UW-SDF: Exploiting Hybrid Geometric Priors for Neural SDF Reconstruction from Underwater Multi-view Monocular Images

Due to the unique characteristics of underwater environments, accurate 3D reconstruction of underwater objects poses a challenging problem in tasks such as underwater exploration and mapping. Traditional methods that rely on multiple sensor data for 3D reconstruction are time-consuming and face challenges in data acquisition in underwater scenarios. We propose UW-SDF, a framework for reconstructing target objects from multi-view underwater images based on neural SDF. We introduce hybrid geometric priors to optimize the reconstruction process, markedly enhancing the quality and efficiency of neural SDF reconstruction. Additionally, to address the challenge of segmentation consistency in multi-view images, we propose a novel few-shot multi-view target segmentation strategy using the general-purpose segmentation model (SAM), enabling rapid automatic segmentation of unseen objects. Through extensive qualitative and quantitative experiments on diverse datasets, we demonstrate that our proposed method outperforms the traditional underwater 3D reconstruction method and other neural rendering approaches in the field of underwater 3D reconstruction.

I. Introduction

Reconstructing high-quality 3D underwater objects is crucial for autonomous underwater vehicles (AUVs) and remotely operated vehicles (ROVs) in tasks like mapping, inspection, and transportation [1]–[4]. However, collecting data underwater is challenging, often requiring skilled divers or specialized equipment. Acoustic and optical sensors [5]–[8] are commonly used underwater, with optical cameras being widely applied due to their passive, cost-effective, and convenient nature in underwater perception. Nevertheless, the absorption and scattering of light in water can result in color distortion and blurring of images, hampering the effectiveness of traditional vision methods in underwater environments. This degradation significantly impacts the quality of 3D reconstruction.