Study On Black Tea Fermentation Based On Near-Infrared Spectroscopy And Machine Vision

Fermentation is a key process in the process of black tea,which has an important effect on the quality of black tea.Traditionally,black tea fermentation monitoring was mainly based on artificial observation of changes in color and odor of tea leaves.However,experience,mental state,environment and other factors will affect the manual evaluation process,making it highly subjective.In order to objectively evaluate black tea fermentation,this paper takes Yinghong No.9 of Yingde black tea as the research object,and near-infrared spectroscopy and machine vision were used instead of artificial sense.A system monitoring black tea fermentation process based on near-infrared spectroscopy and machine vision was built,and some methods of rapid identification of black tea fermentation degree by combining near-infrared and vision strategy were established.By the analyses of mathematical models,the application of near-infrared and vision technology in the rapid detection of tea quality was discussed.Furthermore,the relationship between the metabolites and near-infrared spectroscopy during black tea fermentation was investigated.The main research contents are as follows:(1)A system monitoring black tea fermentation process based on near-infrared spectroscopy and machine vision was established.The near-infrared spectrum acquisition unit uses Y-type optical fiber to collect tea reflection spectrum,and the image acquisition unit uses industrial cameras to collect tea images in real time.The collected data is input to the computer for processing.(2)A total of 204 black tea samples with different fermentation times were collected and information of black tea fermentation was obtained by portable near-infrared spectrometers and industrial camera.Savitzky-Golay smoothing was used to pretreat the original spectra.Competitive Adaptive Reweighted Sampling(CARS),Successive Projections Algorithm(SPA)and Principal Components Analysis(PCA)were used to reduce the dimension of near-infrared spectrum variables.As for image data,9 color feature variables were extracted,and Pearson correlation analysis and principal component analysis were used to extract the feature variables.Finally,Linear Discriminant Analysis(LDA)and Support Vector Machine(SVM)are used to establish classification models based on near-infrared,image and data fusion.The results show that the discriminant model established by SPA extraction of spectral variables combined with Pearson extraction of image variables is effective,and the accuracy of correction set and prediction set are 97.06% and 95.59%,respectively.(3)The Savitzky-Golay smoothing,the first derivative,the second derivative,the standard normal variable and the multiple scattering correction were used to pretreat the original spectra,and the Partial Least Square Regression(PLSR)prediction model was established.The best pretreatment method of near-infrared spectrum was selected by comparing the performance of models.Competitive Adaptive Reweighted Sampling,Successive Projections Algorithm and Iteratively retained Informative Variables(IRIV)are used to reduce the dimension of nearinfrared spectrum variables.Pearson correlation analysis was used to extract feature variables from image data.Finally,Partial Least Squares Regression model and Support Vector Machine were used to build prediction models based on near-infrared,image and data fusion of the two models,and the model performance was analyzed.(4)Ultra high performance liquid chromatography(UPLC)and tandem mass spectrometry(MS/MS)were used to detect the tea samples in the fermentation process,and metabolomics analysis in the fermentation process of black tea was conducted.The results showed that the metabolites in tea showed a certain change rule during the fermentation process.A total of 470 kinds of metabolites were detected,and 381 kinds of metabolites were screened by one-way analysis of variance.Then orthogonal partial least squares discriminant analysis(OPLS-DA)was performed to analyze 381 substances,and significant differential metabolites were screened according to VIP value > 1,and up-regulated metabolite fold change≥1.5 or down-regulated metabolite fold change≤0.667.A total of 114 kinds of differential metabolites were counted,and the changes of tea substances in the fermentation process were analyzed.The feasibility and reliability of monitoring black tea fermentation by near-infrared spectroscopy were verified by analyzing the relationship between near-infrared spectroscopy and different substances.