Motion and Structure Estimation of Non-Cooperative Space Target Based on Tri-EKF Algorithm and Stereo Vision

For the motion state and structure estimation of a non-cooperative space target, the motion parameters are coupled, which leads to the slow convergence problem in the dual extended Kalman filter (Dual EKF) algorithm. In this paper, based on the stereo vision model, a feature line observation model is introduced for decoupling of the translational and rotational motion. Then the triple extended Kalman filter (Tri-EKF) algorithm is proposed to sequentially estimate the rotational, translational motion and structural parameters of the target. Compared with the Dual EKF algorithm, the simulation and experimental results prove that the Tri-EKF has a faster convergence speed on motion estimation and smaller average estimation error given limited observations.