Model-based Vehicular Motion and Structure Estimation

Yi-Sheng Yao, Rama Chellappa

Abstract: This paper presents a model-based algorithm for estimating vehicle motion and structure from a long sequence of images. In addition to employing a simple kinematic law suitable for modeling the steering control, the unstabilized behavior of a vehicular motion due to the uneven terrain is taken into account. A vehicle model which accounts for bounce, pitch and roll is used. With the aid of inertial sensors such as accelerometers and by tracking a set of points over the image sequence, a feature-based approach is proposed. Simulation results for a particular vehicular motion are presented.