Road detection based on bread-first search in urban traffic scenes

This paper presents a new approach of road detection from a moving vehicle. The approach is based on an extension of breadth-first search (BFS) algorithm which provides a testing for bipartite graph. Based on the color difference image of one captured image, the algorithm works first with a road-color-related seed and explores all the neighboring nodes until it finds the possible road boundaries. Then, for the next image, a visual odometry is used to estimate the vehicle pose for further improving the performance of road area discovery. Several experimental results have shown that the proposed approach can effectively extract the road region in real road scenes.