Bibtex:Maidi07c
Un article de Wiki-evr@.
(Différences entre les versions)
| Version du 3 septembre 2010 à 10:38 (modifier) Maidi (Discuter | Contributions) ← Différence précédente |
Version actuelle (6 septembre 2010 à 12:24) (modifier) (défaire) Maidi (Discuter | Contributions) |
||
| Ligne 7 : | Ligne 7 : | ||
| address = {Évry}, | address = {Évry}, | ||
| month = {Novembre}, | month = {Novembre}, | ||
| - | pdf = { | + | pdf = {Thèse_Maidi.pdf}, |
| abstract = {One of the major challenges of augmented reality is the registration of virtual objects on real scene. Several localization techniques were proposed in litterature to improve registration. The hybrid tracking which combines various technologies and methods, seems to open a new way to overcome different problems related to accuracy and robustness. In this work, we developed a multimodal tracking architecture to handle targets occlusions in augmented reality applications. The multimodal approach combines different sensors and techniques which are associated according to the real environment conditions. This architecture consists of a registration module of coded targets based on a hybrid algorithm of pose estimation. However, this tracking algorithm based coded fiducials fails if the visual markers are not completely visible. Therefore, we extended this space-time method of localization by a module of feature points tracking and occlusions management. This step is based on the study of the tracking properties and on the adaptation and the management of various experimental conditions. Finally, a multi-sensors tracking part is included into our software design approach. This module of architecture is composed of a hybrid device which overcomes total targets occlusion problem. Experiments with the overall system and many evaluations showed the interest of the proposed multimodal approach for tracking and occlusion management in augmented reality. This work allowed to define a base of necessary criteria for an effective implementation of a robust and multisensors tracking architecture in augmented reality environments. | abstract = {One of the major challenges of augmented reality is the registration of virtual objects on real scene. Several localization techniques were proposed in litterature to improve registration. The hybrid tracking which combines various technologies and methods, seems to open a new way to overcome different problems related to accuracy and robustness. In this work, we developed a multimodal tracking architecture to handle targets occlusions in augmented reality applications. The multimodal approach combines different sensors and techniques which are associated according to the real environment conditions. This architecture consists of a registration module of coded targets based on a hybrid algorithm of pose estimation. However, this tracking algorithm based coded fiducials fails if the visual markers are not completely visible. Therefore, we extended this space-time method of localization by a module of feature points tracking and occlusions management. This step is based on the study of the tracking properties and on the adaptation and the management of various experimental conditions. Finally, a multi-sensors tracking part is included into our software design approach. This module of architecture is composed of a hybrid device which overcomes total targets occlusion problem. Experiments with the overall system and many evaluations showed the interest of the proposed multimodal approach for tracking and occlusion management in augmented reality. This work allowed to define a base of necessary criteria for an effective implementation of a robust and multisensors tracking architecture in augmented reality environments. | ||
Version actuelle
Madjid Maidi - Suivi Hybride en présence d'Occultations pour la Réalité Augmentée
Pdf