Large area mapping at high resolution underwater continues to be constrained by the mismatch between available navigation as compared to sensor accuracy. In this paper we present advances that exploit consistency and redundancy within local sensor measurements to build high resolution optical and acoustic maps that are a consistent representation of the environment.

We present our work in the context of real world data acquired using Autonomous Underwater Vehicles (AUVs) and Remotely Operated Vehicles (ROVs) working in diverse applications including shallow water coral reef surveys with the Seabed AUV, a forensic survey of the RMS Titanic in the North Atlantic at a depth of 4100 meters using the Hercules ROV and a survey of the TAG hydrothermal vent area in the mid-Atlantic at a depth of 2600m using the Jason II ROV.

Specifically we focus on the related problems of Structure from Motion and Visually Based Navigation from underwater optical imagery assuming pose instrumented calibrated cameras. We present general wide baseline solutions for these problems based on the extension of techniques from the SLAM, photogrammetric and the computer vision communities. We also examine how such techniques can be extended for the very different sensing modality and scale associated with multi-beam bathymetric mapping. For both the optical and acoustic mapping cases we also show how the consistency in mapping can be used not only for better mapping but also to refine navigation estimates.