A Bag-of-Words Speedometer for single camera SLAM

This paper proposes a novel solution to the problem of scale drift in single camera SLAM based on recognising objects of known scale. When reconstructing the trajectory of a camera moving in an unknown environment the scale of the environment, and equivalently the speed of the camera, is obtained by accumulating relative scale estimates over sequences of frames. This leads to scale drift: errors in scale accumulate over time. Our solution is to correct this scale estimate by recognising objects of known size. A bag-of-words-based scheme to learn object classes, to recognise object instances, and to use these observations to correct scale drift is described, and is demonstrated reducing scale drift by 75% while navigating a large indoor environment.