SIA OpenIR  > 光电信息技术研究室
Alternative TitleResearches on Geometric Features Based Visual Pose Estimation Method for Non-cooperative Target in Space
Thesis Advisor朱枫
Keyword位姿测量 椭圆检测 直线特征匹配 几何信息 位姿求解
Degree Discipline模式识别与智能系统
Degree Name博士
Degree Grantor中国科学院沈阳自动化研究所
Place of Conferral沈阳

随着人类对太空探索的不断深入,故障航天器的自动在轨捕获与维修等在轨服务已成为航天技术发展的迫切需要。目标航天器自动捕获的前提是能够在线自动获取两个航天器间的相对位置和姿态。在众多的测量传感器中,视觉已经成为空间近距离测量的主要有时甚至是唯一的测量手段。依据目标航天器是否安装用于测量的人工标志器,视觉测量可分为合作目标视觉测量和非合作目标视觉测量。其中,非合作目标的视觉测量只能利用目标固有的自然特征,相对于合作目标视觉测量来说,技术难度更大;虽然由于其迫切的空间应用需求,已经得到各航天大国的普遍重视,目前仍处于技术攻关阶段。本文以航天器自动捕获为应用背景,开展空间非合作目视觉测量算法研究,具有重要的理论意义与实际应用价值。本文以实际的工程应用需求为背景,针对空间航天器普遍具有的圆、直线等典型几何特征,研究空间非合作目标视觉测量相关算法,包括特征检测、特征匹配和相对位姿求解算法,主要内容包括:(1) 基于几何特性的实时椭圆检测方法研究。星箭对接环是空间航天器普遍具有的典型部件,在图像上表现为椭圆,因此椭圆检测是基于对接环的位姿测量的首要条件。针对复杂空间环境下,椭圆检测准确率低、速度慢的问题,提出基于几何特性的椭圆检测算法,算法制定弧段提取和分类、弧段组合以及后验证策略,在检测准确率、召回率和计算速度上都具有很大的提升。该算法对各类航天器目标具有普遍适用性。(2) 基于共面点线信息的直线特征匹配方法研究。卫星等空间人造目标多为由规则几何形状构成的复杂的结构性目标,目标上包含了大量的直线段特征。针对航天器图像纹理重复导致直线特征匹配准确率低的问题,提出了基于共面点线信息的直线特征匹配方法,利用近似点线不变量和点线方位两种共面点线信息提高直线匹配的准确率。(3) 基于双目图像的位姿求解方法研究。针对复杂空间环境下航天任务对基线距、测量距离范围和测量精度的要求,提出了基于双目图像的非合作航天器位姿估计方法。结合目标几何形状和双目相机结构,通过优化方法提高位姿求解的精度和鲁棒性。(4) 基于双目序列图像的位姿求解方法研究。通过对多帧连续图像分析和处理,可恢复出时间维度上的相对运动信息,可用于相对运动状态估计和预测。针对空间任务对数据处理和存储的限制,提出了一种基于光束平差法的非合作航天器的位姿估计方法。利用特征的几何信息和双目结构建立重投影误差函数。推导和分析测量函数的雅克比矩阵的稀疏形式,通过计算和存储非零值来简化优化过程,减少计算量。

Other Abstract

With the development of space exploration, the on-orbit service of automatic orbital capture and repair mechanism for faulty spacecraft has become an urgent need for space technology development. On-orbit service technologies, such as docking, refueling, repairing, and even upgrading satellites, have become an important issue to be solved in the development of space technology. The automatic capture of the target spacecraft needs the ability to automatically acquire the relative position and attitude between the two spacecraft online. Among the measurement sensors, visual sensors have become the main and sometimes the only measurement method for close-range measurements in space. Depending on whether the target spacecraft is equipped with an artificial marker for measurement, the visual measurement can be divided into cooperative target visual measurement and non-cooperative target visual measurement. Among them, the visual measurement of non-cooperative targets can only utilize the natural features inherent in the target, and the technical difficulty is greater than that of the cooperative target visual measurement; although it has been widely recognized by many countries due to its urgent space application requirements, it is still in the technical research stage. This paper takes the spacecraft automatic capture as the application background, and carries on the research of non-cooperative visual measurement algorithm, which has important theoretical significance and practical application value. Based on the actual engineering application requirements, this paper studies the non-cooperative target visual measurement related algorithms, including feature detection, feature matching and relative pose estimantion algorithms, for the typical geometric features invloving circles, points, and lines. The main contents include: (1) Research on ellipse detection method based on geometric characteristics. The docking ring is a typical component commonly used in spacecraft and appears as an ellipse on the image, so ellipse detection is the primary condition for the pose measurement of the docking ring. Aiming at the problem of low accuracy and slow speed of ellipse detection in complex space environment, an ellipse detection algorithm based on geometric features is proposed. The algorithm develops arc segment extraction and classification, arc segment combination and post-validation strategy, which has great improvement in detection accuracy, recall rate and calculation speed. The algorithm has general applicability to various spacecraft targets. (2) Research on line feature matching method based on coplanar point-line information. Space artificial targets such as satellites are mostly complex structural targets composed of regular geometric shapes, and the target contains a large number of straight line segments. Aiming at the problem that the texture of the spacecraft image is repeated and the accuracy of the linear feature matching is low, a line matching method based on the coplanar point-line information is proposed. The line matching method improves the accuracy rate by using the approximate point-line invariant and the point-line orientation constraints. (3) Research on pose measurement method based on binocular image. Aiming at the requirements of space mission in the complex space environment, such as the baseline distance, the measuring distance range and the measurement accuracy, a non-cooperative spacecraft pose estimation method based on binocular image is proposed. Combining the target geometry and the binocular camera structure, the accuracy and robustness of the pose measurement are improved by an optimization method. (4) Research on pose measurement method based on binocular sequence image. By analyzing and processing multiple frames of continuous images, relative motion information in the time dimension can be recovered, which can be used for relative motion state estimation and prediction. Aiming at the limitations of space tasks on data processing and storage, a pose estimation method for non-cooperative spacecraft based bundle adjustment method is proposed. The reprojection error function is established using the geometric information of the feature and the binocular structure. Deriving and analyzing the sparse form of the Jacobian matrix of the measurement function, simplifying the optimization process and reducing the amount of calculation by calculating and storing non-zero values.

Contribution Rank1
Document Type学位论文
Recommended Citation
GB/T 7714
张丽敏. 基于几何特征的空间非合作目标视觉测量方法研究[D]. 沈阳. 中国科学院沈阳自动化研究所,2018.
Files in This Item:
File Name/Size DocType Version Access License
基于几何特征的空间非合作目标视觉测量方法(4258KB)学位论文 开放获取CC BY-NC-SAApplication Full Text
Related Services
Recommend this item
Usage statistics
Export to Endnote
Google Scholar
Similar articles in Google Scholar
[张丽敏]'s Articles
Baidu academic
Similar articles in Baidu academic
[张丽敏]'s Articles
Bing Scholar
Similar articles in Bing Scholar
[张丽敏]'s Articles
Terms of Use
No data!
Social Bookmark/Share
All comments (0)
No comment.

Items in the repository are protected by copyright, with all rights reserved, unless otherwise indicated.