Submitted to: Applied Engineering in Agriculture
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: September 1, 2006
Publication Date: November 1, 2006
Citation: Pearson, T.C., Brabec, D.L. 2006. Camera attachment for automatic measurement of single-wheat kernel size on a perten skcs 4100. 2006. Applied Engineering in Agriculture. Volume 22(6):927-933. Interpretive Summary: Wheat kernel size and shape is an important quality factor and a characteristic for adjusting milling processes. Measuring kernel size is tedious and time consuming so it cannot be done as often as some wheat millers would like. Automated machines for measuring kernel size suffer from inaccuracies and/or high cost. The Perten Single Kernel Characterization System (SKCS 4100) is an automated instrument which measures several single kernel quality characteristics such as weight, moisture content, hardness, and diameter. Of all of these measurements, the diameter measurement is the least accurate. A low cost color camera was attached to an SKCS 4100 to enable more accurate kernel size determinations. Using image data combined with SKCS data, errors in estimating kernel length and diameter were reduced by 56% and 66%, respectively.
Technical Abstract: A simple camera was attached to a Perten Single-Kernel Characterization System (SKCS) 4100 to measure single kernel morphology as they are fed through the SKCS. The camera and lighting were positioned above the SKCS weigher bucket. Each image of a wheat kernel was captured and processed in 15 to 60 ms. Image measurements included kernel area, length, and width. Samples of durum (DU), hard red winter (HRW), soft red winter (SRW), hard white (HW), and soft white (SW) wheat were used. It was found that the best estimates for kernel length, width, and thickness measurements were obtained by combining image measurements with SKCS measurements. The mean error for kernel-diameter estimates was reduced by 56% from estimates using the SKCS-diameter measurement. Additionally, image measurements reduced mean errors for estimating kernel length by 66% over estimates using SKCS parameters alone.