Prediction of Internal Flesh Browning of “Fuji” Apple Using Visible-Near Infrared Spectra Acquired by a Fruit Sorting Machine

Abstract

Visible-near infrared spectra of 576 “Fuji” apples harvested in 2015 and 2016 were acquired with an apple sorting machine. One month after the spectral acquisition, the cut surface of each samples was scanned, and the occurrence of internal fresh browning was assessed. Various preprocessing methods, including newly proposed brute force differential absorbance, were applied to spectra acquired by the top and bottom spectrometer installed in the sorting machine, and models for the prediction of the occurrence of internal browning were built by partial least squares discriminant analysis. When a “metamodel” was developed by combining models with the lowest error discrimination rate for each of the top and bottom spectrometer, it was possible to predict the occurrence of internal browning with 19.8% classification error, 88.6% sensitivity and 78.1% specificity. In this research, a sorting machine which is installed in actual apple sorting factories was used. Therefore, the results of this research can be easily applied to the apple sorting sites, and it is expected to contribute to the added value improvement of “Fuji” apples.