A Wavelet Transform Based Classification of Bulk Rice Grain

Abstract

This paper deals with the classification of four varieties of bulk rice grain using wavelet decomposition techniques. Features for classification were extracted from a 2 level wavelet decomposed image. Three simple statistical features namely; mean, variance and range were extracted from each of the components (approximation, horizontal, vertical and diagonal) of the wavelet decomposed image. The classification was carried out based on these three features using a linear discriminant classifier. The classification accuracies based on wavelet features for different image planes were compared. It is found that the classification accuracies based on blue channel provides better average classification accuracies of 98.5% as compared to other image planes. It is also found that linear discriminant classifier is able to yield better results as compare to Naïve Bayes classifier and K-Nearest Neighbour classifier for bulk rice classification.