Contrastive analysis of rice grain classification techniques: multi-class support vector machine vs artificial neural network

Shafaf Ibrahim (Faculty of Computer and Mathematical Sciences, Universiti Teknologi MARA)
Saadi Bin Ahmad Kamaruddin (Faculty of Business, Economics and Accounting, HELP University)
Azlee Zabidi (Faculty of Computing (FKOM), College of Computing and Applied Sciences, Universiti Malaysia Pahang)
Nor Azura Md. Ghani (Faculty of Computer and Mathematical Sciences, Universiti Teknologi MARA)

IAES International Journal of Artificial Intelligence (IJ-AI)
Vol 9, No 4: December 2020

Rice is a staple food for 80% of the population in Southeast Asia. Thus, the quality control and classification of rice grain are crucial for more productive and sustainable production. This paper examines the contrastive analysis of rice grain classification performance between multi-class support vector machine (SVM) and artificial neural network (ANN). The analysis has been tested on three types of rice grain images which are Ponni, Basmati, and Brown rice. A digital image transformation analysis based on shape and color features was developed to classify the three types of rice grain. The performance of the proposed study is evaluated to 90 testing images of each rice variation. The ANN is observed to return higher classification accuracy at 93.34% using Level Sweep image transformation technique. Based on the results, it signifies that the ANN performs better classification than the multiclass SVM.