基于点特征的目标跟踪算法研究及其在机器人导航上的应用

【作者】 汪婧；

【导师】 茅耀斌；

【作者基本信息】 南京理工大学 ， 系统工程， 2008， 硕士

【摘要】 目标识别与跟踪属于计算机视觉研究领域的一个重要分支,引起了学者们的高度关注。在机器人研究领域,机器人视觉跟踪被广泛地应用于国防建设、工业监控以及数字医疗等诸多领域。本文首先讨论了目标识别与跟踪相关技术和国内外发展现状,对几种当前流行的识别与跟踪方法进行了较为详细地介绍。然后论文重点研究了基于点特征的目标跟踪算法,在对特征点提取算子进行分析的基础上,着重讨论了SIFT(Scale InvariantFeature Transform)特征点提取算子,并将该算子用于目标跟踪。鉴于SIFT特征点匹配目标跟踪算法实时性差的缺点,论文给出了一种结合SIFT与KLT(Kanade-Lucas-Tomasi)的目标跟踪算法,并基于OpenCV、DirectShow和VC++6.0的软件开发平台,开发了目标跟踪软件。此外,将结合SIFT与KLT的目标跟踪算法用于移动机器人,通过两种控制策略实现了单目标跟踪。在单目标跟踪的基础上,研究了目标导航问题,并将结合SIFT与KLT的目标跟踪算法用于机器人导航系统。实验结果表明,结合SIFT和KLT的算法用于目标跟踪无论从实时性上还是准确度上都比较优越,并通过该算法很好地实现了自主移动机器人的单目标实时跟踪和目标导航。本文提出的算法具有一定的实用性,对今后的移动机器人视觉技术,尤其是相关的导航系统设计与研究有一定的参考价值。更多还原

【Abstract】 Object recognition and tracking is an important research sub-field in computer vision, which has attracted great concerns of researchers. In robotics, the visial tracking of robotics has found extensive applications in national defense, industry monitoring, digital medical service, and many other areas. First, the state-of-the-art of the recognition and tracking related technology is discussed, the several popular approaches of tracking are introduced in detail. Then attention is focused on the study of the object tracking algorithm based on point feature. Based on analyzing the feature extraction operators, the SIFT (Scale Invariant Feature Transform) feature point operator is discussed emphatically and used to object tracking. This paper proposes a kind of unifies SIFT and KLT (Kanade-Lucas-Tomasi) object tracking algorithm. Based on OpenCV, DirectShow and the VC++6.0 software development platform, the software of object tracking is developed.In addition, SIFT and KLT (Kanade-Lucas-Tomasi) object tracking algorithm was used for mobile robot. In context of the mobile robot, the problem of single-target tracking is solved by two kinds of control strategies. In the foundation of single-target tracking, the object navigation problem is studied, and the proposed algorithm is used in the robot navigation system.The experimental result indicated that the algorithm proposed in this paper, to be used in the object tracking, has quite superior in the accuracy and real-time, the single-target real-time tracking and the object navigation of autonomous mobile robot are realized well through this algorithm. The tracking algorithm proposed in this paper has good usability, it will have good value to mobile robot vision technology, especially the related navigation system design.更多还原