An adaptive estimator for registration in augmented reality

Abstract

In augmented reality (AR) systems using head-mounted displays (HMD), it is important to accurately sense the position and orientation (pose) of the user's head with respect to the world, in order that graphical overlays are drawn correctly aligned with real world objects. It is desired to maintain registration dynamically (while the person is moving their head) so that the graphical objects will not appear to lag behind, or swim around, the corresponding real objects. We present an adaptive method for achieving dynamic registration which accounts for variations in the magnitude of the users head motion, based on a multiple model approach. This approach uses the extended Kalman filter to smooth sensor data and estimate position and orientation.