The application of edge detection to obtain salient lane boundaries on road image is popular in the computer vision area. However, edge detectors may be easily distracted by manifold noises that emerged on the road surface such as shadow cast, vehicles or other obstacles. Additionally, sky region can adversely affect the performance of lane detection method due to the presence of horizontal edges in the sky region. In this paper, lane preprocessing approach is proposed to effectively extract lane marks from the traffic scene. Horizon localization is improved to automatically segment the sky and road region through regional minimum search. It is followed by the lane region analysis to adaptively remove the road pixels and obstacles. Evaluations on the proposed lane preprocessing method are performed under different time of the day/night and various road sceneries such as city, highway, rural, suburban etc. Some results and findings are shown in the paper with regards on the observation of the road intensity change respective to the variation of the road conditions.
|Title of host publication||Proceedings - 2010 3rd IEEE International Conference on Computer Science and Information Technology, ICCSIT 2010|
|Publisher||IEEE Computer Society|
|Number of pages||5|
|Publication status||Published - 01 Jan 2010|