Monitoring black tea fermentation quality by intelligent sensors: Comparison of image, e-nose and data fusion

To scieudfically and objectively monitor the fermentatinn gudlity ot bkick tea, a computer vision system (CVS) and elecuonic nose (e-nose) mere employed to analyze the black tea image and odor ei:,envalues of Yinghong No. 9 black tea. First, me variation trends of tea polyphenols, volatile substances, image eigenvalues and odor eigen-alues with the extension of fermentation time were analyzed, and the fennentation process was categorized into three stages for classification. Second. principal component analysis (PCA) was employed on the image and odor eigenvalues obtained by CVS and e-nose. Partial least sguares disaiminant analysis (PLS-DA) mas peiforrned on '.17 volatile components, and 51 differential volatiles were screened out based on variable irnponance in projection (IP >= 1) and one-way analysis of variance (P < 0.05), including geraniol, linalool, nerolidol, and a-ionone. Then, image features and odor features are zsed by using a data fusion strategy. Finally, the image, smell and fusion information were cornbined with random forest (RF), K-nearest neighbor (KNN) and support vector machine (SVM) to establish the classification rnodels of different fermentation stages and to compaie them. The results show that the feature-level fusion strategy integrating the SVM was the most efficient approach, with classification accmacy rates of 100% for the training sets and 95.6% for the testing sets. The perfomiance of Support Vector Regression (SVR) prediction models for tea polyphenol content based on featurelevel fusion data outperfomied data-level models (Rc, RMSEC, Rp and RMSEP of O.96, 0.48 mg/g, 0.94, 0.6 mg/ g)center dot

Automated summary

To monitor the fermentation quality of black tea objectively, a computer vision system and electronic nose were used to analyze the image and odor features of Yinghong No.9 black tea. The fermentation process was divided into three stages based on the trends of tea polyphenols, volatile substances, image eigenvalues, and odor eigenvalues. Principal component analysis and partial least squares discriminant analysis were employed to analyze the data, and a feature-level fusion strategy with support vector machine showed the highest classification accuracy rates of 100% for training sets and 95.6% for testing sets. The performance of support vector regression models for tea polyphenol content prediction based on feature-level fusion data outperformed data-level models.