Abstract: In vehicle motion estimation using visual odometry, drift accumulated from errors in relative measurement is a big problem. This paper put forward a new drift suppression method based on dynamic tracking by B-spline fitting of vertical road profile. When a vehicle is driving tightly on the surface of a road, the motion trajectory will have the same shape as the vertical road profile. Thus tracking of the vertical road profile can be used to suppress the drift value for motion estimation using visual odometry. As vertical road profile tracking and motion estimation are not synchronous, a fixedlag Kalman drift suppressor was established. The experiment validates the effectiveness of the method suggested for drift suppression in visual odometry.  

Key words: visual odometry, motion estmation, drift suppression