Using Visual Information from the Local Environment to Improved Tracking and Position Estimation for Remotely Operated Vehicles used for Underwater Inspections in Reactor Vessels

Speaker:
Bolander, Martin; WesDyne; Sweden

Authors:
Bolander, M.; WesDyne; Sweden
Lee, T.E.; Robotics Institute, Carnegie Mellon University; USA
Michael, N.; Robotics Institute, Carnegie Mellon University; USA
Petrosky, L.J.; Westinghouse; Sweden

ID: ECNDT-0040-2018
Session: Robotic and automation 2
Room: H2
Date: 2018-06-14
Time: 09:40 - 10:00

When using Remotely Operated Vehicle (ROV) for under water inspections global and local positioning feedback is a challenge and affects inspection accuracy and repeatability as well as inspection time.

This paper discusses a visual based tracking system used to locate and track a ROV that performs underwater inspections in the internals of nuclear reactor vessels. The system utilizes an overhead pan-tilt-zoom (PTZ) camera to view the ROV and reactor internals below it. One unique feature of system is that it uses prior knowledge of the reactor vessel geometry to correlate extracted visual features with the true position of the item in the real world. This allows the visual field of the camera to be calibrated to a high accuracy. Visual features extracted from the body of the ROV can then be correlated to in the field of view to yield a very accurate estimate of the position of the ROV. Full scale test in various facilities have confirmed this technique to highly effective for obtaining accurate position of the ROV and thereby remove the concern about positional uncertainty from the inspection campaign.