Tea tree canopy leaf nitrogen estimation method based on hyperspectrum

Abstract

The invention provides a hyperspectrum-based tea tree canopy leaf nitrogen estimation method, which comprises the following steps of: firstly, smoothing original spectral data of a tea tree canopy by using a Savitzky-Golay method so as to reduce random noise and improve a signal-to-noise ratio; then, the original spectrum is converted into 600 wave bands at the interval of 1 nm through spline interpolation, so that the continuity and convenience of calculation are ensured; the method comprises the following steps of: firstly, carrying out first-order derivative processing on a spectrum, then carrying out second-order derivative processing on the processed spectrum, calculating a normalized difference spectrum index (NDSI) by using a derivative spectrum, calculating a Pearson's correlation coefficient (PCC) between the NDSI and a measured nitrogen element, carrying out principal component analysis (PCA) on the highest 20 indexes of the correlation coefficient, and finally constructing a multiple linear regression model by using the principal components with the first 10 PCA scores as parameters. The method is used for predicting the nitrogen content (LNC) of tea tree canopy leaves. Prediction experiments of a plurality of tea gardens show that the model shows a good prediction effect which is close to the actual nitrogen content, and the generalization performance is good.

Description

Technical Field

[0001] The invention relates to the technical field of spectral nondestructive detection of crop biochemical components, and in particular to a method for estimating nitrogen in tea tree canopy leaves based on hyperspectral. Background Art

[0002] Nitrogen is one of the nutrients necessary for the growth of tea trees. The application of appropriate amounts of nitrogen fertilizer can promote the growth of tea trees and increase yields. However, excessive application of nitrogen fertilizer will cause tea trees to absorb too much nitrogen. Excessive growth of tea trees will be prone to diseases and insect pests, and will affect the quality of tea. In addition, excessive nitrogen fertilizer application will also have a negative impact on the ecological environment of tea gardens. Excessive nitrogen will not be absorbed by the soil and will be lost to groundwater and surface water, leading to eutrophication and water pollution. At the same time, excessive nitrogen fertilizer will also destroy the ecological balance of the soil, affect soil quality and microbial communities, and thus affect soil health and productivity. Therefore, timely, accurate and effective monitoring of the nitrogen status of tea trees is an important task in the process of tea tree planting and management, and is also of great significance for maintaining a good tea garden ecological environment.

[0003] Laboratory destructive analysis based on tea samples can obtain more accurate results on the nitrogen content of tea trees, but the process is complicated and it is difficult to monitor the nitrogen level of tea trees at the tea garden scale in real time and quickly. In recent years, with the rise of hyperspectral technology, a new method has been provided for non-destructive monitoring of vegetation nitrogen. The rich band information of hyperspectral data can more completely reflect the spectral characteristics of vegetation and can be used to distinguish the fine spectral differences of the target. Obtaining nitrogen content information and growth conditions of vegetation through hyperspectral technology can greatly reduce manpower and material costs, and has high accuracy. It can also be applied to monitoring and diagnosis at the plot scale, which is an important technical means for the future development of precision agriculture.

[0004] Previous studies on nitrogen content inversion mainly focused on screening sensitive band reflectance and vegetation index, and the spectral data were mainly obtained under ideal laboratory conditions, without considering the interference caused by various factors in the natural environment. The sensitive band reflectance is easily affected by field noise factors, and the vegetation index uses fixed band calculations and is affected by various environmental factors, such as soil type, vegetation type, humidity, etc. These factors may interfere with the relationship between vegetation index and nitrogen content, thereby reducing the accuracy of prediction. Our research shows that the commonly used crop nitrogen inversion vegetation index such as TCARI, MTCARI, MCARI, NDRE, CCCI, OSAVI, etc., are generally less correlated with nitrogen content under natural conditions, generally less than 0.5, so the generalization performance of the constructed model is not high. Summary of the invention

[0005] In view of this, the purpose of the present invention is to provide a method for estimating nitrogen in tea tree canopy leaves based on hyperspectral data, which can make full use of the important information of spectral data. Experimental results in multiple tea gardens show that the model has high prediction accuracy and exhibits good generalization performance.

[0006] To achieve the above object, the present invention adopts the following technical solution: a method for estimating nitrogen in tea tree canopy leaves based on hyperspectral, comprising the following steps:

[0007] Step 1: Collection of tea tree canopy hyperspectral data;

[0008] Step 2: Determination of nitrogen content in tea tree canopy leaves: a certain number of leaves are uniformly collected from the canopy area corresponding to each spectral measurement sample point, packed into bags, sealed and sent to the laboratory, and the nitrogen content of the leaves is determined using the Kjeldahl method;

[0009] Step 3: Smoothing and interpolation of the original spectrum; using the Savitzky-Golay method to smooth it to reduce random noise and improve the signal-to-noise ratio of the data, and then interpolating the original spectrum to a suitable band interval to ensure the continuity and convenience of the calculation;

[0010] Step 4: Spectral derivative transformation and index calculation: perform second-order derivative processing on the processed spectrum and calculate the NDSI index of the full band of the derivative spectrum;

[0011] Step 5: Calculate the correlation between nitrogen content and derivative spectral index; calculate the Pearson correlation coefficient PCC between all NDSI indices and the corresponding nitrogen content, and record the top 20 values ​​with the highest correlation;

[0012] Step 6: Principal component analysis and regression model construction: Use principal component analysis (PCA) to determine the top 10 principal component scores of the top 20 NDSI indices with the highest correlation; then construct a multivariate linear regression model with the above 10 principal components as parameters, evaluate its effect and use it for actual prediction.

[0013] In a preferred embodiment, step 1 is specifically as follows: spectral collection uses the AvaField-1 portable hyperspectral spectrometer produced by Avantes of the Netherlands; data collection operators all wear dark clothing, fix the spectrometer on a tripod, keep the sensor probe horizontal and stable and face down 0.5m above the tea tree canopy, the sensor probe uses a 25° field of view, and each sample point continuously collects spectral data 5 times within one square meter, and the average value is used as the tea tree canopy spectrum of the sample point; before collecting data at each sample point, a standard whiteboard is used for spectral correction.

[0014] In a preferred embodiment, step 2 specifically includes: after collecting spectral data at each spectral sampling point, immediately and evenly collect 15-20 fresh tea leaves within one square meter of the sample point, put them into bags and seal them, and send them to the laboratory to determine the nitrogen content of the leaves using the Kjeldahl method.

[0015] In a preferred embodiment, step 3 specifically includes: converting the smoothed spectral data into 600 bands with an interval of 1 nm by spline interpolation, and the band range is between 400 and 1000 nm, so as to ensure the continuity and convenience of calculation.

[0016] In a preferred embodiment, step 4 specifically includes: calculating the second-order derivatives of all original spectral data, and then further calculating the normalized difference vegetation index NDSI of all bands of the derivative spectrum, and the calculation formula is as follows:

[0017]

[0018] Where x2 and x1 refer to the raw reflectance at any particular wavelength, or the relative reflectance after a series of transformations.

[0019] In a preferred embodiment, step 5 specifically includes: calculating the Pearson correlation coefficient PCC between the NDSI index of all bands of the derivative spectrum and the corresponding nitrogen content, and recording the top 20 values ​​with the highest correlation.

[0020] In a preferred embodiment, step 6 specifically includes: performing principal component analysis on the top 20 correlation values ​​obtained by the above process to achieve further dimensionality reduction, recording the top 10 components of the principal component analysis scores, and using the 10 principal components to construct a multivariate linear regression model; the regression model uses the square of the correlation coefficient R 2 , adjusted square correlation coefficient adjR 2 , root mean square error RMSE and the statistical significance of the model, i.e., p-value, are used to evaluate the goodness of fit of the model; the calculation formulas for each indicator are as follows:

[0021]

[0022] Among them: SSR: residual sum of squares,

[0023] SST: Total sum of squares,

[0024] y i is the actual value, is the predicted value, is the actual average value;

[0025]

[0026] Where: n is the number of samples, p is the number of independent variables

[0027]

[0028] Where: y i is the actual value, is the predicted value, n is the number of samples;

[0029] The regression model fitting formula is as follows:

[0030]

[0031] Among them, X i The first 10 components of the principal components.

[0032] Compared with the prior art, the present invention has the following beneficial effects: first, the original spectral data is smoothed by using the Savitzky-Golay method to reduce random noise, and then the signal-to-noise ratio is improved by performing a second-order derivative change on the processed spectrum so as to more accurately analyze the spectral data and extract sensitive information related to nitrogen. The most important innovation of the present invention is that the normalized difference spectral index (NDSI) is calculated for the entire band, and the Pearson correlation coefficient (PCC) between these indices and the actual nitrogen content is obtained. This method can effectively mine the spectral index characteristics related to nitrogen content. The PCA transformation is performed on the 20 indices with the highest correlation coefficients to further reduce the data dimension, and then the principal components of the top 10 PCA scores are used as parameters to construct a multivariate linear regression model, and multiple spectral features are integrated to improve the accuracy of predicting the nitrogen content of tea tree canopy leaves. In the prediction experiments of multiple tea gardens, the model showed good prediction results, which were close to the actual nitrogen content. The experimental results show the effectiveness and practicality of the method. The method showed stability in experiments in different tea gardens, indicating that it has a certain robustness and can provide reliable prediction results under different environments and conditions.

[0033] In summary, this invention significantly improves the signal-to-noise ratio of the original spectrum through a series of innovative data processing technologies, providing a good foundation for building a high-quality nitrogen prediction model, thereby improving the accuracy and reliability of predicting the nitrogen content of tea tree canopy leaves, and has important application value for tea tree planting and tea garden management. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 A flowchart of a preferred embodiment of the present invention;

[0035] Figure 2 The prediction result of the tea garden JGK42 of the preferred embodiment of the present invention;

[0036] Figure 3This is the prediction result of tea garden HS54 in the preferred embodiment of the present invention;

[0037] Figure 4 It is the prediction result of tea garden YP20 in the preferred embodiment of the present invention;

[0038] Figure 5 This is the prediction result of the tea garden YZX16 in the preferred embodiment of the present invention. DETAILED DESCRIPTION

[0039] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0040] It should be noted that the following detailed descriptions are illustrative and are intended to provide further explanation of the present application. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present application belongs.

[0041] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present application; as used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprise" and / or "include" are used in this specification, they indicate the presence of features, steps, operations, devices, components and / or their combinations.

[0042] A hyperspectral method for estimating nitrogen content in tea tree canopy leaves, referring to Figure 1-5 , including the following steps:

[0043] Step 1: Collection of tea tree canopy hyperspectral data. Sampling points were evenly set at a certain distance in the experimental tea garden, and hyperspectral reflectance data were collected at suitable weather times during the vigorous growth stage of the tea trees.

[0044] Step 2: Determination of nitrogen content in tea tree canopy leaves. A certain number of leaves were uniformly collected from the canopy area corresponding to each spectral measurement sample point, packed into sealed bags and sent to the laboratory for determination of leaf nitrogen content using the Kjeldahl nitrogen method.

[0045] Step 3: Smoothing and interpolation of the original spectrum. The Savitzky-Golay method is used to smooth it to reduce random noise and improve the signal-to-noise ratio of the data. Then the original spectrum of the tea tree canopy is interpolated to a suitable band interval to ensure the continuity and convenience of the calculation.

[0046] Step 4: Spectral derivative transformation and index calculation: Perform second-order derivative processing on the processed spectrum and calculate the NDSI index of the full band of the derivative spectrum.

[0047] Step 5: Calculate the correlation between nitrogen content and NDSI index. Calculate the Pearson correlation coefficient PCC between the NDSI index of the derivative spectrum and the corresponding nitrogen content, and record the top 20 values ​​with the highest correlation.

[0048] Step 6: Principal component analysis and regression model construction. Use principal component analysis (PCA) to determine the first 10 principal component scores. Then build a multivariate linear regression model with the above 10 principal components as parameters, evaluate its effect and use it for actual prediction.

[0049] Specifically:

[0050] 1. Study area and spectral data collection

[0051] The study area is located in a city in the northwest of Fujian Province, China. The study area is 320 meters above sea level, with a subtropical monsoon climate, an average annual temperature of 18-20°C, a relative humidity of about 80-85%, a mild and humid climate, and an annual rainfall of about 2200 mm. The tea tree canopy spectral data collection time is July 2022 and April 15, 2023, from 10:00 to 14:00 on the same day in clear and cloudless weather. The spectrum was collected using the AvaField-1 portable hyperspectral ground feature spectrometer produced by Avantes in the Netherlands, with a spectral range of 200-1100nm, a spectral resolution of 1.5nm, a spectral sampling interval of 0.6nm, and a wavelength accuracy of ±0.3nm.

[0052] Data collection was carried out in 4 different tea gardens (named "YZX16", "YP20", "JGK42", and "HS54" respectively). The sampling points in each tea garden were distributed as evenly as possible, and the distance between the sampling points was about 10 meters. The tea trees at all sampling points grew well and were not covered by other plants. The coverage rate was greater than 90%, the canopy leaves were light green in color, and the canopy structure was almost flat. The operators all wore dark clothing, fixed the spectrometer on a tripod, kept the sensor probe horizontal and stable and facing down 0.5m above the tea tree canopy, and the sensor probe used a 20° field of view. Each sample point collected spectral data 5 times continuously within one square meter, and the average value was used as the tea tree canopy spectrum of the sample point. In order to ensure the measurement accuracy and reduce the impact of environmental and time changes, a standard whiteboard was used for spectral correction before collecting data at each sample point. Due to the different scales of each tea garden, the number of sampling points in each tea garden is also different.

[0053] 2. Obtaining tea tree canopy nitrogen content

[0054] The canopy nitrogen content was determined by artificial destructive laboratory determination. After collecting spectral data at each sample point, 15 canopy leaves were collected evenly within the corresponding range and placed in a corresponding numbered sealed bag, and then brought back to the laboratory for nitrogen content determination. The fresh samples were brought back to the laboratory and washed with tap water and deionized water in turn, and then wiped with clean yarn to remove the surface moisture and placed in a blast oven to dry at 80°C to constant weight, then cooled to 60°C and crushed with a pulverizer, mixed through a 40-mesh sieve, 0.3g of dry sample powder was weighed and placed in a digestive tube, 5ml of sulfuric acid was added and shaken, and hydrogen peroxide was added twice. After the reaction was completed, it was placed on a digestion furnace for heating and digestion to make the solid disappear and become a solution, and the nitrogen content value was finally determined by the Kjeldahl nitrogen analyzer. The standard substance used was GBW07603 (GSV-2), and the N content in the table was 15±0.3g / Kg. The final nitrogen content (g / kg) was calculated by formula 1.

[0055] LNC (g / kg) = sample weight × fixed volume / concentration / 1000 (6)

[0056] 3. Spectral data preprocessing

[0057] First, all spectral data were smoothed using the Savitzky-Golay method to reduce random noise and improve the signal-to-noise ratio. To ensure the continuity and convenience of calculation, the original spectra (1092 bands with 0.6 nm intervals) were converted into 600 bands with 1 nm intervals (ranging from 400 to 1000 nm) by spline interpolation. Both interpolation and smoothing operations were performed in Matlab R2022b. The linear relationship between leaf nitrogen content and spectra in a tea garden was measured using the Pearson correlation coefficient (PCC). Then, the processed spectra were further processed by second-order derivatives.

[0058] 4. Feature Transformation

[0059] The NDSI of the full band of the derivative spectrum was calculated using formulas (1) and (2). This process converted the spectrum into feature space and index form. The linear correlation of these indices with leaf nitrogen content was evaluated using the Pearson correlation coefficient (PCC). For each sampled tea garden, the top 20 PCC values ​​associated with leaf nitrogen content were recorded. For example, in the tea garden named HS54, there were 54 sampling points and their corresponding nitrogen contents and spectra, and the PCC values ​​of the NDSI index relative to these 54 nitrogen contents and spectra were calculated. Other tea gardens were treated in the same way.

[0060] 5. Principal Component Analysis (PCA)

[0061] PCA can reduce the dimensionality while retaining the information of the data to the greatest extent by identifying a new set of coordinate systems (i.e., linear combinations of the original basis). The number of components generated by PCA is the same as the number of input features. In this study, we selected the first 10 principal components that can explain at least 95% of the total variance, and the remaining factors were ignored to achieve dimensionality reduction. The scores of each sample on these principal components were also recorded.

[0062] 6. Multiple linear regression and result analysis

[0063] The PCA score values ​​of the NDSI features were analyzed using a multiple linear regression routine to determine their correlation with nitrogen content. Multiple linear regression uses linear combinations of the independent variables (e.g., 10 scores) to fit the observed dependent data set (e.g., nitrogen concentration). A linear regression model with 10 components was constructed and correlated with nitrogen concentration. The correlation coefficient squared (R 2 ), adjusted square correlation coefficient (adjR 2 ), root mean square error (RMSE) and statistical significance (p value) of the model to evaluate the goodness of fit of the model. The linear equation form determined by regression is shown in formula (6).

[0064] y=50.0000+1.0273x1-0.0262-x2+1.3885x3+0.5974x4+0.4153x5+0.4193x6+0.3116x7+0.0349x8-0.2748x9+0.5204x10 (6)

[0065] Table 1 Statistics of the principal component linear regression model of NDSI in tea garden JGK42 under different spectral treatments

[0066]

[0067] The root mean square error (RMSE) is a measure of the difference between the model's predictions and the actual observed values. The smaller the RMSE, the higher the prediction accuracy of the model. An RMSE of 2.77 means that the average error of the model's predictions is about 2.77 units. An R of 0.88 2 This indicates that approximately 88% of the variation in the dependent variable can be explained by the independent variables in the model. Adjusted R 2 Taking into account the number of independent variables in the model, its value of 0.76 indicates that the model has relatively strong explanatory power. The p-value of 3.29e-07 is much less than 0.01, indicating that the model is very significant.

[0068] The results show that the multivariate linear regression model has high explanatory power (explaining 88% of the variation), is overall significant (the p-value is very small), and has relatively good prediction accuracy (RMSE is 2.77).2 It is also relatively high, indicating that even after considering the number of independent variables, the model still has good explanatory power.

[0069] 7. Prediction Model Validation

[0070] For each sampled tea garden, the top 20 PCC values ​​of the NDSI index related to leaf nitrogen content were also calculated. The results of NDSI are shown in Table 1. Both tables show that the top 20 PCC values ​​of the derivative spectral index are higher. As shown in Table 1, the second-order derivative NDSI often shows the highest top 20 PCC values. This effect is also consistent in other tea gardens, and the PCC values ​​of the derivative index are always higher than the PCC values ​​of the non-derivative index. As shown in Table 2.

[0071] Table 2 Correlation between the top 20 components of NDSI derivatives and nitrogen PCC in different tea gardens

[0072]

[0073] The multivariate linear regression model constructed above was also applied to the prediction of tea canopy nitrogen in other tea gardens and compared with the original spectrum and first-order derivative spectrum NDSI regression model. The specific results are shown in Tables 3 and Figure 2 , 3 , 4, and 5. The results show that the prediction accuracy of the second-order derivative NDSI model is the highest, and the prediction accuracy in the other three tea gardens has achieved good results, R 2 The prediction accuracy is basically above 0.7. Among them, the highest prediction accuracy is the tea garden YZX16. 2 reached 0.96; the lowest was in the tea garden HS54, whose predicted R 2 The value is 0.69, close to 0.7. The prediction results of each tea garden prove that the model has not only high accuracy but also good generalization performance.

[0074] Table 3 Prediction results of different tea gardens

[0075]

[0076]

[0077] The present invention performs noise reduction processing such as smoothing, interpolation and derivative transformation on the hyperspectral data of the canopy of tea trees in the wild tea garden, and then calculates the normalized difference vegetation index (NDSI) of the full spectrum band, extracts the top 20 sensitive index information with the highest correlation with nitrogen content, and then further performs principal component analysis; finally, the principal component of the normalized difference vegetation index is used to construct a multivariate linear regression model. Since the present invention uses the full-band data of the spectral information for change and index calculation, it can make full use of the important information of the spectral data. Experimental results in multiple tea gardens show that the model has high prediction accuracy and exhibits good generalization performance.

[0078] The above is only a preferred embodiment of the present invention and is not intended to limit the present invention. The focus of the present invention is the processing technology of the original spectrum of the tea tree canopy. Through derivative transformation, difference vegetation index calculation and principal component analysis, the signal-to-noise ratio of the original data is effectively improved and the dimension is reduced, thereby providing a prerequisite for building an accurate and reliable tea tree canopy nitrogen inversion model. Any modification, equivalent replacement, improvement and other data processing technologies made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for estimating nitrogen in tea tree canopy leaves based on hyperspectral, characterized in that: The following steps are involved: Step 1: Collection of tea tree canopy hyperspectral data; Step 2: Determination of nitrogen content in tea tree canopy leaves: a certain number of leaves are uniformly collected from the canopy area corresponding to each spectral measurement sample point, packed into bags, sealed and sent to the laboratory, and the nitrogen content of the leaves is determined using the Kjeldahl method; Step 3: Smoothing and interpolation of the original spectrum; using the Savitzky-Golay method to smooth it to reduce random noise and improve the signal-to-noise ratio of the data, and then interpolating the original spectrum to a suitable band interval to ensure the continuity and convenience of the calculation; Step 4: Spectral derivative transformation and index calculation: perform second-order derivative processing on the processed spectrum and calculate the NDSI index of the full band of the derivative spectrum; Step 5: Calculate the correlation between nitrogen content and derivative spectral index; calculate the Pearson correlation coefficient PCC between all NDSI indices and the corresponding nitrogen content, and record the top 20 values ​​with the highest correlation; Step 6: Principal component analysis and regression model construction: Use principal component analysis (PCA) to determine the top 10 principal component scores of the top 20 NDSI indices with the highest correlation; then construct a multivariate linear regression model with the above 10 principal components as parameters, evaluate its effect and use it for actual prediction.

2. The method for estimating nitrogen in tea tree canopy leaves based on hyperspectral according to claim 1, characterized in that: Step 1 is as follows: spectral collection uses the AvaField-1 portable hyperspectral ground feature spectrometer produced by Avantes of the Netherlands; data collection operators all wear dark clothing, fix the spectrometer on a tripod, keep the sensor probe horizontal and stable and face down 0.5m above the tea tree canopy, the sensor probe uses a 25° field of view, and each sample point collects spectral data five times continuously within one square meter, and the average value is used as the tea tree canopy spectrum of the sample point; before collecting data at each sample point, a standard whiteboard is used for spectral correction.

3. The method for estimating nitrogen in tea tree canopy leaves based on hyperspectral according to claim 1, characterized in that: Step 2 specifically includes: after collecting spectral data at each spectral sampling point, immediately and evenly collect 15-20 fresh tea leaves within one square meter of the sample point, put them into bags and seal them, and send them to the laboratory to determine the nitrogen content of the leaves using the Kjeldahl method.

4. The method for estimating nitrogen in tea tree canopy leaves based on hyperspectral according to claim 1, characterized in that: Step 3 specifically includes: converting the smoothed spectral data into 600 bands with an interval of 1 nm through spline interpolation, and the band range is between 400 and 1000 nm to ensure the continuity and convenience of calculation.

5. The method for estimating nitrogen in tea tree canopy leaves based on hyperspectral according to claim 1, characterized in that: Step 4 specifically includes: calculating the second-order derivatives of all original spectral data, and then further calculating the normalized difference vegetation index NDSI of all bands of the derivative spectrum. The calculation formula is as follows: Where x2 and x1 refer to the raw reflectance at any particular wavelength, or the relative reflectance after a series of transformations.

6. The method for estimating nitrogen in tea tree canopy leaves based on hyperspectral according to claim 1, characterized in that: Step 5 specifically includes: calculating the Pearson correlation coefficient PCC between the NDSI index of all bands of the derivative spectrum and the corresponding nitrogen content, and recording the top 20 values ​​with the highest correlation.

7. The method for estimating nitrogen in tea tree canopy leaves based on hyperspectral according to claim 1, characterized in that: Step 6 specifically includes: performing principal component analysis on the top 20 correlation values ​​obtained by the above process to achieve further dimensionality reduction, recording the top 10 components of the principal component analysis scores, and using the 10 principal components to construct a multivariate linear regression model; the regression model uses the square of the correlation coefficient R 2 , adjusted square correlation coefficient adjR 2 , root mean square error RMSE and the statistical significance of the model, i.e., p-value, are used to evaluate the goodness of fit of the model; the calculation formulas for each indicator are as follows: Among them: SSR: residual sum of squares, SST: Total sum of squares, y i is the actual value, is the predicted value, is the actual average value; Where: n is the number of samples, p is the number of independent variables Where: y i is the actual value, is the predicted value, n is the number of samples; The regression model fitting formula is as follows: Among them, X i The first 10 components of the principal components.

Images