Factor Analysis of Nondestructive Measurements of Onion Bulb Internal Rot

Abstract

Some rotten onion bulbs cannot be detected by visual inspection or palpation. Contamination of rotten bulbs into the shipping products diminishes the reliability of production area. In this study, the feasibility of using visible and near-infrared spectroscopy for the nondestructive detection of rotten bulbs on a belt conveyance system in a typical onion packing house was investigated. A partial least squares regression model for bulb rot was developed from the acquired spectra. Datasets including different sample sizes and temperatures with a ratio of sound to rotten bulbs of 3 : 1 supported the development of a robust discriminant model. The model achieved correct onion classification of 95.0% when the bulb rot presence-absence threshold was set to level 1.