Estimation de pose omnidirectionnelle dans un contexte de réalité augmentée
Authors: | Poirier, Stéphane |
Advisor: | Marchand, Mario |
Abstract: | Camera pose estimation is a fundamental problem of augmented reality, and enables registration of a model to the reality. An accurate estimate of the pose is often critical in infrastructure engineering. Omnidirectional images cover a larger field of view than planar images commonly used in AR. This property can be beneficial to pose estimation. However, no existing work present results clearly showing accuracy gains. Our objective is therefore to quantify the accuracy of omnidirectional pose estimation and test it in practice. We propose a pose estimation method for omnidirectional images and have measured its accuracy using automated simulations. Our results show that the large field of view of omnidirectional images increases pose accuracy, compared to poses from planar images. We also tested our method in practice, using data from real environments and discuss challenges and limitations to its use in practice. |
Document Type: | Mémoire de maîtrise |
Issue Date: | 2012 |
Open Access Date: | 18 April 2018 |
Permalink: | http://hdl.handle.net/20.500.11794/23013 |
Grantor: | Université Laval |
Collection: | Thèses et mémoires |
All documents in CorpusUL are protected by Copyright Act of Canada.