Red soil organic matter spectrum inversion method and device for removing influence of free iron

Abstract

The invention discloses a red soil organic matter spectrum inversion method and device for removing influence of free iron, and the method comprises the steps: obtaining soil sample data, storing organic matter data into a set P, storing free iron content data of typical red soil samples into F, storing spectral reflectivity data of red soil survey samples into H, and obtaining a set P; storing spectral reflectivity data of typical red soil samples into D; according to the set H and the set D, utilizing an MODWT algorithm for decomposition to obtain layered discrete detail coefficient sets CH and CD of the red soil survey sample and the typical red soil sample; removing a noise layer based on CD to obtain an effective layer discrete detail coefficient set C; respectively calculating correlation coefficients between each element in the C and the set F and the set P, and screening out a characteristic wave band set CDG; selecting a detail coefficient CHG located on the characteristic wave band from the layered discrete detail coefficient set CH; a prediction model is constructed, and CHG is adopted for training; and predicting the organic matter content by adopting the trained prediction model. The method can eliminate the influence of free iron oxide, and is low in cost and high in efficiency.

Description

Technical Field

[0001] The invention relates to soil spectroscopy technology, and in particular to a red soil organic matter spectral inversion method and equipment for removing the influence of free iron. Background Art

[0002] To prevent soil degradation and achieve sustainable soil management, soil monitoring at different time and space scales is necessary. Soil organic matter is the general term for carbon-containing organic compounds in the soil, which is essential for plant growth. Traditional laboratory methods for measuring organic matter usually involve multiple processing, extraction and analysis steps, which are complex and costly to operate, and it is difficult to obtain the large amount of data required for soil monitoring in a short period of time.

[0003] Soil diffuse reflectance spectroscopy provides an economical, time-saving, repeatable, non-destructive and environmentally friendly alternative to traditional laboratory analysis. The visible and infrared spectra of soil can effectively characterize the chemical, physical and mineralogical composition of soil. Soil spectroscopy analyzes the spectral absorption position, absorption intensity and shape characteristics of different components in the soil, associates the spectral reflectance of known soil samples with their physical and chemical properties to build a calibration model, and uses the calibration model to predict the physical and chemical properties of unknown soil samples. In-depth analysis and simulation of soil visible and infrared reflectance spectral characteristics is the basis for multi-scale and multi-factor soil remote sensing monitoring using multispectral / hyperspectral imaging remote sensing technology.

[0004] The soil spectrum is a mixed spectrum, and the overall reflectance is a complex superposition of the spectral reflectances of its various components. Organic matter and free iron oxide are both important soil components that affect the spectral reflectance characteristics of soil. Organic matter is mainly composed of humus (humic acid and fulvic acid), mostly in the form of CH, CH2, CH3 and C=O. Generally speaking, as the organic matter increases, the soil spectral reflectance decreases. Researchers have found that organic matter has responses in the 500nm-1200nm, 900nm-1220nm, around 800nm, 640nm-720nm, 1702-2052nm, 1726-2426nm and other regions. Soil free iron oxide refers to iron in the form of oxides and their hydrates in soil clay, including hematite (Fe 2 O 3 ), goethite (FeOOH), etc., whose reflectance spectra usually show the same 2+ , Fe 3+ Specific features related to induced absorption, the main absorption features are located near 480nm, 870nm, and 920nm. Multiple studies have shown that in the visible light-short near-infrared band (390-950nm) commonly used by drone-borne and satellite-borne imaging sensors, the spectral characteristics of soil organic matter and free iron oxide have significant mutual influence.

[0005] Red soil is an important soil resource in southern China. The free iron oxide content is usually high, which has an inhibitory effect on the response signals of organic matter and other components in the soil spectral reflectance. In response to this problem, scholars have proposed methods to eliminate the influence of free iron oxide on the estimation of organic matter visible-near infrared spectra, mainly including the External Parameter Orthogonalisation (EPO) method and the Machine Learning (ML) method. Both methods require a large number of soil samples after the free iron content is measured for model training, which requires large investment, high cost and low efficiency. Summary of the invention

[0006] In view of the problems existing in the prior art, the purpose of the present invention is to provide a red soil organic matter spectral inversion method, equipment, storage medium and program product for eliminating the influence of free iron oxide with low cost, high efficiency and strong applicability.

[0007] In order to achieve the above-mentioned object of the invention, the present invention provides the following technical solutions:

[0008] A red soil organic matter spectral inversion method removing the influence of free iron, comprising:

[0009] (1) obtaining sample data of a number of red soil survey samples and typical red soil samples, wherein the sample data of the red soil survey samples include visible-short near-infrared spectral reflectance data and organic matter content data, and the sample data of the typical red soil samples include visible-short near-infrared spectral reflectance data, free iron content data and organic matter content data, storing the organic matter data of the red soil survey samples and the typical red soil samples into a set P, storing the free iron content data of the typical red soil samples into a set F, storing the spectral reflectance data of the red soil survey samples into a set H, and storing the spectral reflectance data of the typical red soil samples into a set D;

[0010] (2) The spectral reflectance data in set H and set D are decomposed using the maximum overlap discrete wavelet MODWT algorithm, thereby generating a hierarchical discrete detail coefficient set CH of the spectral reflectance data of red soil survey samples and a hierarchical discrete detail coefficient set CD of the spectral reflectance data of typical red soil samples;

[0011] (3) Based on the hierarchical discrete detail coefficient set CD, the multi-scale entropy of each layer is calculated, and the noise layer is determined according to the difference in the multi-scale entropy of adjacent layers. The noise layer is removed to obtain the effective layer discrete detail coefficient set C;

[0012] (4) Calculate the correlation coefficients of each element in the effective layer discrete detail coefficient set C with the set F and the set P, screen out the spectral bands whose organic matter correlation coefficient is greater than the free iron correlation coefficient, and remove the spectral bands with significant free iron correlation to obtain the characteristic band set CDG;

[0013] (5) According to the characteristic band set CDG, detail coefficients located on the characteristic band are selected from the hierarchical discrete detail coefficient set CH and stored in the sample set CHG;

[0014] (6) Construct a prediction model for organic matter in red soil survey samples, and use the detail coefficient in the sample set CHG as the independent variable and the organic matter content of the corresponding red soil survey samples as the dependent variable to train the prediction model for organic matter in red soil survey samples;

[0015] (7) Calculate the detail coefficient of the visible-short near-infrared spectral reflectance data of the red soil survey sample to be estimated, and input it into the trained red soil survey sample organic matter prediction model to predict the organic matter content.

[0016] Furthermore, step (1) comprises:

[0017] (1-1) Obtaining sample data of N red soil survey samples, including visible-short near-infrared spectral reflectance data and organic matter content data of the red soil survey samples, where N is a positive integer greater than 1;

[0018] (1-2) obtaining sample data of M typical red soil samples, including visible-short near-infrared spectral reflectance data, free iron content data and organic matter content data of the typical red soil samples, where M is a positive integer greater than 1;

[0019] (1-3) Obtain the organic matter content data of all red soil survey samples and typical red soil samples and store them in the organic matter content data set P = {p i |i=1,...,n},p i is the organic matter content of the ith soil sample, n is the number of soil samples, where n = N + M;

[0020] (1-4) Obtain the spectral reflectance data of all red soil survey samples and store them in the set H = {h i1 |i1=1,...,N}, obtain the spectral reflectance data of all typical red soil samples and store them in the set D={d i2 |i2=1,,..,M}, the i1th data h in the set H i1 ={h i1，j |j=1,...,m}, the i2th data d in set D i2 ={d i2，j |j=1,...,m},h i1，jis the reflectance of the jth band of the i1th red soil survey sample, d i2，j is the reflectance of the jth band of the i2th typical red soil sample, and m is the total number of spectral bands;

[0021] (1-5) Obtain the free iron content data of typical red soil samples and store them in the set F = {f i2 |i2=1,2,...,M},f i2 is the free iron content of the ith typical red soil sample.

[0022] Furthermore, step (2) includes:

[0023] (2-1) According to the total number of spectral bands m, the number of MODWT decomposition levels t is calculated according to the following formula:

[0024] t = log 2 m:

[0025] (2-2) For the spectral reflectance data of each red soil survey sample in set H and the spectral reflectance data of each typical red soil sample in set D, the MODWT algorithm is used to decompose the t layers to obtain the high-frequency detail coefficient sequence of each layer;

[0026] (2-3) The high-frequency detail coefficients of the spectral reflectance data of all red soil survey samples are stored in the hierarchical discrete detail coefficient set CH = {ch i1，j |i1=1,...,N,j=1,...,t},ch i1，j represents the high-frequency detail coefficient sequence of the jth layer of the spectral reflectance of the i1th red soil survey sample, and N is the number of red soil survey samples;

[0027] (2-4) The high-frequency detail coefficient sequence of the spectral reflectance data of all typical red soil samples is stored in the hierarchical discrete detail coefficient set CD = {cd i2,j |i2=1,...,M,j=1,...,t},cd i2,j represents the high-frequency detail coefficient sequence of the jth layer of the spectral reflectance of the i2th typical red soil sample, and M is the number of typical red soil samples.

[0028] Further, step (2-2) includes:

[0029] (2-2-1) Select any data h from the set H i1 , calculate the data h i1 A collection of high-pass filters and low pass filter set

[0030]

[0031] Among them, hl and g l is the lth high-pass filter and low-pass filter in discrete wavelet transform, S is the length of the filter;

[0032] (2-2-2) The high-frequency detail coefficient sequence is calculated using the following formula:

[0033]

[0034]

[0035] v=(ul)mod m

[0036] ch i1，j ={ch i1，j，u}

[0037] In the formula, ch i1，1，u ,ch i1，j，u represents the uth spectral band component of the 1st and jth layers of the high-frequency detail coefficient sequence of the spectral reflectance of the i1th red soil survey sample, wh i1，1，u 、wh i1，j，u represents the uth spectral band component of the approximate coefficient sequence of the 1st and jth layers of the spectral reflectance of the i1th red soil survey sample, wh i1，j-1，v represents the vth spectral band component of the j-1th layer approximate coefficient sequence of the spectral reflectance of the i1th red soil survey sample, h i1，v Indicates h i1 The uth spectral band component, mod is the modulo operation, and v is the remainder;

[0038] (2-2-3) The spectral reflectance data of other red soil survey samples and the spectral reflectance data of all typical red soil samples in set D are calculated according to step (2-2-2) to obtain the high-frequency detail coefficient sequence of the spectral reflectance data of all red soil survey samples and the spectral reflectance data of all typical red soil samples.

[0039] Furthermore, step (3) includes:

[0040] (3-1) Set the band scale τ to different values, for each high-frequency detail coefficient sequence cd in the layered discrete detail coefficient set CD i2，j , recalibrate at different band scales τ according to the following formula:

[0041]

[0042] In the formula, is the detail coefficient sequence cd i2，j The αth element in the coarse-grained sequence obtained after recalibration at the band scale τ, cd i2，j，d Indicates cdi2，j The u-th spectral band component, M is the number of typical red soil samples, and t is the number of decomposition layers;

[0043] (3-2) Based on Calculate the detail coefficient sequence cd at different band scales τ i2，j The sample entropy of

[0044] (3-3) According to the high-frequency detail coefficient sequence cd under different band scales τ i2，j Sample entropy, calculate the high-frequency detail coefficient sequence cd i2，j The multi-scale entropy of

[0045] (3-4) For the layered discrete detail coefficient set CD, the difference in multi-scale entropy between every two adjacent layers is calculated;

[0046] (3-5) If the difference between the multi-scale entropy of the x-th layer and the x-1-th layer is smaller than the difference between the multi-scale entropy of the adjacent layers before the x-1-th layer, then the x-t layer is regarded as the effective layer, and the remaining layers are regarded as the noise layer. The high-frequency detail coefficient sequence of the effective layer is stored in the effective layer discrete detail coefficient set C = {cd i2，j′ |j′=x,...,t},where cd i2，j′ It represents the high-frequency detail coefficient sequence of the j′th layer of the spectral reflectance of the i2th typical red soil sample, and t is the number of decomposition layers.

[0047] Further, step (4) includes:

[0048] (4-1) For each high-frequency detail coefficient sequence cd in the effective layer discrete detail coefficient set C i2，j′ , according to cd i2，j′ The Spearman correlation coefficient is calculated for the free iron content corresponding to the set F as the high-frequency detail coefficient sequence cd i2，j′ The correlation coefficient of free iron fd i2，j′ Store in collection FD:

[0049]

[0050]

[0051] Among them, rg(cd i2，j′，u ) is cd i2，j′ The spectral band components of cd are sorted by numerical value i2，j′，u The rank, cd i2，j′，u Indicates cd i2，j′ The uth spectral band component, rg(f i2 ) is the value of the elements in set F after they are sorted by numerical value. i2 The rank of f i2 represents the free iron content of the i2th typical red soil sample, represents the difference in ranking of free iron, b represents cd i2，j′ The sum of all spectral band components in;

[0052] (4-2) For each high-frequency detail coefficient sequence cd in the effective layer discrete detail coefficient set C i2，j′ , according to cd i2，j′ The Spearman correlation coefficient is calculated with the corresponding organic matter content in set P as the high-frequency detail coefficient sequence cd i2，j′ The organic matter correlation coefficient pd i2，j′ Deposit into collection PD:

[0053]

[0054]

[0055] Among them, rg(p i2 ) is the value of the elements in set P after they are sorted by numerical value. i2 The rank of p i2 represents the organic matter content of the i2th typical red soil sample, It indicates the difference in ranking related to organic matter;

[0056] (4-3) Based on the set FD and the set PD, the spectral bands ccdg whose organic matter correlation coefficient in each layer j′ is greater than the free iron correlation coefficient are screened. i2,j′ , stored in set YZ;

[0057] (4-4) Remove the spectral bands whose free iron correlation coefficient is greater than the preset threshold in the set YZ, and store the remaining spectral bands in the characteristic band set CDG = {cdg i2,j′ |j′=x,…,t}.

[0058] A computer device comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the above method.

[0059] A computer-readable storage medium having a computer program / instruction stored thereon, wherein the computer program / instruction implements the above method when executed by a processor.

[0060] A computer program product comprises a computer program / instruction, wherein the computer program / instruction implements the above method when executed by a processor.

[0061] Compared with the prior art, the present invention has the following beneficial effects: based on the use of discrete wavelet decomposition of spectral reflectance, the present invention screens and determines the bands where the organic matter response is greater than the free iron response in the layered wavelet detail data, and removes the bands where the free iron is significant. Based on the screened characteristic band parameters, the red soil organic matter spectral inversion without the influence of free iron can be achieved. The present invention only needs to analyze a small amount of typical red soil sample data, has low cost, and can significantly improve the spectral inversion accuracy of red soil organic matter content. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Figure 1 : is a schematic flow chart of the red soil organic matter spectrum inversion method for removing the influence of free iron provided in this embodiment;

[0063] Figure 2 is a spectral reflectance curve diagram of a typical red soil sample in this embodiment;

[0064] Figure 3-Figure 11 is the detail coefficient map after decomposition from the 1st layer to the 9th layer in this embodiment;

[0065] Fig.12 is a multi-scale entropy result diagram in this embodiment;

[0066] Fig.13 is a graph of the correlation coefficient between the original spectral reflectance and organic matter in this embodiment;

[0067] Figure 14-18 is a graph of the correlation coefficient between the detail coefficient and the organic matter after decomposition from the 5th to the 9th layer in this embodiment;

[0068] Fig.19 It is a scatter plot comparing the true value and predicted value of the validation set samples in this embodiment. DETAILED DESCRIPTION

[0069] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0070] Embodiment 1

[0071] The red soil samples used in this embodiment come from the red soil profiles of the soil system survey in some southern provinces and the typical red soil profiles collected in Jiangxi Province. The red soil survey samples are soil samples collected from the occurrence layer of 81 red soil profiles, totaling 316 pieces. The visible-near infrared band (350-2500nm) diffuse reflectance spectrum of the red soil survey samples was measured using a Cary 5000 spectrophotometer, and the spectral sampling interval was 1nm. The soil samples were air-dried, ground, sieved (0.25mm) and dried (45°C) before measurement. The organic matter data of the red soil survey samples were obtained from the soil system survey reports of each province. The typical red soil samples include 12 soil samples of 3 fixed depths (0-10cm, 10-20cm and 40-50cm) of 4 profiles developed on different parent materials in a county in Jiangxi Province, and 24 samples of free iron oxide and organic matter removed based on these soil samples, totaling 36 pieces. The visible-near infrared (350-2500nm) diffuse reflectance spectra of typical red soil samples were measured in a dark room using an ASD FieldSpec 3 ground feature spectrometer. The spectral sampling interval was 1nm. The soil samples were air-dried, ground, sieved (2mm), and dried (45°C) before measurement. The organic matter of typical red soil samples was determined by potassium dichromate-sulfuric acid digestion method, and free iron was extracted with sodium dithionite-sodium citrate-sodium bicarbonate solution (DCB) and determined by colorimetry.

[0072] like Figure 1 As shown, the method for estimating the organic matter content of red soil based on visible-short near-infrared (350-950nm) spectral reflectance without the influence of free iron in this embodiment comprises the following steps:

[0073] (1) Obtain sample data of several red soil survey samples and typical red soil samples.

[0074] The sample data of the red soil survey sample includes visible-short near-infrared spectral reflectance data and organic matter content data, and the sample data of the typical red soil sample includes visible-short near-infrared spectral reflectance data, free iron content data and organic matter content data. The organic matter data of the red soil survey sample and the typical red soil sample are stored in set P, the free iron content data of the typical red soil sample is stored in set F, the spectral reflectance data of the red soil survey sample is stored in set H, and the spectral reflectance data of the typical red soil sample is stored in set D. In this embodiment, the spectral reflectance of the typical red soil sample in the visible-short near-infrared 350-950nm range is shown as follows Figure 2 shown.

[0075] Specifically, the step includes:

[0076] Step (1) comprises:

[0077] (1-1) Obtaining sample data of N red soil survey samples, including visible-short near-infrared spectral reflectance data and organic matter content data of the red soil survey samples. In this embodiment, N=316;

[0078] (1-2) obtaining sample data of M typical red soil samples, including visible-short near-infrared spectral reflectance data, free iron content data, and organic matter content data of the typical red soil samples. In this embodiment, M=36;

[0079] (1-3) Obtain the organic matter content data of all red soil survey samples and typical red soil samples and store them in the organic matter content data set P = {p i |i=1,…,n},p i is the organic matter content of the ith soil sample, n is the number of soil samples, where n = N + M;

[0080] (1-4) Obtain the spectral reflectance data of all red soil survey samples and store them in the set H = {h i1 |i1=1,…,N}, obtain the spectral reflectance data of all typical red soil samples and store them in the set D={d i2 |i2=1,…,M}, the i1th data h in the set H i1 ={h i1,j |j=1,…,m}, the i2th data d in set D i2 ={d i2,j |j=1,…,m},h i1,j is the reflectance of the jth band of the i1th red soil survey sample, d i2,j is the reflectance of the jth band of the i2th typical red soil sample, and m is the total number of spectral bands;

[0081] (1-5) Obtain the free iron content data of typical red soil samples and store them in the set F = {f i2 |i2=1,2,...,M},f i2 is the free iron content of the ith typical red soil sample.

[0082] (2) The spectral reflectance data in set H and set D were decomposed using the maximum overlap discrete wavelet (MODWT) algorithm, thereby generating a hierarchical discrete detail coefficient set CH for the spectral reflectance data of red soil survey samples and a hierarchical discrete detail coefficient set CD for the spectral reflectance data of typical red soil samples.

[0083] This step specifically includes:

[0084] (2-1) The number of MODWT decomposition levels t is calculated according to the following formula based on the total number of spectral bands m. In this embodiment, t=9:

[0085] t = log2 m;

[0086] (2-2) For the spectral reflectance data of each red soil survey sample in set H and the spectral reflectance data of each typical red soil sample in set D, the MODWT algorithm (the basis function is "sym5") is used to decompose the t layer to obtain the high-frequency detail coefficient sequence of each layer;

[0087] (2-3) The high-frequency detail coefficients of the spectral reflectance data of all red soil survey samples are stored in the hierarchical discrete detail coefficient set CH = {ch i1，j |i1=1,...,N,j=1,...,t},ch i1，j represents the high-frequency detail coefficient sequence of the jth layer of the spectral reflectance of the i1th red soil survey sample, and N is the number of red soil survey samples;

[0088] (2-4) The high-frequency detail coefficient sequence of the spectral reflectance data of all typical red soil samples is stored in the hierarchical discrete detail coefficient set CD = {cd i2，j |i2=1,...,M,j=1,...,t},cd i2,j represents the high-frequency detail coefficient sequence of the jth layer of the spectral reflectance of the i2th typical red soil sample, and M is the number of typical red soil samples. In this embodiment, the detail coefficient results of each layer after MODWT decomposition of the spectral reflectance of a certain sample are as follows: Figure 3 to Figure 11 shown.

[0089] Step (2-2) includes:

[0090] (2-2-1) Select any data h from the set H i1 , calculate the data h i1 A collection of high-pass filters and low pass filter set

[0091]

[0092] Among them, h l and g l is the th high-pass filter and low-pass filter in discrete wavelet transform, S is the length of the filter;

[0093] (2-2-2) The high-frequency detail coefficient sequence is calculated using the following formula:

[0094]

[0095]

[0096] v=(ul)mod m

[0097] ch i1，j={ch i1，j，u}

[0098] In the formula, ch i1，1，u ,ch i1，j，u represents the uth spectral band component of the 1st and jth layers of the high-frequency detail coefficient sequence of the spectral reflectance of the i1th red soil survey sample, wh i1，1，u 、wh i1，j，u represents the uth spectral band component of the approximate coefficient sequence of the 1st and jth layers of the spectral reflectance of the i1th red soil survey sample, wh i1，j-1，v represents the vth spectral band component of the j-1th layer approximate coefficient sequence of the spectral reflectance of the i1th red soil survey sample, h i1，v Indicates h i1 The uth spectral band component, mod is the modulo operation, and v is the remainder;

[0099] (2-2-3) The spectral reflectance data of other red soil survey samples and the spectral reflectance data of all typical red soil samples in set D are calculated according to step (2-2-2), thereby obtaining a high-frequency detail coefficient sequence of the spectral reflectance data of all red soil survey samples and the spectral reflectance data of all typical red soil samples. In this embodiment, each soil sample has 9 layers of detail coefficients.

[0100] (3) Based on the hierarchical discrete detail coefficient set CD, the multi-scale entropy of each layer is calculated, and the noise layer is determined according to the difference in multi-scale entropy between adjacent layers. The noise layer is removed to obtain the effective layer discrete detail coefficient set C.

[0101] Step (3) includes:

[0102] (3-1) Set the band scale τ to different values, for each high-frequency detail coefficient sequence cd in the layered discrete detail coefficient set CD i2，j , recalibrate at different band scales τ according to the following formula:

[0103]

[0104] In the formula, is the detail coefficient sequence cd i2，j The αth element in the coarse-grained sequence obtained after recalibration at the band scale τ, cd i2，j，d Indicates cd i2，j The u-th spectral band component, M is the number of typical red soil samples, and t is the number of decomposition layers;

[0105] (3-2) Based on Calculate the detail coefficient sequence cd at different band scales τ i2，j The sample entropy of

[0106] (3-3) According to the high-frequency detail coefficient sequence cd under different band scales τ i2，j Sample entropy, calculate the high-frequency detail coefficient sequence cd i2，j In this embodiment, the multi-scale entropy result is as follows: Fig.12 As shown;

[0107] (3-4) For the layered discrete detail coefficient set CD, the difference in multi-scale entropy between every two adjacent layers is calculated;

[0108] (3-5) If the difference between the multi-scale entropy of the x-th layer and the x-1-th layer is smaller than the difference between the multi-scale entropy of the adjacent layers before the x-1-th layer, then the x-t layer is regarded as the effective layer, and the remaining layers are regarded as the noise layer. The high-frequency detail coefficient sequence of the effective layer is stored in the effective layer discrete detail coefficient set C = {cd i2，j′ |j′=x,...,t},where cd i2，j′ represents the high-frequency detail coefficient sequence of the j′th layer of the spectral reflectance of the i2th typical red soil sample, and t is the number of decomposition layers. In this embodiment, x=5.

[0109] (4) Calculate the correlation coefficients of each element in the effective layer discrete detail coefficient set C with the set F and the set P respectively, screen out the spectral bands whose organic matter correlation coefficient is greater than the free iron correlation coefficient, and remove the spectral bands with significant free iron correlation to obtain the characteristic band set CDG.

[0110] Step (4) comprises:

[0111] (4-1) For each high-frequency detail coefficient sequence cd in the effective layer discrete detail coefficient set C i2，j′ , according to cd i2，j′ The Spearman correlation coefficient is calculated for the free iron content corresponding to the set F as the high-frequency detail coefficient sequence cd i2，j′ The correlation coefficient of free iron fd i2，j′ Store in collection FD:

[0112]

[0113]

[0114] Among them, rg(cd i2，j′，u ) is cd i2，j′ The spectral band components of cd are sorted by numerical value i2，j′，u The rank, cd i2，j′，u Indicates cd i2，j′ The uth spectral band component, rg(f i2 ) is the value of the elements in set F after they are sorted by numerical value. i2 The rank of f i2represents the free iron content of the i2th typical red soil sample, represents the difference in ranking of free iron, b represents cd i2，j′ The sum of all spectral band components in;

[0115] (4-2) For each high-frequency detail coefficient sequence cd in the effective layer discrete detail coefficient set C i2，j′ , according to cd i2，j′ The Spearman correlation coefficient is calculated with the corresponding organic matter content in set P as the high-frequency detail coefficient sequence cd i2，j′ The organic matter correlation coefficient pd i2，j′ Deposit into collection PD:

[0116]

[0117]

[0118] Among them, rg(p i2 ) is the set P after the elements are sorted by numerical size. i2 The rank, P i2 represents the organic matter content of the i2th typical red soil sample, Indicates the difference in organic matter related ranking; in this embodiment, the original spectral reflectance of a soil sample and the organic matter correlation coefficient are as follows Fig.13 As shown, the correlation coefficient between the detail coefficient of layers 5-9 and the organic matter is as follows Figure 14-18 As shown;

[0119] (4-3) Based on the set FD and the set PD, the spectral bands ccdgi whose organic matter correlation coefficient in each layer j′ is greater than the free iron correlation coefficient are screened 2 ,j, ,stored in set YZ;

[0120] (4-4) Remove the spectral bands whose free iron correlation coefficient is greater than the preset threshold in the set YZ, and store the remaining spectral bands in the characteristic band set CDG = {cdg i2,j′ |j′=x,...,t}.

[0121] (5) According to the characteristic band set CDG, detail coefficients located on the characteristic band are selected from the hierarchical discrete detail coefficient set CH and stored in the sample set CHG.

[0122] (6) A prediction model for organic matter in red soil survey samples was constructed, and the detail coefficient in the sample set CHG was used as the independent variable, and the organic matter content of the corresponding red soil survey samples was used as the dependent variable to train the prediction model for organic matter in red soil survey samples.

[0123] The organic matter prediction model for red soil survey samples is specifically a linear support vector machine, and may also be other types of neural networks.

[0124] The sample set CHG was divided into a training set and a test set. The linear support vector machine regression method was used to invert the organic matter content in the training set. The model parameters were optimized through ten-fold internal cross-validation to obtain the red soil organic matter prediction model. The above model was then used to estimate the organic matter content of each sample in the validation set, and the estimation accuracy was calculated. The prediction root mean square error (RMSE) and the coefficient of determination (R 2 ), the ratio of standard deviation to RMSE (RPD) of the validation set standard deviation and the predicted root mean square error (RMSE, RPD) as the accuracy evaluation method. In this embodiment, the true value and predicted value results obtained by each independent validation set are as follows Fig.19 shown.

[0125] (7) Calculate the detail coefficient of the visible-short near-infrared spectral reflectance data of the red soil survey sample to be estimated, and input it into the trained red soil survey sample organic matter prediction model to predict the organic matter content.

[0126] The original spectral reflectance (Raw), continuum removal spectral reflectance (Continue Removal, CR), maximum overlap discrete wavelet reconstruction spectral reflectance (noise layer removal, MODWT) and organic matter characteristic band screening based on maximum overlap discrete wavelet elimination of free iron influence (MODWT-SR) preprocessing method claimed in this patent were used to build organic matter estimation models and compare the verification accuracy. The model accuracy obtained by different preprocessing methods is shown in Table 1:

[0127] Table 1 Prediction accuracy of different spectral preprocessing methods

[0128]

[0129] From Table 1, we can see that relative to the original spectrum, the verification R 2 The MODWT pre-processing validation R 2 Improved by 3%, validation of MODWT-SR preprocessing R 2 Increased by 40%.

[0130] Embodiment 2

[0131] The embodiment of the present invention provides a computer device, and the embodiment of the present invention provides services for the implementation of the method of the above-mentioned embodiment 1 of the present invention. The device may include: a memory storing a computer executable program; a processor coupled to the memory 301; the processor calls the computer executable program stored in the memory to execute the steps in the method described in the embodiment 1. The memory may include a computer system readable medium in the form of a volatile memory, such as a random access memory (RAM) and / or a cache memory. The processor executes various functional applications and data processing by running the program stored in the memory, such as implementing the method provided in the embodiment 1 of the present invention. The code of the computer executable program can be written in one or more programming languages ​​or a combination thereof, and the programming language includes an object-oriented programming language, such as Java, Smalltalk, C++, and also includes a conventional procedural programming language, such as "C" language or similar programming languages.

[0132] Embodiment 3

[0133] The embodiment of the present invention provides a storage medium containing a computer executable program, which is used to perform the method of embodiment 1 when executed by a computer processor. The storage medium of the embodiment of the present invention may adopt any combination of one or more computer-readable media. The computer-readable medium may be a computer-readable signal medium or a computer-readable storage medium. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above.

[0134] Embodiment 4

[0135] The embodiment of the present invention also provides a computer product, such as an app on a mobile phone or tablet, an installation program on a computer, etc., which includes a computer program / instruction, and when the computer program / instruction is executed by a processor, the method described in Embodiment 1 is implemented. The code of the computer executable program for performing the operation of the present invention can be written in one or more programming languages ​​or a combination thereof, and the programming language includes an object-oriented programming language, such as Java, Smalltalk, C++, and also includes a conventional procedural programming language, such as "C" language or similar programming languages.

[0136] It should be understood that the above embodiments and descriptions only describe the principles, main features and advantages of the present invention. Without departing from the spirit and scope of the present invention, the present invention may be subject to various changes and improvements, and these changes and improvements all fall within the scope of protection of the present invention.

Claims

1. A red soil organic matter spectrum inversion method for removing the influence of free iron, characterized in that: include: (1) obtaining sample data of a number of red soil survey samples and typical red soil samples, wherein the sample data of the red soil survey samples include visible-short near-infrared spectral reflectance data and organic matter content data, and the sample data of the typical red soil samples include visible-short near-infrared spectral reflectance data, free iron content data and organic matter content data, storing the organic matter data of the red soil survey samples and the typical red soil samples into a set P, storing the free iron content data of the typical red soil samples into a set F, storing the spectral reflectance data of the red soil survey samples into a set H, and storing the spectral reflectance data of the typical red soil samples into a set D; (2) The spectral reflectance data in set H and set D are decomposed using the maximum overlap discrete wavelet MODWT algorithm, thereby generating a hierarchical discrete detail coefficient set CH of the spectral reflectance data of red soil survey samples and a hierarchical discrete detail coefficient set CD of the spectral reflectance data of typical red soil samples; (3) Based on the hierarchical discrete detail coefficient set CD, the multi-scale entropy of each layer is calculated, and the noise layer is determined according to the difference in multi-scale entropy between adjacent layers. The noise layer is removed to obtain the effective layer discrete detail coefficient set C; (4) Calculate the correlation coefficients of each element in the effective layer discrete detail coefficient set C with the set F and the set P, screen out the spectral bands whose organic matter correlation coefficient is greater than the free iron correlation coefficient, and remove the spectral bands with significant free iron correlation to obtain the characteristic band set CDG; (5) According to the characteristic band set CDG, detail coefficients located on the characteristic band are selected from the hierarchical discrete detail coefficient set CH and stored in the sample set CHG; (6) Construct a prediction model for organic matter in red soil survey samples, and use the detail coefficient in the sample set CHG as the independent variable and the organic matter content of the corresponding red soil survey samples as the dependent variable to train the prediction model for organic matter in red soil survey samples; (7) Calculate the detail coefficient of the visible-short near-infrared spectral reflectance data of the red soil survey sample to be estimated, and input it into the trained red soil survey sample organic matter prediction model to predict the organic matter content.

2. The red soil organic matter spectrum inversion method for removing the influence of free iron according to claim 1 is characterized in that: Step (1) comprises: (1-1) Obtaining sample data of N red soil survey samples, including visible-short near-infrared spectral reflectance data and organic matter content data of the red soil survey samples, where N is a positive integer greater than 1; (1-2) obtaining sample data of M typical red soil samples, including visible-short near-infrared spectral reflectance data, free iron content data and organic matter content data of the typical red soil samples, where M is a positive integer greater than 1; (1-3) Obtain the organic matter content data of all red soil survey samples and typical red soil samples and store them in the organic matter content data set P = {p i |i=1,…,n},p i is the organic matter content of the ith soil sample, n is the number of soil samples, where n = N + M; (1-4) Obtain the spectral reflectance data of all red soil survey samples and store them in the set H = {h i1 |i1=1,…,N}, obtain the spectral reflectance data of all typical red soil samples and store them in the set D={d i2 |i2=1,…,M}, the i1th data h in the set H i1 ={h i1,j |j=1,…,m}, the i2th data d in set D i2 ={d i2,j |j=1,…,m},h i1,j is the reflectance of the jth band of the i1th red soil survey sample, d i2,j is the reflectance of the jth band of the i2th typical red soil sample, and m is the total number of spectral bands; (1-5) Obtain the free iron content data of typical red soil samples and store them in the set F = {f i2 |i2=1,2,…,M},f i2 is the free iron content of the ith typical red soil sample.

3. The red soil organic matter spectrum inversion method for removing the influence of free iron according to claim 1 is characterized in that: Step (2) comprises: (2-1) According to the total number of spectral bands m, the number of MODWT decomposition levels t is calculated according to the following formula: t = log2m; (2-2) For the spectral reflectance data of each red soil survey sample in set H and the spectral reflectance data of each typical red soil sample in set D, the MODWT algorithm is used to decompose the t layers to obtain the high-frequency detail coefficient sequence of each layer; (2-3) The high-frequency detail coefficients of the spectral reflectance data of all red soil survey samples are stored in the hierarchical discrete detail coefficient set CH = {ch i1,j |i1=1,…,N,j=1,…,t},ch i1,j represents the high-frequency detail coefficient sequence of the jth layer of the spectral reflectance of the i1th red soil survey sample, and N is the number of red soil survey samples; (2-4) The high-frequency detail coefficient sequence of the spectral reflectance data of all typical red soil samples is stored in the hierarchical discrete detail coefficient set CD = {cd i2,j |i2=1,…,M,j=1,…,t},cd i2,j represents the high-frequency detail coefficient sequence of the jth layer of the spectral reflectance of the i2th typical red soil sample, and M is the number of typical red soil samples.

4. The red soil organic matter spectrum inversion method for removing the influence of free iron according to claim 3 is characterized in that: Step (2-2) includes: (2-2-1) Select any data h from the set H i1 , calculate the data h i1 A collection of high-pass filters and low pass filter set Among them, h l and g l is the lth high-pass filter and low-pass filter in discrete wavelet transform, S is the length of the filter; (2-2-2) The high-frequency detail coefficient sequence is calculated using the following formula: v=(ul)mod m ch i1,j ={ch i1,j,u } In the formula, ch i1,1,u ,ch i1,j,u represents the uth spectral band component of the 1st and jth layers of the high-frequency detail coefficient sequence of the spectral reflectance of the i1th red soil survey sample, wh i1,1,u 、wh i1,j,u represents the uth spectral band component of the approximate coefficient sequence of the 1st and jth layers of the spectral reflectance of the i1th red soil survey sample, wh i1,j-1,v represents the vth spectral band component of the j-1th layer approximate coefficient sequence of the spectral reflectance of the i1th red soil survey sample, h i1,v Indicates h i1 The uth spectral band component, mod is the modulo operation, and v is the remainder; (2-2-3) The spectral reflectance data of other red soil survey samples and the spectral reflectance data of all typical red soil samples in set D are calculated according to step (2-2-2) to obtain the high-frequency detail coefficient sequence of the spectral reflectance data of all red soil survey samples and the spectral reflectance data of all typical red soil samples.

5. The red soil organic matter spectrum inversion method for removing the influence of free iron according to claim 1 is characterized in that: Step (3) includes: (3-1) Set the band scale τ to different values, for each high-frequency detail coefficient sequence cd in the layered discrete detail coefficient set CD i2,j , recalibrate at different band scales τ according to the following formula: In the formula, is the detail coefficient sequence cd i2,j The αth element in the coarse-grained sequence obtained after recalibration at the band scale τ, cd i2,j,d Indicates cd i2,j The uth spectral band component in , M is the number of typical red soil samples, and t is the number of decomposition layers; (3-2) Based on Calculate the detail coefficient sequence cd at different band scales τ i2,j The sample entropy of (3-3) According to the high-frequency detail coefficient sequence cd under different band scales τ i2,j Sample entropy, calculate the high-frequency detail coefficient sequence cd i2,j The multi-scale entropy of (3-4) For the layered discrete detail coefficient set CD, the difference in multi-scale entropy between every two adjacent layers is calculated; (3-5) If the difference between the multi-scale entropy of the x-th layer and the x-1-th layer is smaller than the difference between the multi-scale entropy of the adjacent layers before the x-1-th layer, then the x-t layer is regarded as the effective layer, and the remaining layers are regarded as the noise layer. The high-frequency detail coefficient sequence of the effective layer is stored in the effective layer discrete detail coefficient set C = {cd i2,j′ |j′=x,…,t}, where c i2,j′ It represents the high-frequency detail coefficient sequence of the j′th layer of the spectral reflectance of the i2th typical red soil sample, and t is the number of decomposition layers.

6. The red soil organic matter spectrum inversion method for removing the influence of free iron according to claim 5 is characterized in that: Step (4) comprises: (4-1) For each high-frequency detail coefficient sequence cd in the effective layer discrete detail coefficient set C i2,j′ , according to cd i2,j′ The Spearman correlation coefficient is calculated with the corresponding free iron content in set F as the high-frequency detail coefficient sequence cd i2,j′ The correlation coefficient of free iron fd i2,j′ Store in collection FD: Among them, rg(cd i2,j′,u ) is cd i2,j′ The spectral band components of cd are sorted by numerical value i2,j′,u The rank, cd i2,j′,u Indicates cd i2,j′ The uth spectral band component, rg(f i2 ) is the value of the elements in set F after they are sorted by numerical value. i2 The rank of f i2 represents the free iron content of the i2th typical red soil sample, represents the difference in ranking of free iron, b represents cd i2,j′ The sum of all spectral band components in; (4-2) For each high-frequency detail coefficient sequence cd in the effective layer discrete detail coefficient set C i2,j′ , according to cd i2,j′ The Spearman correlation coefficient is calculated with the corresponding organic matter content in set P as the high-frequency detail coefficient sequence c i2,j′ The organic matter correlation coefficient pd i2,j′ Deposit into collection PD: Among them, rg(p i2 ) is the value of the elements in set P after they are sorted by numerical value. i2 The rank of p i2 represents the organic matter content of the i2th typical red soil sample, It indicates the difference in ranking related to organic matter; (4-3) Based on the set FD and the set PD, the spectral bands ccdg whose organic matter correlation coefficient in each layer j′ is greater than the free iron correlation coefficient are screened. i2,j′ , stored in set YZ; (4-4) Remove the spectral bands whose free iron correlation coefficient is greater than the preset threshold in the set YZ, and store the remaining spectral bands in the characteristic band set CDG = {cdg i2,j′ |j′=x,…,t}.

7. The red soil organic matter spectrum inversion method for removing the influence of free iron according to claim 1 is characterized in that: The red soil survey sample organic matter prediction model is specifically a linear support vector machine.

8. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: The processor executes the computer program to implement the method according to any one of claims 1 to 7.

9. A computer-readable storage medium having a computer program / instruction stored thereon, characterized in that: The computer program / instructions, when executed by a processor, implement the method of any one of claims 1-7.

10. A computer program product comprising a computer program / instructions, characterized in that When the computer program / instructions are executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images