Robust pose estimation for outdoor mixed reality with sensor fusion

Abstract

We present a sensor fusion based technique for outdoor augmented reality system for mobile devices using GPS, gyroscope, and geo-referenced 3D models of the urban environment. Geo-spatial interaction not only provides overlays of the existing environment but compliments with other data such as location-specific photos, videos and other information from different time periods enhancing the overall user experience of augmented reality. To provide robust pose estimation of the camera relative to the world coordinates, firstly, GPS and gyroscope are used to obtain the rough estimation. Secondly, model based silhouette tracking and sensor fusion approach is used to refine the rough estimation and to provide seamless media rich augmentation of 3D textured models. © 2009 Springer Berlin Heidelberg.