Visual Odometry for Velocity Estimation of UGVs

Abstract

An accurate and robust velocity estimation method based on an optical flow technique is presented in this paper. Using image sequences captured by a monocular camera mounted under an UGV (unmanned ground vehicle), image velocities are obtained from the optical flow technique. Combining with a camera model, the velocities of the UGV are directly estimated. This velocity estimation method is validated over various types of terrain surfaces, such as coarse sand, fine sand and mixture of coarse sand and gravel. Experimental results show that estimated velocities have very good agreement with measured velocities. Height between the projection center of camera and the terrain surface is proved to be a key parameter in velocity estimation. Height compensation is implemented to give accurate velocity estimation results. Velocity estimation method proposed has many potential applications including localization and slip estimation for UGVs.