A RGBD SLAM algorithm combining ORB with PROSAC for indoor mobile robot

摘要

In order to enhance the instantaneity of SLAM for indoor mobile robot, a RGBD SLAM method based on Kinect was proposed. In the method, oriented FAST and rotated BRIEF(ORB) algorithm was combined with progressive sample consensus(PROSAC) algorithm to execute feature extracting and matching. More specifically, ORB algorithm which has better property than many other feature descriptors was used for extracting feature. At the same time, ICP algorithm was adopted for coarse registration of the point clouds, and PROSAC algorithm which is superior than RANSAC in outlier removal was employed to eliminate incorrect matching. To make the result more accurate, pose-graph optimization was achieved based on g2o framework. In the end, a 3D volumetric map which can be directly used to the navigation of robots was created.