A stereo vision-based augmented reality system with an inertial sensor

This paper proposes a robust registration method with stereo cameras and an inertial sensor for augmented reality. The geometric registration is one of the most important problems because virtual objects should be superimposed on the right place as if they really exist in the real world. The vision-based registration is achieved by estimating a position and an orientation of camera(s) by tracking markers in the real world. The proposed method realizes a robust tracking of markers using stereo images and a camera orientation obtained by the inertial sensor.