ABSTRACT
In this research paper, we have classified the eight different classes of rice grains using six types of color combinations. It is mentioned that the performance of self-learning neural network model (SLNNM) as a classifier is the most consistent and accurate among others; hence, it is best suited for rice classification. (Such as LDA and BPNN) also achieve satisfactory performance. Feature sets with Matrix-like structure give, compared to other engineered expression features, promising classification accuracy with less number of required set of expressions which makes them very effective. To summarize, the result suggests that those color feature with contrast-related information should be preferred to HSI color feature. This highlights the necessity of meticulous selection of features and classifier for better performance. In future studies, rice grain classification or similar tasks can obtain higher accuracy and efficiency in the aspect of research reformulation and methodology adjustment by drawing lessons from these conclusions.
