Non-invasive anticipation of infusion taste in fine-manipulated green teas through hyperspectral appearance analysis guided by ECG content

The high catechin content in summer-to-autumn tea leaves often results in strong, unpleasant tastes, leading to significant resource waste and economic losses due to lignification of unpicked leaves. This study aims to improve the taste quality of summer-to-autumn green teas by combining fine manipulation techniques with hyperspectral observation. Fine manipulation notably enhanced infusion taste quality, particularly in astringency and its aftertaste (aftertasteA). Using Partial Least Squares Discriminant Analysis (PLSDA) on hyperspectral data, 100% prediction accuracy was achieved for dry tea appearance in the near-infrared spectrum. Astringency and aftertasteA correlated with hyperspectral data, allowing precise estimation with over 90% accuracy in both visible and near-infrared spectrums. Epicatechin gallate (ECG) emerged as a key taste compound, enabling noninvasive taste prediction. Practical applications in processing and quality control are demonstrated by the derived equations (Astringency = -0.88 x ECG + 45.401, AftertasteA = -0.353 x ECG + 18.609), highlighting ECG's role in shaping green tea taste profiles.