Robust, real-time motion estimation from long image sequences using Kalman filtering

Abstract

This paper presents how to estimate the left and right monocular motion and structure parameters of two stereo image sequences including direction of translation, relative depth, observer rotation and rotational acceleration, and how to compute absolute depth, absolute translation and absolute translational acceleration parameters at each frame. For improving the accuracy of the computed parameters and robustness of the algorithm, A Kalman filter is used to integrate the parameters over time to provide a “best” estimation of absolute translation at each time.