An Underwater Features Matching System Integrated with Image Enhancement

Sainan Zhang,Zhibin Yu

DOI: https://doi.org/10.1109/ICEICT57916.2023.10245163

2023-01-01

Abstract:With the increasing demand for ocean exploration and underwater observation, video-based 3D reconstruction tasks and simultaneous localization and mapping (SLAM) tasks in underwater scenarios attract great attention. Accurate point correspondence extraction from consecutive underwater video frames is crucial for estimating camera poses and motion trajectories in video-based 3D vision tasks. However, the poor quality of underwater images, caused by light absorption and scattering in water, makes it challenging to obtain accurate matches solely relying on a robust feature-matching network. This paper proposes a matching system designed explicitly for achieving more precise matching between underwater video frames. The system integrates an image enhancement model, a feature extraction model, and a feature-matching model. To address the lack of ground truth depth and camera poses for underwater images, we introduce a unique homographic transformation to simulate camera motion-induced viewpoint changes. The transformed images, paired with the original ones, are utilized to train the feature matching network, which is a graph neural network with self- and cross-attention. And we replace dot-product attention with linear attention to reduce computational complexity. The experimental results validate the effectiveness of underwater feature-matching of the proposed method.