Non-destructive detection of Tieguanyin adulteration based on fluorescence hyperspectral technique

Abstract

Tieguanyin is one of the top ten famous teas in China, due to its brand effect and market value, illegal businessmen often use adulterated Tieguanyin to make high profits. Tea adulteration detection becomes especially important to eliminate tea fraud in the market. This study developed a non-destructive testing method to detect adulterated Tieguanyin. Benshan was used as adulterated tea and adulterated in the proportion of 0, 5, 10, 20, 30, 45, 60, 75, 90, and 100% (w/w) in Tieguanyin. The fluorescence hyperspectral data of the samples were collected to establish a two-class discrimination model and a prediction model of the degree of adulteration. The two-class discrimination model used support vector classification (SVC) for classification and it worked best when using derivative pre-processing, with 100% recall, precision, and accuracy. In the adulteration degree detection, the support vector regression (SVR) was used for adulteration prediction, and the second derivative (2ndDer)-principal component analysis (PCA)-SVR model predicted the best results with R_c² and R_p² of 0.9298 and 0.9124, respectively, and RMSEC and RMSEP of 0.09 and 0.1044, respectively. Results showed that fluorescence hyperspectral technology has wide application prospects and feasibility in the non-destructive detection of adulterated tea.