VISO: Robust Underwater Visual-Inertial-Sonar SLAM with Photometric Rendering for Dense 3D Reconstruction
Shu Pan, Simon Archieri, Ahmet Fatih Cinar, Jonatan Scharff Willners, Ignacio Carlucho, Yvan Petillot
AI summary
Problem
Underwater visual sensing degrades rapidly in turbid or dark conditions, while sonar data alone is too sparse and noisy for accurate pose estimation and high-fidelity mapping.
Approach
The system tightly couples a 3D sonar, stereo camera, and IMU in a single optimization framework, using coarse-to-fine online extrinsic calibration and robust feature association to fuse modalities for real-time dense mapping.
Key results
- Coarse-to-fine online extrinsic calibration between 3D sonar and camera
- Robust sonar point cloud association with 2D-2D RANSAC outlier rejection
- Real-time dense 3D reconstruction via photometric rendering of sonar clouds
- Lower translation and rotation errors than SOTA baselines in tank and lake tests
Why it matters
Provides a reliable perception solution for autonomous underwater vehicles operating in visually challenging environments where traditional cameras or sonar alone fail.
Abstract
Visual challenges in underwater environments sig- nificantly hinder the accuracy of vision-based localisation and the high-fidelity dense reconstruction. In this paper, we propose VISO, a robust underwater SLAM system that fuses a stereo camera, an inertial measurement unit (IMU), and a 3D sonar to achieve accurate 6-DoF localisation and enable efficient dense 3D reconstruction with high photometric fidelity. We introduce a coarse-to-fine online calibration approach for extrinsic pa- rameters estimation between the 3D sonar and the camera. Additionally, a photometric rendering strategy is proposed for the 3D sonar point cloud to enrich the sonar map with visual information. Extensive experiments in a laboratory tank and an open lake demonstrate that VISO surpasses current state-of- the-art underwater and visual-based SLAM algorithms in terms of localisation robustness and accuracy, while also exhibiting real-time dense 3D reconstruction performance comparable to the offline dense mapping method.