视觉导航具有采用其他传感器导航方式所无法比拟的优点，是目前移动机器人智能导航的主要发展方向。目前，如何实时可靠地提取环境中的视觉信息作为路标仍然是制约移动机器人进行智能视觉导航的主要瓶颈。本文在国家863计划和国家自然科学基金的资助下，以室内服务机器人的视觉导航为应用背景，对于人工路标的设计、识别以及基于该路标的视觉定位方法进行了探讨和研究。本论文的主要工作和创新之处如下： 第一，分析比较了当前存在的一些人工路标的设计图案的优缺点，在此基础上结合二维码技术，设计了一种用于机器人室内视觉导航的人工路标——MR二维码，在设计过程中对所设计路标的颜色、形状、模块以及采用的编解码方式都做了具体分析。 第二，针对所设计的人工路标，提出了一种鲁棒的识别算法。该算法主要基于形状不变量理论，通过识别路标的外轮廓进行检测，然后进行识别。在提取外轮廓上的直线时针对实际情况未使用hough变换而采用了计算曲率的方式。为减少计算量以及增加检测的鲁棒性，对轮廓的跟踪也进行了分析。实验结果表明该识别方法的可靠性很好。 第三，为了能在很弱的光照条件下进行视觉导航，设计了一种具有二维码性质的有源路标并设计了提出了相应的识别方法。实验结果表明该路标以及识别方法可以满足一般的室内视觉导航需求。 最后，对如何利用MR二维码进行视觉定位以及视觉导航进行了分析。在视觉定位方面根据路标位置的不同采用了不同的计算方式，并对如何进行全局定位进行了单独分析，在定位的基础上提出了用路径跟随和视觉伺服结合的方式完成视觉导航任务的方案。在机器人平台上对所提方案进行了实验验证，结果表明所设计的人工路标能够满足自主移动机器人室内视觉导航的要求，完成一般的室内视觉导航任务。; Visual navigation，superior to the navigation system without vision, has become the main direction of research on intelligent navigation for mobile robots. However, extracting the vision information as landmark robustly and real-time is still difficult to process the intelligent visual navigation for mobile robots. Supported by the 863 National High-tech Project and National Natural Scientific Foundation, we have developed to solve the problems of designing artificial landmark, recognition of the landmark and visual locating method based on the landmark for navigation of indoor service robot. The novel work and contribution of this thesis can be summarized as follows: Firstly, we compare the existing patterns of artificial landmarks and analyse the merits and drawbacks of them. Furthermore, combining the popular technique of 2D Code, a novel artificial landmark for mobile robots-MR 2D Code is presented, and its color, shape, modules, encoding and decoding method are analysed in the process of designing. Secondly, directing against the artificial landmark we presented, a robust recognition algorithm is presented. The algorithm detects the landmark by its outline which is mainly based on the invariant theory, and then recognizes it. A method of computing curvature is adopted instead of the Hough transform to extract the lines of the outline. In order to reduce computing cost and improve robustness, tracking outline is analysed too. The experiments show the adequate reliability of the algorithm. Thirdly, in order to navigate under weak illumination, an active artificial landmark and the corresponding recognition algorithm are presented, and they are different from MR 2D Code. The result in this thesis suggest that the active artificial landmark can meet the ordinary requirement of indoor visual navigation. Finally, the visual location and visual navigation based on the artificial landmark are analysed. The pose estimating methods are different according to the attached positions, and the global locating is analysed independently. Based on the visual location, a visual navigation scheme combining path following and visual servo is presented. The scheme is verified on the robot platform developed by our laboratory, and the result shows that the artificial landmark presented in this thesis can meet the ordinary requirement of indoor visual navigation.
修改评论