Motion segmentation and scene classification from 3D LIDAR data

We propose a hierarchical data segmentation method from a 3D high-definition LIDAR laser scanner for cognitive scene analysis in context of outdoor vehicles. The proposed system abstracts the raw information from a parallel laser system (velodyne system). It extracts essential information about drivable road segments in the vicinity of the vehicle and clusters the surrounding scene into point clouds representing static and dynamic objects which can attract the attention of the mission planning system. The system is validated on real data acquired from our experimental vehicle in urban, highway and cross-country scenarios.