Relative localization from image sequences
Brian L. Stuart · 2005
This paper presents a technique for vehicle localization. The technique is based on processing images taken by one or more downward pointing cameras. Motion represented by shift between successive images allows us to directly measure the velocity at two points on either side of the axis of motion. From these possibly di erent velocities, we can infer the vehicle speed and turning radius and from those, the vehicle orientation, position, velocity and acceleration vectors.