Wheat from different producing areas has different flavor and properties, and thus the identification of producing area of wheat is significant to assure the quality of wheat. The traditional method of producing area of wheat determination is time-consuming, complex and needs a lot of pretreatment. The purpose of this research is to develop a new method for the determination of wheat producing areas by terahertz time domain spectroscopy in combination with chemometrics. Firstly, a total of 240 wheat samples from Shandong Province, Shaanxi Province, Henan Province, Hebei Province and Anhui Province of China were collected to analyze and obtain the time-domain spectral signals, frequency-domain spectral signals, and absorption coefficient spectral signals of the samples were obtained. Then, four different preprocessing methods of Savitzky-Golay (S-G), multiplicative scatter correction (MSC), mean centering, and standard normal variate (SNV) were applied to preprocess the absorption coefficient spectral signals, and the uninformative variable elimination (UVE) was used for variable selection of THz spectra data, for developing an effective prediction model. Finally, chemometrics methods, including the partial least squares discriminant analysis (PLS-DA), back propagation neural network (BPNN) and least squares support vector machines (LS-SVM) qualitative models were used for model building and discrimination results obtained through such models were compared. According to the test results, the comprehensive discrimination accuracy of wheat from different origins by the SNV-LS-SVM model reached 96.76%, Furthermore, these results demonstrated that an accurate qualitative analysis of producing area of wheat samples could be achieved by terahertz time-domain spectroscopy combined with chemometrics, which can provide a fast and accurate solution for grain security detection and origin tracing.
Keywords: Chemometrics; SVM model; Terahertz spectroscopy; Wheat.
Copyright © 2021 Elsevier B.V. All rights reserved.