Classification of Wheat Rootstock and Their Hybrids According to Color Features by Machine Learning Algorithms

Abstract

Accurate classification of wheat varieties has a large economic market in the world is enabled both high income in the market and the development of new fertile hybrids for changing weather conditions due to global warming. In this study, instead of using the conventional classification method, we extracted color features of the 1400 durum wheat grain samples, consisting of Ahmetbugdayi, Cesare and their hybrids BC1F6 and BC2F5, by using image processing techniques. For the color features, every twelve channels of four different color spaces were used and square-shaped samples were taken from the center of all the grains in these channels of images. the averages of the channel pixels values were used as color features. Then six different machine learning algorithms were employed for the classification task. ANN, SVM and DT models achieved more than 0.99 accuracies. On the other hand, k-NN and RF model reached approximately 0.99 accuracies. According to our results, in addition to different wheat varieties, also sibling hybrid seeds can be classified with high accuracy according to their color characteristics by the methods we proposed.