ENHANCED VINS-BASED UNDERWATER LOCALIZATION WITH IMAGE ENHANCEMENT

The underwater visual-inertial navigation system (VINS) confronts significant challenges due to the adverse underwater visual environment, including light absorption and scattering, tiny suspended particles, color distortion, and image blurring. To address these issues, this paper introduces a multi-scale fusion-based image enhancement algorithm, integrating it into the front-end of underwater image enhancement techniques. This integration effectively enhances the performance of underwater localization. Experimental results on the Aqualoc underwater dataset demonstrate that the proposed method increases the number of extracted feature points and achieves more stable tracking, thereby reducing localization errors compared to traditional VINS approaches.