Estimation of the degree of red coloration in flesh of a red-fleshed apple cultivar ‘Kurenai no Yume’ with a UV–vis-NIR interactance device

Abstract

This study presents a new rapid and non-destructive technique for detecting the degree of red coloration in the flesh of a red-fleshed apple cultivar ‘Kurenai no Yume’. Using a UV–vis-NIR interactance device, interactance data ranging from 190 to 1800nm was collected non-destructively from the surface of 180 apple fruits. Levels of anthocyanins in the flesh of each sample were determined with a spectrophotometer. Partial least squares (PLS) regression, multiple linear regression (MLR) and artificial neural networks (ANN) modeling analysis were performed to develop models for estimating anthocyanin content from interactance data. Results showed that all the developed models achieved reasonable predictive accuracy. Particularly, the ANN model based on 11 key wavelengths, which achieved consistent performance in predictions with calibration (R2=0.5798) and test (R2=0.5324) datasets, were the most promising from the perspective of practical applications. This study strongly suggests that the interactance technique combined with the developed models might provide a new rapid and non-destructive approach for assessing the quality of red-fleshed apples.