Image Cluster Features Shape and Texture Determinants of Rice Quality Using the K Means Algorithm

Abstract

There is a lot of fraud case in the forgery of ricequality by mixing good quality rice with low quality rice for increasing price. To protect the community from counterfeiting, we conduct research to detect the quality of rice which can later help the community to be able to distinguish good and bad quality. This paper presents a low-cost image processing system for assessing the quality of rice. Many factors affect the quality of rice such as grain fragments, non-uniform color, odor and other factors. This study uses procentage of broken rice grains and color uniformity to determine the quality of rice. We propose texture feature with Otsu segementation for determining the number of broken grains and color distribution for specifying the color uniform. The classification results using K Fold validation on the original data show the results of  K-Nearest Neighbor have 99.70% accuracy.