Hierarchical segment learning method for road objects extraction and classification

In this paper, we propose a new hierarchical segment learning approach for extraction and classification of roadside objects. The proposed approach is based on hierarchical segment extraction and classification of segmented objects using a neural network. In this approach, we extract different road objects such as sky, road, sign and vegetation in hierarchical stages and classify them using a neural classifier. The approach improves the overall classification accuracy while extracting different road objects from the road images. The proposed approach has been applied to a set of images extracted from video data collected by Transport and Main Roads Queensland. The experimental results indicate that this approach can extract and classify road objects with a reasonable high accuracy.