| Mobile robot is an offshoot of the robotics, it is widely applied in room-service, security, medical-care, sea-refloatation, space-exploration, mine-exploration and military-service. To accomplish the task in the environment freely, autonomous navigation is necessary for the mobile robots. If the environment is unknown, the mobile robot needs to build the environment map and locates itself at the same time, that is simultaneous localization and mapping(SLAM).More and more SLAM applications use camera because it has many advantages, such as low-cost and rich-information. This dissertation is focused on the SLAM based on vision.This paper reviews some of the key issues of SLAM, some typical SLAM algorithms and their advantages and disadvantages are introduced respectively. As the image information from the vision sensor is different from the traditional distance sensor information, first this paper presents both the point-feature and line-feature extraction methods, then the depth calculations of the features for both monocular vision and stereo vision are given. Focused on accessed point-features, a vision SLAM algorithm based on EKF is well presented, including the system model, data-association and state update. Because of the limitations of the point-feature in the vision SLAM realization, this paper proposes a vision SLAM algorithm using line-feature and a map representation method combined with points and lines is given. Experiments based on Pineer3DX mobile robot platform verified that both the two algorithms are effective, and the results show that the vision SLAM algorithm using line-feature has faster convergence in structural environment. Finally the algorithm is used in an office-like environment... |
PDF Full Text Request