空间非合作目标的运动参数估计与三维重建

发布时间：2018-12-11 22:16

【摘要】：随着航天技术的飞速发展以及空间活动的不断增多,以卫星在轨装配、故障维修等为目的的空间目标在轨捕获技术已经成为航天技术领域重要的研究方向。空间非合作目标的运动参数估计与三维重建是空间目标在轨捕获技术领域的主要关键技术之一,具有重要研究价值与意义。本课题主要研究工作是基于双目立体相机的图像信息,实现对空间非合作目标的运动参数估计及三维重建。首先,构建了基于双目立体相机的三维点云获取系统。对相机成像模型进行描述并完成相机的标定;采用Triclops立体视觉库设计了初始三维点云的获取方案,通过对点云进行预处理和立体处理,获取目标的视差信息,并根据三角测量的方法计算出目标的深度信息,进而获取目标初始三维点云;研究了基于PCL的三维点云后处理方法,完成对目标点云去噪、拼接和采样。其次,利用由目标点云计算出的粗糙空间目标位姿作为输入,研究了一种基于无迹Kalman滤波器的运动参数估计算法,可以同时估计出空间目标平移与旋转的运动参数。包括质心的位移及速度、旋转角速度、惯性主轴的姿态以及目标的主惯量相对值。通过数值仿真结果表明,该方法对空间非合作目标运动参数的估计具有较高的鲁棒性及精度。然后,研究了一种提高空间非合作目标三维重建速度与准确性的方法。研究了基于二次栅格的点云简化算法,通过二次栅格对点云进行空间划分,采用k近邻搜索算法搜寻数据点的k近邻,进而计算出点云的法向量信息,根据法向量之间的夹角来进行点云的选择性采样,能够保留目标的几何特征,最后利用Power Crust算法对目标进行表面重建。针对空间目标中典型的喷嘴特征进行了仿真研究,验证该方法能够在去除点云中大量冗余数据的同时,保留模型表面的基本几何特征,实现对空间目标点云快速、准确的三维重建。最后,搭建了实验平台对空间非合作目标运动参数估计与三维重建算法进行实验验证。使用机械臂抓取卫星模型模拟其在空间中的运动,通过对不同视角下采集到的目标图像信息进行处理,最终获取目标的运动参数,并对获取到的卫星模型点云进行简化及三维重建,重塑其几何外形。对实验结果分析表明:本文的方法对空间非合作目标运动参数的估计具有较高的鲁棒性及精度,并能够实现目标点云快速、准确的三维重建。
[Abstract]:With the rapid development of space technology and the increasing of space activities, satellite on-orbit assembly, fault maintenance and other space target on-orbit acquisition technology has become an important research direction in the field of space technology. Estimation of motion parameters and 3D reconstruction of space non-cooperative targets is one of the key technologies in the field of on-orbit acquisition of space objects, which has important research value and significance. Based on the image information of binocular stereo camera, the motion parameter estimation and 3D reconstruction of non-cooperative objects in space are realized in this paper. Firstly, a three-dimensional point cloud acquisition system based on binocular stereo camera is constructed. The camera imaging model is described and the camera calibration is completed. The acquisition scheme of initial 3D point cloud is designed by using Triclops stereo vision library. The parallax information of the target is obtained by preprocessing and stereoscopic processing of the point cloud, and the depth information of the target is calculated according to the method of triangulation. Then the initial 3D point cloud of the target is obtained. A 3D point cloud post-processing method based on PCL is studied. The target point cloud is de-noised, stitched and sampled. Secondly, a motion parameter estimation algorithm based on the unscented Kalman filter is proposed, which can estimate the motion parameters of the spatial target translation and rotation simultaneously by using the rough space target position and pose of the target cloud computing as the input. It includes the displacement and velocity of the center of mass, the angular velocity of rotation, the attitude of the inertial spindle and the relative value of the main inertia of the target. The numerical simulation results show that the proposed method is robust and accurate to the estimation of motion parameters of space non-cooperative targets. Then, a method to improve the speed and accuracy of 3D reconstruction of space non-cooperative targets is studied. In this paper, the point cloud simplification algorithm based on quadratic grid is studied. The point cloud is partitioned by quadratic grid, and k nearest neighbor search algorithm is used to search k-nearest neighbor of data point, and then the normal vector information of point cloud is calculated. According to the angle between normal vectors, the point cloud can be sampled selectively, and the geometric features of the target can be preserved. Finally, the Power Crust algorithm is used to reconstruct the surface of the target. The typical nozzle features in space target are simulated. It is verified that this method can remove a large amount of redundant data from the point cloud, while preserving the basic geometric features of the model surface, so that the point cloud of the space target can be quickly implemented. Accurate 3D reconstruction. Finally, an experimental platform is built to verify the motion parameter estimation and 3D reconstruction algorithm. The robot arm grabbing satellite model is used to simulate its motion in space, and the target motion parameters are obtained by processing the target image information collected from different angles of view. The point cloud of the satellite model is simplified and reconstructed, and its geometric shape is reconstructed. The experimental results show that the proposed method is robust and accurate for estimating the motion parameters of non-cooperative objects in space, and can realize fast and accurate 3D reconstruction of target point clouds.
【学位授予单位】：哈尔滨工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41

【参考文献】