Partially Calibrated Camera Based Pose Estimation of Mobile Robots with Application to Visual Tracking

In this paper, a new method is developed for relative pose estimation of a monocular camera mounted on a wheeled mobile robot. Different from most of the previous works, the proposed method requires the camera be partially calibrated instead of fully calibrated, and it works for both coplanar and non-planar scenes. Specifically, two image frames are utilized to extract the geometric relationship by considering the planar motion constraint of the mobile robot. Then the relative pose information of the camera is derived based on the developed geometric relationship. To evaluate the effectiveness of the proposed method, the comparison between the proposed method and the classical multi-view geometry based approaches is presented via simulation. Furthermore, practical relevance of the proposed method is illustrated by the application example for the visual trajectory tracking of the mobile robot. Both simulation and experimental results indicate that the proposed method can achieve sufficient pose estimation accuracy to accomplish the visual tracking task.