Classification of wheat cultivars using image processing and artificial neural networks.

Classification of wheat grains has significant importance in determining the market value of wheat varieties. Wheat class identification is also necessary for plant breeders to predict yield and quality. In this study, classification of four Iranian wheat cultivars was carried out using morphological features and artificial neural networks. After preparing samples, 164 images of grains were acquired for each cultivar in a lighting chamber. Ten morphological features were extracted from images using image processing techniques. For classifying wheat varieties, various topologies of artificial neural networks (ANN) with different number of neurons in the hidden layers were developed. The nine important morphological features extracted from images were used as input for developed ANN. 60% of all samples were used for training networks. Validation of the developed ANN structures was done by 15% of samples while 25% of samples were used for evaluation of the networks. The best topology for ANN was 9-26-4. Results showed that overall classification accuracy of 85.72% was obtained for classification of wheat cultivars.