A machine learning approach for grain crop's seed classification in purifying separation

Abstract

The paper presents a study of the machine learning ability to classify seeds of a grain crop in order to improve purification processing. The main seed features that are hard to separate with mechanical methods are resolved with the use of a machine learning approach. A special training image set was retrieved in order to check if the stated approach is reasonable to use. A set of tests is provided to show the effectiveness of the machine learning for the stated task. The ability to improve the approach with deep learning in further research is described.