A hybrid pose tracking approach for handheld augmented reality

摘要: With the rapid advances in mobile computing, handheld Augmented Reality draws increasing attention. Pose tracking of devices is fundamental importance to register virtual information with real world and still a crucial challenge. In this paper, we present low-cost, accurate robust approach combining fiducial inertial sensors for pose tracking. Two LEDs are used as markers indicate position device. They detected by an adaptive thresholding method which illumination changes, then tracked Kalman filter. By inclination provided on-device accelerometer, 6 degree-of-freedom (DoF) estimated. Handheld freed from computer vision processing, leaving most computing power available applications. When one LED occluded, system able recover 6-DoF pose. Performance evaluation proposed carried out comparing ground truth data generated state-of-the-art commercial motion OptiTrack. Experimental results show that has achieved accuracy 1.77 cm estimation 4.15 degrees orientation estimation.