Robust Monocular Egomotion Estimation Based on an IEKF

In this contribution a robust approach for the estimation of the camera motion is presented. For this purpose, features from a monocular image sequence are extracted and evaluated so that the threedimensional path of a moving camera can be calculated. The algorithm gives robust results even in the presence of noise and independently moving objects. The two different categories of constraint equations used in the proposed algorithm are the epipolar constraint and the trilinear constraints. The optimization of the constraints with respect to the motion parameters is implemented as a robust Iterated Extended Kalman Filter. Test results are presented from real data, captured from a moving vehicle in an urban scenario.