X-ray image analysis to detect infestation due to Cryptolestes ferrugineus in stored wheat

The Canada Grain Act imposes a zero tolerance for stored-product insects in grain. Incoming and export grain in the primary and terminal elevators, respectively is inspected for the presence of insects using Berlese funnels. This method takes 5 to 6 h to extract the larval stages of the rusty grain beetle, Cryptolestes ferrugineus (Stephens), the most common stored-grain insect in Canada. During this time the grain may have been binned in the elevator or loaded on to ships. This results in manifestation of infestation and cross contamination of stored-grain in the grain handling system. The feasibility of using real-time soft X-ray images to detect insect infestations in wheat was determined in this study. Uninfested and infested Canadian Western Red Spring wheat kernels fed on by different life stages of C. ferrugineus were X-rayed at 15 kV potential and 65 /spl mu/A current. Five hundred uninfested and 440 infested kernels were X-rayed at different times for the four larval instars, pupae, and adult stages of the insect infesting wheat kernels. Histogram groups, histogram and shape moments, and textural features using co-occurrence and run length matrices were extracted from the X-ray images. The 57 extracted features were used to identify uninfested and infested kernels by the non-parametric classifier and multi-layer feed-forward backpropagation neural network (BPNN). The non-parametric classifier correctly identified 83.3% of the sound kernels. The BPNN identified 75.7% of sound kernels and classed 24.3% as infested. More than 87% of wheat kernels infested by larvae were identified as infested by the nonparametric classifier and BPNN. More than 96% of kernels infested by the pupal and adult stages of C. ferrugineus were correctly classified by the nonparametric classifier and BPNN methods.