Clustering based neural network approach for classification of road images

Abstract

This paper presents a new approach of combining clustering and neural network classifier for the classification of road images into road and sky segments. The proposed approach first creates clusters for each available class and then uses these clusters to form subclasses for each extracted road image segment. The integration of clusters in the classification process is designed to increase the learning abilities and improve the accuracy of the classification system. The experiments using clustering based neural network classifier have been conducted on the set of images obtained from Transport and Main Roads Queensland. The results have been analysed and presented in this paper.