August 25, 2025Open Access

A Robust Framework Fusing Visual SLAM and 3D Gaussian Splatting with a Coarse-Fine Method for Dynamic Regions Segmentation

View Full Paper

Ask AI

Bookmark

Camera pose estimation improved by excluding features on dynamic regions, achieving up to 95% reduction in absolute trajectory error.
Dense optical flow clustering provides detailed analysis for ambiguous targets while ensuring robust mapping performance.
Cascaded coarse-to-fine approach reduces motion analysis computation time by 91.7% compared to dense-only methods.
Validations on diverse datasets highlight superior robustness and accuracy for localization and mapping in dynamic scenarios.

Discussion

No takes yet. Share an insight, caveat, or question.

Proposed framework enhances camera pose estimation in dynamic environments, indicating significant performance improvement.

Chen et al. (2025) studied this question.

View Full Paper

Ask AI

Bookmark