X-ray detection and sorting of olives damaged by fruit fly

Abstract

An algorithm using a Bayesian classifier was developed to automatically detect olive fruitfly infestations in x-ray images of olives. The data set consisted of 249 olives with various degrees ofinfestation and 161 non-infested olives. Each olive was x-rayed on film and digital images wereacquired with a film scanner at a resolution of 59 pixels per cm. Features extracted from the imageswere submitted to the classification algorithm and error rates for detection of the infestationsobtained. Feature selection involved pixel intensity values and pixel derivative values at each pixellocation in the image. The ability of the algorithm to differentiate infested and non-infested olives wastested. Internal damage to the olive was a factor in detection, with slight damage correctly identified50% of the time and severe damage correctly identified 86% of the time. Non-infested olives werecorrectly identified with 90% accuracy.