Towards Robust Global VINS: Innovative Semantic-Aware and Multi-Level Geometric Constraints Approach for Dynamic Feature Filtering in Urban Environments In this study, we propose an innovative real-time global visual-inertial navigation system that features a novel front-end and an adaptive backend, designed to enhance localization accuracy and state estimation robustness in dynamic outdoor environments. Our global VINS effectively addresses the challenges posed by dominant and ambiguous dynamic objects, surpassing the limitations of traditional deep learning and vision-based methods. It efficiently manages moving objects in the scene, overcoming the issues of cautious or excessive dynamic feature removal, ensuring more reliable and accurate state estimation.
Our paper has been published, please refer to the Link