中国科学院沈阳自动化研究所机构知识库
Advanced  
SIA OpenIR  > 机器人学研究室  > 学位论文
题名: 面向托卡马克维护的多视觉遥操作机器人环境感知技术研究
其他题名: Research on environmental perception for tele-operation maintenance of Tokamak using multi-vision approach
作者: 肖锡臻
导师: 韩建达
关键词: 特征点提取 ; 相对位姿测量 ; 特征点跟踪 ; 视觉里程计 ; 三维重建
索取号: TP242/X47/2015
页码: 54页
学位专业: 模式识别与智能系统
学位类别: 硕士
答辩日期: 2015-05-26
授予单位: 中国科学院沈阳自动化研究所
学位授予地点: 中国科学院沈阳自动化研究所
作者部门: 机器人学研究室
中文摘要: 伴随着全球人口的增长,能源危机问题将变得日益严峻,ITER计划为解决全球性的能源问题提供了希望。我国于2006年加入ITER计划,从加入计划以来便开始对和平利用核能遇到的系列问题进行独立研究。核聚变的高温高辐射使得需要定期对内部的真空室进行维护,由于环境恶劣不适合人类直接进行维护。本文的研究工作源于国家科技部“热核聚变反应计划专项”中的子课题“精细遥操作主从机器人与人机交互控制系统”(Y2D1153101)。本文主要研究了该项目中在遥操作条件下机械手维护真空室时对周围环境感知定位的问题,主要解决相机与被操作瓦块的相对位姿测定以及操作环境重建问题。主要有以下几个方面的内容:1. 通过介绍热核聚变反应计划专项的背景以及以遥操作方式控制机械手作业的必要性。分析遥操作过程中遇到的两个关键需求:机械手跟瓦块的相对位姿测量和机械手作业环境的三维重建。根据这两个关键需求,分析、比较了国内外相关算法的研究现状。2. 给出了适合热核聚变反应真空室的瓦块顶点的精确提取算法。由于瓦块角点不是图像中的显著特征,使用一般性的角点提取算法往往失效。通过分析,我们发现一般性角点提取算法只是利用的灰度信息未能利用对象目标的形状等先验知识。据此,我们提出了通过利用瓦块形状这些先验知识,结合角点以及边的图像信息,通过优化梯度最大值的方式计算瓦块的角点,提高了角点提取的鲁棒性和精度。该算法为测量瓦块与相机位的相对位姿测量提供了基础。3. 给出了基于非线性优化的实时位姿测量算法。由于真空室中可用于测量的对应点较少,使用传统方法在瓦块角点提取和瓦块角点三维坐标计算中所存在误差会极大影响测量精度。为了减少误差对位姿测量的影响,我们提出了基于非线性优化的相对位姿测量算法。该算法可以在一定程度上减少误差对位姿求解的影响。同时我们使用李群表示位姿矩阵,解决了非线性优化中收敛速度的问题,在速度方面可以满足实时测量的要求。4. 给出了两种用于连续估计瓦块跟相机相对位姿的方法。一是基于图像金字塔的跟踪算法跟踪瓦块的顶点。该方法通过优化瓦块的顶点处图像灰度差的平方和最小得到瓦块的角点图像坐标,再根据位姿估计算法得到位姿。二是基于视觉里程计的算法。该方法通过相机移动带来的视差变化直接计算相机的位姿变化,根据位姿变化以及之前瓦块跟相机的相对位姿得到新的相对位姿。5. 给出了该专项计划精细遥操作主从机器人与人机交互控制系统中使用的结构光仪器进行三维重建的框架。介绍了项目中实际使用的点云获取、清理、拼接、曲面化等一系列方法及三维重建中用于拼接的算法原理。
英文摘要: With the world's population grows, the energy crisis will become increasingly severe, ITER plans provide a way to solve the global energy problem totally. From the begining of China's join of the ITER project in 2006, Chinese govement has started its own research programs of research that focus on a series of problems encountered in the peaceful use of nuclear energy. The high temperature and high radiation of Fusion make it necessary to periodically perform maintenance inside the vacuum chamber.Due to that the harsh environment is not suitable for direct human maintenance, the teleoperation with robot becomes an effective way to substitute human to complete maintenance inside the vacuum chamber. In the teleoperation, it is important to measure the relative pose-position between the operating arms and the blanket, and to reconstruct the environment arounding operating arms. In this paper, the research contents includes two topics: how to accurately measure the relative pose between the operating arms and how to reconstruct 3D environment around the operating arms. The main research work, achievements and software development on these two topics in this paper are prensented as follows.1. The robot teleoperation on the vacuum chamber of nuclear fusion and the corresponding vision based navigation are introduced. In the vision based navigation for teleoperation, the current research status and development on measurement of relative pose and position and the reconstruction of the robot operating environment are analyzed and reviewed.2. the precise extraction algorithm of the vertices of blankets that are applied in the teleoperation system is proposed. The imaging of the vertices of blankets is not as salient as can be extracted by general corner extraction algorithms. An algorithm with priori knowledge of the shape, and also combining the information of corner and edge is proposed.It can calculate the corner points of blanket in subpixel level. In the algorithm, the sum of the gradient of blanket’s edget is taken as the cost function. By maximizing of the cost function the accurate and robust extraction of vertices of blankets is realized. This algorithm provides a basis for measuring the relative pose position between camera and blanket.3. an algorithm that measure of the pose and position between camera and blanket with nonlinear optimization approach is proposed , which can run in real time. Because far fewer corresponding points of blanket can be used in measurement, the application of traditional methods often leads to large error due to the errors in the image point and three-dimensional coordinates of corners of blanklet. In order to reduce the impact of the errors of measurement, an algorithm that use nonlinear optimization is proposed. The algorithm can largely reduce the impact of errors of measurement. In the algorithm, Lie group theory is applied to represent a pose-position matrix, which can largely increase the convergence speed to meet the requirements of real-time measurements.4. two methods are proposed, which can be used for continuously estimating the relative pose and position between camera and blanket. The first method can track vertexes based on image pyramid algorithm. By optimizing the difference of intensity of current frame and last frame of corners the algorithm can calculate the image point of corners in current frame, then get the relative pose and position between camera and blanket by above algorithm. The second method is based on visual odometry approach. The changes of camera’s pose and position generate different view of the scene. By catching the difference VO algorithms are applied to determine the relative pose and position change of camera. By combining the relative pose and position change of camera and the last pose and position between camera and blanket, the current pose and position between camera and blanket can be calculated.5. a framework of three-dimensional reconstruction process with the structured light instruments in the teleoperation is presented. An pipelines for the acquisition of point cloud, the filter of point clouds, the registration of point clouds, the generation of surface from point clouds and point clouds construction are introduced, and basic algorithm behind the software is presented.
语种: 中文
产权排序: 1
内容类型: 学位论文
URI标识: http://ir.sia.cn/handle/173321/16800
Appears in Collections:机器人学研究室_学位论文

Files in This Item:
File Name/ File Size Content Type Version Access License
面向托卡马克维护的多视觉遥操作机器人环境感知技术研究.pdf(1895KB)----限制开放 联系获取全文
Service
Recommend this item
Sava as my favorate item
Show this item's statistics
Export Endnote File
Google Scholar
Similar articles in Google Scholar
[肖锡臻]'s Articles
CSDL cross search
Similar articles in CSDL Cross Search
[肖锡臻]‘s Articles
Related Copyright Policies
Null
Social Bookmarking
Add to CiteULike Add to Connotea Add to Del.icio.us Add to Digg Add to Reddit
所有评论 (0)
暂无评论
 
评注功能仅针对注册用户开放,请您登录
您对该条目有什么异议,请填写以下表单,管理员会尽快联系您。
内 容:
Email:  *
单位:
验证码:   刷新
您在IR的使用过程中有什么好的想法或者建议可以反馈给我们。
标 题:
 *
内 容:
Email:  *
验证码:   刷新

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

 

 

Valid XHTML 1.0!
Copyright © 2007-2016  中国科学院沈阳自动化研究所 - Feedback
Powered by CSpace