Quantitative analysis of soybean protein content by terahertz spectroscopy and chemometrics

Abstract

The protein content in soybeans is one of the important indicators to measure the quality of soybeans. This paper studied the feasibility of terahertz (THz) spectroscopy and chemometrics for quantitative detection of protein content in soybeans. Firstly, the THz absorption spectra of samples were processed by eight pre-processing. Secondly, through the calibration set, partial least squares regression (PLS), principal component analysis-radial basis function neural network (PCA-RBFNN) and artificial bee colony algorithm support vector regression (ABC-SVR) were used to establish quantitative detection models of soybean protein. Samples from the prediction set were used to verify the models. Finally, the related coefficient of prediction set (Rp), root mean square error of prediction set (RMSEP) and relative standard deviation (RSD) of ABC-SVR model were respectively 0.9659, 1.3085% and 3.5334%. The experimental results show that after proper pre-processing, THz spectroscopy and chemometrics can be used to quickly detect the protein content in soybeans.