Date of Award

12-2015

Document Type

Master Thesis

Degree Name

Masters of Science (Research)

Department

Computing

First Advisor

Dr Robert Sheehy

Second Advisor

Dr Pat Doody

Abstract

Landing a helicopter on a ship in high seas can be a dangerous endeavour. This thesis proposes to examine the possible uses of Artificial Neural Networks (A.N.N.) in the aiding and/or the landing of an Unmanned Aerial Vehicle (U.A.V.). It proposes that this procedure can be segregated into three distinct phases. The data for the A.N.N. training and testing sets is generated through simulation in the Unity cross-platform game engine. Phase 1 is intended to convert video images from an on-board camera to a set of numeric outputs suitable for use in Phase 2. Phase 2 estimates the current relative orientation and distance of the camera to the platform. Phase 3 determines when a future landing window may occur.

Phase 1 takes live video feed of the helipad and a corner recognition algorithm is applied to images captured from it. The co-ordinates of the vertices have been measured to within +/- 0.3%. Phase 2 required normalized points representing positions on a screen of specific elements on the landing pad. Orientation has been determined to within 3.60 and distance correct to within 2%. Phase 3 takes the orientations calculated from Phase 2 over a given time period and predicts whether at a specific, fixed, time into the future landing would be possible based on a maximum deviation of the orientation from the ideal.

Access Level

info:eu-repo/semantics/openAccess

Coverage

July 2024

Download

Share

COinS