Semihierarchical Reconstruction and Weak-area Revisiting for Robotic Visual Seafloor Mapping

TL;DR

The paper tackles deep-sea visual mapping by integrating navigation data with a hierarchical, hybrid SfM framework that combines SLAM-like local reconstructions and global pose-graph optimization. It introduces weak-area revisiting to target unreconstructed or poorly connected regions, uses refraction-aware calibration and color normalization, and performs chunked MVS for dense reconstructions. The approach demonstrates improved geometric consistency and reduced computation time (average $76.7\%$ faster) across multiple real-world datasets, outperforming vanilla COLMAP in challenging underwater conditions. This work advances scalable, offline, high-quality seafloor mapping and provides practical guidance for large-area deep-sea surveys.

Abstract

Despite impressive results achieved by many on-land visual mapping algorithms in the recent decades, transferring these methods from land to the deep sea remains a challenge due to harsh environmental conditions. Images captured by autonomous underwater vehicles (AUVs), equipped with high-resolution cameras and artificial illumination systems, often suffer from heterogeneous illumination and quality degradation caused by attenuation and scattering, on top of refraction of light rays. These challenges often result in the failure of on-land SLAM approaches when applied underwater or cause SfM approaches to exhibit drifting or omit challenging images. Consequently, this leads to gaps, jumps, or weakly reconstructed areas. In this work, we present a navigation-aided hierarchical reconstruction approach to facilitate the automated robotic 3D reconstruction of hectares of seafloor. Our hierarchical approach combines the advantages of SLAM and global SfM that is much more efficient than incremental SfM, while ensuring the completeness and consistency of the global map. This is achieved through identifying and revisiting problematic or weakly reconstructed areas, avoiding to omit images and making better use of limited dive time. The proposed system has been extensively tested and evaluated during several research cruises, demonstrating its robustness and practicality in real-world conditions.

Figures (19)

• Figure 1: Example underwater images. Due to light attenuation and scattering effects, the images appear to have a strong blue and green hue. Left and center also show the heterogeneous illumination, which is caused by the moving light source mounted on the robot. Image courtesy: GEOMAR and Schmidt Ocean Institution.
• Figure 2: From left to right: The Girona 500 AUV Anton and Luise; The CoraMo Mk ii camera system; The LED light ring.
• Figure 3: Overview of the visual seafloor mapping pipeline.
• Figure 4: First stage of the underwater camera calibration: centering the camera in the dome port. Sample images have shown that the vertical edges of the calibration target are consistent above and below the water surface, which suggests that the refraction effect has been effectively avoided.
• Figure 5: The notations of different coordinate systems. $^p {{{\bf T}}}_w$ denotes the prior coordinate frame, which represents the vehicle body and $^c {{{\bf T}}}_w$ represents the camera coordinate frame.