Pose estimation from visual measurements using the epipolar constraint

This paper presents a new approach for estimating the pose, or position and orientation, of a target object relative to a camera. The approach assumes that the object geometry is known and that identifiable feature points on the object are tracked. Using the epipolar constraint, the problems of estimating orientation and position are decoupled. The orientation estimation problem is formulated and conditions under which the problem is well-posed are derived. An iterative numerical technique for solving the orientation estimation problem is presented, and the method is demonstrated in numerical simulations. A similar approach for simultaneously estimating position and orientation is also presented.