Monocular Vision-Based SLAM for Autonomous Underwater Vehicles: An Application to Autonomous Ship Hull Inspection

Authors: Ryan Eustice.

Time: 15:35-16:00

This talk discusses the application of vision-based SLAM to navigation and mapping in underwater marine environments. In particular, we will look at the the problem of autonomous ship hull inspection by an underwater vehicle. The goal of this work is to automatically map, and navigate an autonomous underwater vehicle with respect to the underwater surface area of a ship-hull for foreign object detection and maintenance inspection. We employ a pose-graph SLAM representation and use an extended information filter for inference. For perception, we use a calibrated monocular camera. A combination of SIFT and Harris interest features are used within a pairwise image registration framework to extract pose-constraints. Because the ship-hull surface can range from being locally planar to highly 3D structured, we use a model selection framework to appropriately choose a homography versus essential matrix motion constraint. This allows the image registration engine to exploit geometry information at the early stages of estimation, which results in better navigation and structure reconstruction. Preliminary results are reported for mapping a relatively small 1,300 image data set covering a 30 m by 8 m section of the hull of a USS aircraft carrier. The result validates the algorithm's potential to provide in-situ navigation in the underwater environment, building a texture-mapped 3D model of the ship hull as a byproduct.