Classification of cereal grains using wavelet, morphological, colour, and textural features of non-touching kernel images

Abstract

Images of non-touching kernels of Canada Western Red Spring (CWRS) wheat, Canada Western Amber Durum (CWAD) wheat, barley, oats, and rye were acquired using an area scan camera. Morphological, colour, textural, and wavelet features were extracted from colour images of cereal grains for classification. A total of 51 morphological features, 93 colour features, 56 textural features, and 135 wavelet features were extracted from each kernel. Linear and quadratic statistical classifiers were used for classification using individual types of features and their combinations to find the best feature set and classification method for improved classification of cereal grains. Combining all morphological, colour, textural and wavelet features gave the best classification using the linear discriminant classifier with a classification accuracy of 99.4% for CWRS wheat, followed by 99.3%, 98.6%, 98.5%, and 89.4% for rye, barley, oats, and CWAD wheat, respectively.