ARS | Publication request: DETERMINING VITREOUSNESS OF DURUM WHEAT USING TRANSMITTED AND REFLECTED IMAGES

Submitted to: Transactions of the ASAE
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: October 1, 2004
Publication Date: January 1, 2005
Citation: Wang, N., Zhang, N., Dowell, F.E., Pearson, T.C. 2005. Determining vitreousness of durum wheat using transmitted and reflected images. Transactions of the ASAE. Vol 48(1):219-222.

Interpretive Summary: Durum wheat production accounts for approximately 8% of the wheat production worldwide, and is mainly used to make semolina for macaroni, spaghetti, and other pasta products. The best durum wheat for pasta products should appear hard, glassy and translucent, and have excellent amber color, good cooking quality, and high protein content. Nonvitreous (starchy) kernels are opaque and softer, and result in decreased yield of coarse semolina. Thus, vitreousness of durum wheat has been used as one of the major quality attributes in grading. Traditionally, grain grading has been primarily done by visual inspection by trained personnel. This method is subjective and tedious. It also produces great variations in inspection results between inspectors. The objective of this research was to examine the use of digital imaging technology for determining durum vitreousness. Results showed that 100% of non-vitreous kernels and 92.6% of mottled kernels, which is one of the hardest defect categories to consistently detect visually, could be correctly classed. Results of the study also indicated that using transmitted illumination may greatly reduce the hardware and software requirements for the inspection system while providing faster and more accurate results for inspection of vitreousness of durum wheat.

Technical Abstract: Digital imaging technology has found many applications in grain industry. In this study, images of durum wheat kernels acquired under three illumination conditions ' reflected, side-transmitted, and transmitted ' were used to develop artificial neural network (ANN) models to classify durum wheat kernels by their vitreousness. The results showed that the models trained using transmitted images provided the best classification for the nonvitreousness class ' 100% for non-vitreous kernels and 92.6% for mottled kernels. Results of the study also indicated that using transmitted illumination may greatly reduce the hardware and software requirements for the inspection system, while providing faster and more accurate results, for inspection of vireousness of durum wheat.