Monocular simultaneous localization and mapping with a modified covariance Extended Kalman Filter

摘要

A Modified Covariance Extended Kalman Filter (MVEKF) algorithm is proposed to the monocular simultaneous localization and mapping (SLAM) in this paper. Recent literatures have shown that it is possible to solve the monocular SLAM using the Extended Kalman Filter (EKF) and the inverse-depth parameterization. However, the EKF algorithm has its intrinsic disadvantage such as the divergence. Here we propose the use of MVEKF algorithm to improve the performance of monocular SLAM. Experiments were carried out on indoor image sequences and result show that the MVEKF algorithm could improve the convergence of landmarks.