Vehicle Model Based Visual-Tag Monocular Orb-Slam

Monocular ORB-SLAM has been proved to be one of the best open-source SLAM method. However, it is still unsatisfying especially in low illumination indoor environment, which is caused by scale recovery and wrong feature matching. In this paper, we proposed a vehicle model based monocular ORB-SLAM method supplemented by April-Tag to improve the performance of original algorithm. This approach is practical when autonomous driving in low-light and less-feature environment like garages and tunnels. We achieve this by proposing a vehicle model based initialization method fusing April-Tag measurement to recover scale. During tracking procedure, the outliers ORB feature points will be removed by checking reprojection error calculated from April-Tag. In addition, considering vehicle model can only obtain 2D motion, the vertical transition is estimated from camera model. Afterwards, a local Bundle Adjustment(BA) is applied to optimize camera pose both from frame to frame and frame to keyframe which will reduce accumulative error of the vehicle model. Finally, a convincing result is obtained from the testing drive in a garage.