CASIA OpenIR  > 毕业生  > 博士学位论文
增强现实中的跟踪问题研究
其他题名Research on Tracking for Augmented Reality
姬靖
学位类型工学博士
导师郁文生
2009-05-28
学位授予单位中国科学院研究生院
学位授予地点中国科学院自动化研究所
学位专业控制理论与控制工程
关键词增强现实 跟踪 姿态估计 平面图案 P4p问题 Pde方法 Augmented Reality Tracking Pose Estimation Planar Pattern P4p Problem Pde Method
摘要增强现实(AR)是虚拟现实技术的一个重要发展方向,是新一代的人机交互技术,具有广阔的应用前景和重要的研究价值。增强现实技术通过计算机产生虚拟信息与真实世界融合,增强了人们对真实世界的感受。为了实现虚实世界无缝隙融合的目的,增强现实需要实时精确地跟踪用户视角的变化,因此跟踪技术是增强现实的一项关键技术,跟踪问题的解决将会极大促进增强现实技术的发展和应用。本文主要围绕增强现实中的跟踪问题进行研究,论文的主要内容如下: 1.分析了增强现实中基于平面标记的跟踪系统所存在的问题。利用增强现实中广泛使用的ARToolKit开发工具,进行了真实的增强现实实验,并在Matlab环境下进行了仿真,证实了姿态跳变问题和虚拟物抖动问题的存在,并对这些问题进行了分析。分析结果对改进基于平面标记的增强现实系统具有一定的参考价值。 2.利用不等式机器证明理论及不等式机器证明软件DISCOVERER工具箱,分析了增强现实中基于4个基准点的跟踪系统P4P问题。首先对PnP问题进行了系统总结,推导化简得到P4P问题的系统方程组,最后详细分析了一个特殊情况下的P4P问题,给出了该情况下解的完全分类。研究结果对实际应用具有一定的参考价值。基于这些分类判别条件,在实时的增强现实系统中,可以很好地避开不稳定点,为实时决策提供依据,从而提高系统的稳定性和准确性。 3.总结了无标记增强现实系统中的跟踪方法和策略,通过比较各种不同的跟踪方法,选择Sift不变量特征算法作为跟踪器设计的基础,然后通过估计单应矩阵来得到姿态估计,给出了一种简单的跟踪器设计思路。针对增强现实中图像噪声的干扰问题,考虑边缘和纹理特征的保护要求,在增强现实的图像预处理中采用基于偏微分方程(PDE)的图像复原方法。实验结果表明此方法在去除图像中噪声的同时,成功保留了边缘和纹理等重要图像特征,为解决增强现实跟踪问题打下了较好的基础。
其他摘要As the one of new generations of human-machine interaction technologies, Augmented Reality (AR) is an important direction of virtual reality technology. With the computer-generated information, it can enhance the people’s perception to the real world and can be used in many fields. In order to align the virtual objects with real objects in the user’s view seamlessly, AR systems need accurately and quickly tracking the position and orientation of the camera. Thus, tracking is a very important issue in real-time AR systems. The resolution of tracking problem will promote the development and application of AR technology. This dissertation focuses on tracking problem in AR. The main content is as following: First, the problems arising from planar pattern-based tracking systems in AR are analyzed. By using the widely used AR toolkit-ARToolKit, we carry out the real experiments of augmented reality. Also simulation to the tracking in the AR systems is implemented in Matlab. The results of both two experiments reveal the existence of pose ambiguity problem and jitter problem of virtual object. An analysis to these problems is presented, which does some good to the development of planar-based AR systems. Second, based on the theory of automated inequality-type theorem proving and the relevant toolkit named “DISCOVERER”, P4P problem of 4 control point-based AR tracking systems is investigated. The PnP problem is briefly reviewed. To utilize the method of the inequality automated discovering, an equation system for P4P problem in AR is formulated. Moreover, a special case of P4P problem is computed and the complete distribution of real solutions is given. Based on these determinant conditions, the unstable points for AR systems can be avoided in practice, which can enhance the robustness and accuracy of AR systems. Third, various methods and strategies for tracking technology in marker-less AR systems are summarized. By comparing the different tracking methods, we choose the “Sift” invariant feature transform algorithm as the basis of tracker design. Furtherly taken advantage of the homography matrix, the pose of camera can be estimated. And this is our simple idea for the design of tracker. On the other hand, since edge and texture features have important roles in the tracking for AR systems, the denoising procedure in AR image preprocessing must be well-designed as to protect these valuable features. In this thesis, the PDE-based denoisin...
馆藏号XWLW1310
其他标识符200518014628032
语种中文
文献类型学位论文
条目标识符http://ir.ia.ac.cn/handle/173211/6177
专题毕业生_博士学位论文
推荐引用方式
GB/T 7714
姬靖. 增强现实中的跟踪问题研究[D]. 中国科学院自动化研究所. 中国科学院研究生院,2009.
条目包含的文件
文件名称/大小 文献类型 版本类型 开放类型 使用许可
CASIA_20051801462803(11223KB) 暂不开放CC BY-NC-SA请求全文
个性服务
推荐该条目
保存到收藏夹
查看访问统计
导出为Endnote文件
谷歌学术
谷歌学术中相似的文章
[姬靖]的文章
百度学术
百度学术中相似的文章
[姬靖]的文章
必应学术
必应学术中相似的文章
[姬靖]的文章
相关权益政策
暂无数据
收藏/分享
所有评论 (0)
暂无评论
 

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。