Tea Grade Discrimination Research Based On Spectral Technology

With the rapid development of economy and the improvement of people's living standard,the quality and safety of food are increasingly concerned by consumers.The tea as a kind of health drink is becoming more and more popular among consumers in many countries around the world.At present,the criterion of the tea has not been unified,resulting in uneven quality of tea in the tea market.Usually,consumers judge the quality and grade of tea by the color and smell.This subjective judgment cannot accurately distinguish the real quality of tea.To solve the above problems,I firstly collected a variety of spectral data of tea through near-infrared,mid-infrared and laser-induced breakdown spectroscopy.Secondly,partial least squares discriminant analysis(PLS-DA)model was established based on spectral data,and then the predicted results were obtained.The results showed that this method can make a contribution for Chinese tea evaluation standard.The main research works of this thesis are as follows:1)Taking tea as the research object,I firstly collected near-infrared spectral data as the original data;Secondly,four spectral pretreatments(SG smoothing,baseline correction,standard normal transformation and normalization)were performed on the band data.Meanwhile,continuous projection algorithm(SPA)was used to select the optimal spectral data from the selected band data.Finally,the PLS-DA model was established.Compared with the prediction results using the original data,the model established by selecting the band after normalization processing achieve the optimal prediction.2)Collecting of tea spectral data by middle infrared spectroscopy as the original spectrum;Four spectral pretreatment methods and SPA were used to select spectral data from original spectral data.And then PLS-DA model was established to obtain the predicted results.Compared with the prediction results of original data,the prediction model after SNV processing has the best performance.3)Collecting tea spectral data by laser-induced breakdown spectroscopy as the original spectrum,and the spectral data were selected from the original spectral data and the pre-processed spectral data by two pre-processing methods(smoothing,SNV);then SPA.PLS-DA model was established to obtain the predicted results.Compared with the prediction results of original data,the prediction model after SNV processing has the best performance.4)PLS-DA models of four NIR-MIR,NIR-LIBS,MIR-LIBS and NIR-MIR-LIBS were established,and compared predictive effects.The results showed that the prediction model based on the fusion data of NIR-LIBS and NIR-MIR-LIBS achieve the goal of better accuracy and stability.