A neural ensemble approach for segmentation and classification of road images

This paper presents a novel neural ensemble approach for classification of roadside images and compares its performance with three recently published approaches. In the proposed approach, an ensemble neural network is created by using a layered k-means clustering and fusion by majority voting. This approach is designed to improve the classification accuracy of roadside images into different objects like road, sky and signs. A set of images obtained from Transport and Main Roads Queensland is used to evaluate the proposed approach. The results obtained from experiments using proposed approach indicate that the new approach is better than the existing approaches for segmentation and classification of roadside images.