Identifying structural anomalies in image reconstructions of underwater ship hulls

This paper reports on an algorithm enabling an autonomous underwater vehicle (AUV) to localize into a 3D computer aided design (CAD) model of a ship hull in situ using an optical camera and Doppler velocity log (DVL). The precision of our localization algorithm allows the identification of structural deviations between 3D structure inferred from bundle-adjusted camera imagery and the CAD model. These structural deviations are clustered into shapes, which allow us to fuse camera-derived structure into a CAD-derived 3D mesh. This augmented CAD model can be used within a 3D photomosaicing pipeline, providing a visually intuitive 3D reconstruction of the ship hull. We evaluate our algorithm on the Bluefin Robotics Hovering Autonomous Underwater Vehicle (HAUV) surveying the SS Curtiss, and provide a 3D reconstruction that fuses the CAD mesh with 3D information corresponding to underwater structure, such as biofouling.