Comparison Performance of Visible-NIR and Near-Infrared Hyperspectral Imaging for Prediction of Nutritional Quality of Goji Berry (<i>Lycium barbarum</i> L.)

Abstract

The potential of hyperspectral imaging for the prediction of the internal composition of goji berries was investigated. The prediction performances of models obtained in the Visible-Near Infrared (VIS-NIR) (400–1000 nm) and in the Near Infrared (NIR) (900–1700 nm) regions were compared. Analyzed constituents included Vitamin C, total antioxidant, phenols, anthocyanin, soluble solids content (SSC), and total acidity (TA). For vitamin C and AA, partial least square regression (PLSR) combined with different data pretreatments and wavelength selection resulted in a satisfactory prediction in the NIR region obtaining the R2pred value of 0.91. As for phenols, SSC, and TA, a better performance was obtained in the VIS-NIR region yielding the R2pred values of 0.62, 0.94, and 0.84, respectively. However, the prediction of total antioxidant and anthocyanin content did not give satisfactory results. Conclusively, hyperspectral imaging can be a useful tool for the prediction of the main constituents of the goji berry (Lycium barbarum L.).

Keywords

• prediction
• vitamin C
• phenols
• soluble solids
• acidity