Rapid classification of tef [Eragrostis tef (Zucc.) Trotter] grain varieties using digital images in combination with multivariate technique

Varieties of a single crop type may vary in several attributes affecting the choice at different spots of the food supply chain. This paper demonstrates a rapid classification of ten tef [Eragrostis tef (Zucc.) Trotter] grain va-rieties based on image processing and multivariate data analysis. Extreme Gradient Boosted Tree Discriminant Analysis (EGBDA) was applied for the variety-based classification. The developed classification model achieved a remarkable classification performance with 97% of prediction accuracy and 99% of precision. A less complex classification model using eighteen selected variables also achieved similar classification performance. The developed technique can authenticate tef varieties at the research and industrial level. Although the finding of this study is remarkable, it is essential to incorporate additional tef varieties into the model and consider other sources of variation such as agroecology as an extension of this finding.