Feasibility of NIR spectroscopy to detect olive fruit infested by Bactrocera oleae

Abstract

Olive fruit fly infestation is a significant problem for the milling process. In most cases, damage from insects is 'hidden', i.e. not visually detectable on the fruit surface. Consequently, traditional visual sorting techniques are generally inadequate for the detection and removal of olives with insect damage. In this study, the feasibility of using NIR spectroscopy to detect hidden insect damage is demonstrated. Using a genetic algorithm for feature selection (from 2 to 6 wavelengths) in combination with linear discriminant analysis (LDA), quadratic discriminant analysis (QDA) or k-nearest-neighbors (kNN) routines, classification error rates as low as 0.00% false negative, 12.50% false positive, and 6.25% total error were achieved, with an AUC value of 0.9766 and a Wilk's λ of 0.3686 (P<. 0.001). Multiplicative scatter correction, Savitzky-Golay spectral pre-treatment with 13 smoothing points and mean centering spectral pre-treatments were used. The optimal features corresponded to Abs[1108. nm], Abs[1232. nm], Abs[1416. nm], Abs[1486. nm] and Abs[2148. nm]. © 2014 Elsevier B.V.