Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

MotionGS : Compact Gaussian Splatting SLAM by Motion Filter

Xinli Guo, Weidong Zhang, Ruonan Liu, Peng Han, Hongtian Chen

TL;DR

MotionGS tackles dense SLAM by integrating a compact 3D Gaussian Splatting representation with deep feature extraction and a dual keyframe strategy, enabling real-time tracking and high-fidelity mapping with low memory. The system consists of a tracking thread that uses motion-filtered keyframes and a mapping thread that jointly optimizes poses and Gaussians within a diff-Gaussian rasterization framework. Key contributions include a compact 3DGS loss with a mask-based pruning term, a dual keyframe strategy (motion and information keyframes), and a direct, pixel-level pose optimization that leverages Lie algebra for efficient updates. Experiments on TUM RGB-D and Replica demonstrate state-of-the-art tracking and rendering quality with substantially reduced storage, highlighting the practical potential of 3DGS-based SLAM for real-time dense mapping.

Abstract

With their high-fidelity scene representation capability, the attention of SLAM field is deeply attracted by the Neural Radiation Field (NeRF) and 3D Gaussian Splatting (3DGS). Recently, there has been a surge in NeRF-based SLAM, while 3DGS-based SLAM is sparse. A novel 3DGS-based SLAM approach with a fusion of deep visual feature, dual keyframe selection and 3DGS is presented in this paper. Compared with the existing methods, the proposed tracking is achieved by feature extraction and motion filter on each frame. The joint optimization of poses and 3D Gaussians runs through the entire mapping process. Additionally, the coarse-to-fine pose estimation and compact Gaussian scene representation are implemented by dual keyframe selection and novel loss functions. Experimental results demonstrate that the proposed algorithm not only outperforms the existing methods in tracking and mapping, but also has less memory usage.

MotionGS : Compact Gaussian Splatting SLAM by Motion Filter

TL;DR

MotionGS tackles dense SLAM by integrating a compact 3D Gaussian Splatting representation with deep feature extraction and a dual keyframe strategy, enabling real-time tracking and high-fidelity mapping with low memory. The system consists of a tracking thread that uses motion-filtered keyframes and a mapping thread that jointly optimizes poses and Gaussians within a diff-Gaussian rasterization framework. Key contributions include a compact 3DGS loss with a mask-based pruning term, a dual keyframe strategy (motion and information keyframes), and a direct, pixel-level pose optimization that leverages Lie algebra for efficient updates. Experiments on TUM RGB-D and Replica demonstrate state-of-the-art tracking and rendering quality with substantially reduced storage, highlighting the practical potential of 3DGS-based SLAM for real-time dense mapping.

Abstract

With their high-fidelity scene representation capability, the attention of SLAM field is deeply attracted by the Neural Radiation Field (NeRF) and 3D Gaussian Splatting (3DGS). Recently, there has been a surge in NeRF-based SLAM, while 3DGS-based SLAM is sparse. A novel 3DGS-based SLAM approach with a fusion of deep visual feature, dual keyframe selection and 3DGS is presented in this paper. Compared with the existing methods, the proposed tracking is achieved by feature extraction and motion filter on each frame. The joint optimization of poses and 3D Gaussians runs through the entire mapping process. Additionally, the coarse-to-fine pose estimation and compact Gaussian scene representation are implemented by dual keyframe selection and novel loss functions. Experimental results demonstrate that the proposed algorithm not only outperforms the existing methods in tracking and mapping, but also has less memory usage.
Paper Structure (18 sections, 16 equations, 3 figures, 5 tables)

This paper contains 18 sections, 16 equations, 3 figures, 5 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. The Proposed Approach
  4. Compact 3DGS Scene Representation
  5. Dual Keyframe Strategy
  6. 3DGS-based Direct pose optimization
  7. SLAM System
  8. Tracking
  9. Mapping
  10. Experiment
  11. Experimental Setting
  12. Metrics
  13. Baseline Methods
  14. Evaluation
  15. Track
  16. ...and 3 more sections

Figures (3)

  • Figure 1: Overview of MotionGS. The input to MotionGS at each timestep is the current RGB-D/RGB image. After the motion filter, the directly pose optimization of the motion keyframe is done based on the photometric error between the GT and render result. After information filter, the joint optimization of keyframe poses and 3D scene geometry on sliding windows and random historical frames is carried out in the mapping thread. Finally, the scene is refined.
  • Figure 2: Render Performance in Replica.
  • Figure 3: Render Performance in TUM.