SIIS-SLAM: A Vision SLAM Based on Sequential Image Instance Segmentation

Simultaneous localization and mapping (SLAM) is a fundamental function of intelligent robots. To reduce the influence dynamic objects on SLAM in environments, this study pro-poses visual based sequential image segmentation, referred to as SIIS-SLAM. Based ORB-SLAM3, SIIS-SLAM integrates instance segmentation optical flow detection module. The module designed eliminate effectiveness estimation relative pose between frames. Specifically, coarse estimated by ORB-SLAM3 box coordinates instances detected Mask R-CNN, effectively improves speed accuracy segmentation. Dynamic can be combining results Filtering feature points improve robustness SLAM. Experimental demonstrate that achieves better environments compared ORB SLAM3 other advanced methods.