Towards 3D karst underwater scene reconstruction from rotating sonar data

2026-06-18

AI Summary

This study addresses the challenges of sparse, noisy sonar data and navigation drift in underwater 3D reconstruction of karst aquifers by proposing a novel framework that integrates continuous-time SLAM with a two-stage deep learning pipeline. The approach first employs continuous-time SLAM to correct trajectory drift and enhance pose accuracy. Subsequently, a newly designed two-stage deep learning model efficiently generates high-fidelity surface meshes from rotating sonar point clouds. The resulting immersive and navigable 3D model significantly improves reconstruction quality and geometric completeness, thereby providing a robust and reliable data foundation for hydrogeological analysis.

Abstract

Karst aquifers provide critical freshwater resources but pose significant hazards due to their complex and poorly understood subsurface geometry. Mapping these environments is challenging because sonar data from underwater exploration is sparse and noisy, while navigation estimates suffer from drift limiting standard 3D reconstruction methods. We present a pipeline for reconstructing underwater karst conduits from a sonar profiler. We combine a continuous-time SLAM approach to correct trajectory drift with a novel two-stage deep learning method for surface reconstruction, producing an immersive and navigable 3D mesh for hydrogeological analysis.

Problem

• 3D reconstruction
• karst aquifers
• sonar data
• trajectory drift
• underwater mapping

Innovation

• continuous-time SLAM
• two-stage deep learning
• 3D mesh reconstruction
• rotating sonar
• karst aquifer mapping