The purpose of localization study is to enable mobile robots to determine their position in environment in real-time and reliably, which is necessary for mobile robots to navigate autonomously. Vision based localization has many advantages that other methods are unsurpassable and has become an important research field. Supported by Innovation Foundation of IACAS “Service and Education Robots”, some key techniques of vision based localization for indoor mobile robots are studied in this thesis. The main contents of this paper are listed as follows: Firstly, the history, application background and development status of robots and mobile robots are briefly reviewed. The main study focuses on mobile robots are addressed in details. Thereafter, a survey on vision based localization of mobile robot is presented. Secondly, feature tracking of monocular image sequences is discussed considering the indoor environment which the mobile robots work in. With respect to point feature tracking, the KLT algorithm is introduced. A method is proposed to improve KLT algorithm by means of distance constraints between feature points. Then the Hough-based tracking method is depicted, followed with the introduction of a light adaptive, novel feature tracking method which combines both the KLT method and the Hough-based method. Thirdly, a real-time visual odometry algorithm with monocular sequences is presented based on two planar assumptions: indoor robots are moving on planar ground and the camera view is planar structure. Experimental results show that the proposed method achieve good precision in short distance run as well as long distance run. The validity and error of the estimation are also analyzed. Fourthly, there are sparse and similary features in most indoor environment, such as office. So correlation based matching methods may not work well under this circumstance. A new method that uses the SIFT descriptor of the feature to realize stereo matching is proposed and confirmed to be feasible by experiments. Stereo image sequence based localization is also discussed and analysed. Finally, the achieved research results are summarized and future work is addressed.
修改评论