基于立体视觉Tri-EKF算法的空间非合作目标运动与结构估计

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

针对空间非合作目标的运动状态及外形结构估计的双重扩展卡尔曼滤波(Dual EKF)算法存在运动状态变量耦合及收敛速度慢的问题。基于立体视觉观测模型,引入特征线作为新观测量分析了目标平移与旋转运动估计的解耦性,提出三重扩展卡尔曼滤波(Tri-EKF)算法,实现了对目标旋转、平移运动以及结构参数的串行估计。本文对两种算法进行仿真比较,通过试验校验Tri-EKF算法。结果表明:有限次观测下,Tri-EKF算法对目标运动估计收敛速度更快,平均估计误差更小。

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.