Multi class Support Vector Machines classifier for machine vision application

Abstract

Classification of objects has been a significant area of concern in machine vision applications. In recent years, Support Vector Machines (SVM) is gaining popularity as an efficient data classification algorithm and is being widely used in many machine vision applications due to its good data generalization performance. The present paper describes the development of multi-class SVM classifier employing one-versus-one max-wins voting method and using Radial Basis Function (RBF) and Linear kernels. The developed classifiers have been applied for color-based classification of apple fruits into three pre-defined classes and their performance is compared with conventional K-Nearest Neighbor (KNN) and Naïve Bayes classifiers. The multi-class SVM classifier with RBF kernel has shown superior classification performance.