The combined terahertz time-domain spectroscopy (THz-TDS) and chemometric technology is used to detect the adulteration of similar substances in Panax notoginseng powder. Four kinds of samples are prepared in the experiment, three kinds of adulterated samples are Panax notoginseng powder adulterating with zedoary turmeric powder, Panax notoginseng powder adulterating with wheat flour, and Panax notoginseng powder adulterating with rice flour, respectively. The values of adulterated concentration are from 5% to 60%, the interval of adulterated concentration is 5%, and the other sample is pure Panax notoginseng powder. The modeling and prediction sets are divided by 3 : 1 by class. The feature information of models is extracted by elimination of uninformative variable (UVE) method and successive projection algorithm (SPA); combining with back propagation neural network (BPNN), the UVE-BPNN and SPA-BPNN qualitative models are established, respectively. The model’s results show that the UVE-BPNN model is better; the classification accuracy of the prediction set of UVE-BPNN is 95%. Then, the least square support vector machine (LS-SVM) algorithm and partial least square (PLS) algorithm are used to establish the quantitative analysis model. The model’s results show that the LS-SVM model is better among the quantitative analysis models of zedoary turmeric powder and wheat flour, the correlation coefficient of prediction (RP) is 0.90 and 0.93 of LS-SVM, respectively, and the root mean square error of prediction (RMSEP) of LS-SVM is 0.072 and 0.068, respectively. Among the quantitative analysis models for rice noodles, the PLS model is better, with the RP of 0.94 and RMSEP of 0.06. The results show that the combined THz-TDS and chemometric technology can be used to determine the adulteration of similar substances in Panax notoginseng powder quickly, accurately, and nondestructively.

In this paper, the combined terahertz time-domain spectroscopy (THz-TDS) and chemometrics method is proposed to identify four different parts of Panax notoginseng rapidly and nondestructively. The research objects of the taproot, scissor, rib, and hairy root of P. notoginseng are taken. The refractive index, absorption coefficient, time-domain, and frequency-domain spectra of the samples are analyzed. It is found that the terahertz spectra of different parts of P. notoginseng are significantly different, so the absorption coefficient of samples is selected to establish models. Firstly, the baseline correction, multiple scattering correction, and normalization algorithms are used to preprocess the absorption coefficient in 0.5–2.0 THz to remove noise. Then, the Kennard–Stone (KS) algorithm is used to divide the model set and the prediction set at the ratio of 3:1, and the successive projection algorithm (SPA) is used to select the characteristic frequency points of the samples. Finally, the chosen characteristic variables are input into the support vector machine (SVM) and linear discriminant analysis (LDA) algorithm to establish the qualitative analysis models, respectively. In the SPA-SVM models, the performance of the SPA-SVM model under the linear kernel function by baseline is best, the accuracy of the training set of it is 99.50%, and the accuracy of the test set of it is 99.25%. In the SPA-LDA models, the performance of the SPA-LDA model by baseline is best, and the accuracy of the training set of it is 100%, and the accuracy of the test set of it is 100%. And the value of cumulative variance contribution is proposed to assess whether the variable is good or bad to model. The results show that the combined THz-TDS and chemometrics method can be used to realize rapid, accurate, and nondestructive identification of different parts of P. notoginseng.