Abstract:
Wheat is one of the most important food sources in the world. Classification of wheat grains is essential for controlling the quality of wheat for flour industry, wheat traders and wheat researchers. The main focus of this study is on the classification bread and durum wheat grain varieties grown in similar and different agro meteorological conditions. Images of twelve wheat varieties (six wheat varieties from AARC and the remaining six wheat varieties from Haramaya University, College of Agriculture and Environmental Science) of durum wheat and bread wheat classes were captured using camera and stored for processing. Important morphological and color features were extracted for classification. Artificial neural network was used to classify the images into two categories. The classification results were organized using morphological features, color features and combinations of color and morphological features. Six morphological features and twelve color features were extracted from each wheat grain samples. The same classification organizations were used for seed samples grown in the similar or different agro meteorological conditions. Of the total 600 images, 15% of total images (90 images) were used for testing, 70% (420 images) for training and 15% (90 images) for validation. The overall accuracy of classification using only morphological feature was 96.0%, color features were 92.5% and combination of morphological and color feature was 100% for seeds grown under the same agro metrology. The accuracy of classification using morphology was 87.5%; color feature was 85.7% and combinations of color and morphology 90.5% grown under different environmental condition. This seems like the agro meteorological variability has affected either morphological or color features of wheat varieties
