|Delwiche, Stephen - Steve|
Submitted to: Cereal Chemistry
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 11/5/1999
Publication Date: 2/1/2000
Citation: N/A Interpretive Summary: Protein content of a wheat lot has a direct bearing on the suitability of the wheat as an ingredient for a specific product (bread, noodles, crackers, etc.). Because of this, value of the wheat lot is influenced by protein content. Currently, on a worldwide basis, protein content of wheat and other small grains is determined by near-infrared (NIR) reflectance on lot subsamples. At the request of U.S. Congress several years ago, USDA scientists (Manhattan, KS) developed an instrument (dubbed SKCS, for Single Kernel Characterization System) that destructively examines small samples of wheat for hardness, size, and moisture content on an individual kernel basis. The intention of single kernel analysis is to provide information on the distribution of the physical and chemical properties that affect class and grade. Recent research has been oriented toward the development of an NIR-based device that can measure the protein content of individual kernels. This device will be placed on the front end of the SKCS. A drawback in implementation of single kernel protein analysis is the time and expense of development of a calibration equation. The current study demonstrates that if single kernel protein content is not needed, an NIR model for protein content of a traditional wheat sample (a conglomeration of thousands of kernels) can still be based on the spectra of individual kernels, but without the need for a reference analysis on every kernel. We show that an NIR protein model that spectrally averages single kernel spectra from the same number of kernels (300) used in a SKCS hardness determination will have an accuracy that is approximately equivalent to that from a bulk sample NIR instrument.
Technical Abstract: Protein content of wheat by near-infrared (NIR) reflectance of bulk samples is routinely practiced. New instrumentation that permits NIR analysis of individual kernels is presently under development, with the potential for rapid NIR-based determinations of color, disease, and protein content, all on a single kernel (sk) basis. In the event that the protein content of the bulk sample is needed rather than that of the individual kernels, the present study examines the feasibility of estimating bulk sample protein from sk spectral readings. On the basis of a diverse set of 318 wheat samples of 10 kernels per sample, encompassing five U.S. wheat classes, the study demonstrates that with as few as 300 kernels bulk sample protein content may be estimated by sk NIR reflectance spectra at an accuracy (standard error < 0.25%) equivalent to conventional bulk kernel NIR instrumentation.