Study On Grades And Biochemical Components Of Green Tea Based On Visible-Near-Infrared Spectroscopy Technology

Green tea is one of the three most-consumed kinds of drink in the world with its refreshing and health-caring features.Tea polyphenols and free amino acids are main components whose content directly affects the color,aroma and taste of green tea.High-quality green tea is not only a hot product,but also the direction for the sound development of the tea.Therefore,it is highly necessary to explore the rapid and non-destructive analysis methods for identification of green tea grades and its quality components.This paper,which takes Meitan Cuiya tea as a research object,measures the visible-near-infrared spectra of 150 crushed green tea samples through spectrograph to analyze the correlations between green tea grades,polyphenols and free amino acids on the one hand and their visible-near-infrared spectroscopy on the other hand.Additionally,the spectral classification method for green tea,and the spectral detection methods for the content of green tea polyphenols and free amino acids are also discussed.The main research findings and conclusions are as follows:(1)The original spectral data of the samples are pre-processed by five treatments including Savitzky-Golay smoothing(SG),multiple scattering correction(MSC),standard normal variable transformation(SNV),first derivative and detrending.Then the partial least squares regression(PLSR)model and the support vector regression(SVMR)model are established based on different preprocessing methods and raw data to study the influence of different spectroscopic pretreatment methods on different models.The results show that the SG method of the PLSR model works the best in green tea grade identification,with the correlation coefficient of the model prediction set at 0.9565,and the minimum root mean square error at 0.2958.In the quantitative detection of tea polyphenol content,the SG method in SVMR model has the best pretreatment effect,with the correlation coefficient of the model prediction set at 0.9838,and the minimum root mean square error at 0.3691.And in the quantitative detection of free amino acid content,the SG method in SVMR model also shows the best pretreatment effect,with the correlation coefficient of the model prediction set at 0.9138,and the minimum root mean square error at 0.1902.Therefore,the pretreatment of the original spectrum can improve the accuracy of modeling.(2)Stepwise regression analysis method(SWR),continuous projection algorithm(SPA),and competitive adaptive reweighting algorithm(CARS)are adopted to reduce the dimension of the pretreated spectral data,and the PLSR models for the five types of Cuiya tea grades identification are discussed based on three different characteristic wavelengths and full-wave band data.The results show that all four models have good prediction effects.In green tea grades identification,the CARS-PLSR model works the best in prediction,with the correlation coefficient of the model prediction set at 0.9739,and the root mean square error at 0.2250.In the quantitative detection of tea polyphenol content,the SG Smoothing-SVMR model shows the best prediction effect,with the correlation coefficient of the model prediction set at 0.9838,and the root mean square error at 0.1256.In the quantitative detection of free amino acid content,the SPA-SVMR model works the best,with the correlation coefficient of the model prediction set at 0.9392,and the root mean square error at 0.1532.Therefore,it is feasible to use the characteristic wavelength screening to identify the grade,tea polyphenols content and free amino acid content of Meitan Cuiya tea.(3)It can be known from the predicted effects of the set that the standard deviation of all 40 samples within the margin of error.Therefore,the near-infrared spectroscopy proves viable for effective quantitative analysis of the grades and biochemical components(tea polyphenols and free amino acids)of Meitan Cuiya tea.