|Brabec, Daniel - Dan|
Submitted to: Applied Engineering in Agriculture
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 7/1/2003
Publication Date: 11/1/2003
Citation: Pearson, T.C., Brabec, D.L. 2003. Automated detection of internal insect infestations in whole wheat kernels using a perten skcs 4100. Applied Engineering in Agriculture. 2003. 19(6):727-733. Interpretive Summary: Insect infestations of wheat degrade the end use quality and international competitiveness of U.S. wheat. However, insect infested kernels remains as one of the most difficult defects to detect. In this study, software was developed to work in conjunction with a commercial instrument, the Perten SKCS 4100, to detect wheat kernels infested with insects. The detection accuracy of this method is competitive with other detection methods under development but has the advantage of being automated, commercially available, and only needs the software developed in this study to be implemented. This technology should aid wheat millers and handlers in detecting loads of wheat infested with insects so these loads can be handled and treated appropriately.
Technical Abstract: The wheat industry is in need of an automated, economical, and rapid means of detecting whole wheat kernels internally infested with insects. The feasibility of the Perten Single Kernel Characterization System (SKCS) to detect internal insect infestations was studied. The SKCS monitors compression force and electrical conductance as individual kernels are crushed. Samples of hard red winter wheat (HRW) and soft red winter wheat (SRW) infested with rice weevil and lesser grain borer were run through the SKCS and the conductance/force signals saved for post-run processing. Algorithms were developed to detect kernels with live internal insects, kernels with dead internal insects, and kernels from which insects have emerged. The conductance signal was used to detect live infestations and the force signal for dead and emerged infestations. Live insect detection rates were 24.5% for small-sized larvae, 62.2% for medium-sized larvae, 87.5% for large-sized larvae, and 88.4% for pupae. The predicted, and observed, false positive (sound kernels classified as infested) rate was 0.01%. Dead insect detection rates were 60.7% for large-sized larvae, 65.1% for pupae, and 72.6% for kernels where the insect emerged. The false positive rate of the dead insect detection algorithm ranged from 0.2% for SRW to 0.5% for HRW. In all cases, insect detection rates were higher for rice weevil than lesser grain borer. The classification algorithms were robust for a wide range of moisture contents.