Attitude/Position Estimation of Monocular Vision Based on Multiple Model Kalman Filter

摘要

In this paper, a multiple model Kalman filter (MMKF) is proposed for attitude/position estimation of monocular vision, in which the noise disturbance and data loss during camera imaging process is considered. Firstly, by establishing an acceleration model of artificial markers in the camera image-plane, a Kalman filtering (KF) strategy is adopted to deal with the disturbance and data loss problem of markers. Especially for the data loss situation, an extended Kalman filter (EKF) with equality constraints, taking into account the prior information between the artificial markers, is designed. Furthermore, after comprehensive consideration of the computational complexity and accuracy of the above two strategies, we present a novel MMKF estimation scheme to ensure the speed and accuracy of attitude/position estimation. The effectiveness of the proposed MMKF estimation strategy is evaluated by numerical simulation experiments.