Analysis of head pose accuracy in augmented reality

摘要

A method is developed to analyze the accuracy of the relativenhead-to-object position and orientation (pose) in augmented realitynsystems with head-mounted displays. From probabilistic estimates of thenerrors in optical tracking sensors, the uncertainty in head-to-objectnpose can be computed in the form of a covariance matrix. The positionalnuncertainty can be visualized as a 3D ellipsoid. One useful benefit ofnhaving an explicit representation of uncertainty is that we can fusensensor data from a combination of fixed and head-mounted sensors innorder to improve the overall registration accuracy. The method wasnapplied to the analysis of an experimental augmented reality system,nincorporating an optical see-through head-mounted display, anhead-mounted CCD camera, and a fixed optical tracking sensor. Thenuncertainty of the pose of a movable object with respect to thenhead-mounted display was analyzed. By using both fixed and head mountednsensors, we produced a pose estimate that is significantly more accuratenthan that produced by either sensor acting alone