Rice Classification and Quality Analysis using Deep Neural Network

Abstract

Rice is one of the most extensively cultivated grain cereals in the world and comes in a vast range of genetic variants. It is expensive and time consuming. In this research, five different kinds of rice grains were used. The types were Arborio rice, Basmati rice, Ipsala rice, Jasmine rice, and Karacadag rice. The collection includes 75,000 grain samples and 17 features were extracted, namely 13 morphological as well as 4 shape features. Models for classifying procedures as well as their Aspect ratio for quality analysis efficiency were established by ResNet50 and Xception. Canny Edge Detection is used for preprocessing. Focusing on thresholds, rice quality is divided into three categories: best, good, and fine. The systems’ confusion matrix data were also used to produce summary statistics for sensitivity, specificity, F1 score, and accuracy, and the findings for the two models are shown in the table. The systems’ classifying efficiency scores are 98.90 percent for ResNet50 as well as 98.32 percent for Xception. The findings show that systems employed in this research for rice variety identification and quality assessment can be implemented successfully in this area.