The hybrid outdoor tracking extension for the daylight blocker display

From UMPCs to SmartPhones, we witness the emergence of highly integrated mobile computing platforms which boast higher performance than any of their preceeding systems. However, due to the equally growing demand for ever more complex applications such as mixed reality applications for outdoor scenarios, the need for an efficient distribution of resources for the different application tasks remains. In particular pose estimation in outdoor environments still presents a major ongoing challenge. Since mobile platforms have limited performance, the best approach for real-time pose estimation is using sensor fusion combining optical and inertial sensors. However different algorithms using different sensors require varying amounts of processing power (e.g. there are peaks when processing key frames for feature point detection). In this paper we introduce a novel sensor fusion pose estimation approach which we combine with the first compact daylight blocking optical stereo see-through display for mixed reality we presented a year ago [Santos et al. 2008a] [Santos et al. 2008b]. Through two new feature matching algorithms and appropriate sequencing of tracking algorithms using different sensors we attempt to achieve time-constant tracking update rates while keeping efforts for pose estimation at a fixed share of the overall available computing performance. By doing so we are able to guarantee the main mixed reality application a fixed share of the remaining computing performance on the mobile platform while preserving high tracking stability.