Remote sensing index-based prediction method for soil organic carbon distribution

Abstract

The application relates to the technical field of remote sensing detection, in particular to a soil organic carbon distribution prediction method based on a remote sensing index, which comprises the following steps: acquiring the organic carbon content in soil samples and soil hyperspectral data, extracting the correlation coefficient matrix of any two waveband spectrums under multiple sensitive indexes; determining optimal double-waveband sensitive hyperspectral indexes according to the multiple correlation coefficient matrices; performing cross-region verification on the multiple optimal sensitive indexes, selecting the optimal sensitive index with the best robustness as the organic carbon sensitive hyperspectral index; performing wide-waveband spectral resampling on the hyperspectral data, calculating the resampled organic carbon sensitive wide-waveband index according to the optimal hyperspectral index correlation coefficient matrix, and evaluating the spatial and temporal distribution characteristics of soil organic carbon based on remote sensing data based on the index. The method provided by the application has good reproducibility and generalization ability in cross-region and cross-time application.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of remote sensing detection, and particularly relates to a soil organic carbon distribution prediction method based on a remote sensing index. BACKGROUND

[0002] The existing remote sensing quantitative estimation method of soil organic carbon (SOC) has recognized limitations in model universality, and it is extremely difficult to realize cross-time and cross-region reproduction. Although in the development of the past 20 years, with the development of satellite remote sensing technology and its combination with machine learning algorithms, rich technical methods have been proposed, but there are still key technical bottlenecks in the large-scale dynamic remote sensing monitoring method of soil physical and chemical properties. Soil itself is a complex system, and the transmission of electromagnetic waves in this system has strong uncertainty, and there is a complex "reciprocity effect" among different functional groups, for example, the spectral feature band of SOC will be weakened or shifted due to the participation of iron oxide. Therefore, it becomes very difficult to effectively predict (and invert) the soil spectral characteristics by applying physical theory models. The modeling-verification strategy is a widely used quantitative inversion technology method of soil physical and chemical properties, and multiple linear models such as partial least squares regression have been proved to be effective in realizing quantitative analysis of organic carbon content based on soil spectrum. With the development of machine learning technology, artificial neural networks, random forests, support vector machines and other algorithms have also been introduced to realize the quantitative estimation of SOC content at a local scale. However, the modeling-verification inversion process highly depends on ground sampling and the model has poor cross-region universality, which is not suitable for developing SOC monitoring at a large spatial scale and high time efficiency, and the main technical bottlenecks are as follows:

[0003] (1) The sample dependence of the traditional soil parameter remote sensing quantitative estimation model based on statistical process is obvious, and it cannot be used cross-regionally.

[0004] (2) The least squares regression and other statistical models, as well as machine learning or deep learning models such as artificial neural networks, cannot provide explicit mathematical expressions for specific wave bands, and it is extremely difficult to be applied. SUMMARY

[0005] Therefore, the present application aims to provide a soil organic carbon distribution prediction method based on a remote sensing index, which combines data mining algorithms to construct soil remote sensing indexes, and through cross-region verification and cross-time verification, a new type of organic carbon sensitive wide band remote sensing index is proposed, which provides a reliable and easy-to-promote technical method for realizing cultivated land soil quality monitoring and agricultural ecological research, and fills the gap of cultivated land SOC cross-region and cross-time monitoring technical method.

[0006] To achieve the above purpose, the technical scheme of the present application is as follows:

[0007] A soil organic carbon distribution prediction method based on remote sensing index, comprising:

[0008] S1: obtaining the organic carbon content in the soil sample and the soil hyperspectral data, calculating a plurality of sensitive indexes between any two waveband spectra; based on the relationship between the organic carbon content and the soil hyperspectral data, extracting the correlation coefficient matrix of each sensitive index and the soil organic carbon content;

[0009] S2: determining two optimal spectral wavebands under each sensitive index and the optimal sensitive index corresponding to the two optimal spectral wavebands according to the plurality of correlation coefficient matrices obtained in step S1;

[0010] S3: cross-region verification of a plurality of optimal sensitive indexes, selection of the most robust optimal sensitive index as the organic carbon sensitive hyperspectral index;

[0011] S4: according to the spectral response function of remote sensing data, wide-band resampling of the soil hyperspectral data in step S1, calculation of the corresponding relationship of the resampled wide-band in the correlation coefficient matrix in step S2, extraction of the organic carbon sensitive wide-band remote sensing index according to the organic carbon sensitive hyperspectral index in step S3;

[0012] S5: using the organic carbon sensitive wide-band remote sensing index in step S4 to predict the organic carbon distribution of the remote sensing data of the region to be detected.

[0013] Further, in step S1: using Pearson correlation analysis algorithm, the correlation coefficient is calculated by the following formula:

[0014] ;

[0015] Wherein, r i,j represents the correlation coefficient of the spectrum of the i-th waveband and the j-th waveband of the soil sample, X i,j represents the sensitive index between the i-th waveband and the j-th waveband spectrum, and Y represents the organic carbon content; the obtained correlation coefficient is arranged in the order of waveband to form a correlation coefficient matrix.

[0016] Further, the sensitive index is a difference index, a ratio index and a normalized index.

[0017] Further, in step S2: selecting the two spectral wavebands corresponding to the maximum correlation coefficient in the correlation coefficient matrix as the two optimal spectral wavebands; each sensitive index under the two optimal spectral wavebands is the corresponding optimal sensitive index.

[0018] Further, between the step S2 and the step S3, further comprising: verifying the optimal spectral band determined in the step S2 by using the ICO algorithm and / or the SPA algorithm, comprising: calculating each sensitive index for any two spectral bands in the soil hyperspectral data, constructing a spectral index set corresponding to each sensitive index, the size of the spectral index set being N*N*n, wherein N represents the total number of bands in the soil hyperspectral data, and n represents the number of soil samples; in each spectral index set, taking a sub-spectral index set with a size of N*n corresponding to any one band, introducing the organic carbon content of the corresponding sample to calculate the ICO algorithm and / or the SPA algorithm, so that the optimal sub-spectral index set is selected from the N sub-spectral index sets, and the band spectrum corresponding to the optimal sub-spectral index set constitutes a verification band spectrum set; if the optimal band spectrum determined in the step S2 is in the verification band spectrum set, it is indicated that the optimal band spectrum determined in the step S2 is correct.

[0019] Further, in the step S4, the wide-band spectral resampling of the soil hyperspectral data comprises: setting a moving average window according to the full width at half maximum of the spectral response function of the remote sensing sensor, and resampling the spectral data in the effective spectral interval in the soil hyperspectral data to obtain the wide-band spectral data.

[0020] Further, the step S4 further comprises: expanding the resampled wide-band spectral data based on a regression fitting conversion equation.

[0021] Further, the organic carbon sensitive wide-band remote sensing index is as follows:

[0022] RSI=R SWIR / R NIR ;

[0023] Wherein, RSI represents the organic carbon sensitive wide-band remote sensing index, R SWIR represents the spectral reflectivity of the short-wave infrared band in the remote sensing data of the to-be-detected area, and R NIR represents the spectral reflectivity of the near-infrared band in the remote sensing data of the to-be-detected area.

[0024] Compared with the prior art, the application can achieve the following beneficial effects:

[0025] The soil organic carbon distribution prediction method based on the remote sensing index has the explicit physical meaning and the concise expression, and can quickly and quantitatively evaluate the spatial and temporal distribution characteristics of soil organic carbon in a large area by combining the spectral characteristics of soil organic carbon and multi-temporal remote sensing data, develop a multi-year, cross-regional and large spatial scale soil organic carbon remote sensing index, generate a spatial distribution map of soil organic carbon with a resolution of more than 10 meters, and realize accurate tracking and dynamic analysis of the spatial and temporal variation trend of soil organic carbon. In addition, the method has good cross-regional application universality, and has a more concise expression (short-wave infrared spectrum and near-infrared spectrum ratio) compared with the traditional model, and is suitable for multi-type satellite remote sensing data (with short-wave infrared and near-infrared channels), and has strong popularization ability. BRIEF DESCRIPTION OF DRAWINGS

[0026] The accompanying drawings, which form a part of the present application, are included to provide a further understanding of the application and are incorporated herein for reference. The illustrative embodiments of the present application and their description serve to explain the application. They do not, however, limit the present application, which is defined only by the appended claims.

[0027] Figure 1 The flowchart of the soil organic carbon distribution prediction method based on the remote sensing index according to the embodiment of the present application;

[0028] Figure 2 The flowchart of the soil organic carbon distribution prediction method based on the remote sensing index according to the embodiment of the present application;

[0029] Figure 3 The process diagram of obtaining the correlation coefficient matrix in step S1 according to the embodiment of the present application;

[0030] Figure 4 The correlation image under the ratio index according to the embodiment of the present application;

[0031] Figure 5 The verification process diagram of the optimal waveband spectrum according to the embodiment of the present application;

[0032] Figure 6 The result map of cross-regional verification according to the embodiment of the present application. DETAILED DESCRIPTION

[0033] In order to make the purpose, technical scheme and advantages of the present application clearer, the present application will be further described in detail below with reference to the drawings and specific embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application, and do not limit the present application.

[0034] It should be noted that the embodiments in the present application and the features in the embodiments can be combined with each other without conflict.

[0035] In the description of the present application, it should be understood that the terms "center", "longitudinal", "transverse", "upper", "lower", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer" and the like indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, and are only for the purpose of facilitating the description of the present application and simplifying the description, and do not indicate or imply that the device or element referred to must have a particular orientation, be constructed and operated in a particular orientation, and therefore cannot be understood as a limitation of the present application. In addition, the terms "first", "second" and the like are only for the purpose of description and cannot be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Therefore, the features defined with "first", "second" and the like can explicitly or implicitly include one or more of the features. In the description of the present application, unless otherwise specified, the meaning of "a plurality of" is two or more.

[0036] In the description of the present application, it should be noted that unless otherwise explicitly specified and limited, the terms "mounting", "connection", "connection" should be understood broadly, for example, it can be fixedly connected, or it can be detachably connected, or integrally connected; it can be mechanically connected, or it can be electrically connected; it can be directly connected, or it can be indirectly connected through an intermediate medium, or it can be the communication inside two elements. For those skilled in the art, the specific meaning of the above terms in the present application can be understood through specific circumstances.

[0037] The present application will be described in detail below with reference to the accompanying drawings and in conjunction with the embodiments.

[0038] As shown in Figure 1 and Figure 2 The soil organic carbon distribution prediction method based on remote sensing index according to the embodiments of the present application comprises:

[0039] S1: Obtain the organic carbon content in the soil sample and the soil hyperspectral data, calculate a plurality of sensitive indexes between any two waveband spectra; based on the relationship between the organic carbon content and the soil hyperspectral data, extract the correlation coefficient matrix of each sensitive index and the soil organic carbon content.

[0040] In the embodiment of the present application, the organic carbon content and the soil hyperspectral data in the soil sample are obtained in the environment of a standardized laboratory. Specifically, after the soil sample is sieved through a 0.075 mm standard soil sieve, the soil sample is dried at a temperature of 50°C for 24 hours, and then the standardized spectrum measurement is performed. In the process of spectrum measurement, environmental light interference needs to be avoided as much as possible, so the spectrum measurement is performed in a darkroom, and an ASD portable hyperspectral instrument is used to collect soil sample spectrum reflectance data (401 nm-2400 nm) by using a high-density reflection probe (2901±10°%K, 4.05 W halogen light source). The probe is fixed on an iron stand, the soil sample is placed in a clean culture dish, and the probe is as close as possible to the surface of the soil sample but does not need to contact the soil sample to avoid contaminating the probe. The instrument and the probe are preheated for more than 30 minutes before the spectrum measurement starts. After the spectrum measurement starts, the instrument needs to be calibrated every 30 minutes using a standard diffuse reflection reference white board. Each soil sample is measured 5 times, and the average value is taken as the spectrum data of the sample. Standardized spectrum measurement is performed on 17,730 soil samples from different regions to construct a hyperspectral index. The organic carbon content of all soil samples is measured by using potassium dichromate volumetric method to analyze the organic carbon reflectance spectrum characteristics. In the embodiment of the present application, 10,000 soil samples are selected from 17,730 soil samples as calculation soil samples, and the remaining soil samples are verification soil samples. The soil samples are divided into 7 parts according to the distribution area. In the present application, the soil sample hyperspectral data are obtained in a controllable laboratory environment, the soil sample processing and preparation process, the laboratory environmental conditions, and the spectrum measurement operation specification are strictly constrained to ensure the quality of the spectrum data acquisition, greatly reduce the data error brought into further spectrum analysis, and effectively solve the technical problem that the spectrum variation caused by the difficulty in realizing the collection, preparation and spectrum test of the soil sample under standardized conditions in the current soil spectrum analysis technology is an important factor restricting the model sharing and cross-application.

[0041] In some embodiments, the process of obtaining the correlation coefficient matrix is as shown in Figure 3

[0042] The Pearson correlation analysis algorithm is used to calculate the correlation coefficient by the following formula:

[0043] ;

[0044] wherein r i,j represents the correlation coefficient of the i th wave band and the j th wave band of the soil sample, X i,j represents the sensitive index between the spectrum of the i th wave band and the j th wave band, and Y represents the organic carbon content. The obtained correlation coefficient is arranged in the order of the wave band to form a correlation coefficient matrix.

[0045] ​The sensitive index of any two wave bands of soil spectrum data is calculated by traversing the full wave band of 401-2400 nm, and then the correlation coefficient between any sensitive index and the organic carbon content is calculated based on correlation analysis, a three-dimensional correlation coefficient matrix is constructed, and the analysis and extraction of the organic carbon sensitive spectrum index are realized.

[0046] Specifically, in some embodiments, the sensitive index is a difference index SSDI i,j i j , a ratio index SSRI i,j i j , and a normalized index SSNI i,j i j i j , wherein R i and R j represent the spectral reflectivity of the i-th and j-th wave bands in the wave band interval of 401-2400 nm, respectively. In the embodiments of the present application, the wave band interval of 401-2400 nm is divided according to a 1 nm wave band interval, and a total of 2000 spectral wave bands are obtained. It can be understood that for the difference index SSDI i,j , the corresponding correlation coefficient is:

[0047] ;

[0048] For the ratio index SSRI i,j , the corresponding correlation coefficient is:

[0049] ;

[0050] For the normalized index SSNI i,j , the corresponding correlation coefficient is:

[0051] ;

[0052] In the embodiments of the present application, based on the Pearson correlation coefficient, the correlation coefficient between any two wave bands of the sensitive index and the organic carbon content is filled in the corresponding position of the correlation coefficient matrix, all wave bands (401-2400 nm) are traversed, and finally a full wave band correlation coefficient matrix with a size of 2000x2000 can be obtained.

[0053] According to the correlation coefficient matrix of the ratio index, the embodiments of the present application provide a correlation image under the ratio index as shown in Figure 4 .​​​​​​​​Figure 4 Figures (a) through (e) show the changes in the correlation index with sample sizes of 100, 500, 1000, 5000, and 10000, respectively. Figure 4 The colors, from blue to red, represent the transition from negative correlation to positive correlation. Figure 4 It can be seen that as the sample size included in the analysis increases, the high-low distribution of the correlation coefficient between the dual-band index and organic carbon content gradually converges and stabilizes. This distribution characterizes the correlation coefficient relationship between any dual-band index (difference index, ratio index, and normalized index) and organic carbon content in the 1 nm interval spectral range. Specifically, the correlation coefficient matrix characterizes the significant response of the dual-band spectral index to organic carbon content in the near-infrared and short-wave infrared ranges. As organic carbon content increases, the reflectance in the near-infrared band decreases more rapidly due to changes in organic matter, associated minerals, or structure, while the decrease in the short-wave infrared band is relatively slower or less affected. As organic carbon content increases, the ratio of the short-wave infrared band to the near-infrared band increases, resulting in a positive correlation between the sensitivity index and organic carbon content.

[0054] S2: Based on the multiple correlation coefficient matrices obtained in step S1, determine the two optimal spectral bands for each sensitivity index, and the optimal sensitivity index corresponding to the two optimal spectral bands. In some embodiments, the two spectral bands corresponding to the largest correlation coefficient in the correlation coefficient matrix are selected as the two optimal spectral bands; each sensitivity index under the two optimal spectral bands is the corresponding optimal sensitivity index. In this embodiment of the invention, step S2 specifically includes: for each sensitivity index among the difference index, ratio index, and normalization index, selecting the two spectral bands corresponding to the largest correlation coefficient in the correlation coefficient matrix as the two optimal spectral bands; each sensitivity index under the two optimal spectral bands is the corresponding optimal sensitivity index. Specifically, taking the ratio index as an example, the largest correlation coefficient in the correlation coefficient matrix corresponding to the ratio index (e.g., the largest correlation coefficient in the ratio index) is selected as the largest correlation coefficient in the correlation coefficient matrix (e.g., the largest correlation coefficient in the ratio index) is selected as the largest correlation coefficient in the correlation coefficient matrix. Figure 4 (e) The two bands corresponding to the point with the deepest red color (i.e., λ) i =1406, λ j =1281) represents two optimal spectral bands; based on the two optimal spectral bands λ i =1406 and λ j =1281 Calculate the optimal sensitivity index corresponding to the ratio index, that is:

[0055] R-SSRI=R 1406 / R 1281 ;

[0056] Where R-SSRI represents the optimal sensitivity index corresponding to the ratio index, R 1406 and R 1281respectively represent spectral reflectance of wavelengths of 1406 nm and 1281 nm. In the above manner, the optimal sensitive indices R-SSDI and R-SSNI corresponding to the difference index and the normalized index of the following formula respectively can be obtained:

[0057] R-SSDI=R 1399 -R 1353 ;

[0058] R-SSNI=(R 1406 -R 1281 ) / (R 1406 +R 1281 ).

[0059] In some embodiments, after step S2, further comprising: verifying the optimal waveband spectrum determined in step S2 by using an ICO algorithm (Interval Combination Optimization) and / or an SPA algorithm (Successive Projections Algorithm). The SPA algorithm is a forward optimization algorithm based on a successive projection strategy. When extracting spectral feature bands, the successive projection algorithm starts from a waveband, and through continuous iteration, filters out waveband data with the smallest correlation with the initial waveband spectral data to form a data subset, and then performs step-by-step multiple regression modeling and verification based on the data subset. Since the data subset selected by the algorithm is composed of waveband data with the smallest correlation, the SPA algorithm can effectively reduce the multicollinearity between feature bands, and is the main algorithm for constructing remote sensing indices. The ICO algorithm is a spectral feature band analysis method. The algorithm realizes spectral interval combination optimization by using a soft shrinkage and local search method, and has the advantages of few parameters, soft shrinkage, and fast convergence. Similar to the SPA algorithm, the ICO algorithm is also based on multi-band spectral data combination, and is a multi-band combination screening strategy with the highest model accuracy in multiple regression modeling and verification as the optimization condition. Unlike the SPA algorithm, the ICO is a spectral interval analysis method, which replaces the point-by-point analysis strategy as in the SPA algorithm, thereby reducing the calculation pressure of the soft shrinkage strategy in the optimization calculation process, effectively reducing the overfitting risk in optimization, and being the main algorithm for remote sensing index optimization. The present application further verifies the optimal waveband spectrum determined in step S2 by using the two algorithms, to ensure the accuracy of the optimal waveband spectrum.

[0060] The verification process of the optimal waveband spectrum is as follows Figure 5As shown, specifically includes: the calculation of each sensitive index of any two spectral bands in the soil hyperspectral data, construct each sensitive index corresponding to the spectral index set, the size of the spectral index set is N×N×n, wherein N represents the total number of bands in the soil hyperspectral data, N=2000 in the embodiment of the present application, n represents the number of soil samples, n represents the number of soil samples calculated in the embodiment of the present application, that is, n=10000;In each spectral index set, take any one band corresponding to the sub-spectral index set IDX with the size of N×n, introduce the organic carbon content of the corresponding sample to calculate the ICO algorithm and / or SPA algorithm, so that the optimal sub-spectral index set Opt IDX The optimal sub-spectral index set Opt IDX The corresponding band spectrum composition verifies the wave band spectrum set; if the optimal wave band spectrum determined in step S2 is in the wave band spectrum set, it indicates that the optimal wave band spectrum determined in step S2 is correct.

[0061] S3: Cross-region verification is performed on the plurality of optimal sensitive indexes, and the optimal sensitive index with the best robustness is selected as the organic carbon sensitive hyperspectral index.

[0062] In the embodiment of the present application, the difference index, the ratio index and the normalized index are respectively verified by 7 verification soil samples from different regions, and the stability of the three optimal sensitive indexes in cross-region application is tested, and the test results are as shown in Figure 6 As shown, Figure 6 (a)-(f) in the table respectively show the accuracy performance of the optimal difference index (R-SSDI), the optimal ratio index (R-SSRI) and the optimal normalized index (R-SSNI) in the 7 test sample sets, and the 7 verification soil samples are respectively from different regions, and the verification results are represented by different color scatter points and fitting results in Figure 6 The results show that the optimal sensitive index corresponding to the ratio index is the most robust in cross-region verification, which is the organic carbon sensitive hyperspectral index selected in the embodiment of the present application.

[0063] S4: according to the spectral response function of the remote sensing data, the soil hyperspectral data in step S1 is resampled in a wide band, the corresponding relationship of the resampled wide band in the correlation coefficient matrix in step S2 is calculated, and the organic carbon sensitive wide band remote sensing index is extracted according to the organic carbon sensitive hyperspectral index in step S3. The soil hyperspectral data is resampled in a wide band by using the spectral response function of the remote sensing sensor in the present application, so that the sampling rate of the soil hyperspectral data in the laboratory environment is matched with the remote sensing data, the hyperspectral data in the laboratory environment is mapped into the remote sensing data, so that the measured data in the laboratory environment is migrated to the remote sensing data, so that the organic carbon sensitive wide band remote sensing index obtained in step S4 reflects the distribution of soil organic carbon when applied in remote sensing data, so that the prediction of soil organic carbon distribution based on remote sensing index is realized. In some embodiments, resampling the soil hyperspectral data in a wide band spectrum includes: setting a moving average window according to the full width at half maximum of the spectral response function of the remote sensing sensor, resampling the spectral data in the effective spectral interval in the soil hyperspectral data to obtain wide band spectral data. Step S4 further includes: based on the regression fitting conversion equation, the resampled wide band spectral data is extended.

[0064] To verify the effectiveness of the method provided by the present application, the embodiments of the present application use the published Landsat, Sentinel and MODIS remote sensing data to predict the distribution of organic carbon content in a provincial demonstration area, at this time, step S4 specifically includes: using the full width at half maximum of the spectral response function of the corresponding remote sensing sensor of the Landsat, Sentinel and MODIS data respectively, setting a moving average window, using spectral resampling technology, using the moving average window to resample the hyperspectral data in the effective spectral interval of different remote sensing data, and then realizing the conversion of hyperspectral data to multispectral data based on the regression fitting conversion equation; based on the scale converted spectral data, the correlation coefficient matrix of the wide band spectral index and the organic carbon content is reconstructed; according to the organic carbon sensitive hyperspectral index in step S3, the organic carbon sensitive wide band remote sensing index is extracted. At this time, the following is obtained:

[0065] The organic carbon sensitive wide band remote sensing index of the Landsat remote sensing data is:

[0066] RSI Landsat = R B7 / R B4 ;

[0067] Wherein, RSI Landsat represents the organic carbon sensitive wide band remote sensing index of the Landsat remote sensing data, R B7 represents the spectral reflectance of the 7th band of the Landsat remote sensing data, corresponding to the short wave infrared band of the Landsat remote sensing data, R B4R4 represents the spectral reflectance of the 4th band in Landsat remote sensing data, corresponding to the near-infrared band in Landsat remote sensing data;

[0068] The organic carbon sensitive wide band remote sensing index of Sentinel remote sensing data is:

[0069] RSI Sentinel = R B12 / R B8 ;

[0070] wherein RSI Sentinel represents the organic carbon sensitive wide band remote sensing index of Sentinel remote sensing data, R B12 represents the spectral reflectance of the 12th band in Sentinel remote sensing data, corresponding to the short-wave infrared band in Sentinel remote sensing data, R B8 represents the spectral reflectance of the 4th band in Sentinel remote sensing data, corresponding to the near-infrared band in Sentinel remote sensing data;

[0071] The organic carbon sensitive wide band remote sensing index of MODIS remote sensing data is:

[0072] RSI MODISl = R B5 / R B2 ;

[0073] wherein RSI MODISl represents the organic carbon sensitive wide band remote sensing index of MODIS remote sensing data, R B5 represents the spectral reflectance of the 5th band in MODIS remote sensing data, corresponding to the short-wave infrared band in MODIS remote sensing data, R B2 represents the spectral reflectance of the 2nd band in MODIS remote sensing data, corresponding to the near-infrared band in MODIS remote sensing data;

[0074] The organic carbon sensitive wide band remote sensing index of the above three remote sensing data can be summarized as follows:

[0075] RSI = R SWIR / R NIR ;

[0076] wherein RSI represents the organic carbon sensitive wide band remote sensing index (Ratio Soil Index), R SWIR represents the spectral reflectance of the short-wave infrared band in the remote sensing data of the region to be detected, R NIR represents the spectral reflectance of the near-infrared band in the remote sensing data of the region to be detected.

[0077] S5: using the organic carbon sensitive wide-band remote sensing index in step S4, predicting the organic carbon distribution of the remote sensing data of the region to be detected.

[0078] It should be understood that the various forms of flow shown above can be used to reorder, add or delete steps. For example, the steps described in the present disclosure can be performed in parallel, sequentially or in a different order, as long as the desired results of the technical solutions of the present disclosure can be achieved, which is not limited herein.

[0079] The above detailed description does not constitute a limitation on the protection scope of the present application. Those skilled in the art should understand that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modifications, equivalent replacements and improvements made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A method for predicting soil organic carbon distribution based on remote sensing indices, characterized in that, include: S1: Obtain the organic carbon content and soil hyperspectral data from soil samples, and calculate multiple sensitivity indices between any two spectral bands; based on the relationship between organic carbon content and soil hyperspectral data, extract the correlation coefficient matrix between each sensitivity index and soil organic carbon content; the sensitivity indices are the difference index, the ratio index, and the normalization index. S2: Based on the multiple correlation coefficient matrices obtained in step S1, determine the two optimal spectral bands for each sensitivity index, and the optimal sensitivity index corresponding to the two optimal spectral bands; S3: Cross-regional verification was performed on multiple optimal sensitivity indices, and the optimal sensitivity index with the best robustness was selected as the organic carbon sensitive hyperspectral index; S4: Based on the spectral response function of the remote sensing data, perform broadband resampling of the soil hyperspectral data in step S1, calculate the correspondence of the resampled broadband in the correlation coefficient matrix in step S2, and extract the organic carbon sensitive broadband remote sensing index based on the organic carbon sensitive hyperspectral index in step S3. S5: Using the organic carbon-sensitive broadband remote sensing index from step S4, predict the distribution of organic carbon in the remote sensing data of the area to be detected.

2. The method for predicting soil organic carbon distribution based on remote sensing indices according to claim 1, characterized in that, In step S1: Using the Pearson correlation analysis algorithm, the correlation coefficient is calculated using the following formula: ； Where, r i,j X represents the correlation coefficient between the spectral bands of the i-th and j-th bands of the soil sample. i,j represents the sensitivity index between the i-th and j-th spectral bands, and Y represents the organic carbon content; The obtained correlation coefficients are arranged in band order to form a correlation coefficient matrix.

3. The method for predicting soil organic carbon distribution based on remote sensing indices according to claim 1, characterized in that, In step S2: Select the two spectral bands corresponding to the largest correlation coefficient in the correlation coefficient matrix as the two optimal spectral bands; each sensitivity index under the two optimal spectral bands is the corresponding optimal sensitivity index.

4. The method for predicting soil organic carbon distribution based on remote sensing indices according to claim 1, characterized in that, Between steps S2 and S3, the method further includes: verifying the optimal spectral band determined in step S2 using a spectral interval combination optimization algorithm and / or a continuous projection algorithm, including: For any two spectral bands in the soil hyperspectral data, calculate each sensitivity index and construct a spectral index set corresponding to each sensitivity index. The size of the spectral index set is N×N×n, where N represents the total number of bands in the soil hyperspectral data and n represents the number of soil samples. In each spectral index set, take any band and select a sub-spectral index set of size N×n. Introduce the organic carbon content of the corresponding sample and perform spectral interval combination optimization algorithm and / or continuous projection algorithm to select the optimal sub-spectral index set from the N sub-spectral index sets. The band spectrum corresponding to the optimal sub-spectral index set constitutes the verification band spectrum set. If the optimal band spectrum determined in step S2 is in the band spectrum set, it indicates that the optimal band spectrum determined in step S2 is correct.

5. The method for predicting soil organic carbon distribution based on remote sensing indices according to claim 1, characterized in that, Step S4, which involves broadband spectral resampling of the soil hyperspectral data, includes: Based on the full width at half maximum (FWHM) of the spectral response function of the remote sensing sensor, a moving average window is set to resample the spectral data within the effective spectral range of the soil hyperspectral data to obtain broadband spectral data.

6. The method for predicting soil organic carbon distribution based on remote sensing indices according to claim 5, characterized in that, Step S4 also includes: expanding the resampled wideband spectral data based on the regression fitting transformation equation.

7. The method for predicting soil organic carbon distribution based on remote sensing indices according to claim 1, characterized in that, The organic carbon-sensitive broadband remote sensing index is as follows: RSI=R SWIR / R NIR ; Wherein, RSI represents the organic carbon sensitive broadband remote sensing index, R SWIR R represents the spectral reflectance in the shortwave infrared band of the remote sensing data of the area to be detected. NIR This represents the spectral reflectance in the near-infrared band of the remote sensing data of the area to be detected.

Images