Visual SLAM with keyframe selection for underwater structure inspection using an autonomous underwater vehicle

Abstract

This study introduces an efficient visual simultaneous localization and mapping (SLAM) algorithm that can be applied to autonomous inspection of underwater structures, such as ship hulls, dams, and marine structures. Considering that visual features on the surface of typical underwater structures are not uniformly distributed, the proposed visual SLAM algorithm includes an intra-image analysis scheme that evaluates whether each image obtained from the surface is informative before extracting the features. By using only potentially effective images for feature-based image registration, the computational efficiency of the visual SLAM can be greatly improved, compared with the conventional exhaustive approach. Experimental results using a hover-capable unmanned underwater vehicle verify the practical feasibility and performance of the proposed methodology.