面向空间飞行器逼近停靠的非合作目标 视觉测量关键技术研究
文献类型：学位论文
| 作者 | 王敏燕
|
| 学位类别 | 硕士 |
| 答辩日期 | 2016-05-25 |
| 授予单位 | 中国科学院沈阳自动化研究所 |
| 导师 | 郝颖明 |
| 关键词 | 计算机视觉 非合作目标 位姿测量 典型部件 特征提取 鲁棒性 |
| 其他题名 | Research on Key Technologies for Vision Pose Measurement of Space Non-cooperative Target Specific to the Rendezvous in the Approaching and Docking state |
| 学位专业 | 模式识别与智能系统 |
| 中文摘要 | 本文以空间飞行器逼近停靠为应用背景，选择星箭对接环和太阳能帆板两种典型部件，分别从基于典型部件的位姿解算技术、典型部件的识别与特征提取技术，空间飞行器逼近停靠视觉测量系统实现与验证三方面展开研究，为工程应用提供技术支持。取得的主要研究成果包括：1.在基于典型部件的位姿解算方法方面，面向圆形的星箭对接环和矩形的太阳能帆板，分别进行了基于圆特征和直线特征的位姿解算方法研究。对于星箭对接环，根据工程需求，在姿态角已知的条件下，提出了基于精确相机模型的相对位置直接解算方法和对相机模型误差不敏感的插值位置求解方法，实现了基于圆和圆外一点的相对位姿求解方法。对于矩形太阳能帆板，实现了基于门型三线的相对位姿求解方法。对上述算法分别进行了仿真实验。2. 对于图像中能够检测到较多可以用于视觉位姿计算的特征量的情况，提出了多特征组合位姿优化求解技术。针对不同种类的特征，设计量纲统一的目标函数，并选取粒子群算法进行优化求解，使用圆和圆外一点进行初始位姿计算，比较了LM算法和粒子群算法的优化结果，实验结果表明，粒子群算法能够取得比较显著的优化效果。3.针对典型部件的识别与特征提取，结合实际工程需求选择了合适的点、线、椭圆提取算法。点特征提取进行的是亮斑检测，主要采用区域生长和椭圆检测提取长短轴比与面积值均满足一定阈值条件的亮斑。线特征检测，主要采用LSD算法，并在检测结果基础上进行了目标提取处理。椭圆检测采用的是基于最小二乘法的迭代拟合法。将各种算法分别进行工程化，利用工程图像以及实验图像进行测试，均能取得比较理想的结果。4. 面向实际工程应用，研究了不具备精确标定条件下的摄像机标定技术，比较了三种可能的标定方法：直接标定法，已知fu,fv进行标定以及已知fu，fv和R矩阵进行标定，找到了适合实际工程应用的标定方法。5.为完成空间飞行器逼近停靠视觉测量技术的验证，编写了视觉位姿测量软件并搭建了地面演示实验系统。视觉测量软件实现了典型部件的识别与位姿解算功能，利用地面演示实验系统以及实际工程图像对软件进行测试，实验结果表明软件能够实现较高精度的位姿解算功能。本文提出的方法，已经成功应用到实际工程项目中，达到了项目技术指标要求。 |
| 英文摘要 | This topic is based on the application background of space vehicle approaching, Select satellite rocket docking ring and its solar panels two typical parts as the research object, This article pursue research from three aspects: pose solving technique based on typical parts, recognition and feature extraction of typical parts, realization and verification of the space vehicle approaching the vision measurement system. This article is expected to provide technical support for engineering applications. The main results obtained include: 1. In terms of the position and attitude calculation method based on the typical parts, for circular satellite rocket docking ring and rectangular solar panels, respectively the calculation method research based on the pose of the round features and linear features. For star arrow docking ring , according to the engineering requirements, a direct method of resolving pose is proposed based on the accurate camera model, an interpolation method? of resolving pose is proposed which is not sensitive to the error of camera model. A method of resolving pose based on a circle and a point is implemented. For rectangular solar panels, A method of resolving pose based on three door-like lines is implemented. Simulation experiments are carried out on the above algorithm. 2. For the situation that multiple features can be detected in the image, A method of pose optimization was proposed. The method chooses the appropriate solution to get initial value of pose according to existing features , builds a normalization goal function utilizating. point , line, surface features, and obtains optimal solution by Particle Sarm Optimization. The optimization results of LM algorithm and particle swarm optimization are compared. The experimental results show that the particle swarm optimization algorithm can achieve a significant effect. 3. For typical pars recognition and feature extraction, combined with the actual engineering needs, suitable method of points, lines and ellipse features extraction is chosen. point extraction is a bright spot detection, bright spot detection mainly uses the regional growth and ellipse detection to determine the ratio of two axis and the area value to meet a certain threshold condition of the bright spot. Line feature detection, mainly uses LSD algorithm. Because the LSD has some defects, the detection results are adjusted. Ellipse detection is based on the iterative detection of the least square method. All kinds of algorithms are used to carry out the engineering, and the engineering images and the experimental images are used to test the algorithm. The effect of various algorithms is ideal. 4. For practical engineering applications, The camera calibration technique is studied without the condition of accurate calibration, Three possible calibration methods are compared: direct calibration method, calibration method with fu 、fvknown, calibration method with fu,fv?,R?known, We find a suitable calibration method for practical engineering applications. 5. For the verification of the space vehicle approaching the vision measurement technology, Visual position and attitude measurement software was programmed and the ground demonstration experiment system was set up. The recognition of typical parts and the calculation function of the position and orientation are realized by the visual measurement software. The software is tested by the ground demonstration experiment system and the actual engineering image, The experimental results show that the software can achieve a higher accuracy of the position and attitude calculation function. The method proposed in this paper has been successfully applied to practical engineering projects, the project can meet the requirements of technical indicators. |
| 语种 | 中文 |
| 产权排序 | 1 |
| 页码 | 69页 |
| 源URL | [http://ir.sia.cn/handle/173321/19627]  |
| 专题 | 沈阳自动化研究所_光电信息技术研究室 |
| 推荐引用方式 GB/T 7714 | 王敏燕. 面向空间飞行器逼近停靠的非合作目标 视觉测量关键技术研究[D]. 中国科学院沈阳自动化研究所. 2016. |
入库方式： OAI收割
来源：沈阳自动化研究所
浏览0
下载0
收藏0
其他版本
除非特别说明，本系统中所有内容都受版权保护，并保留所有权利。