A Fast Closed-Form Solution for Single-View Pose Determination

Abstract

In this paper, we present an extremely efficient closed-form method which yields results comparable to iterative methods, at a minor fraction of the computational cost, and without camera calibration. The algorithm solves the single view pose determination of a square marker with a straightforward, single-stage calculation. Further, the cost of the method is independent of image resolution. Our technique has two major distinctive points: an assumption of near-constant depth for the marker, which allows the splitting of the projection matrix into a perspective division and a rotation matrix, and a heuristic method using the perspective effect to disambiguate similar marker orientations.