A new 3D LIDAR-based lane markings recognition approach

A new active vision based lane markings recognition approach is presented. An efficient method for separating road surface data from a 3D LIDAR data frame is proposed which delimitates the drivable region by its profile. A new approach for extracting significant points of lane markings is developed using digital signal processing technology. And the significant points are clustered based on an orientation guided scheme. The tracking problem of lane markings in sequential frames is investigated. The experimental results show that the presented approach has a global perspective of surrounding road environments and is not sensitive to different natural illuminations.