Dissertation Defense

Place Recognition and Localization for Multi-Model Underwater Navigation with Vision and Acoustic Sensors

Jie Li


Place recognition and localization are important topics in both robotic navigation and computer vision. This problem is more challenging in the underwater environment due to factors including long-term changes in the physical appearance of features in the aqueous environment attributable to biofouling and the natural growth, low density of salient features and low visibility in a turbid environment.

In this thesis, we propose place recognition and localization approaches for underwater navigation systems and address the challenges present in extreme underwater environments when using both optical and acoustic modalities. In particular, we report on a place recognition approach that leverages optical cameras in the presence of dramatic appearance changes, a real-time localization technique using Forward-looking Sonar imagery, and a novel approach to explicitly modeling Forward-looking Sonar image features in a SLAM systems with other sensor modalities.

All the proposed contributions are evaluated with real-data collected for ship hull inspection. The culmination of these contributions is a system capable of performing underwater SLAM with both optical and acoustic imagery gathered across years under challenging imaging conditions.

Sponsored by

Ryan Eustice and Matthew Johnson- Roberson