Predicting Brix values of stored apples using near infrared spectra

Abstract

Sweetness sorting machines, based on near infrared (NIR) spectroscopy, have been introduced to packinghouses for peaches, apples, oranges, and other fruits for nondestructive quality evaluation. However, the NIR method for Brix values is not always calibrated well, so that a big bias occurs occasionally. In this study, the means of developing a universal calibration for Brix using non-stored apples, which is also suitable for stored samples, was explored. The calibration developed, using NIR spectra [not the multiplicative scatter correction (MSC) one], predicted the Brix values of non-stored samples well but not that of stored ones. With the 5-month stored apples, a bias of 2.66° Brix occurred. However, there was no bias when the spectra of non-stored and stored samples were used to formulate the calibration equation. When calibration and prediction using MSC-treated spectra were performed, the drawback described above could be mitigated. More accurate results were obtained using NIR spectra within 758 to 1161 nm.