Accurate Determination of Geographical Origin of Tea Based on Terahertz Spectroscopy

This paper proposes a structured model for the identification of green tea, as well as tracing its geographical origins. Considering that the features of different types of green tea are similar under THz time-domain spectroscopy, we designed a program to perform principal component analysis (PCA) of the spectroscopic data of various green tea samples and to determine the data sequences of principal components. We then established a training set for the principal components to train a support vector machine (SVM) model via a genetic algorithm (GA). We used this model to optimize the parameters and develop a GA-based SVM model with an identification rate of 96.25% for the tested samples. Taken together, our results confirm that THz time-domain spectroscopy combined with GA-SVM can be effectively applied to rapidly identify types of green tea with different geographical origins.