Logo image
Localisation for Augmented Reality at Sports Events
Graduate Thesis/Dissertation   Open access

Localisation for Augmented Reality at Sports Events

Patrick Skinner
Master of Science - MSc, University of Otago
University of Otago
2021
Handle:
https://hdl.handle.net/10523/12145

Abstract

Sports Augmented Reality Camera Localisation AR Mobile Computer Vision
The use of computer generated graphics and visualisations has become increasingly common in professional sports broadcasting over the past two decades. Overlaid graphics are designed to convey information about the game to the spectator in a way that is contextualised spatially, such as virtual offside lines being overlaid on soccer footage. We aim to take these graphics, that are only seen in professional broadcasting, and provide them to on-site sports spectators using an augmented reality mobile application. Allowing fans in the stadium to view the same pitch-aligned visualisations as those watching at home, all rendered on their phone in real time. In sports broadcasting the alignment of these graphics with the environment is often achieved by using professional cameras equipped with high accuracy sensors or elaborate manual calibration techniques to measure the broadcasting camera's position and orientation. These measurements allow the graphics to be accurately matched to the camera view. The goal of our research was to design a method of camera localisation that will allow spectators to enjoy accurately aligned augmented reality content with minimal prior calibration and without a complicated manual initialisation process. We investigated whether line pitch markings could be used to estimate an on-site user's position and orientation with sufficient accuracy to align augmented reality visualisations with the pitch. Our evaluation found that our approach performed with sufficient speed and accuracy to form the basis of a mobile augmented reality application for on-site spectators.
pdf
Localisation_for_Augmented_Reality_at_Sports_Events_r.pdfDownloadView
Thesis Open Access All Rights Reserved

Metrics

90 File views/ downloads
167 Record Views

Details

Logo image