Estimating the catechin concentrations of new shoots in green tea fields using ground-based hyperspectral imagery. (15th February 2022)

• Record Type: Journal Article
• Authors: Kang, Ye Seong; Ryu, Chanseok; Suguri, Masahiko; Park, Si-bum; Kishino, Shigenobu; Onoyama, Hiroyuki
• Is Part Of: Food chemistry. Volume 370(2022)
• Publication Date: 2022-02-15
• Subjects: CF Commercial fertilizer -- EC Epicatechin -- EGC Epigallocatechin -- ECG Epicatechin gallate -- EGCG Epigallocatechin gallate -- NC Nitrogen concentration -- OF Organic fertilizer -- PGP Prism-grating-prism -- PLSR Partial least squares regression -- RE Relative error -- RMSE Root mean square error; Epicatechin (PubChem CID 72276) -- Epigallocatechin (PubChem CID 72277) -- Epicatechin gallate (PubChem CID 107905) -- Epigallocatechin gallate (PubChem CID 65064); Hyperspectral imagery -- Catechin -- Fertilizer -- Green tea -- Partial least squares regression; Food -- Analysis -- Periodicals; Food -- Composition -- Periodicals; 664
• DOI: 10.1016/j.foodchem.2021.130987
• Languages: English
• ISSNs: 0308-8146

Abstract

Highlights: This study is to non-destructively estimate catechin concentrations in green tea shoots using spectral imagery. Hyperspectral image sensor is used to extract the spectral attributes of green tea shoots. Catechin concentrations grown in commercial or organic fertilizers are estimated. The possibility of developing a year-invariant model is presented with mutual prediction. The direction for model improvement is proposed by comparing the weights of variables. Abstract: Hyperspectral imagery was applied to estimating non-galloyl (EC, EGC) and galloyl (ECG, EGCG) types of catechins in new shoots of green tea. Partial least squares regression models were developed to consider the effects of commercial fertilizer (CF) and organic fertilizer (OF). The models could explain each type of catechin with a precision of more than 0.79, with a few exceptions. When the CF model was applied to the OF hyperspectral reflectance and the OF model was applied to the CF hyperspectral reflectance for mutual prediction, the prediction accuracy was better with the OF models than CF models. The prediction models using both CF and OF data (hyperspectral reflectances, and concentrations of catechins) had a precision of more than 0.76 except for the non-galloyl-type catechins as a group and EGC alone. These results provide useful data for maintaining and improving the quality of green tea.