基于点云的空间非合作目标初始相对位姿获取

针对空间非合作目标相对位姿解算对点云迭代最近点(ICP)算法迭代初值要求较高，易产生误匹配等问题，提出了一种基于点云分割与点云关键点的ICP初始迭代位姿获取方法。首先，考虑到非合作目标可能具有的高度对称外观，基于LCCP点云分割算法与点云Harris关键点，在已知目标外轮廓点云基础上，设计了点云ICP初始迭代位姿获取方法。该方法具体流程为：点云降采样处理与关键部分提取、关键部分点云多角度预变换处理、最佳点云匹配及初始迭代位姿获取。最后，以某通信卫星模型搭建虚拟验证平台，使用本文提出的方法给出ICP迭代初始位姿并进行对比实验。结果表明，本文提出的点云ICP迭代初始位姿获取方法可有效规避非合作目标高度对称带来的误匹配，可给出较精确的相对位姿。

Point Cloud Based Initial Relative Pose Acquiring Process for RelativePose Estimation of Non Cooperative Object in Space

To solve the problem brought by ICP process requiring high accuracy of initial iterative pose while estimating the relative pose of a non cooperative object in space, we propose an innovative initial pose acquiring process for point cloud ICP, based on point cloud segmentation and point cloud keypoints. Firstly, we take the potentially highly symmetrical appearance of a non cooperative object into consideration, and propose the initial pose acquiring algorithm using LCCP point cloud segmentation and Harris keypoints. The proposed algorithm requires a point cloud model of target contour, and consists of three steps, i.e. point cloud down sampling process and crucial parts extraction, multiple angel pre transformation process of crucial point cloud, and optimal matching extraction and initial pose calculation. Then, we build a simulation platform using a communication satellite model, and the contrast experiment is carried out. Simulation results show that the proposed point cloud ICP iterative initial pose acquiring method can effectively solve the mismatch problem caused by the highly symmetrical appearance of the non cooperative target, and thus derive a more accurate relative pose.