基于点集配准的立方体卫星非合作姿态估计

Non-cooperative pose estimation for cubesat based on point set registration

非合作目标相对姿态估计是空间在轨任务的重要问题。目前视觉成像系统通常采用特征点实现姿态测量,对单个特征点依赖性强,因此存在鲁棒性差的缺点。针对空间非合作目标中的立方体卫星在自主交会对接等任务中的姿态估计问题,采用双目视觉立体匹配并根据三角测量原理获取目标卫星的部分三维点集数据,利用点集配准解算该测量三维数据相对于已知模型的变换参数,最终实现立方体卫星的相对姿态估计。该方法基于稠密点集保证了较高的测量鲁棒性,并可扩展用于解决其他非合作目标的相对姿态估计问题。利用高精度转台的对比测量实验,验证了本文方法的精度和有效性。

The non-cooperative target relative pose estimation technique is the essential issue for spatial on-orbit service missions.The vision based imaging systems usually measure the required attitudes with the help of special feature points,have strong dependence on single feature point and poor robustness in the complex space environment.In this paper,aiming at the pose estimation task for the well-known cubesat in space non-cooperative target in the space missions such as automatic rendezvous and docking,the binocular stereo vision matching is adopted;the partial three-dimensional （3D）point set data of the target satellite are obtained based on triangular measurement principle.Then,the point set registration is utilized to solve the transformation parameters between the measured 3D data and the known model.Finally,the relative attitude estimation of the target cubesat is achieved.This method is based on dense point set, hence ensures stronger measurement robustness.This method can also be extended to solve the relative attitude estimation of other non-cooperative targets.The comparison measurement experiment with high precision turntable verifies the accuracy and effectiveness of the proposed method.