Detection of infestations by Cryptolestes ferrugineus inside wheat kernels using a soft X-ray method

摘要

The Canada Grain Act imposes a zero tolerance for stored-product insects in grain for human consumption. The Berlese funnel method currently used to detect insect infestations in terminal elevators and grain inspection offices is slow and unreliable.The potential of a soft X-ray method to detect infestations caused by Cryptolestes ferrugineus (Stephens), the most common stored-grain insect in Canada, was determined in this study. Canada Western Red Spring wheat kernels uninfested and infested by four larval instars, pupae, and adults of C. ferrugineus were scanned using soft X-rays at 15 kV potential and 65 (MU.A current. Five hundred sound kernels and 443 kernels infested by different life stages of C. ferrugineus were used as the grain samples. Algorithms were developed to extract a total of 57 features using histogram, histogram and shape moments, and textural features using co-occurrence and run length matrix methods. The extracted features were used to identify uninfested and infested kernelsusing the statistical classifiers and a 4-layer backpropagation neural network (BPNN). More than 75.3, 86.5, and 95.7% sound, kernels infested by larvae, and pupae-adults, respectively, were correctly identified by the parametric classifier, non-parametric classifier, and BPNN using all 57 features. There were no significant differences between the identification percentages of sound kernels by the three classifiers but the parametric classifier and BPNN identified significantly higher percentages of infested kernels Identification percentages of infested kernels were higher using textural features or all 57 features than using histogram features. kw:wheat; Cryptolestes ferrugineus; X-ray image; insect infestation detection; histogram features; textural features; statistical classifier; neural networks; blé; Cryptolestes ferrugineus; image rayon X; détection d'infestation d'insectes; caractéristiques histogramme; caractéristiques de texture; classificateur statistique; réseau neuronal