VISO: Robust Underwater Visual-Inertial-Sonar SLAM with Photometric Rendering for Dense 3D Reconstruction

TL;DR

VISO addresses the key challenge of robust underwater localisation and high-fidelity dense reconstruction by tightly fusing a stereo camera, an IMU, and a 3D sonar. It introduces a coarse-to-fine online calibration between the camera and sonar, along with a dense sonar-based mapping pipeline that renders photometric information into the sonar map. The approach achieves superior localisation robustness and accuracy compared with state-of-the-art underwater and visual SLAM methods and enables real-time dense 3D reconstruction that rivals offline dense mapping. Extensive tank and open-lake experiments demonstrate strong performance in turbid and visually challenging conditions, highlighting the practical impact for underwater inspection and navigation.

Abstract

Visual challenges in underwater environments significantly 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 parameters 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.

Figures (10)

• Figure 1: (a) Dense mapping result in the lake. (b) The corresponding camera view of colour-dotted boxes in the areas of interest on the dense sonar map.
• Figure 2: Overview of VISO, where a 3D sonar is fused with an IMU and a stereo camera to enable accurate localisation and real-time dense mapping.
• Figure 3: The visualisation of sparse and noisy sonar point clouds with corresponding camera images in different viewpoints. The point clouds vary in different perspectives.
• Figure 4: The key sonar frame (a) and the current sonar frame (b) data association results visualised in the XY plane. Green points are the backprojected 3D sonar point cloud, red points are the outliers after initial matching, blue points are the matched features after outlier rejection.
• Figure 5: (a) The robot platform. (b) The laboratory tank environment with the motion capture system setup. (c) Visualisation of the tank sequence 1. (d) Visualisation of the tank sequence 2.