Submitted to: American Association of Cereal Chemists Meetings
Publication Type: Abstract Only
Publication Acceptance Date: 4/12/2007
Publication Date: 10/10/2007
Citation: Delwiche, S.R. 2007. Enhancements in color-based detection of mold-damaged wheat kernels [abstract]. American Association of Cereal Chemists Meetings. Available: http://www.aaccnet.org/meetings/2007/abstract/O07ma22.htm.
Technical Abstract: One of the most common molds that infect the seeds of small cereals, such as wheat, throughout the world is Fusarium Head Blight (FHB). A metabolite, known as deoxynivalenol (DON or vomitoxin), often occurs with the FHB and is cause for concern because of a moderate toxicity to humans and non-ruminant livestock. Government regulatory bodies of most countries impose limits on the allowable levels of DON in raw and finished food/feed products, which adds burden onto grain exporting countries, such as the United States, that must ensure that grain sent overseas is safe and meets the recipient country’s specifications. Over the years, the Beltsville USDA laboratory has explored the use of visible and near-infrared spectroscopy as a basis for identification and removal of kernels infected with FHB. Whereas laboratory tests have demonstrated a >95% rate at detecting mold-damaged kernels, this accuracy drops off precipitously as kernels are processed at speeds approaching those of commercial optical sorters. Current work is underway at Beltsville that utilizes alternate methods of lighting at high frequencies (> 1 kHz) that allow for multiple measurements of reflected broadband light such that, when processed with measurements from other bands, category assignment (normal vs. mold-damaged) improves. For example, multispectral information is collected and processed on single kernels in freefall at the sub-millisecond level. Aspects of the selection of wavelength bands, lighting, timing, and decision algorithms are addressed.