Non-destructive determination of four tea polyphenols in fresh tea using visible and near-infrared spectroscopy

Tea is one of the drinks with the largest consumption in the world. It is widely welcomed by people due to its unique taste and its great nutritive value. And tea polyphenol is a crucial indicator to evaluate the quality of tea, so the detection of tea polyphenols have a great significance. In this paper, a total of 159 tea samples of Juhuachun (J), Zhenong25 (Z) and Yingshuang (Y) were selected. The visible and near-infrared spectrum data of the samples and the physical and chemical values of four tea polyphenols were collected, and three pretreatment methods were used to preprocess the spectra. The optimal pretreatment method was selected by comparative analysis of modeling methods. Subsequently, three feature selection methods of successive projections algorithm, competitive adaptive reweighted sampling and random frog were used to extract the characteristic wavelengths. Based on the characteristic wavelength, the prediction models for partial least squares, multiple linear regression, and least squares-support vector machines were established respectively. After comparative analysis, the results show that the LS-SVM model based on SPA is the most appropriate for detecting tea polyphenols. And its RP2 was 0.996, 0.991, 0.997 and 0.988, respectively. This research shows that Vis-NIR spectroscopy can be used to rapidly detect the content of tea polyphenols in tea. The results of the research provide technical guidelines for non-destructive detection of the quality of tea.