Detection Method Of Nitrogen Content In Rubber Tree Leaves Based On Visible/near Infrared Spectroscopy

Rubber trees have always been the commercial source of natural rubber,and the growth of rubber trees is significantly influenced by the region,making rubber trees one of the most important crops in the tropics.At the same time,the content of nitrogen directly affects the growth trend of rubber trees.Therefore,rational application of nitrogen fertilizer is of great significance to maintain the normal growth of rubber trees and increase the production of natural rubber.Based on the visible/near-infrared spectroscopy combined with chemometrics principle,the nitrogen content of rubber tree leaves was tested rapidly,non-destructively and accurately.The main research work is as follows:1、Modeling set diversity method and preprocessing method after removing singular samples.For the 176 rubber tree leaf samples collected,the Monte Carlo cross validation(MCCV)method was used to detect the singular samples.Two samples with numbers 21 and 136 were removed,and the remaining 174 samples were calibration and predicted.Ten-fold cross-validation(10-fold CV),KS,XY three diversity methods on the nitrogen content model.The results show that the model established by the SPXY method is better than the other two models.After SPXY diversity,the spectral data is then subjected to standard normal variate(SNV),Savitzky-Golay,and multiplicative scatter correction(MSC),Smoothing,Wavelet Transform(WT),Normalize preprocessing methods are compared with the modeling effects of the original spectra.The results show that the optimal prediction model of nitrogen content is obtained by MSC,and 13 principal components are extracted.The calibration set Rc=0.1072,RMSEC=0.9438%,the prediction set Rp=0.9123,RMSEP=0.1120%.The experiments were performed using spectral data after MSC processing.2、The wavelength interval and single wavelength selection method for the quantitative analysis model of nitrogen content in rubber tree leaves were studied.Firstly,the interval random frog algorithm(iRF),interval partial least squares(iPLS)and backward interval partial least squares(BiPLS)were compared and analyzed.The ability of the wavelength interval method to predict the nitrogen content of rubber tree leaves determines the optimal selection of the band selection results of the visible/near infrared spectral data by the iRF algorithm,and then uses the wavelength selection method genetic algorithm(GA)and the competitive reweighting adaptive selection algorithm(CARS)screens the full-band spectral data and the optimal interval combination selected by the iRF algorithm.The results show that the PLS nitrogen is established after the rough selection by the iRF algorithm and then by the CARS and GA algorithms(iRF-CARS and iRF-GA).The effect of the content prediction model is better than that of the single-wavelength selection method,and the model is greatly simplified.3、Comparison of prediction accuracy of two wavelength combination methods under different models.Multivariate linear regression(MLR),artificial neural network(ANN)and least squares support vector machine(LSSVM)were established for the two wavelength combination methods of iRF-CARS and iRF-GA.The results show that iRF-CARS-PLS and iRF-CARS-MLR model has the same prediction effect on rubber tree leaves,and is better than other models.The predictions under the four models of iRF-CARS are higher than those of iRF-GA,which proves that the iRF-CARS method is superior to iRF-GA.The iRF-CARS method only selects 16 points as the characteristic wavelength point.Compared with the full-band data modeling,the wavelength significance is reduced by 99.07%.The accuracy correction set of the iRF-CARS-PLS and iRF-CARS-MLR models is Rc=0.9542,RMSEC=0.0971%,the predicted set Rp=0.9345,RMSEP=0.1015%.The iRF-CARS method can indeed be used as a strategy for wavelength selection to detect the nitrogen content of rubber tree leaves.