Location: Stored Product Insect Research Unit
Title: Imaging and Automated Detection of Sitophilus Oryzae L. (Coleoptera: Curculionidae) Pupae in Hard Red Winter Wheat Authors
Submitted to: Journal of Economic Entomology
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: October 24, 2005
Publication Date: March 1, 2006
Citation: Toews, M.D., Pearson, T.C., Campbell, J.F. 2006. Imaging and automated detection of Sitophilus oryzae L. (Coleoptera: Curculionidae) pupae in hard red winter wheat. Journal of Economic Entomology 99: 583-592. Interpretive Summary: Detection of insects that live and develop inside kernels of grain is difficult because there are no signs of infestation on the outside of the grain kernels, however, these infested kernels can lead to insect fragments in milled products such as wheat flour. A novel method of insect detection utilizing computed tomography and a custom software program were developed to rapidly recognize and count the insect-infested kernels. The detection accuracy using the new procedures was 95% for samples containing an average of 5 infested kernels per 100 g of wheat. These results are the first step toward developing next generation computerized equipment to rapidly detect insects in many types of cereal grains and beans.
Technical Abstract: Computed tomography, an imaging technique commonly used for diagnosing internal human health ailments, utilizes multiple x-rays and sophisticated software to re-create a cross sectional representation of a subject. The use of this technique to image hard red winter wheat samples infested with pupae of Sitophilus oryzae was investigated. A software program was developed to rapidly recognize and quantify the infested kernels. Samples were imaged in a 7.6 cm (OD) plastic tube containing 0, 50, or 100 infested kernels in one kg of wheat. Inter-kernel spaces were filled with corn oil to increase the contrast between voids inside kernels and voids among kernels. Automated image processing, utilizing a custom C language software program, was conducted separately on each 100 g portion of the prepared samples. The average detection accuracy in the 5 infested kernels per 100 g samples was 94.4 ± 7.3% (mean ± SD, n = 10) while the average detection accuracy in the 10 infested kernels per 100 g sample was 87.3 ± 7.9%. Detection accuracy in the 10 infested kernels per 100 g samples was slightly less than the 5 infested kernels per 100 g samples due to some infested kernels overlapping with each other or air bubbles in the oil. A mean of 1.2 ± 0.9 (n = 10) bubbles (per tube) were incorrectly classed as infested kernels in replicates containing no infested kernels. In light of these positive results, future studies should be conducted utilizing additional grains, insect species, and life stages.