Study On Hyperspectral Inversion Of Main Biochemical Indexes Affecting Tea Quality

Tea is a kind of complex creatures,through the experiment concluded that some chemical composition of tea closely affects the quality of the tea,they mainly includes:tea polyphenols,free amino acidss,soluble sugar,etc.In this article,we selected different varieties of fresh tea leaves from Fujian province Anxi county,including Tieguanyin,Huangdan,Maoxie,Benshan.Each species collected 35 samples.The content of tea polyphenols,free amino acidss,soluble sugar of each sample were determinated by chemistry method,the spectral information were also measured.Adopting the method of statistical modeling.The estimate model for tea samples from different varieties and the estimation model of high spectral of all tea samples were established for accurate analysis and evaluate the content of biochemical components.Through the study it can be concluded as follows:(1)The original spectral data of different varieties of tea samples showed the main similar spectral features,then,analyzed the data of spectral information from different varieties of samples in different methods,extracted the characteristic spectral variables of different varieties of tea samples spectrum data.To measure the correlation between the content of biochemical component and different forms of spectral variables,it can conclude that the correlation of the content of biochemical component with the original spectrum,smooth spectral variables,characteristic spectrum variables was nonsignificant than with the derivative spectra,derivative +SG smooth spectral variables.(2)From the estimation model of the biochemical component can be concluded that compare the effect of the estimated model with derivative spectra and derivative spectra+SG smooth spectrum variables as independent variables was better than with spectrum characteristic variables as independent variables,and the effect of the multiple regression model is better than a single regression model.(3)From the optimal model for hyperspectral estimation different biochemical components of different varieties to analyze,Benshan's optimal estimation models for the content of biochemical components in the tea leaves,the best estimation model of the soluble sugar model's R2 is larger than that of tea polyphenols and free amino acidss,the R2 are 0.912,0.848 and 0.863,respectively.Maoxie's best estimation model of free amino acidss'R2 is higher than tea polyphenols and soluble sugar's,the R2 are 0.919,0.914 and 0.834 respectively.Tieguanyin's optimal estimation models for the content of biochemical components in the tea leaves,the best estimation model of the tea polyphenols model's R2 is larger than that of soluble sugar and free amino acidss,the R2 are 0.867,0.774 and 0.838,respectively.The three biochemical components'estimation models' R2 value of Huangdan were all higher than 0.9.The R2 value of Huangdan's tea polyphenol estimation model is 0.913,and the other two biochemical components were 0.911.(4)From the biochemical component content to analyze estimation models,the best tea polyphenol content estimation models of different tea varieties is Maoxie,the R2 value is 0.914.The best estimation models for free amino acidss content is still for Maoxie,the R2 is 0.919.The best estimation models of soluble sugar content is Benshan and the R2 is 0.912.The best estimation models of different biochemical components in different tea varieties,the tea polyphenol,free amino acidss and soluble sugar's R2 are 0.747,0.794 and 0.812,respectively.(5)In this paper,researched the estimation models of the content and spectral variability of different biochemical components in each tea variety and the estimation models of different biochemical components and spectral variables of mixed variety.In general,the accuracy of hyperspectral estimation models of different biochemical components of single variety is higher than that of mixed varieties,which may be related to the difference of different varieties.