空间非合作目标基于点云模型的视觉与惯性融合相对导航方法与实验研究

项目名称： 空间非合作目标基于点云模型的视觉与惯性融合相对导航方法与实验研究

项目编号： No.61573115

项目类型： 面上项目

立项/批准年度： 2016

项目学科： 自动化技术、计算机技术

项目作者： 宋申民

作者单位： 哈尔滨工业大学

项目金额： 16万元

中文摘要： 非合作目标的交会对接是航天器在轨操控的关键技术之一。在近距离逼近段，视觉测量是其主要的导航手段。由于非合作目标没有已知的光点标志，同时也由于空间光学污染、遮挡等因素的影响，使得非合作目标的位姿测量成为一个挑战性的课题。本项目将基于点云模型的视觉测量方法与惯性技术融合，提出一种高精度的非合作目标相对导航策略。 利用双目立体视觉的测量方式来获取非合作目标的散乱点云数据，通过对散乱点云数据的处理，获得非合作目标的点云模型，进而完成对其对接机构的识别。通过将极大验后估计、高斯条件分布等性质融合进高斯滤波器设计的方式，结合视觉、惯导数据约束关系模型，给出高精度的非合作目标相对导航参数。最后，通过仿真验证所提出方法的有效性。

中文关键词： 相对导航；点云模型；非线性滤波；高斯滤波框架；数值稳定性

英文摘要： The rendezvous and docking of non-cooperative target is one of the key technologies of on-orbit operation. Vision measurement is the main navigation approach during the stage of proximity operations. For the reasons of no luminous markers on the target, the disturbance of light pollution and occlusion, the estimation of position and attitude for non-cooperative target becomes a challenge problem. This project fuses the method of point-sets-based vision measurement and the inertial technology. A high accurate algorithm of relative navigation for non-cooperative target is presented. It takes advantage of binocular vision to get the scattered points data about non-cooperative target, and its model of point sets and docking mechanism are obtained by processing these points data.Maximum a posteriori estimation and conditional Gaussian distributions are exploited to the design of Gaussian filter. And the high-precision relative state with respect to non-cooperative target is estimated by combining with the constraint relationship between the data of vision and inertial navigation system. Finally, the validity of the proposed method is verified by simulation.

英文关键词： Relative navigation；Point cloud model；Nonlinear filtering；Guassian filtering framework；Numerical stability