Intelligent evaluation of black tea fermentation degree by FT-NIR and computer vision based on data fusion strategy

Abstract

Fermentation is the most crucial step in black tea processing. In this study, we used a combination of Fourier transform near-infrared spectroscopy (FT-NIR) and computer vision system (CVS) for comprehensively evaluating the black tea fermentation degree. 110 samples of black tea at different fermentation degrees were collected in this research. The tea polyphenols (TP) content was measured using a UV–Vis spectrometer, and TP degradation during black tea fermentation was measured for classifying the degree of fermentation. Two mid-level strategies were used to analyze the fusion signals of FT-NIR and CVS. K-nearest neighbors, linear discriminant analysis, and a support vector machine (SVM) were applied for classification modeling. The advantages of FT-NIR and CVS were integrated based on mid-level fusion, and the results obtained were better than those obtained using independent methods for evaluating black tea fermentation. The mid-level fusion strategy that combined FT-NIR and CVS for principal component analysis feature extraction was the most effective in the SVM model. Classification accuracies of calibration and prediction sets were each 100%. Therefore, in this study, it was demonstrated that a combination of FT-NIR and CVS at mid-level fusion strategy can be used as a method to rapidly evaluate the black tea fermentation degree.