Grading of Green tea and quantitative determination of beta-carotene and lutein based on hyperspectral imaging

Abstract. Green tea comes from leaves of Camellia sinensis after a series of special processing technology such as fixation, rolling and drying. With the different leaves and process, there are large number of green tea varieties. Therefore, the different green tea varieties and grades which contain different chemical and medical characteristics are need to be classified and graded. In this paper, hyperspectral technology coupled with four kinds of classification methods were applied to detect the variety and grade of five green teas (Biluochun, Jingshan, Longjing, Queshe, Sanbeixiang). Besides, some quantitative analysis had been used on the spectra to fast predict two lipid-soluble pigments beta-carotene and lutein. Robust and universal models were built for green tea classification and grading. The results showed that support vector machine (SVM) obtained relative better results on separating three grades of each green tea with the classification accuracy higher than 93%. Moreover, the quantitative results showed that the partial least squares (PLS) model combined with competitive adaptive reweighted sampling (CARS) was efficient to extract the characteristic variables for beta-carotene and lutein. Determination coefficient of SVM based model was above 0.97 for pigment prediction. This study introduces a simple method for green tea classification and grading and also accomplishes quantitative determination of two pigments. Besides, the visualization of pigment distribution was tried, which embodied the image information. This methodology is potential for online detection and grading on green tea processing.