Real Time Tracking Using Stereo Vision for Augmented Reality

Correct tracking of the user position is crucial for an Augmented Reality (AR) application. For this reason, tracking became a very important research subject in the field of real-time AR. The most common solution to implement an AR system is based on the use of an Head Mounted Display (HMD) and two cameras for the stereoscopic video see-through visualization. Almost all the systems presented in literature employ one camera to obtain a marker based video-tracking. This technique is widespread because it offers sufficient precision and accuracy. This paper describes an innovative tracking solution based on a stereo vision approach which aims to use the information coming from both cameras in order to improve tracking accuracy. The proposed solution is also able to switch itself into a standard single camera visual tracking when the stereo pair correspondence is missing. Further, a comparison of the performance of both mono and stereo tracking is reported.