Ability of hyperspectral imaging to assess physicochemical and phytochemical quality parameters of raisins

Abstract

The possibility to apply the hyperspectral imaging (HSI) technique for the evaluation of some physicochemical and phytochemical quality parameters of raisins was examined. Italia grapes from conventional and organic production and dried using different conditions (temperatures and pretreatments) were studied. Neural network method was used to test the data set and good raisin classification were obtained. The selection of the most relevant wavelengths was achieved using the Lasso and the Genetic Algorithm-Partial Least Squares (GAPLS) methods. The selections of wavelengths made with the Lasso method were interesting only for a few quality parameters, while those done by the GAPLS method were powerful, generating consistent models (generally R² > 0.80). This latter method resulted in better models for color indices (0.94 < R² < 0.99) and for phenolics (0.78 < R² < 0.97) and particularly for the flavonols (quercetine-3-O-glucoside and rutin). The common wavelengths for physicochemical features were between about 380 and 1015 nm. The main bands characteristic of color ranged from 440 to 730 nm. Phenolic compounds presented bands between 660 and 1000 nm, whereas texture parameters were between 390 and 997 nm. This study suggests that HSI combined with chemometrics may be a non-destructive tool able to rapidly assess quality parameters of dried grapes. Consequently, HSI could be used for monitoring different process like drying, assessing the composition of raisins at any stage from production to sales, classifying raisins to get different quality clusters for commercial purpose, authenticating the variety or the type of production.