Submitted to: Applied Engineering in Agriculture
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 6/27/2022
Publication Date: 7/27/2022
Citation: Brabec, D.L., Pordesimo, L.O. 2022. Estimating chalkiness in endosperm of typical and bleached durum kernels from transmission scanned images. Applied Engineering in Agriculture. 38(4):651-658. https://doi.org/10.13031/aea.15023.
Interpretive Summary: The vitreousness of durum wheat (Triticum durum Desf.) is regarded by the wheat industry worldwide as an important quality factor for this grain. Vitreous kernels are glassy or translucent in appearance, while non-vitreous kernels contain floury or chalky regions. USDA Federal Grain Inspection Service (FGIS) has three sub-class of durum samples: hard-vitreous amber durum (HVAC) with 75% or more vitreous kernels, amber-durum with 60%-75% vitreous kernels, and durum with less than 60% vitreous kernels. The vitreous kernels are harder and yield more semolina and have more desirable pasta color and cooking quality and are preferred by both domestic and international markets. Trained technicians using manual methods and visual inspection evaluate the sub-class of a grain sample at inland and export grain facilities. This is a labor intensity process and can have some inconsistencies due to variation in bran appearance that can make it the outer coat appear cloudy. Image analysis or machine vision techniques have the potential to identify vitreous and non-vitreous kernels. Using a scanner and back illuminating wheat kernels, this technique reveals the internal structure of the kernel. The resulting image analysis has the potential to assess vitreousness or chalkiness. Using transmitted lighting from a flatbed scanner a set of durum market samples were evaluated for varied levels of chalkiness. Levels of chalkiness allow the seeds to be sorted into low, medium, and high levels of chalkiness. The imaging method worked well for identifying and sorting for the high and medium levels chalkiness and for identify the totally vitreous kernel. However, the imaging and analysis has some inherent lower limits, because of occasional lighting issues within kernel features such as the germ or crease. Further research is needed to improve the accuracy of classifying kernels having the lowest levels of chalkiness. Still, the use of flatbed scanners shows good potential for a more objective and consistent method of evaluating durum samples at multiple locations in the market chain.
Technical Abstract: The vitreousness of durum samples is regarded by the worldwide wheat industry as an important quality factor for durum shipments. One issue with grading durum occurs on occasional years of unfavorable harvest conditions which result in the kernel outer bran becoming cloudy. Imaging scanners can use either reflective or transmitted lighting. Back illuminating wheat kernels reveals some internal composition features of the kernels. All scanned images of our durum samples were collected using transmissive lighting. Although, “vitreousness” is the usual term applied to durum samples, for this study, “chalkiness” or the inverse property was featured. Both typical and bleached durum kernels were imaged and analyzed. Bleaching with 6% hypochlorite and 22 grams KOH was done to some samples to remove the outer seed coat to help expose any chalkiness in kernels. With both the reference samples and durum seeds presort at several levels of chalkiness, the typical kernels had higher magnitudes of %chalk and bleaching kernels did not improve the computer image analysis of the scanned seed images. Seed orientation is a source of measurement variability. Seeds were manually and carefully oriented into two distinctly different positions; Crease-Down and Crease-Side. 20% of the seeds test at the two orientations produced large differences in chalkiness, over 10%. The Crease-Side position tended to produce more seeds with higher chalkiness values. Thirdly, the computer analysis developed by USDA-ARS researchers was compared to computer analysis by a commercial seed analysis program using the same durum seed images, since the USDA-ARS computer processing was slower. The two methods correlated with an R-square of 0.95. Several durum market samples were provided by Federal Grain Inspection Service. The market samples contained levels of non-vitreous seeds for each of 3 sub-classes of durum; 80%, 65%, 50% vitreous seed. The samples were imaged and analyzed, and chalkiness distributions were plots showing the variation of individual seed chalkiness. These plots characterize the %chalk for the 3 classes of wheat. The 80% vitreous sample contained the highest number of seeds with low chalkiness while the 50% sample had the highest number of seeds with high chalkiness. Imaging and analysis has some inherent lower limits, because of computer errors from kernel features such as the germ or crease. The imaging methods worked satisfactorily for the high and medium levels %chalk. But, the very-low level of chalkiness proved to be more challenging and was not consistent.