Mesh-based Photorealistic and Real-time 3D Mapping for Robust Visual Perception of Autonomous Underwater Vehicle

TL;DR

This work tackles robust underwater localization and photorealistic dense mapping for AUVs in hazy, low-contrast environments. It combines a learning-based underwater image enhancement (Joint-ID) with a mesh-based dense mapping pipeline and a real-time sliding-window expansion to maintain lightweight, textured maps. The authors validate their approach qualitatively on real-world FLSea data and quantitatively on a distorted indoor synthetic dataset, demonstrating improved localization accuracy and human-interpretable maps. The proposed framework offers practical benefits for inspection tasks, enabling reliable perception and realistic visualization under challenging underwater conditions.

Abstract

This paper proposes a photorealistic real-time dense 3D mapping system that utilizes a learning-based image enhancement method and mesh-based map representation. Due to the characteristics of the underwater environment, where problems such as hazing and low contrast occur, it is hard to apply conventional simultaneous localization and mapping (SLAM) methods. Furthermore, for sensitive tasks like inspecting cracks, photorealistic mapping is very important. However, the behavior of Autonomous Underwater Vehicle (AUV) is computationally constrained. In this paper, we utilize a neural network-based image enhancement method to improve pose estimation and mapping quality and apply a sliding window-based mesh expansion method to enable lightweight, fast, and photorealistic mapping. To validate our results, we utilize real-world and indoor synthetic datasets. We performed qualitative validation with the real-world dataset and quantitative validation by modeling images from the indoor synthetic dataset as underwater scenes.

Figures (15)

• Figure 1: Illustration of the proposed method. We utilize a learning-based image enhancement and apply mesh-based mapping to generate a photorealistic map.
• Figure 2: The overall pipeline of the proposed system.
• Figure 3: Illustration of Joint-ID and image enhancement result.
• Figure 4: Point sampling and triangulation: The red points are extracted feature points. We performed Delaunay triangulation on the extracted points. The red triangle mesh is the one that is removed from the image plane. The purple triangle mesh is removed after the 3D projection. The yellow triangle mesh is the correct one on the map.
• Figure 5: Comparison of 3D mesh maps. (a) Before outlier rejection. (b) After outlier rejection.