A new approach for rice quality analysis and classification using thresholding-based classifier

Abstract

Quality detection and analysis of grains like wheat, rice, etc., is an important process normally followed using manual perceptions. The decision depends and varies from person to person and even involves human error with the same person analyzing at different times. This paper presents a novel approach to provide, simple and efficient classification of rice grains using thresholding method. This approach would be further used to automate the process and maintain the uniformity in decisions. In this research, India Gate basmati rice varieties such as— brown basmati rice, classic, super, tibar, dubar, rozana, and mogra were used for the analysis. The process involved identification of region of interest (ROI) and extraction of six morphological features—area, major axis length, minor axis length, aspect ratio, perimeter, and eccentricity. Further three color features—max hue, max saturation, and max value features from each rice grain of sample image were extracted. Through exhaustive experimentation and analysis the rice qualities were defined into seven classes such as; best, good, fine, 3/4 broken, 5/8 broken, 1/2 broken, and 1/4 broken based on threshold values. Through varying logical conditions and relationships among these quality features, aforesaid rice brands were classified. Three of such varying techniques were used out of which one method is being presented here. The results of the presented approach were 100 % accurate for some classes while 95 % for others.