A monocular ORB-SLAM in dynamic environments

Abstract

The well-known ORB-SLAM is capable to deal with the localization and mapping tasks in static scenes. But in a dynamic environment, it is hard to perform well as the positions of 3D points in the map may change over time. In order to solve the problem, we propose a new slam framework, i.e. DORB-SLAM, which extends the application range of the ORBSLAM in dynamic circumstances. We compared the appearance of the image blocks centered at the projection of the 3D points in current frame and the reference frame to decide whether the points are dynamic points or not. We test the DORB-SLAM in a public dataset and the results indicate that the algorithm is more robust than the origin ORB-SLAM in dynamic enviroments.