Qualitative and quantitative analysis of the pile fermentation degree of Pu-erh tea

During the pile fermentation (PF) process, changes to the polyphenol composition directly affect the quality of Pu-erh tea. In this study, we have applied laboratory-made computer vision system (CVS) and miniature near-infrared spectroscopy (NIRS) to the at-line rapid detection of the PF degree of Pu-erh tea at an industrial scale. High-performance liquid chromatography was used for determining the content of catechins (EGCG, EGC ...) and gallic acid (GA). Based on the least-square support vector machine (LSSVM) qualitative model analysis, the texture features and color information extracted by CVS better predicted the degree of PF with a prediction set of 99.30% and calibration set of 100.00% compared to the spectral information extracted by NIRS. The best quantitative models for total catechins (TC), GA/TC, and red and green values of the tea infusion (R-TI, G-TI) were obtained with residual prediction deviations (RPD) of 4.76, 2.36, 5.18, and 4.71, respectively, based on CVS fusion data. And the optimal PF degree of Pu-erh tea was defined when the predicted values of TC, GA/TC, R-TI, and G-TI were in the ranges of 0.46 +/- 0.08 mg/g, 19.04 +/- 6.67, 74.81 +/- 6.37, and 29.81 +/- 2.46, respectively. In-situ quality monitoring of Pu-erh tea PF was realized.

Automated summary

In this study, a laboratory-made computer vision system (CVS) and miniature near-infrared spectroscopy (NIRS) were used for the rapid detection of pile fermentation (PF) degree of Pu-erh tea at an industrial scale. The CVS extracted texture features and color information were found to better predict the PF degree compared to the spectral information extracted by NIRS. By fusing the CVS data, quantitative models for total catechins (TC), GA/TC, and the red and green values of the tea infusion (R-TI, G-TI) were obtained, enabling in-situ quality monitoring of Pu-erh tea PF.