GCORB SLAM: a VSLAM system graph-cut dynamic point removal in complex environments

Simultaneous Localization and Mapping (SLAM) is essential for unmanned devices to navigate complex environments accurately. Traditional methods falter in dynamic settings due to reliance on geometric features. This study proposes addressing dynamic point culling as a joint probabilistic inference problem, enhanced with Gcransac and graph optimization to refine data sampling. By integrating polar coordinate constraints with geometric error for dynamic point rejection, and upgrading to joint optimization with multimodal information, this approach facilitates dynamic SLAM sensing. Consequently, it improves the system's robustness and accuracy over traditional techniques, making significant strides in reliable navigation for unmanned devices in challenging environments.