GEMOR-GS: Global Extrinsic-calibration-refined Mapping and Online Relocalization for 3D Gaussian-Splatting SLAM
GEMOR-GS: Global Extrinsic-calibration-refined Mapping and Online Relocalization for 3D Gaussian-Splatting SLAM
This research introduces GEMOR-GS, a system designed to create photorealistic and metrically consistent world representations using 3D Gaussian splatting for LiDAR-inertial-visual SLAM. The method refines the extrinsic calibration between LiDAR and cameras offline, then alternately refines keyframe poses and Gaussian parameters using differentiable rendering and projection objectives. Experiments on self-collected and public datasets show that GEMOR-GS improves map geometric accuracy by 14.3-19.9 dB over online maps, increases PSNR by approximately 9 dB compared to Gaussian-LIC2, and achieves 86-100% relocalization success, serving as a dependable world model for robot perception and decision-making.
The paper presents GEMOR-GS, a method for refining 3D Gaussian Splatting maps in LiDAR-inertial-visual SLAM systems to improve geometric accuracy and relocalization capabilities.
Source: arXiv Robotics — research abstracts · Read original article ↗
Loading article text…
What the source reports
Publisher-reported claims, with original evidence. These results have not been independently verified by RoboSignal.
Reported numbers
PSNR improvement dB
≈9
View original evidence
alization, and Gaussian RGB-D warp refinement. Experiments on one self-collected and five public sequences quantify each element: the offline reconstruction wins all paired keyframes against the online map by 14.3--19.9 dB, improves PSNR by approximately 9 dB over Gaussian-LIC2, and achieves 86--100% held-out relocalization success tied to extrinsic quality. The results show that GEMOR-GS improves the geometric accur
Open source S5offline reconstruction PSNR gain dB
14.3
View original evidence
alization, and Gaussian RGB-D warp refinement. Experiments on one self-collected and five public sequences quantify each element: the offline reconstruction wins all paired keyframes against the online map by 14.3--19.9 dB, improves PSNR by approximately 9 dB over Gaussian-LIC2, and achieves 86--100% held-out relocalization success tied to extrinsic quality. The results show that GEMOR-GS improves the geometric accur
Open source S5offline reconstruction PSNR gain dB
19.9
View original evidence
alization, and Gaussian RGB-D warp refinement. Experiments on one self-collected and five public sequences quantify each element: the offline reconstruction wins all paired keyframes against the online map by 14.3--19.9 dB, improves PSNR by approximately 9 dB over Gaussian-LIC2, and achieves 86--100% held-out relocalization success tied to extrinsic quality. The results show that GEMOR-GS improves the geometric accur
Open source S5
- paper: not_reported
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
alization, and Gaussian RGB-D warp refinement. Experiments on one self-collected and five public sequences quantify each element: the offline reconstruction wins all paired keyframes against the online map by 14.3--19.9 dB, improves PSNR by approximately 9 dB over Gaussian-LIC2, and achieves 86--100% held-out relocalization success tied to extrinsic quality. The results show that GEMOR-GS improves the geometric accur
Open source S5
Source：arXiv Robotics — research abstracts · arxiv.org