Advancing Underwater Vision for 3D (AUV3D)

Safe and efficient construction, operation and decommissioning of subsea
assets is critically important to UK and worldwide energy production. This is
particularly true for offshore renewable energy where cost efficiencies are
necessary to deliver clean power power that is cost competitive with other
low carbon systems and at an affordable scale. From construction to
decommissioning, underwater survey provides the data to monitor condition,
predict asset life and ensure the environment is protected. We aim to deliver
a step change in efficiency and safety by delivering live, dense, 3D point
cloud data from small, Remotely Operated Underwater Vehicles. This will
enable smaller vessels to be used with fewer crew, no divers, and removing
the need to put people at risk. Compared to traditional visual survey, 3D
data allows accurate measurement and repeatable, reliable metrics for asset
condition monitoring. Ultimately, live 3D enables accurate navigation for fully
autonomous underwater vehicles reducing manpower and increasing
efficiency yet further. Quality 3D visual data is also a prerequisite to applying
artificial intelligence and deep learning solutions to 3D images thereby
enabling greater autonomy and reliably repeatable measurements.

The key objective of the AUV3D project is to prototype and demonstrate the
feasibility of a high-quality underwater, intelligent, stereo camera system.
This system will enable innovative, real-time processing of underwater 3D
from ROV video survey. To do this we will exploit recent advances in both
camera technology and embedded GPU computing, and together these
technologies enable Artificial Intelligence to be used to accurately to assess
underwater 3D scenes.

By demonstrating the feasibility of the software and hardware necessary to
produce live 3D data from cameras in the challenging and extreme subsea
environment we enable the development of a complete vision based
underwater Robotic Artificial Intelligence (RAI) survey solution. This has the
potential to create small, capable, intelligent autonomous vehicles and allow
more efficient survey with fewer people in harm's way.