TOWARD BICHROMATIC OPTICAL SORTING OF SCAB-DAMAGED WHEAT

Authors: Delwiche, Stephen - Steve; Pearson, Thomas; Brabec, Daniel - Dan

Submitted to: National Fusarium Head Blight Forum
Publication Type: Abstract Only
Publication Acceptance Date: 12/11/2005
Publication Date: 12/20/2005
Citation: Delwiche, S.R., Pearson, T.C., Brabec, D.L. 2005. Toward bichromatic optical sorting of scab-damaged wheat. In: Proceedings of the 2005 National Fusarium Had Blight Forum, December 11-13, 2005, Milwaukee, WI. East Lansing, Michigan State University. p. 176.

Interpretive Summary:

Technical Abstract: Our previous optical work exhaustively examined the best single wavelength and best two-wavelength combination that maximized class separation between normal and scab-damaged wheat kernels, using linear discriminant analysis. This was accomplished by single kernel visible (410-865 nm) and near-infrared (1031-1674 nm) scanning of more than 4500 kernels from 100 commercial varieties, equally divided between normal and scab-damaged categories. Generally, the best two-wavelength models were approximately 95% accurate in classification. From there, we have studied the application of a good-performing wavelength pair (675 nm and 1480 nm, as selected from a handful of manufacturers stock combinations) in a high-speed commercial sorter. Unknown until this study was the effectiveness of such classification models when applied under real-time, high-speed sorting conditions. More than 40 samples of soft red winter and soft white wheat, each approximately 5 kg, obtained from commercial mills in the eastern United States, primarily from the 2003 harvest. Of these, 35 samples were from regular process streams (selected because of a priori knowledge of elevated Fusarium damage), with the remaining 7 samples taken from discard piles of cleaners. The sorter, which used detectors and interference filters, one in the visible region and the other in the near-infrared region (see above), consisted of a series of parallel inclined channels. The level of DON ranged from 0.6 to 20 mg/kg. Samples were processed by the sorter operating at a throughput of 0.33 kg/(channel-min) (four of ten available channels used) and a kernel rejection rate of 10%. Visual measurements of the proportion of Fusarium-damaged kernels were collected on incoming and sorted specimens. Sort effectiveness was assessed by two means: percentage reduction of Fusarium-damaged kernels and percentage reduction in DON concentration. Results indicated that the fraction of DON contaminant level in the sorted wheat to that in the unsorted wheat ranged from 18 to 112 percent, with an average of 51 percent. Nine of the 35 regular samples and all 7 of the discard pile samples underwent a second sort, with 5 from this second set undergoing a third sort. Multiple sorting was effective in yielding product whose DON concentration was between 16 and 69 percent of its original, unsorted value. In summary, sorting resulted in a reduction of DON concentration by approximately one half on average, with further reduction arising from the resorting of accepted material. This study is fully described in our recent publication [Delwiche, et al., Plant Disease, vol. 89, p. 1214-1219 (2005)].