 200) at each band position and the conductivity data ECi (i = 1, 2, 3, ..., 200), determine the band with the highest correlation coefficient, extract the spectral reflectance data of all sampling points in this band, and use it as the one-dimensional spectral index set D1-1 for all sample points. For all sampling points Pi (i = 1, 2, 3, ..., 200), select the reflectance of any two bands and calculate six different types of two-dimensional soil salinity spectral indices under all band combinations to highlight the quantitative relationship with soil conductivity. The six different types of two-dimensional soil salinity spectral indices include: DI (R λ1 ,R λ2 )=R λ1 -R λ2

[0044] NDI(R λ1 ,R λ2 )=(R λ1 -R λ2 ) / (R λ1 +R λ2 )

[0045] RI(R λ1 ,R λ2 )=R λ1 / R λ2

[0046] PSI(R λ1 ,R λ2 )=(R λ1 +M*R λ2 ) / (M 2 +1) 0.5

[0047] PI(R λ1 ,R λ2 )=R λ1 *R λ2

[0048] SRI(R λ1 ,R λ2 )=(R λ1 2 +R λ2 2 ) 0.5

[0049] Among them, R λ1 and R λ2Represents the reflectance data of any two bands between 350nm and 2500nm, M represents the slope of the soil line, and the M value in the present invention is 1.0563. For each type of two-dimensional soil salinity spectral index, a two-dimensional correlation coefficient diagram is plotted between the calculation results under different band combinations and the electrical conductivity, and the optimal spectral band combination corresponding to the highest value of the correlation coefficient between the two-dimensional spectral index and the electrical conductivity is determined. Based on the optimal spectral band combination of each type of two-dimensional spectral index, six different types of two-dimensional spectral index sets corresponding to all sampling points are calculated, including the DI two-dimensional spectral index set D2-1, the NDI two-dimensional spectral index set D2-2, the RI two-dimensional spectral index set D2-3, the PSI two-dimensional spectral index set D2-4, the PI two-dimensional spectral index set D2-5, and the SRI two-dimensional spectral index set D2-6.

[0050] For all sampling points Pi (i = 1, 2, 3, ..., 200), the reflectance of any three bands is calculated. Seven different types of three-dimensional soil salinity spectral indices are calculated for all band combinations to highlight the quantitative relationship with soil electrical conductivity. The seven different types of three-dimensional soil salinity spectral indices include:

[0051] TI1(R λ1 ,R λ2 ,R λ3 )=R λ1 / (R λ2 *R λ3 )

[0052] TI2(R λ1 ,R λ2 ,R λ3 )=R λ1 / (R λ2 +R λ3 )

[0053] TI3(R λ1 ,R λ2 ,R λ3 )=(R λ1 -R λ2 ) / (R λ1 +R λ3 )

[0054] TI4(R λ1 ,R λ2 ,R λ3 )=(R λ1 -R λ2 ) / (R λ1 -R λ3 )

[0055] TI5(R λ1 ,R λ2 ,R λ3 )=(Rλ1 +R λ2 ) / R λ3

[0056] TI6(R λ1 ,R λ2 ,R λ3 )=(R λ1 -R λ2 ) / [(R λ1 -R λ2 )-(R λ2 -R λ3 )]

[0057] TI7(R λ1 ,R λ2 ,R λ3 )=(R λ1 -R λ2 )-(R λ2 -R λ3 )

[0058] Among them, R λ1 , R λ2 and R λ3 Reflectance data for any three bands between 350nm and 2500nm are represented. For each type of three-dimensional soil salinity spectral index, a three-dimensional correlation coefficient diagram is plotted between the calculated results and electrical conductivity under different band combinations. The optimal spectral band combination corresponding to the highest correlation coefficient between the three-dimensional spectral index and electrical conductivity is determined. Based on the optimal spectral band combination for each type of three-dimensional spectral index, seven different types of three-dimensional spectral index sets corresponding to all sampling points are calculated, including the three-dimensional spectral index set D3-1 for TI1, the three-dimensional spectral index set D3-2 for TI2, the three-dimensional spectral index set D3-3 for TI3, the three-dimensional spectral index set D3-4 for TI4, the three-dimensional spectral index set D3-5 for TI5, the three-dimensional spectral index set D3-6 for TI6, and the three-dimensional spectral index set D3-7 for TI7.

[0059] Step 3: Using the SHAP analysis method, calculate the importance of the six 2D spectral indices at all sampling points and select the two most important 2D spectral indices. Calculate the importance of the seven 3D spectral indices at all sampling points and select the three most important 3D spectral indices. A multivariate linear regression model was established using the 1D spectral indices, the two most important 2D spectral indices, and the three most important 3D spectral indices as independent variables and the conductivity measurements as the dependent variable.

[0060] Download the hyperspectral remote sensing image of the area to be measured, perform radiation correction, geometric correction and spectral resampling on the hyperspectral remote sensing image, extract the band with the highest correlation with conductivity in the hyperspectral image as the one-dimensional spectral index band, extract and calculate the two most important two-dimensional spectral index bands, and extract and calculate the three most important three-dimensional spectral index bands. Substitute the value of each pixel in the above spectral index bands into the conductivity prediction model to predict the conductivity of the pixel, traverse and calculate the conductivity prediction results of all pixels in the area to be measured, and realize the synchronous measurement of field soil conductivity in the area to be measured.

[0061] Step 4: Use the SHAP analysis method to determine the importance of the six two-dimensional spectral index sets (D2-1, D2-2, D2-3, D2-4, D2-5, and D2-6), as well as the seven three-dimensional spectral indices (D3-1, D3-2, D3-3, D3-4, D3-5, D3-6, and D3-7). For the two-dimensional spectral index set, denote the two indices with the highest SHAP importance as X1 and X2. For the three-dimensional spectral index set, denote the three indices with the highest SHAP importance as X3, X4, and X5. Denote the one-dimensional spectral index set D1-1 as X6. Denote the dataset of all sample conductivity data ECi as Y. The six spectral index sets of soil samples from all sampling points, namely X1i (i = 1, 2, 3, ..., 200), X2i (i = 1, 2, 3, ..., 200), X3i (i = 1, 2, 3, ..., 200), X4i (i = 1, 2, 3, ..., 200), X5i (i = 1, 2, 3, ..., 200), and X6i (i = 1, 2, 3, ..., 200) of different dimensions, were used as independent variables, and the conductivity data set Yi (i = 1, 2, 3, ..., 200) of soil samples from all sampling points was used as the dependent variable. A multiple linear regression model Yi = a1*X1+a2*X2+a3*X3+a4*X4+a5*X5+a6*X6+a0 was established. The partial least squares method is used to determine the model coefficients a0, a1, a2, a3, a4, a5, and a6 to complete the modeling of the soil electrical conductivity hyperspectral prediction model.

[0062] Download the hyperspectral remote sensing image (HSI) for the area to be measured. After preprocessing the image through radiometric and geometric corrections, resample the HSI to a 10nm spectral width, the same band setting as the field-measured spectral data. Save the resampled image as HSI-1(I,J), where I is the total number of rows and J is the total number of columns. Extract the band with the highest correlation with conductivity from the hyperspectral image HSI-1(I,J) as the one-dimensional spectral index band. The data extracted from the two two-dimensional spectral index bands are denoted as X1(I,J) and X2(I,J), the data extracted from the three three-dimensional spectral index bands are denoted as X3(I,J), X4(I,J), and X5(I,J), and the data extracted from the one-dimensional spectral index band is denoted as X6(I,J). The conductivity prediction model is used to derive the value of each pixel in the spectral index bands described above. This prediction is expressed as: Y(I,J) = a1*X1(I,J) + a2*X2(I,J) + a3*X3(I,J) + a4*X4(I,J) + a5*X5(I,J) + a6*X6(I,J) + a0. This method is used to traverse and calculate the conductivity prediction results for all pixels in the area to be measured, enabling simultaneous measurement of soil conductivity in the field.

Claims

1. A method for rapid measurement of soil electrical conductivity using hyperspectral remote sensing technology, characterized in that: The following steps are involved: Step 1: Set up a regular square grid sampling area in the area to be measured according to the systematic grid point method. Use a 350-2500nm ground object spectrometer to perform 10 repeated spectral measurements at the center point of each sampling grid area at a measurement angle of 25° and a height of 50cm vertically above the ground. Calculate the mean of the 10 measured spectral data by band and resample to 10nm. Step 2: In each grid sample area where spectral measurements have been completed, determine an equilateral triangle with a side length of 2m in the north, east, and west directions, with the center of the spectrometer surface as the triangle center. Collect 0-20 cm surface soil samples at the three vertices of the triangle. Bring them back to the laboratory for mixing, drying, grinding, and passing through a 2mm sieve. Prepare a standard soil suspension with a water-soil mass ratio of 5:1, and measure the soil electrical conductivity value using the electrode method; Step 3: Based on the Pearson correlation coefficient between the reflectance of each band and the electrical conductivity at the sampling point, the spectral reflectance of the band with the highest correlation coefficient is extracted as the one-dimensional spectral index set; the reflectance of any two bands is used to calculate six two-dimensional soil salinity spectral indices, and a two-dimensional correlation coefficient diagram of the indices with electrical conductivity is plotted. The optimal spectral band combination corresponding to the highest correlation coefficient between each type of two-dimensional spectral index and electrical conductivity is determined, and the six two-dimensional spectral index sets for all sampling points under this combination are calculated; the reflectance of any three bands is used to calculate seven three-dimensional soil salinity spectral indices, and a three-dimensional correlation coefficient diagram of the indices with electrical conductivity is plotted. The optimal spectral band combination corresponding to the highest correlation coefficient between each type of three-dimensional spectral index and electrical conductivity is determined, and the seven three-dimensional spectral index sets for all sampling points under this combination are calculated; Step 4: Using the SHAP analysis method, the two most important two-dimensional spectral indices and three three-dimensional spectral indices were selected. A multivariate linear regression model was established with the one-dimensional spectral index, the selected two-dimensional spectral index, and the three-dimensional spectral index as independent variables, and the conductivity measurement value as the dependent variable. The hyperspectral remote sensing image of the area to be measured was downloaded, and radiation correction, geometric correction, and spectral resampling were performed. The band with the highest correlation with conductivity was extracted as the one-dimensional spectral index band. The two most important two-dimensional spectral index bands and three three-dimensional spectral index bands were extracted and calculated. The pixel values ​​of each spectral index band were introduced into the conductivity prediction model, and the conductivity prediction results of all pixels were traversed and calculated to achieve synchronous measurement of soil conductivity in the field.

2. The method for rapid measurement of soil electrical conductivity using hyperspectral remote sensing technology according to claim 1, characterized in that: The ground object spectrometer described in step 1 includes a VNIR detector and a SWIR detector. The VNIR detector has a measurement range of 350-1000 nm, a sampling interval of 1.4 nm, and a spectral resolution of 3 nm; the SWIR detector has a measurement range of 1000-2500 nm, a sampling interval of 2 nm, and a spectral resolution of 10 nm.

3. The method for rapid measurement of soil electrical conductivity using hyperspectral remote sensing technology according to claim 1, wherein: The six two-dimensional soil salinity spectral indices described in step 3 include: DI(R λ1 ,R λ2 )=R λ1 -R λ2 AND(R λ1 ,R λ2 )=(R λ1 -R λ2 ) / (R λ1 +R λ2 ) RI(R λ1 ,R λ2 )=R λ1 / R λ2 PSI(R λ1 ,R λ2 )=(R λ1 +M*R λ2 ) / (M 2 +1) 0.5 PI(R λ1 ,R λ2 )=R λ1 *R λ2 SRI(R λ1 ,R λ2 )=(R λ1 2 +R λ2 2 ) 0.5 Among them, R λ1 and R λ2 Respectively represent the reflectance data of any two bands between 350nm and 2500nm, M represents the slope of the soil line, M1.0563; For each type of two-dimensional soil salinity spectral index, a two-dimensional correlation coefficient diagram between its calculation results and electrical conductivity under different band combinations was drawn, and the optimal spectral band combination corresponding to the highest correlation coefficient between the two-dimensional spectral index and electrical conductivity was determined. According to the optimal spectral band combination of each type of two-dimensional spectral index, six different types of two-dimensional spectral index sets corresponding to all sampling points were calculated, including DI two-dimensional spectral index set D2-1, NDI two-dimensional spectral index set D2-2, RI two-dimensional spectral index set D2-3, PSI two-dimensional spectral index set D2-4, PI two-dimensional spectral index set D2-5, and SRI two-dimensional spectral index set D2-6.

4. The method for rapid measurement of soil electrical conductivity using hyperspectral remote sensing technology according to claim 1, wherein: The seven three-dimensional soil salinity spectral indices described in step 3 include: TI1(R λ1 ,R λ2 ,R λ3 )=R λ1 / (R λ2 *R λ3 ) TI2(R λ1 ,R λ2 ,R λ3 )=R λ1 / (R λ2 +R λ3 ) TI3(R λ1 ,R λ2 ,R λ3 )=(R λ1 -R λ2 ) / (R λ1 +R λ3 ) TI4(R λ1 ,R λ2 ,R λ3 )=(R λ1 -R λ2 ) / (R λ1 -R λ3 ) TI5(R λ1 ,R λ2 ,R λ3 )=(R λ1 +R λ2 ) / R λ3 TI6(R λ1 ,R λ2 ,R λ3 )=(R λ1 -R λ2 ) / [(R λ1 -R λ2 )-(R λ2 -R λ3 )] TI7(R λ1 ,R λ2 ,R λ3 )=(R λ1 -R λ2 )-(R λ2 -R λ3 ) Among them, Rλ1, Rλ2 and Rλ3 represent the reflectance data of any three bands between 350nm and 2500nm. For each type of three-dimensional soil salinity spectral index, the three-dimensional correlation coefficient diagram between its calculation results and conductivity under different band combinations is plotted, and the optimal spectral band combination corresponding to the highest correlation coefficient between the three-dimensional spectral index and conductivity is determined. According to the optimal spectral band combination of each type of three-dimensional spectral index, seven different types of three-dimensional spectral index sets corresponding to all sampling points are calculated, including TI1 three-dimensional spectral index set D3-1, TI2 three-dimensional spectral index set D3-2, TI3 three-dimensional spectral index set D3-3, TI4 three-dimensional spectral index set D3-4, TI5 three-dimensional spectral index set D3-5, TI6 three-dimensional spectral index set D3-6, and TI7 three-dimensional spectral index set D3-7.

5. The method for rapid measurement of soil electrical conductivity using hyperspectral remote sensing technology according to claim 1, wherein: Step 4: Use the SHAP analysis method to determine the importance of the six two-dimensional spectral index sets D2-1, D2-2, D2-3, D2-4, D2-5, and D2-6, and the importance of the seven three-dimensional spectral indices D3-1, D3-2, D3-3, D3-4, D3-5, D3-6, and D3-7; For the two-dimensional spectral index set, the two indices with the highest importance index of SHAP analysis are recorded as X1 and X2; For the three-dimensional spectral index set, the three indices with the highest importance index in SHAP analysis are recorded as X3, X4, and X5; The one-dimensional spectral index set D1-1 is recorded as X6; the data set of all sample conductivity data ECi is recorded as Y; the X1i of all sampling point soil samples is recorded as The six spectral index sets of different dimensions are used as independent variables, and the conductivity data set Yi (i = 1, 2, 3, ..., 200) of soil samples at all sampling points is used as the dependent variable to establish a multiple linear regression model. The formula is: Yi=a1*X1+a2*X2+a3*X3+a4*X4+a5*X5+a6*X6+a0 The partial least squares method is used to determine the model coefficients a0, a1, a2, a3, a4, a5, and a6 to complete the modeling of the soil conductivity hyperspectral prediction model; the hyperspectral remote sensing image HSI in the area to be measured is downloaded, and after preprocessing operations such as radiation correction and geometric correction, the hyperspectral remote sensing image HSI is resampled to the same band setting as the field measured spectral data, and the resampled hyperspectral remote sensing image is saved as HSI-1(I,J), where I is the total number of pixel rows of the remote sensing image, and J is the total number of pixel columns of the remote sensing image; The band with the highest correlation with conductivity in the hyperspectral image HSI-1(I,J) is taken as the one-dimensional spectral index band. The data results extracted from the two two-dimensional spectral index bands are recorded as X1(I,J) and X2(I,J), the data results extracted from the three three-dimensional spectral index bands are recorded as X3(I,J), X4(I,J) and X5(I,J), and the data result extracted from the one-dimensional spectral index band is recorded as X6(I,J). The value of each pixel in the spectral index band is brought into the conductivity prediction model to predict the conductivity of the pixel. The formula is: Y(I,J)=a1*X1(I,J)+a2*X2(I,J)+a3*X3(I,J)+a4*X4(I,J)+a5*X5(I,J)+a6*X6(I,J)+a0 traverses and calculates the conductivity prediction results of all pixels in the area to be measured, realizing the synchronous measurement of field soil conductivity in the area to be measured.

6. The method for rapid measurement of soil electrical conductivity using hyperspectral remote sensing technology according to claim 1 or 5, characterized in that: The spectral width of the hyperspectral remote sensing image after resampling in step 4 is 10 nm.

Images