Nondestructive Testing Model of Tea Polyphenols Based on Hyperspectral Technology Combined with Chemometric Methods

Nondestructive detection of tea’s internal quality is great significance for the processing and storage tea. In this study, hyperspectral imaging technology adopted to quantitatively detect content tea polyphenols in Tibetan teas by analyzing features spectrum wavelength ranging from 420 1010 nm. The samples are divided with joint x-y distances (SPXY) Kennard-Stone (KS) algorithms, while six algorithms used preprocess spectral data. Six other Random Forest (RF), Gradient Boosting (GB), Adaptive boost (AdaBoost), Categorical (CatBoost), LightGBM, XGBoost, carry out feature extractions. Then based on a stacking combination strategy, new two-layer prediction model constructed, which compare four individual regressor models: RF Regressor (RFR), CatBoost (CatBoostR), LightGBM (LightGBMR) XGBoost (XGBoostR). experimental results show that newly-built Stacking predicts more accurately than models. coefficients determination Rc2 andRp2 0.9709 0.9625, root mean square error RMSEC RMSEP 0.2766 0.3852 model, respectively, shows can be determined precision.