S3LAM: Surfel Splatting SLAM for Geometrically Accurate Tracking and Mapping
Ruoyu Fan, Yu-Hui Wen, Tao Zhang, Long Zeng, Yong-Jin Liu
AI summary
Problem
Existing 3D Gaussian Splatting SLAM systems compromise geometric accuracy and tracking stability because unoriented ellipsoids lack explicit surface normals and prioritize color consistency over fine geometry.
Approach
S3LAM represents scenes using oriented 2D Gaussian surfels for explicit surface modeling, coupled with an adaptive rendering strategy to handle uncertainty and a novel analytic pose Jacobian that aligns camera orientation with reconstructed surfaces.
Key results
- State-of-the-art mapping and tracking performance on Replica, TUM-RGBD, and ScanNet++ datasets
- Online adaptive surface mapping strategy using depth distortion to dynamically refine uncertain geometry
- Novel surfel-based pose Jacobian with radial gradients that improves tracking convergence under large viewpoint changes
- Superior geometric fidelity and reduced visual artifacts compared to NeRF and 3DGS-based SLAM baselines
Why it matters
Enables reliable, high-fidelity scene reconstruction and robust camera tracking for real-time robotics, autonomous navigation, and augmented reality applications.
Abstract
We propose S3LAM, a novel RGB-D SLAM sys- tem that leverages 2D surfel splatting to achieve geometrically accurate scene representations for simultaneous tracking and mapping. Unlike existing 3DGS-based SLAM approaches that rely on 3D Gaussian ellipsoids, we utilize 2D Gaussian surfels as primitives for more efficient scene representation. By focusing on the surfaces of objects in the scene, this design enables S3LAM to reconstruct high-quality geometry, benefiting both mapping and tracking. To address inherent SLAM challenges including real-time optimization under limited viewpoints, we introduce a novel adaptive surface rendering strategy that improves mapping accuracy while maintaining computational efficiency. We further derive camera pose Jacobians directly from 2D surfel splatting formulation, highlighting the im- portance of our geometrically accurate representation that improves tracking convergence. Extensive experiments on both synthetic and real-world datasets demonstrate that S3LAM achieves state-of-the-art performance. Our code is available at https://github.com/FanryZ/S3LAM.