The Evaluation of the Nutritional and Sensory Properties of Apples Using Non-destructive Methods and Sensory Evaluation

Abstract

Non-destructive methods have many advantages over conventional chemical analysis and sensory evaluation. In this research, two non-destructive methods (machine vision and hyperspectral imaging) were applied to investigate the nutritional (pigment contents) and sensory properties (colour) of four popular apple cultivars in New Zealand (Braeburn, Granny Smith, Envy and Royal Gala). These methods were compared with conventional chemical analyses of pigment content, as well as sensory evaluation of colour, to determine their feasibility in measuring these nutritional and sensory properties. Conventional pigment analysis was carried out using UV-Vis Spectrophotometry, while sensory evaluation was conducted by an experienced trained panel. The chemical analysis results show that all four apple cultivars had differing pigment profiles. The peel of Granny Smith contained high levels of chlorophylls, which have high antioxidant and anticancer properties. The peels of the other red cultivars were also good sources of phytonutrients, as their anthocyanin contents were high. Sensory evaluation also showed distinct sensory attribute profiles for the apples. The red cultivars had radar plots (Figure 25) that are different to Granny Smith’s. Machine vision analysis using the LensEye program was able to give mean L*a*b* values, which provided indications of the apple hues. Information regarding colour intensity and colour homogeneity were also derived from the Colour Contours and Colour Primitive functions. Though significant information was obtained, it was found that the red cultivars had similar L*a*b* profiles. Furthermore, huge biological variations exist within cultivars, therefore correlations with chemical and sensory results were not excellent. Better correlations may be found if apple cultivars that have more significantly different colour profiles were compared. Much information could also be derived from hyperspectral imaging, though for the analyses in this study, the Partial Least Square Regression models were not suitable in prediction of pigment contents. This was likely due to the non-homogeneous nature of the samples. Improvements in methodology can yield better results.