Lane detection using histogram-based segmentation and decision trees

Abstract

A vision system for intelligent vehicles is proposed here. The system exploits the characteristics of the gray level histogram of the road to detect lane markers. Each lane marker is then analyzed using a decision tree, and finally the relations between lane markers are analyzed to create structures defining the lane boundaries. The resulting system also generates images that can be used as preprocessing stages in lane detection, lane tracking or obstacle detection algorithms. The system runs in realtime at rates of about 30 Hz.