Determination of Tibetan Tea Quality by Hyperspectral Imaging Technology and Multivariate Analysis

Abstract

Tibetan tea is a unique dark tea in Ya'an, its taste and quality are closely related to the content of tea polyphenols (TPs) and free amino acids (FAAs). Firstly, TPs and FAAs were determined using the Folin-ciocalteu colourimetric and ninhydrin colourimetry methods, respectively. The collected hyperspectral data are then entered into the machine learning model through preprocessing and feature dimensionality reduction to predict tea quality. Results showed that SG-SNV-PCA-Extratree had the best predictive ability for TPs with Rp2=0.9248, RMSEP=0.4842, and RPD=3.6460. In detecting FAAs, SG-MSC-PCA-Extratree had the best predictive ability with Rp2=0.8736, RMSEP=0.1590, and RPD=2.8130. SG-MSC-PCA-SVM classified the tea grade with 100% accuracy. These findings highlight the potential of hyperspectral imaging technology (HSI) as an alternative for rapid non-destructive testing of tea quality, which could broaden its applications in other fields.

Keywords: Hyperspectral imaging technology, Multivariate Analysis, TPs, FAAs, Tea grade