Soil texture classification-based soil humidity index set construction method and system

Abstract

The invention relates to the technical field of soil humidity calculation, in particular to a soil humidity index set construction method and system based on soil texture classification. According to the method, spectral reflection characteristic changes caused by composition and structure differences of soil particles when the soil textures are different are fully considered, soil types are divided, spectral differences of different soil textures are fully considered, and a wave band most sensitive to soil moisture is screened, so that a soil humidity index set is constructed based on the wave band; according to the method, soil samples with part of attribute information missing and corresponding spectral data are eliminated, and a plurality of spectrums with the highest soil water content in each group of data are further eliminated, so that the stability of the spectrums is ensured, and the representativeness of the samples is ensured, and a more accurate soil humidity index set can be conveniently obtained subsequently; the technical problem that the prediction precision is reduced due to the fact that an existing index generally adopts a dual-band form and neglects the difference of different soil textures is solved.

Description

Technical Field

[0001] The present invention relates to the technical field of soil moisture calculation, and in particular to a method and system for constructing a soil moisture index set based on soil texture classification. Background Art

[0002] The soil moisture index is a mathematical calculation method used to estimate soil moisture content. Its basic principle is that soils with different water contents have different reflectivity in different bands of light. When the soil moisture is higher, light in the short-wave infrared band is absorbed more and the reflectivity decreases, while the reflectivity of some near-infrared bands changes less. Using this characteristic, we can combine the reflectivity of these bands through mathematical calculations to form a soil moisture index, thereby indirectly estimating the soil moisture content.

[0003] Existing indices usually adopt a dual-band format, failing to fully utilize the multi-band characteristics of the short-wave near-infrared bands of satellite remote sensing, limiting the accuracy and sensitivity of soil moisture monitoring. Secondly, the differences in different soil textures (such as clay, sand, and loam) are generally ignored during the construction process of existing indices, resulting in poor adaptability of the index to different soil textures and the prediction accuracy may be significantly reduced. Summary of the Invention

[0004] In response to the shortcomings of the existing technology, the present invention provides a method and system for constructing a soil moisture index set based on soil texture classification, which solves the technical problem that the existing index usually adopts a dual-band form and ignores the differences between different soil textures, resulting in reduced prediction accuracy.

[0005] To solve the above technical problems, the present invention provides the following technical solution: a method for constructing a soil moisture index set based on soil texture classification, the construction method specifically comprising the following steps: S1. Obtain several different types of soil samples and spectral data; S2. Analyze the correlation between the water content of different types of soil and the reflectance of several single bands, and select the bands with the highest correlation as the first sensitive bands; S3. Analyze the importance of each band's influence on soil moisture content, and select several bands with the largest importance index to obtain the second most sensitive band; S4. Calculate the average of the reflectance differences between soil samples with zero water content and soil samples with the highest water content in each band among all soil types, and select several bands with the largest average values ​​as the third sensitive bands; S5. Calculate the intersection of the first sensitive band, the second sensitive band, and the third sensitive band to obtain a fourth sensitive band, and construct several soil moisture index formulas applicable to different soil types based on the fourth sensitive band; S6. Based on the soil index formula, a fitting model between the index of different soil types and soil moisture content is constructed, and the soil index formula corresponding to the best fitting model is selected to obtain a soil moisture index set suitable for different soil moisture content analysis.

[0006] Preferably, in step S1, the following steps are specifically included: S11, obtaining a number of soil samples and corresponding spectral data; S12. Divide the soil samples into several types, and each type of soil sample contains at least one soil sample with a water content of 0 and several soil samples with a non-zero water content; S13, removing soil samples and corresponding spectral data with missing attribute information from each type of soil samples; S14. Eliminate several soil samples with the highest water content and the corresponding spectral data from each type of soil samples.

[0007] Preferably, in step S12, the soil types include silt loam, clay, loam, clay loam, sandy soil, silt clay loam, silt clay, sandy loam, sandy clay loam and loamy sand, and the attribute information of the soil samples in each soil type is the same.

[0008] Preferably, in step S2, the following steps are specifically included: S21, setting a first band range for different types of soil, and dividing the first band range into a plurality of single bands; S22. Using the reflectance of several single bands as the independent variable and the measured soil moisture content as the dependent variable, linear regression was used to calculate the correlation between each band and soil moisture content; S23, RMSE, nRMSE, As the evaluation index of correlation, several bands with the highest index values ​​are selected as the first sensitive bands.

[0009] Preferably, in step S3, the following steps are specifically included: S31, setting a second band range for different types of soil, and dividing the second band range into a plurality of single bands; S32. Random forest analysis was used to calculate the importance index of each band to soil moisture content; S33. Select several single bands with the highest importance as the second sensitive bands.

[0010] Preferably, in step S4, the following steps are specifically included: S41. Select soil samples with zero water content and the highest water content from different soil types; S42. subtract the spectral reflectance of a soil sample with zero water content from the spectral reflectance of the soil sample with the highest water content in the same soil type to obtain a spectral reflectance difference; S43. Calculate the average of the spectral reflectance differences of all soil types; S44. Select several bands with the largest average value of spectral reflectance difference as the third sensitive band.

[0011] Preferably, in step S5, the following steps are specifically included: S51, calculating the intersection of the first sensitive band, the second sensitive band, and the third sensitive band to obtain a fourth sensitive band; S52. Based on the fourth sensitive band and GPT-4o, several soil moisture index formulas suitable for different soil types are constructed.

[0012] Preferably, in step S6, the following steps are specifically included: S61. Calculate the soil moisture index according to the soil moisture index formula and the soil spectrum data of the soil sample in the database; S62. Based on the soil moisture index and water content of the soil samples, construct a fitting model between the soil moisture index and soil water content of different soil types; S63, setting a number of evaluation indicators, and marking the soil moisture index formula corresponding to the fitting model with the highest value of the evaluation indicators as the standard soil moisture index formula; S64. Obtain standard soil moisture index formulas corresponding to different soil types to establish a soil moisture index set suitable for different soil moisture content analyses.

[0013] The present invention also provides a soil moisture index set construction system, comprising a processor and a memory, wherein the memory is used to store a computer program, and when the computer program is executed by the processor, any one of the soil moisture index set construction methods is implemented.

[0014] Through the above technical solution, the present invention provides a method and system for constructing a soil moisture index set based on soil texture classification, which has at least the following beneficial effects: 1. This invention fully considers the changes in spectral reflectance characteristics caused by differences in soil particle composition and structure when the soil texture is different, classifies soil types, and fully considers the spectral differences of different soil textures to screen the bands most sensitive to soil moisture. Based on this, a soil moisture index set is constructed to improve the accuracy of estimating soil moisture content from spectral data.

[0015] 2. The present invention eliminates soil samples and corresponding spectral data with missing attribute information, and further eliminates several spectra with the highest soil moisture content in each set of data to ensure the stability of the spectrum, thereby ensuring the representativeness of the sample and facilitating the subsequent acquisition of a more accurate soil moisture index set.

[0016] 3. The present invention selects the first sensitive band, the second sensitive band and the third sensitive band to obtain a fourth sensitive band that is highly sensitive to soil moisture changes, so as to obtain a more accurate soil moisture index set based on the fourth sensitive band. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings: Figure 1 A flowchart of a method for constructing a soil moisture index set based on soil texture classification according to the present invention; Figure 2 Schematic diagram of the single-band regression analysis results of the present invention; Figure 3 Schematic diagram of the random forest analysis results of the present invention; Figure 4 Schematic diagram of the difference analysis results of dry and wet soil spectral reflectance according to the present invention; Figure 5 Thermal map for estimating soil moisture content for existing soil moisture index formula; Figure 6 A thermal map of soil moisture content estimated using the soil moisture index formula of the present invention; Figure 7 Comparison of the lowest RMSE of the soil moisture index formula of the present invention and the existing soil moisture index formula in different soil types; Figure 8 The soil moisture index formula of the present invention and the existing soil moisture index formula are different in different soil types. Increase rate; Figure 9 Radar chart comparing nRMSE of the soil moisture index formula of the present invention and the existing soil moisture index formula in different soil types; Figure 10 Schematic diagram of the fitting results of SiCLMI and SMC. DETAILED DESCRIPTION

[0018] To make the above-mentioned objectives, features, and advantages of the present invention more clearly understood, the present invention is further described below in detail with reference to the accompanying drawings and specific embodiments. This will enable a full understanding of how this application uses technical means to solve technical problems and achieve technical effects, and to implement the invention accordingly.

[0019] The specific prediction process for estimating soil moisture content (SMC) includes two key steps: first, using experimental data, the relationship between a specific soil moisture index and the actual soil moisture content (SMC) is calculated, and statistical regression methods (such as linear regression and nonlinear fitting) are used to establish an index-moisture content fitting model, that is, to find the mathematical relationship between the index and SMC. Second, in practical applications, soil spectral data is obtained through satellites, drones, or ground sensors, and the corresponding soil moisture index is calculated. Then, the calculated index is substituted into the fitting formula to estimate the soil moisture content. This method is widely used in agriculture, meteorology, water resources management, and ecological monitoring, helping scientists and farmers accurately monitor soil moisture changes and improve agricultural production efficiency. It can also be used for drought monitoring and water management, providing a scientific basis for climate research and natural disaster warning.

[0020] In order to solve the problem that the existing technology usually adopts a dual-band form, which limits the accuracy and sensitivity of soil moisture monitoring and ignores the differences in different soil textures, resulting in poor adaptability of the index to different soil textures and reduced soil moisture prediction accuracy, this embodiment proposes a soil moisture index set construction method and system based on soil texture classification, such as Figure 1 As shown in FIG, the method fully considers the changes in spectral reflectance characteristics caused by differences in soil particle composition and structure when the soil texture is different, so as to improve the versatility and reliability of the soil moisture index. The method includes the following steps: S1. Obtain several different types of soil samples and spectral data. After obtaining the soil samples, in order to ensure their representativeness, they need to be preprocessed. For example, experimental data containing soil samples and corresponding spectral data can be obtained from the MARMIT2020 data repository. The database contains 222 sets of soil spectral data, a total of 1894 spectra, covering multiple soil textures and regions. The spectral coverage range is 350nm-2500nm, and the spectral resolution is 1nm, which is highly extensive and representative. Each set of soil spectral data contains a dry soil spectrum and several soil spectra with different soil moisture contents. Their soil composition and properties are the same. The attribute information of soil data includes clay, silt, sand, soil organic matter, CaCO, Mg, Al, P, Ca, Mn, Fe, Zn and Ni content, pH value, cation exchange capacity, etc., but some soil The attribute information of soil samples was missing. Therefore, to ensure the accuracy of classification, this study eliminated 26 groups of soil samples with missing clay, silt or sand content and the corresponding spectral data (a total of 384 data, each data for a soil sample and the corresponding spectral data). During the data processing process, it was found that the spectra of some high-water content soils showed abnormal spectral fluctuations due to saturation effects and other reasons. Therefore, this study further eliminated several spectra with the highest soil water content in each group of data, and finally retained 196 groups, totaling 1314 spectral data. The soil samples and spectral data were then classified according to soil texture. Considering that the 1314 soil samples and spectral data belonged to 196 groups, and the attributes of each type of soil sample were exactly the same, the classification work was carried out based on the spectral data of 196 groups, and the soil samples used in this study were classified into 10 soil texture types, as shown in Table 1 below:

[0021] The following is a further description of the classification method for soil samples and corresponding spectral data, which specifically includes the following steps: S11, obtaining a number of soil samples and corresponding spectral data; S12. Divide the soil samples into several types, and each type of soil sample contains at least one soil sample with a water content of zero and several soil samples with a non-zero water content; the soil types include silt loam, clay, loam, clay loam, sandy soil, silt clay loam, silt clay, sandy loam, sandy clay loam, and loamy sand, and the attribute information of the soil samples in each soil type is the same.

[0022] S13, removing soil samples and corresponding spectral data with missing attribute information from each type of soil samples; S14. Eliminate several soil samples with the highest water content and the corresponding spectral data from each type of soil samples.

[0023] S2. Analyze the correlation between the moisture content of different types of soil and the reflectance of several single bands, and select the bands with the highest correlation as the first sensitive bands. Before the start of this study, we first conducted a systematic evaluation of existing soil moisture indices and found that most existing indices are based on the SWIR (shortwave infrared) band (1300nm–2500nm). However, the band selection is relatively simple. Many indices do not fully consider the spectral differences of different soil textures, resulting in poor prediction accuracy under certain soil textures. The index formula structure is mostly a simple ratio or normalization form, which does not fully utilize multi-band combinations and various mathematical functions. To overcome the above limitations, this study adopted the following strategy for sensitive band screening: single-band regression analysis, random forest feature importance analysis, and dry-wet soil spectral reflectance difference analysis were used to screen the bands most sensitive to soil moisture. The results showed that 1900-2100nm, 1450nm, 2200nm, 1640nm, and 2130nm are the most representative soil moisture sensitive bands. The sensitive band screening process specifically includes the following steps: S21. Set the first band range for different types of soil and divide the first band range into several single bands, as shown in Table 2 below. In the evaluation of existing soil moisture indices, this paper counted the bands used. The results showed that most soil moisture indices mainly rely on the short-wave infrared band (SWIR, 1300nm–2500nm), and the formulas of a few indices also include the near-infrared band (NIR, 780nm–1300nm) and the red light band (Red, 620nm–700nm). Previous studies have shown that these bands are soil moisture-sensitive bands that have been widely verified by scholars and can well reflect changes in soil moisture content. Among them, the short-wave infrared band is particularly sensitive to soil moisture absorption characteristics and is widely used in soil moisture remote sensing monitoring. The near-infrared and red light bands also show certain moisture response capabilities under specific conditions.

[0024]

[0025] In order to explore the band that is most sensitive to soil moisture changes, this paper conducted a fitting analysis of single band reflectance and soil moisture content (SMC). Specifically, the reflectance of multiple single bands was selected as independent variables in the 1300nm-2300nm band range (i.e., the first band range) at 20nm intervals, and the measured soil moisture content (SMC) was used as the dependent variable. Linear regression was used to calculate the correlation between each band and SMC, and RMSE, nRMSE, As evaluation indicators, such as Figure 2As shown in Table 3 below, for ease of observation, the 10 bands with the best fitting effects are summarized, and the results of the fitting analysis of all the data are presented. The results show that the bands with higher SMC fitting accuracy are concentrated around 1900-2100nm, among which the bands 1920nm, 1940nm and 1960nm rank in the top three. In addition, there is also a peak around 1440nm, indicating that the nearby bands also have a certain correlation with SMC.

[0026]

[0027] S22. Using the reflectance of several single bands as the independent variable and the measured soil moisture content as the dependent variable, linear regression was used to calculate the correlation between each band and soil moisture content; S23, RMSE, nRMSE, As the evaluation index of correlation, several bands with the highest index values ​​are selected as the first sensitive bands.

[0028] S3. Analyze the importance of each band on soil moisture content, and select several bands with the largest importance index to obtain the second sensitive band. This paper also uses random forest analysis to select the reflectance of multiple single bands in the 1200nm-2300nm band range (the second band range mentioned below) at intervals of 20nm to evaluate the impact of each band on soil moisture and explore its importance ranking. Random forest is an integrated learning method based on decision trees that can effectively identify the importance of feature variables and reduce the overfitting problem of a single model. Through this method, this paper quantifies the sensitivity of different spectral bands to soil moisture changes, providing data support for subsequent index construction, such as Figure 3 As shown in the figure, the feature importance index of each band is obtained by performing random forest analysis on all soil data. Obviously, the feature importance index of the 1920 band and the 1940 band is the highest. The specific analysis process includes the following steps: S31, setting a second band range for different types of soil, and dividing the second band range into a plurality of single bands; S32. Random forest analysis was used to calculate the importance index of each band to soil moisture content; S33. Select several single bands with the highest importance as the second sensitive bands.

[0029] S4. Calculate the average value of the reflectance difference between soil samples with zero water content and the highest water content in each band for all soil types, and select several bands with the largest average value as the third sensitive band. Specifically, select the spectral reflectance corresponding to dry soil (water content is 0) and wet soil (highest water content in the group) from 196 groups of soil data, and calculate their spectral reflectance difference, and calculate the average value of the spectral reflectance difference between dry and wet soil for various soil types. This method can highlight the influence of soil moisture on spectral reflectance, thereby identifying moisture-sensitive bands, such as Figure 4 As shown in the figure, the average value image of the difference in spectral reflectance between dry and wet soil for all soil data is displayed. To ensure the clarity of the data, only the 1000-2300nm band is displayed. The results show that the image has two peaks, a short peak at around 1450nm and a high peak at around 1960nm. The band with the highest difference in spectral reflectance between dry and wet soil is around 1900-2100nm. The specific calculation process includes the following steps: S41. Select soil samples with zero water content and the highest water content from different soil types; S42. subtract the spectral reflectance of a soil sample with zero water content from the spectral reflectance of the soil sample with the highest water content in the same soil type to obtain a spectral reflectance difference; S43. Calculate the average of the spectral reflectance differences of all soil types; S44. Select several bands with the largest average value of spectral reflectance difference as the third sensitive band.

[0030] S5. Calculate the intersection of the first sensitive band, the second sensitive band, and the third sensitive band to obtain the fourth sensitive band, and construct several soil moisture index formulas suitable for different soil types based on the fourth sensitive band. The first sensitive band, the second sensitive band, and the third sensitive band screened out by steps S2-S4 above are all concentrated between 1900-2100nm, especially the bands in the range of 1900-2000nm. Specifically, in the single-band regression analysis, the fitting accuracy of the 1920nm, 1940nm, and 1960nm bands is the highest. In the random forest analysis, the importance index of the 1920nm and 1940nm bands is the highest, and The difference analysis of the spectral reflectance between dry and wet soil also shows that the spectral reflectance difference is the largest near 1960nm. This result indicates that the band in the range of 1900-2000nm is most sensitive to changes in soil moisture and can more accurately reflect the changing trend of soil moisture. However, it is worth noting that most of the existing soil moisture indices do not fully utilize this band range. Traditional indices tend to use bands such as 2200nm and 1650nm, while ignoring the key sensitive range of 1900-2000nm. Therefore, when constructing the new index, this article specifically introduces the 1900-2000nm band range to fully utilize the high sensitivity of this band to changes in soil moisture.

[0031] This paper uses GPT-4o to generate and optimize large-scale indices and constructs soil moisture index formulas suitable for different soil textures. GPT-4o has strong mathematical reasoning and complex pattern recognition capabilities, and can provide better solutions in multi-band combination and nonlinear function construction. Specifically, this study inputs key information such as soil moisture sensitive band screening results, existing index formulas and their performance into GPT-4o, and instructs it to combine multiple band combinations such as two-band, three-band, and four-band, and use logarithmic, power, exponential, square root and other functions to generate a large number of soil moisture index formulas (for example, the logarithmic function helps to smooth the spectral reflectance changes and improve the index stability, the power function can enhance the index's ability to distinguish between high and low moisture soils, and the exponential function amplifies different The researchers then calculated these new indices based on the classified soil spectral database and established a fitting model between the indices and soil moisture content (SMC). They then screened and analyzed the indices using metrics such as R², RMSE, nRMSE, and bias. The results of these screening analyses were fed back to GPT-4o, allowing it to construct a new round of indices based on the analysis results. The researchers then continued to calculate and evaluate the results. After multiple rounds of iterations, they ultimately determined 11 new indices with the best performance, enabling them to outperform the traditional soil moisture index in predicting each soil texture and unclassified soil data. The specific steps involved are as follows: S51, calculating the intersection of the first sensitive band, the second sensitive band, and the third sensitive band to obtain a fourth sensitive band; S52. Based on the fourth sensitive band and GPT-4o, several soil moisture index formulas suitable for different soil types are constructed.

[0032] S6. Based on the soil index formula, a fitting model is constructed between the index of different soil types and the soil moisture content, and the soil index formula corresponding to the best fitting model obtained through comprehensive evaluation is selected to obtain a soil moisture index set suitable for analyzing different soil moisture contents. The specific steps include the following: S61. Calculate the soil moisture index according to the soil moisture index formula and the soil spectrum data of the soil sample in the database; S62. Based on the soil moisture index and water content of the soil samples, construct a fitting model between the soil moisture index and soil water content of different soil types; S63, setting a number of evaluation indicators, and marking the soil moisture index formula corresponding to the fitting model with the highest value of the evaluation indicators as the standard soil moisture index formula; S64. Obtain standard soil moisture index formulas corresponding to different soil types to establish a soil moisture index set suitable for different soil moisture content analyses.

[0033] The existing soil moisture index formula is shown in Table 4 below:

[0034] The soil moisture index formula obtained by the present invention is shown in Table 5 below:

[0035] In Tables 4 and 5, Reflectivity is expressed as For example, it represents the reflectivity of the 1450nm band, such as Figure 5-Figure 9 The figure shows a comparison of evaluation indicators when calculating soil moisture using the existing soil moisture index formula and the soil moisture index formula of the present invention.

[0036] Example of calculating soil moisture content using the soil moisture index: Steps to calculate soil moisture content of Silty Clay Loam type soil using the index SiCLMI: (1) Based on soil spectral data, select the required bands and use the formula The SiCLMI index value is calculated.

[0037] (2) According to the index-water content fitting model formula Calculate the soil moisture content of the soil data, and the result is as follows Figure 10 shown.

[0038] The present invention also provides a soil moisture index set construction system, comprising a processor and a memory, wherein the memory is used to store a computer program, and when the computer program is executed by the processor, the soil moisture index set construction method is implemented.

[0039] Those skilled in the art will appreciate that all or part of the steps in the above-mentioned embodiment methods can be accomplished by instructing the relevant hardware through a program. Therefore, the present application may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Furthermore, the present application may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0040] Each embodiment in this specification is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to in detail. For the above embodiments, since they are basically similar to the method embodiments, the description is relatively simple. For relevant parts, please refer to the partial description of the method embodiments.

[0041] The above embodiments provide a detailed introduction to the present invention. Specific examples are used herein to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core ideas. At the same time, for those skilled in the art, according to the ideas of the present invention, there may be changes in the specific implementation methods and application scopes. In summary, the contents of this specification should not be understood as limiting the present invention.

Claims

1. A method for constructing a soil moisture index set based on soil texture classification, characterized in that: The construction method specifically includes the following steps: S1. Obtain several different types of soil samples and spectral data; S2. Analyze the correlation between the moisture content of different types of soil and the reflectance of several single bands, and select the bands with the highest correlation as the first sensitive bands; S3. Analyze the importance of each band's influence on soil moisture content, and select several bands with the largest importance index to obtain the second most sensitive band; S4. Calculate the average of the reflectance differences between soil samples with zero water content and soil samples with the highest water content in each band among all soil types, and select several bands with the largest average values ​​as the third sensitive bands; S5. Calculate the intersection of the first sensitive band, the second sensitive band, and the third sensitive band to obtain a fourth sensitive band, and construct several soil moisture index formulas applicable to different soil types based on the fourth sensitive band; S6. Based on the soil index formula, a fitting model between the index of different soil types and soil moisture content is constructed, and the soil index formula corresponding to the best fitting model is selected to obtain a soil moisture index set suitable for different soil moisture content analysis.

2. The construction method according to claim 1, characterized in that In step S1, the following steps are specifically included: S11, obtaining a number of soil samples and corresponding spectral data; S12. Divide the soil samples into several types, and each type of soil sample contains at least one soil sample with a water content of 0 and several soil samples with a non-zero water content; S13, removing soil samples and corresponding spectral data with missing attribute information from each type of soil samples; S14. Eliminate several soil samples with the highest water content and the corresponding spectral data from each type of soil samples.

3. The construction method according to claim 2, characterized in that In step S12, the soil types include silt loam, clay, loam, clay loam, sandy soil, silt clay loam, silt clay, sandy loam, sandy clay loam and loamy sand, and the attribute information of the soil samples in each soil type is the same.

4. The construction method according to claim 1, characterized in that In step S2, the following steps are specifically included: S21, setting a first band range for different types of soil, and dividing the first band range into a plurality of single bands; S22. Using the reflectance of several single bands as the independent variable and the measured soil moisture content as the dependent variable, linear regression was used to calculate the correlation between each band and soil moisture content; S23, RMSE, nRMSE, As the evaluation index of correlation, several bands with the highest index values ​​are selected as the first sensitive bands.

5. The construction method according to claim 1, characterized in that In step S3, the following steps are specifically included: S31, setting a second band range for different types of soil, and dividing the second band range into a plurality of single bands; S32. Random forest analysis was used to calculate the importance index of each band to soil moisture content; S33. Select several single bands with the highest importance as the second sensitive bands.

6. The construction method according to claim 1, characterized in that In step S4, the following steps are specifically included: S41. Select soil samples with zero water content and the highest water content from different soil types; S42. subtract the spectral reflectance of a soil sample with zero water content from the spectral reflectance of the soil sample with the highest water content in the same soil type to obtain a spectral reflectance difference; S43. Calculate the average of the spectral reflectance differences of all soil types; S44. Select several bands with the largest average value of spectral reflectance difference as the third sensitive band.

7. The construction method according to claim 1, characterized in that In step S5, the following steps are specifically included: S51, calculating the intersection of the first sensitive band, the second sensitive band, and the third sensitive band to obtain a fourth sensitive band; S52. Based on the fourth sensitive band and GPT-4o, several soil moisture index formulas suitable for different soil types are constructed.

8. The construction method according to claim 1, wherein: In step S6, the following steps are specifically included: S61. Calculate the soil moisture index according to the soil moisture index formula and the soil spectrum data of the soil sample in the database; S62. Based on the soil moisture index and water content of the soil samples, construct a fitting model between the soil moisture index and soil water content of different soil types; S63, setting a number of evaluation indicators, and marking the soil moisture index formula corresponding to the fitting model with the highest value of the evaluation indicators as the standard soil moisture index formula; S64. Obtain standard soil moisture index formulas corresponding to different soil types to establish a soil moisture index set suitable for different soil moisture content analyses.

9. A system for implementing the method for constructing a soil moisture index set according to any one of claims 1 to 8, characterized in that: The method comprises a processor and a memory, wherein the memory is used to store a computer program, and when the computer program is executed by the processor, the method for constructing a soil moisture index set according to any one of claims 1 to 8 is implemented.

Images