Optical Tracking using Line Pencil Fiducials

Abstract

In this paper, a new pattern based optical tracking method is presented for the recognition and pose estimation of input devices for virtual or augmented reality environments. The method is based on pencils of linefiducials, which reduces occlusion problems and allows for single camera pattern recognition and orientation estimation. Pattern recognition is accomplished using a projective invariant property of line pencils: the cross ratio. Orientation is derived from single camera line-plane correspondences, and position estimation is achieved using multiple cameras. The method is evaluated against a related point based tracking approach. Results show our method has lower latency and comparable accuracy, and is less sensitive to occlusion.