Detection of anthocyanin content in fresh Zijuan tea leaves based on hyperspectral imaging

Abstract

Anthocyanins are characteristic substances that determine the leaf color and sensory quality of Zijuan tea. The qualitative determination of anthocyanin content in fresh Zijuan tea leaves by artificial vision may lead to the uneven quality of tea products. Hyperspectral images at 400–956 nm were obtained. K-nearest neighbor (KNN) and support vector machine (SVM) models were used to identify the tenderness grade, and partial least squares regression (PLSR) and support vector regression (SVR) models were established based on principal component analysis for prediction. In short, the classification accuracy of the SVM model was better than 90%; variable combination population analysis (VCPA) and variables combination population analysis combined with iterative retained information variable (VCPA-IRIV) can effectively simplify the model and the prediction accuracy is over than 0.92. The best predictive models of total anthocyanins, Cya-3,5-O-diglucoside, Cya-3-O-glucoside, and petunidin were VCPA + SVR, VCPA-IRIV + SVR, VCPA-IRIV + PLSR, and VCPA + PLSR, and the relative percent deviation were 3.233, 2.868, 3.529 and 3.298, respectively. The total anthocyanins were visualized by the spatial distribution of anthocyanins in samples with different tenderness. This study provided a rapid nondestructive method of fresh Zijuan tea leaf tenderness grade and quality components to control the quality of fresh leaf picking of tea with specific leaf color.