Research on ORB-SLAM3 algorithm adapting to dynamic environment

Abstract

The visual SLAM algorithm based on the assumption of static environment will be affected by the motion feature points in dynamic environment, resulting in the system robustness and accuracy degradation. To solve this problem, a visual SLAM algorithm for dynamic feature removal based on instance segmentation is proposed. Firstly, Yolact network is used to detect potential moving objects, and the obtained semantic information is combined with epipolar geometric constraints to accurately eliminate dynamic feature points. Considering the failure of example segmentation network, this algorithm combines the semantic information in the historical key frame with the depth map and uses the region growth algorithm to segment the dynamic objects that are not detected by Yolact, which further improves the robustness of the system. The experimental results show that compared with ORB-SLAM3, the absolute trajectory error and relative trajectory error of the algorithm in the dynamic environment are significantly reduced.