B-spline modeling of road surfaces for freespace estimation

In this paper, we propose a general technique for robust modeling of non-planar ground surfaces. It is based on representing the ground plane as a parametric B-Spline surface and on robust tracking of the surface parameters using stereo disparities and a Kalman filter. We adopt a recently published real-time optimization algorithm for finding the road-obstacle boundary to integrate disparity values. The results show that in situations where a planar road assumption yields wrong results due to undulating roads, our approach is able to distinguish between free space and obstacles.