Simultaneous Optimization of SG Smoothing Parameters and PLS Factor Was Applied to NIRS Analysis of Soil Organic Matter

Abstract

The rapid and simple analytical method for soil organic matter by near infrared spectroscopy (NIRS) is very significant in precision agriculture. In this paper, the rapid determination method and the analysis model of soil organic matter were established by using the NIRS technology, partial least squares (PLS) regression and Savitzky-Golay (SG) smoothing method. Based on the prediction effect of the optimal single wavelength model, calibration set and prediction set were divided. By extending the number of smoothing points and the degree of polynomial, 483 smooth modes were calculated. The PLS models corresponding to all combinations of 483 SG smoothing modes and 1-30 PLS factor were established respectively. The optimal smoothing parameters were the second order derivative smoothing, 2 or 3 degree polynomial, 61 smoothing points, the optimal PLS factor, root mean squared error of predication (RMSEP) and correlation coefficient of predication (RP) were 19, 0.197 (%) and 0.925 respectively, which was obviously superior to the direct PLS model without SG smoothing and the optimal SG smoothing model within 25 smoothing points (the original smoothing method). This demonstrates that the extending of SG smoothing modes and large-scale simultaneous optimization selection of SG smoothing parameters and PLS factor was all very necessary, and can be effectively applied to the model optimization of NIRS analysis.