A method for separating spectral information aggregation

Abstract

The application discloses a spectrum information separation and aggregation method, which comprises the following steps: smoothing and denoising spectrum data by using a spectrum denoising method; processing the spectrum data by using a spectrum data processing method, and obtaining spectrum information with at least two dimensions; quantitatively analyzing the correlation between spectrum information with different dimensions by using a correlation analysis algorithm, and calculating a spectrum information coupling coefficient according to the correlation between the dimensions; coupling spectrum information with different dimensions by using a coupling technology, and obtaining final spectrum information; calculating the correlation between the coupled spectrum information and a ground object parameter by using the correlation analysis algorithm, and evaluating the coupling effect; and the application can realize effective aggregation of spectrum data by taking hyperspectral technology as a main means, improve the sensitivity and estimation capability of spectrum data to a detection target, has higher detection precision, better robustness and universality, and can effectively couple available information of spectrum.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of processing and separation of spectral data, and particularly relates to a spectral information separation and aggregation method. BACKGROUND

[0002] Remote sensing technology is a new non-destructive, accurate, real-time, fast information detection method, which has been widely used in county resource, land resource, environmental pollution, food safety, clothing quality and other fields of key indicators detection, and has become a key technology of future intelligent information. With the continuous development of spectroscopy, various new spectra are discovered, and different spectral analysis methods are also established, and corresponding spectral analysis instruments appear. Spectral analysis has been studied for a long time in principle, and has been almost perfect in theory. Spectral analysis has become one of the most modern analytical chemistry methods, which is most widely used and has the strongest function. Spectral analysis method has superior performance in qualitative, quantitative and structural analysis, and has been applied to soil information detection, life science, medicine, food, chemical industry, medicine, environment, commercial inspection, space exploration and other fields. However, the current spectral data processing and analysis technology mainly focuses on the surface, and the research and development of deep separation and aggregation of spectral data are relatively less.

[0003] In order to develop spectral data separation and aggregation technology, and provide basic technical support for the application of remote sensing technology in county remote sensing information detection and various industries, the present application breaks through the conventional thinking inertia, takes the spectral data separation and aggregation method as the breakthrough point, and develops a spectral information separation and aggregation method by means of hyperspectral technology, so as to provide a new basic technical support for the application of remote sensing technology in county remote sensing information and various fields of national economy. SUMMARY

[0004] The present application aims to provide a spectral information separation and aggregation method, which can realize the aggregation and separation of the available information in the spectral data, so as to effectively improve the sensitivity of spectral information to the detection index.

[0005] The detection principle adopted by the present application is as follows: the spectral data can extract available spectral information from the spectral data by using spectral processing technology, and each processing method can obtain different information contained in the original spectrum. The spectral information extracted based on a specific processing method can be regarded as the expression of the original spectrum in a certain dimension. When different methods are used to extract the rich information in the original spectrum, multiple dimensional spectral information can be obtained. If the correlation between different dimensional spectral information is weak, there is strong complementarity between different dimensional spectral information, so coupling different dimensional spectral information with weak correlation has great significance for improving the sensitivity of spectral information to the ground object parameter, and can provide basic technical support for the application and popularization of spectral technology.

[0006] The technical scheme adopted to achieve the above purpose is:

[0007] A spectral information separation-aggregation method, comprising the following steps:

[0008] a1. Smoothing and denoising the spectral data by using a spectral denoising method to improve the signal-to-noise ratio of the spectral data and weaken the interference of noise information on the separation-aggregation of spectral information;

[0009] b2. Processing the spectral data after step a1 by using a spectral data processing method to obtain at least two-dimensional spectral information, and recording each spectral information as SI n ;

[0010] c3. Quantitatively analyzing the correlation between the spectral information of different dimensions obtained in step b2 by using a correlation analysis algorithm, and calculating the spectral information coupling coefficient according to the correlation between the dimensions;

[0011] d4. Coupling the spectral information of different dimensions extracted in step b2 by using coupling technology based on the spectral information coupling coefficient of each dimension in step c3 to obtain the final spectral information;

[0012] e5. Calculating the correlation between the coupled spectral information in step d4 and the ground object parameter by using a correlation analysis algorithm, and evaluating the coupling effect.

[0013] Further, the spectral denoising method in step a1 is to smooth the spectral data by using a low-pass filter, and the coefficient of the low-pass filter is:

[0014] SM = [0.0800, 0.2147, 0.5400, 0.8653, 1.0000, 0.8653, 0.5400, 0.2147, 0.0800].

[0015] Further, the spectral data processing method in step b2 is a traditional mathematical transformation or wavelet transformation or spectral absorption feature algorithm.

[0016] Further, the correlation analysis algorithm in step c3 is:

[0017]

[0018] In the formula, cor j is the correlation coefficient, X i is the spectral reflectance of the i-th sample in the j waveband, Y i is the measured parameter of the sample, and n is the total number of samples; X A is the average value of the spectral reflectance of all samples in the j waveband, Y A is the average value of the parameters of all samples in the j waveband.

[0019] The calculation method of the coupling coefficient of spectral information is:

[0020] cor max_j = max(cor j )

[0021] cor max = [cor max_1 , cor max_2 , …, cor max_n ]

[0022] cor max_all = max(cor max )

[0023]

[0024] Wherein, cor max_j is the maximum correlation coefficient of the jth transform method (or the maximum correlation coefficient of the jth scale based on wavelet decomposition); cor max is a data set constructed based on the maximum correlation coefficient of each transform algorithm; cor max_all is the maximum correlation coefficient of all transform algorithms; coef j is the aggregation coefficient of the jth transform algorithm; max is the maximum value of the vector array.

[0025] Further, in step d4, the spectral data obtained by wavelet transform processing is coupled,

[0026] (1) Based on the same wavelet basis decomposition, the coupling formula of each decomposition scale information is as follows:

[0027]

[0028] Wherein, S c is the coupling result of each decomposition scale information after decomposition with the same wavelet basis; SW j is the jth scale information after decomposition; coef j is the coupling coefficient of the jth scale information after decomposition; n is the decomposition scale.

[0029] (2) Based on different wavelet basis decomposition, the coupling formula of each wavelet basis decomposition information is as follows:

[0030]

[0031] Wherein, S wb_c is the coupling result of each decomposition scale information after decomposition with the same wavelet basis; SW wb_j is the jth scale information after decomposition with wb wavelet basis; coefwb_j is the coupling coefficient of the coupling result of each decomposition scale information after decomposition by the wb wavelet base; m is the number of the wavelet bases used.

[0032]

[0033] wherein S wb_all is the coupling result of each wavelet base; SW wb_c is the coupling result of each decomposition scale information after decomposition by the wb wavelet base; coef wb_c is the coupling coefficient of the coupling result of each decomposition scale information after decomposition by the wb wavelet base; m is the number of the wavelet bases used.

[0034] When the spectral data obtained by processing using the traditional mathematical transform or the spectral absorption feature extraction algorithm is coupled, it is required that one dimension of spectral information is obtained after processing by each method, i.e. the data volume does not increase after processing by each method, and the formula is as follows:

[0035]

[0036] wherein S c is the coupling result of each method; ST i is the spectral information obtained after processing the spectrum by the i th method; coef i is the coupling coefficient of the spectral information obtained by the i th processing method; n is the number of the methods used.

[0037] Further, the maximum value and the average value of the correlation coefficient are used to preliminarily evaluate the coupling effect in the step e5, and the maximum value and the average value of the estimation model accuracy are used to finally evaluate the coupling effect, wherein the maximum value can reflect the optimal effect, and the average value can reflect the overall effect.

[0038] Further, the maximum value and the average value of the correlation coefficient are used to preliminarily evaluate the coupling effect in the step e5, and the maximum value and the average value of the estimation model accuracy are used to finally evaluate the coupling effect, wherein the maximum value can reflect the optimal effect, and the average value can reflect the overall effect.

[0039] (1) Based on the coupling result after decomposition by the same wavelet base, the same wavelet base should be used to separate the coupling result again, and then the evaluation index is used for evaluation, and the overall evaluation or the scale-by-scale evaluation should be performed on all scales to maintain the consistency of the evaluation target;

[0040] (2) Based on the coupling result after decomposition by different wavelet bases, each wavelet base should be used to separate the coupling result again, and then the evaluation index is used for evaluation, and the overall evaluation or the scale-by-scale evaluation should be performed on all scales to maintain the consistency of the evaluation target;

[0041] The coupling result of the spectral information processed by the traditional mathematical transformation method or the spectral absorption characteristic algorithm can be directly analyzed in correlation with the detection target, and an evaluation index is used for evaluation,

[0042] The correlation analysis algorithm calculation formula is as follows:

[0043]

[0044] In the formula, cor j is a correlation coefficient, X i is the spectral reflectivity of the i-th sample in the j waveband, Y i is the measured parameter of the sample, and n is the total number of samples; X A is the average value of the spectral reflectivity of all samples in the j waveband, Y A is the average value of the parameters of all samples in the j waveband.

[0045] Further, the modeling method in step e5 is a random forest or a neural network or a partial least squares algorithm.

[0046] The present application has the beneficial effects that:

[0047] The present application uses hyperspectral technology as the main means, can complete the aggregation and separation of the available information in the spectral data, realizes the effective aggregation of the spectral data, improves the sensitivity and estimation ability of the spectral data to the detection target, provides basic technical support for the application of spectral technology in various fields of national economy, and the detection precision of the present application is higher, the robustness and universality are better, and the available information of the spectrum can be effectively coupled. BRIEF DESCRIPTION OF DRAWINGS

[0048] Figure 1 It is a method flowchart of the present application;

[0049] Figure 2 It is a correlation coefficient comparison chart before and after coupling in the present application;

[0050] Figure 3 It is a modeling precision comparison and analysis chart of the estimation model before and after coupling in the present application;

[0051] Figure 4 It is a verification precision comparison and analysis chart before and after coupling in the present application. DETAILED DESCRIPTION

[0052] The present application will be further described below in combination with the drawings.

[0053] A spectral information separation and aggregation method, as shown in Figure 1 , comprising the following steps:

[0054] a1. Smoothing and denoising the spectral data by using a spectral denoising method to improve the signal-to-noise ratio of the spectral data and weaken the interference of noise information on the separation and aggregation of spectral information;

[0055] b2. Processing the spectral data after step a1 by using a spectral data processing method to obtain at least two dimensions of spectral information, and recording each spectral information as SI n ;

[0056] c3. Quantitatively analyzing the correlation between the spectral information of different dimensions obtained in step b2 by using a correlation analysis algorithm, and calculating the coupling coefficient of the spectral information according to the correlation between the dimensions;

[0057] d4. Coupling the spectral information of different dimensions extracted in step b2 by using a coupling technology based on the coupling coefficient of the spectral information of each dimension in step c3 to obtain the final spectral information;

[0058] e5. Calculating the correlation between the coupled spectral information in step d4 and the ground object parameter by using a correlation analysis algorithm, and evaluating the coupling effect.

[0059] Preferably, the spectral denoising method in step a1 is to smooth the spectral data by using a low-pass filter, and the coefficient of the low-pass filter is:

[0060] SM = [0.0800, 0.2147, 0.5400, 0.8653, 1.0000, 0.8653, 0.5400, 0.2147, 0.0800].

[0061] Taking a corn sample as an example, the specific calculation formula is as follows:

[0062] Let the spectrum of the corn sample be a function f(x) of wavelength, f s (x) be the smoothed spectrum, and i be the wavelength, then:

[0063]

[0064] Preferably, the spectrum data processing method in step b2 is a traditional mathematical transform or a wavelet transform or a spectral absorption characteristic algorithm. The traditional mathematical transform can be a logarithmic transform, a differential transform, a curvature difference transform, etc., the wavelet transform can be a continuous wavelet transform, a discrete wavelet transform, a complex wavelet transform, etc., and the spectral absorption characteristic algorithm can be a de- envelope algorithm, an absorption peak depth algorithm, etc. The present application takes the discrete wavelet transform as an example, the wavelet base selected is Coif2, db5, Meyer, rbio3.7, and sym2, and the decomposition scale is 10. The spectral information based on the discrete wavelet transform algorithm is denoted as SWwb_n, wherein xbj is the wavelet base selected, and n is the decomposition scale (for example, SW db5 _1 is the 1-scale spectral information based on the db5 wavelet base).

[0065] Discrete wavelet algorithm: The discrete wavelet algorithm is a new type of signal processing technology, has a multi-scale analysis and singular point detection function, can utilize a low-pass and a high-pass filter to decompose a signal into a series of high-frequency and low-frequency signals, wherein the spectral resolution of the low-frequency information is reduced by a factor of two with an increase in the decomposition number, and the present application mainly utilizes the low-frequency information of the discrete wavelet. The formula of the discrete wavelet algorithm is as follows:

[0066]

[0067] In the formula, the parameter a represents a scale coefficient, is the inverse of the frequency; the parameter b represents a time shift (or a translation), λ is a wavelength, , and ψ (t) is a mother wavelet.

[0068] In order to obtain the correlation coefficient and the coupling coefficient, the correlation analysis algorithm calculation formula in step c3 is:

[0069]

[0070] In the formula, cor j is the correlation coefficient, X i is the spectral reflectivity of the i-th sample located in the j-th wave band, Y i is the measured parameter of the sample, and n is the total number of samples; X A is the average value of the spectral reflectivity of all samples located in the j-th wave band, Y A is the average value of the parameters of all samples located in the j-th wave band.

[0071] The calculation method of the spectral information coupling coefficient is:

[0072] cor max_j = max(cor j )

[0073] cor max = [cor max_1 , cormax_2 cor max_n ]

[0074] cor max_all = max(cor max )

[0075]

[0076] wherein cor max_j is the maximum correlation coefficient of the jth transform method (or the maximum correlation coefficient of the jth scale based on wavelet decomposition) ; cor max is the data set constructed based on the maximum correlation coefficients of each transform algorithm; cor max_all is the maximum correlation coefficient of all transform algorithms; coef j is the aggregation coefficient of the jth transform algorithm; max is the maximum value of the vector array.

[0077] In the actual calculation of the correlation coefficient and the coupling coefficient, programming can be realized by using languages such as matlab, C, C++, etc.

[0078] Further, in the step d4, when the spectral data processed by wavelet transform is coupled,

[0079] (1) Based on the decomposition of the same wavelet base, the coupling formula of each decomposition scale information is as follows:

[0080]

[0081] wherein S c is the coupling result of each decomposition scale information after decomposition of the same wavelet base; SW j is the jth scale information after decomposition; coef j is the coupling coefficient of the jth scale information after decomposition; n is the decomposition scale.

[0082] (2) Based on the decomposition of different wavelet bases, the coupling formula of each wavelet base decomposition information is as follows:

[0083]

[0084] wherein S wb_c is the coupling result of each decomposition scale information after decomposition of the same wavelet base; SW wb_j is the jth scale information after decomposition of the wb wavelet base; coef wb_j is the coupling coefficient of the jth scale information after decomposition of the wb wavelet base; n is the decomposition scale.

[0085]

[0086] Among them, S wb_all The coupling results for each wavelet basis; SW wb_c This represents the coupling results of information at each decomposition scale after applying the WB wavelet basis decomposition; coef wb_c denoted as the coupling coefficients of the coupling results of information at each decomposition scale after applying the wb wavelet basis decomposition; m represents the number of wavelet bases used.

[0087] When coupling spectral data obtained by traditional mathematical transformations or spectral absorption feature algorithms, it is required that each method yields spectral information in one dimension, meaning that the amount of spectral data does not increase after processing by each method. The formula is as follows:

[0088]

[0089] Among them, S c The coupling results after processing by each method; ST i The spectral information obtained after processing the spectrum using the i-th method; coef i is the coupling coefficient of the spectral information obtained by the i-th processing method; n is the number of methods used.

[0090] In special cases, the coupling coefficient can be considered as 1, but the improvement effect is limited. The following formula can be used instead.

[0091] Row coupling:

[0092]

[0093]

[0094] Among them, S wb_c This represents the coupling results of information from different decomposition scales after applying the same wavelet basis decomposition; SW wb_j This refers to the j-th scale information after decomposition using the wb wavelet basis; coef wb_j Let be the coupling coefficient of the j-th scale information after decomposition using the WB wavelet basis. The special case here refers to situations where rapid calculation is required, or when the correlation between spectral information of different dimensions is weak, but there is strong complementarity between them.

[0095] In this patent, step e5 uses the maximum and average values ​​of the correlation coefficients to perform a preliminary evaluation of the coupling effect, and uses the maximum and average values ​​of the accuracy of the estimation model to perform a final evaluation of the coupling effect. The maximum value reflects the optimal effect, while the average value reflects the overall effect.

[0096] Preferably, in step e5, the coupling result of the spectral information after wavelet transform processing is...

[0097] (1) Based on the coupling results of the same wavelet basis decomposition, the same wavelet basis should be used to separate the coupling results, and then the evaluation index should be used for evaluation. The overall evaluation or scale-by-scale evaluation should be performed on all scales to maintain the consistency of the evaluation target;

[0098] (2) Based on the coupling results of different wavelet basis decomposition, each wavelet basis should be used to separate the coupling results, and then the evaluation index should be used for evaluation. The overall evaluation or scale-by-scale evaluation should be performed on all scales to maintain the consistency of the evaluation target;

[0099] For the coupling results of the spectral information processed by the traditional mathematical transformation method or spectral absorption characteristic algorithm, the correlation analysis can be directly performed with the detection target, and the evaluation index can be used for evaluation,

[0100] The correlation analysis algorithm calculation formula is as follows:

[0101]

[0102] In the formula, cor j is the correlation coefficient, X i is the spectral reflectivity of the i-th sample at the j waveband, Y i is the measured parameter of the sample, and n is the total number of samples; X A is the average value of the spectral reflectivity of all samples at the j waveband, and Y A is the average value of the parameters of all samples at the j waveband.

[0103] Preferably, the modeling method in step e5 is random forest or neural network or partial least squares algorithm.

[0104] Figure 2 is the correlation coefficient comparison chart of the spectral information before and after coupling of different dimensions according to the method of the present patent. In the figure, the R2 curve chart of the spectral information before and after coupling at 1-10 scales and the soil organic matter content, wherein the dotted line is after coupling, and the solid line is before coupling. As can be seen from the figure, in the 1-3 scale, the spectral information after coupling can significantly improve the sensitivity of the spectrum to the soil organic matter content; after coupling, the number of wavebands sensitive to the soil organic matter content is significantly increased, and the distribution in the 350-2500 nm interval is relatively uniform. In the 4-6 scale, the spectral information after coupling can significantly improve the sensitivity of the spectrum to the soil organic matter content in the local area, and can make up for the deficiency of the spectrum before coupling in the local sensitivity. In the 7-9 scale, the sensitivity of the spectrum to the soil organic matter content before and after coupling has no obvious enhancement or weakening. In the 10 scale, the sensitivity of the spectrum to the soil organic matter content after coupling is significantly higher than that before coupling.

[0105] Figure 3 andFigure 4 The figure for comparison and analysis of modeling precision and verification precision of the coupling front and back estimation model in the application; as shown in the figure, overall, the modeling precision and verification precision after coupling are obviously higher than those before coupling, which shows that the coupling algorithm proposed in the application can obviously improve the estimation ability of the spectrum to the soil organic matter content.

[0106] The embodiments are not intended to limit the shape, material, structure, etc. of the application in any form, and any simple modification, equivalent change and modification of the above embodiments according to the technical essence of the application are all within the protection scope of the technical solutions of the application.

[0107] Finally, it should be pointed out that: the above embodiments are only used to illustrate the technical solutions of the application, but not to limit them; although the application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: they can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement for part of the technical features, but these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the application.

Claims

1. A method for separating and aggregating spectral information, characterized in that, Includes the following steps: a1. Use spectral denoising methods to smooth and denoise spectral data, improve the signal-to-noise ratio of spectral data, and reduce the interference of noise information on the separation and aggregation of spectral information; b2. Process the spectral data obtained in step a1 using spectral data processing methods to acquire spectral information with at least two dimensions, and denote each spectral information as SI. n ; c3. For the spectral information of multiple dimensions obtained in step b2, use the correlation analysis algorithm to quantitatively analyze the correlation between spectral information of different dimensions, and calculate the coupling coefficient of spectral information based on the correlation between each dimension. d4. Based on the coupling coefficients of the spectral information of each dimension in step c3, the coupling technology is used to couple the spectral information of different dimensions extracted in step b2 to obtain the final spectral information; e5. Use the correlation analysis algorithm to calculate the correlation between the coupled spectral information and the ground object parameters in step d4, and evaluate the coupling effect.

2. The method for separating and aggregating spectral information according to claim 1, characterized in that, The spectral denoising method in step a1 involves smoothing the spectral data using a low-pass filter. The coefficients of the low-pass filter are: SM=[0.0800, 0.2147, 0.5400, 0.8653, 1.0000, 0.8653, 0.5400, 0.2147,0.0800].

3. The method for separating and aggregating spectral information according to claim 2, characterized in that, The spectral data processing method in step b2 is either traditional mathematical transformation, wavelet transform, or spectral absorption feature algorithm.

4. The method for separating and aggregating spectral information according to claim 3, characterized in that, The correlation analysis algorithm in step c3 is as follows: ； In the formula, X is the correlation coefficient. i Y is the spectral reflectance of the i-th sample located in band j. i Here, n represents the parameter measured in the sample, and n is the total number of samples. The average spectral reflectance of all samples located in band j is given. This represents the average value of all sample parameters located in band j; The method for calculating the coupling coefficient of spectral information is as follows: ； ； ； ； Among them, cor max_j The maximum correlation coefficient of the j-th transformation method, or cor max_j The maximum correlation coefficient at the j-th scale based on wavelet decomposition; This is a dataset constructed based on the maximum correlation coefficient of each transformation algorithm; cor max_all The maximum correlation coefficient for all transformation algorithms; coef j is the aggregation coefficient of the j-th transformation algorithm; max is the maximum value of the vector array.

5. The spectral information separation and aggregation method according to claim 4, characterized in that, In step d4, when coupling the spectral data obtained by wavelet transform processing... (1) The coupling formulas for information at each decomposition scale after decomposition based on the same wavelet basis are as follows: ； in, This represents the coupling results of information from each decomposition scale after applying the same wavelet basis decomposition. This represents the information at the j-th scale after decomposition. is the coupling coefficient of the j-th scale information after decomposition; n is the decomposition scale. (2) The coupling formulas for the decomposition information of each wavelet basis after different wavelet basis decompositions are as follows: ； in, This represents the coupling results of information from each decomposition scale after applying the same wavelet basis decomposition. This refers to the j-th scale information after decomposition using the WB wavelet basis; The coupling coefficients are the j-th scale information after decomposition using the WB wavelet basis; n is the decomposition scale. ； in, The coupling results of each wavelet basis; This is the coupling result of information from each decomposition scale after applying the WB wavelet basis decomposition. The coupling coefficients are the coupling results of the information at each decomposition scale after applying the wb wavelet basis decomposition; m is the number of wavelet bases used. When coupling spectral data obtained by traditional mathematical transformations or spectral absorption feature extraction algorithms, it is required that each method yields spectral information in one dimension, meaning that the amount of spectral data does not increase after processing by each method. The formula is as follows: ； in, The coupling results after processing by each method; This refers to the spectral information obtained after processing the spectrum using the i-th method. is the coupling coefficient of the spectral information obtained by the i-th processing method; n is the number of methods used.

6. The method for separating and aggregating spectral information according to claim 5, characterized in that, In step e5, the maximum and average values ​​of the correlation coefficients are used to initially evaluate the coupling effect, and the maximum and average values ​​of the accuracy of the estimation model are used to finally evaluate the coupling effect. The maximum value reflects the optimal effect, while the average value reflects the overall effect.

7. The method for separating and aggregating spectral information according to claim 6, characterized in that, In step e5, the coupling result of the spectral information after wavelet transform processing is described. (1) Based on the results of decomposition and coupling of the same wavelet basis, the same wavelet basis should be used to separate the coupling results again, and then the evaluation index should be used for evaluation. When evaluating, all scales should be evaluated as a whole or scale by scale to maintain the consistency of the evaluation objectives. (2) Based on the results of decomposition and recoupling of different wavelet bases, the coupling results should be separated again using each wavelet base, and then the evaluation index should be used for evaluation. When evaluating, an overall evaluation or a scale-by-scale evaluation should be performed on all scales to maintain the consistency of the evaluation objectives. The coupling results of spectral information processed using traditional mathematical transformation methods or spectral absorption characteristic algorithms can be directly correlated with the detection target, and evaluation indicators can be used for evaluation. The correlation analysis algorithm uses the following formula for calculation: ； In the formula, X is the correlation coefficient. i Y is the spectral reflectance of the i-th sample located in band j. i Here, n represents the parameter measured in the sample, and n is the total number of samples. The average spectral reflectance of all samples located in band j is given. This is the average value of all sample parameters located in band j.

8. The method for separating and aggregating spectral information according to claim 7, characterized in that, The modeling method in step e5 is random forest, neural network, or partial least squares algorithm.

Images