|Delwiche, Stephen - Steve|
Submitted to: Applied Engineering in Agriculture
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 2/4/2005
Publication Date: 7/1/2005
Citation: Delwiche, S.R., Gaines, C.S. 2005. Wavelength selection for monochromatic and bichromatic sorting of fusarium-damaged wheat. Applied Engineering in Agriculture. 21(4):681-688. Interpretive Summary: Fusarium head blight (FHB) is a fungal disease that affects small cereal grains and is becoming increasingly problematic in North America. The occurrence and severity of the disease varies with season and geographical location, with warm humid weather during flowering and the presence of crop residue from the previous year as exacerbating factors. In wheat, FHB causes the seed to be shriveled, light in weight, and, in many circumstances, infused with the mycotoxin called deoxynivalenol, or DON, which is considered a human health hazard and has been shown to cause disease in non-ruminant animals, such as swine. The U.S. Food and Drug Administration specifies that finished wheat products do not possess DON concentrations in excess of 1 part per million. The European Union is considering even more stringent tolerance levels, thus U.S. grain processors and exporters are under increasing pressure in domestic and overseas markets to reduce contaminant levels at a time when the problem appears to becoming worse. Mills can reduce mycotoxin concentrations through conventional milling practices for light infections. At heavier infection rates, mills traditionally resort to using the stock as animal feed (higher tolerance levels, but lower economic value) or blending with non-contaminated wheat, an option that could disappear if the regulatory agencies outlaw this practice. Therefore, a study was undertaken to examine a third option, this being the possibility of removal of scab-damaged kernels by optical sorting. Four thousand eight hundred kernels from 100 commercial varieties, equally divided between normal and scab-damaged categories, were individually scanned in the wavelength regions that range from that being sensitive to the human eye (400 to 750 nm) to that which is often used for constituent analysis (1000 to 1700 nm). While laboratory-grade spectrometers were used to collect data, thus making it possible to develop mathematically complex classification functions (a subject of our previous studies), we restricted the functions to contain no more than one or two wavelengths, so as to mimic the design of monochromatic and bichromatic commercially available high-speed sorters. Best classification accuracies for two-wavelength models occurred at wavelengths close to 500 nm and 550 nm for the visible region (at 94% accuracy) and 1150 nm and 1250 nm for the longer wavelength, near-infrared region (97% accuracy). Other models, such as those that used a wavelength from each region (750 nm and 1480 nm, for 86% accuracy), also demonstrated reasonably high accuracies. This study has demonstrated the feasibility of optical sorting for wheat scab and has established wavelength filter criteria for high-speed sorting (the subject of present research). Grain terminals and mills are the intended beneficiaries of this work.
Technical Abstract: Fusarium head blight (FHB) is a fungal disease that affects small cereal grains such as wheat, causing the seed to be shriveled, light in weight, and, in many circumstances, infused with the mycotoxin, deoxynivalenol (DON), an FDA-recognized human health hazard. The occurrence and severity of the disease varies with season and geographical location, with warm humid weather during flowering and the presence of crop residue from the previous year as exacerbating factors. A study was undertaken to determine the most suitable visible or near-infrared wavelengths that could be used in high-speed sorting for removal of FHB-infected soft red winter wheat kernels. Four thousand eight hundred kernels from 100 commercial varieties, equally divided between normal and scab-damaged categories, were individually scanned in the extended visible (410-865 nm) and near-infrared (1031-1674 nm) regions. Single- and all combinations of two-wavelength linear discriminant analysis models were developed and characterized through cross-validation by the average correctness of classification percentages. Short visible (~420 nm) and moderate near-infrared (1450-1500 nm) wavelengths produced the highest single-term classification accuracies (at approximately 77% and 83%, respectively). The best two-term models occurred near the wavelengths of 500 and 550 nm for the visible region alone (94% accuracy), 1152 and 1248 nm for the near-infrared region alone (97%), and 750 and 1476 nm for the hybrid region (86%). These wavelengths are, therefore, considered of importance in the design of monochromatic and bichromatic high-speed sorters for scab-damage reduction.