Skip to main content
ARS Home » Research » Publications at this Location » Publication #134696
Title: DETERMINING WHEAT VITREOUSNESS USING IMAGE PROCESSING AND A NEURAL NETWORK

Author
item WANG, NING - KSU, MANHATTAN, KS
item Dowell, Floyd
item ZHANG, NAIQIAN - KSU, MANHATTAN, KS

Submitted to: Transactions of the ASAE
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 4/1/2003
Publication Date: 7/1/2003
Citation: Wang, N., Dowell, F.E., Zhang, N. 2003. Determining wheat vitreousness using image processing and a neural network. Transactions of the ASAE. 2003. 46(4):1143-1150

Interpretive Summary: Durum wheat is used by semolina millers and producers for pasta products and couscous. Vitreousness of durum wheat is a measure of its quality and is related to the protein content. Currently, the vitreousness of durum wheat kernels is determined by visual inspection, which is subjective and tedious. And results are variable between inspectors. An objective grading and classification system would reduce inspector subjectivity and labor and benefit producers, grain handlers, wheat millers, and processors. The Grain Check 310 is a real-time, image-based wheat quality inspection machine that can replace tedious visual inspections for purity, color, and size characteristics of grains. It also has the potential for measuring the vitreousness of durum wheat. In this study, different neural network calibration models were developed to classify vitreous and nonvitreous kernels and evaluated using samples from GIPSA and from fields in North Dakota. Model transferability between different inspection machines was also tested. The results show that the machine tends to be more consistent than human inspectors.

Technical Abstract: The Grain Check 310 is a real-time, image-based wheat quality inspection machine that can replace tedious visual inspections for purity, color, and size characteristics of grains. It also has the potential for measuing the vitreousness of durum wheat. Different neural network calibration models were developed to classify vitreous and non vitreous kernels and evaluated using samples from GIPSA and from fields in North Dakota. Model transferability between different inspection machines was also tested.