Underwater 3D Reconstruction by Interleaving Multimodal SLAM and Incremental Gaussian Splatting

Daniel Yang, Jungseok Hong, John J. Leonard, Yogesh Girdhar

Keywords: 3D reconstruction, gaussian splatting, multimodal slam, coral reefs

TL;DR: achieve underwater 3d reconstruction with noisy inputs through multimodal slam and 3dgs

Abstract:

We propose ReefMapGS, an incremental 3DGS reconstruction framework that replaces SfM with pose-graph optimization-based SLAM, leveraging multi-modal sensors (e.g., acoustic, inertial, pressure, and visual) to estimate camera poses and their associated uncertainties.
ReefMapGS builds an initial model from a high-certainty region and progressively expands by interleaving local tracking of new observations with 3DGS scene optimization. Refined poses are fed back into the pose graph for global trajectory optimization, tightly coupling reconstruction and localization. We show COLMAP-free 3D reconstruction of two geometrically complex underwater reef sites and improved global pose estimation of our AUV over complex foveated survey trajectories spanning up to 700 m.

Email Sharing: We authorize the sharing of all author emails with Program Chairs.

Data Release: We authorize the release of our submission and author names to the public in the event of acceptance.

Paper Acceptance: Yes

Submission Number: 7