|Brabec, Daniel - Dan|
Submitted to: Sensing and Instrumentation for Food Quality and Safety
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 3/24/2008
Publication Date: 10/30/2008
Publication URL: http://www.springerlink.com/content/2l32443135850209/fulltext.html
Citation: Pearson, T.C., Brabec, D.L., Dogan, H. 2008. Improved discrimination of soft and hard white wheat using the SKCS and imaging parameters. Sensing and Instrumentation for Food Quality and Safety. 3:89-99. Interpretive Summary: Wheat kernel hardness is a measure of the kernel texture and an important indication of baking qualities of flour produced from the wheat. While wheat can have a broad range of hardness values, there are two main categories, or classes, of wheat based on hardness: soft and hard. It is desirable to market and trade wheat of a pure hardness class as it will have more predictable end use qualities. One of the most commonly used methods for measuring wheat hardness and determining purity of hardness classes in loads of wheat is the Single Kernel Characterization System (SKCS). However, for some varieties of wheat, particularly those grown in the Pacific Northwest, the SKCS has trouble distinguishing kernels from hard and soft classes. This leads to errors in determining if a sample is pure hard wheat, pure soft wheat, or a mixture. This research focused on improving the accuracy of the SKCS for wheat grown in the Pacific Northwest by use of more modern digital signal processing of the data that the SKCS already produces and by combining images with the SKCS. It was found that integrating new signal processing techniques into the SKCS software can reduce the errors made by the SKCS in half. By adding data extracted from images of kernels, the errors can be reduce by over 70%. This technology should aid wheat inspectors to determine the proper quality of a load of wheat, especially at export terminals. This will help improve the quality and international competitiveness of wheat produced in the United States.
Technical Abstract: Natural variation in the hardness of wheat kernels often results in an overlap between hard and soft classes in the distribution of hardness indices (HI) as measured with the single kernel characterization system (SKCS) and is a major contributor to classification errors. This is particularly true for the case of the hard white and soft white wheat classes. To address this problem, a color camera was incorporated into the SKCS system so that color and kernel size data could be combined with SKCS measurements for classification purposes. Samples of hard red (HR), soft red (SR), hard white (HW), and soft white (SW) wheat were classified using the SKCS system with and without the camera and results compared. Using the camera system, errors for separating HW from SW classes were reduced to less than 5%, as compared to 17.1% using SKCS alone. Furthermore, improved data processing applied to the low-level data currently produced by the SKCS system led to greater than 50% reduction in classification errors between SW and HR as compared to using HI data alone. Similar improvements in classification accuracies for 300-kernel mixtures of SW and HW were also achieved. This should aid grain inspectors in properly identifying mixtures of these two classes. Unfortunately, for the SR and HR classes, incorporating the camera data decreased classification accuracy while increasing the complexity of the system.