Robust vehicle environment reconstruction from point clouds for irregularity detection

Understanding the surrounding environment including both still and moving objects is crucial to the design and optimization of intelligent vehicles. Knowledge about the vehicle environment could facilitate reliable detection of moving objects, especially irregular events (e.g., pedestrians crossing the road, vehicles making sudden lane changes,) for the purpose of avoiding collisions. Inspired by the analogy between point cloud and video data, we propose to formulate a problem of reconstructing the vehicle environment (e.g., terrains and buildings) from a sequence of point cloud sets. Built upon existing point cloud registration tool such as iterated closest point (ICP), we have developed an expectation-maximization (EM)-ICP technique that can automatically mosaic multiple point cloud sets into a larger one characterizing the still environment surrounding the vehicle. Moreover, we propose to address the issue of irregularity detection from the extracted moving objects. Our experimental results have shown successful reconstruction of a variety of challenging vehicle environments (including rural and urban, road and intersection, etc.) and simultaneous tracking/segmentation of multiple moving objects.