基于高光谱成像技术的木姜叶柯产地识别及品质成分含量预测研究OA

Lithocarpus litseifolius(Hance)Chun.,hailed as the'Chinese Caterpillar Fungus'among teas,is rich in dihydrochalcone glycosides such as phloridzin and 3-hydroxy phloridzin,but their content exhibits significant regional variations.This study proposes a method for geographical origin discrimination and active compound content prediction of Lithocarpus litseifolius(Hance)Chun.based on hyperspectral imaging and chemometrics.Hyperspectral images of samples from four pro-duction regions—Jiangxi,Guizhou,Hunan,and Yunnan,were collected to obtain raw spectral da-ta.Second derivative(SD),multiplicative scatter correction(MSC),and standard normal variate(SNV)were applied for noise reduction.For origin discrimination,partial least squares-discrimi-nant analysis(PLS-DA),support vector machine(SVM),and k-nearest neighbor classification(KNN-Class)models were established.For phloridzin and 3-hydroxy phloridzin content prediction,partial least squares regression(PLSR)and KNN regression(KNN-Reg)models were developed.Ad-ditionally,the successive projections algorithm(SPA)and competitive adaptive reweighted sampling(CARS)were employed for feature wavelength selection.The results demonstrated that the optimal model for origin discrimination was SD-CARS-SVM,achieving a prediction accuracy of 100%.The best models for phloridzin and 3-hydroxy phloridzin content prediction were SD-CARS-PLSR and MSC-CARS-PLSR,respectively,with prediction set determination coefficients(Rp²)of 0.93 and 0.83,and residual prediction deviations(RPD)of 3.87 and 2.45.This study provides a rapid quali-ty assessment solution for Lithocarpus litseifolius(Hance)Chun.and lays the foundation for develop-ing specialized miniaturized detection instruments.