| Mobile Robot Mapping is an important research direction in the field of robot navigation. Map representing the operating environment of the robot has very important affect on the later localization and plan. As a positive sensor, Vision sensor has advantages of low cost, much information and so on. Especially the binocular stereo vision system, by which the stereo information can be renewed directly, is more and more used in mobile robot mapping. Therefore, it is of great significance to research on mobile robot mapping. The thesis focused on the method of indoor mobile robot mapping based on binocular stereo vision, and the main work is as following:Firstly, the presented study status and methods of mobile robot mapping at home and abroad are summarized, and several weight questions are indicated. The hardware structure of mobile robot platform MT-R and the Bumblebee2 binocular stereo vision system are introduced.Secondly, the technology of feature extraction and stereo match is researched on, and the algorithm of extracting and matching SIFT feature points, which are invariant to image rotation, scale zoom and affine, is focused on. According to the experiment, some key parameters, which have an impact on feature extraction and match property, are analyzed, and the balance between the quantity of feature points and the match efficiency is established. Euclidean distance is used to be the similarity measure, and to some context, the mismatch is eliminated by the use of some constraint criteria.Finally, the method of indoor visual landmarks mapping is described in detail. The SIFT feature points extracted from the environment are employed to be as the visual landmarks and the 3D coordinates of the landmarks are computed under the global frame. The landmarks database containing the 3D information of visual landmarks is established, and the 2D visual landmark map is constructed in the horizon plane. The experiment result demonstrates the effectiveness of the plan. |
PDF Full Text Request