Recognition of the Duration and Prediction of Insect Prevalence of Stored Rough Rice Infested by the Red Flour Beetle (Tribolium castaneum Herbst) Using an Electronic Nose.

The purpose of this research is to explore the feasibility of applying an electronic nose for the intelligent monitoring of injurious insects in a stored grain environment. In this study, we employed an electronic nose to sample rough rice that contained three degrees of red flour beetle ( Herbst) infestation for different durations-light degree (LD), middle degree (MD), and heavy degree (HD)-and manually investigated the insect situation at the same time. Manual insect situation investigation shows that, in all three rice treatments, the insect amounts gradually decreased after infestation. When the insect population of stored rough rice was under 13 insects per 60 g of rough rice, the natural speed of decrease of the insect population became very slow and reached the best artificial insect killing period. Linear discriminant analysis (LDA) provided good performance for MD and HD insect harm duration identification, but performed poorly for LD insect harm duration identification. Both k-means clustering analysis (K-means) and fuzzy c-means analysis (FCM) effectively identified the insect harm duration for stored rough rice. The results from the back-propagation artificial neural network (BPNN) insect prevalence prediction for the three degrees of rough rice infestation demonstrated that the electronic nose could effectively predict insect prevalence in stored grain (fitting coefficients were larger than 0.89). The predictive ability was best for LD, second best for MD, and least accurate for HD. This experiment demonstrates the feasibility of electronic noses for detecting both the duration and prevalence of an insect infestation in stored grain and provides a reference for the intelligent monitoring of an insect infestation in stored grains.