A Three-Point-Frame-Difference SLAM Algorithm for Indoor Environment with Dynamic Objects.

Existing visual SLAM methods usually assume that the environment is static, which limits the application of most visual SLAM systems in the vast majority of real-world environments. The feature points of dynamic objects generally result in low positioning accuracy therefore leads to relocation failure. To address this problem, this paper presents a Three-Point-Frame-Difference (TPFD) method based on the ORB-SLAM2 to improve the system localization accuracy. The proposed TPFD method is constructed based on the principle that the relative position of static points remains unchanged, and the frame difference method is utilized to implement an intersection operation on the adjacent three image frames, thereby distinguishing the dynamic and static points. The effectiveness of the proposed method is validated by the public TUM RGB-D dataset. The robust of the TPFD on trajectory error is compared with the object detection method based on ORB-SLAM2. Results indicate that both the TPFD method and object detection method can effectively remove dynamic points in high and low dynamic environments, and both methods perform better in high dynamic environments. And also, the proposed TPFD method is more efficient compared with object detection method.