Monocular pose estimation algorithm based on the bundle adjustment method

摘要

To determine the relative pose between an object and a single camera using correspondences between 3D feature points of the object and their corresponding 2D projections in the image, this paper proposes a target's pose measurement algorithm based on the bundle adjustment method. This iterative algorithm can be divided into two steps: first, a reliable initial pose is computed by using only three non-collinear points; second, the optimal rotation and translation matrix of the target is estimated based on the bundle adjustment method. Experiments on simulated data and real data show that this method can get high precision with its rotation angles error within 20 arc-second and the translation error within 20 micron in ideal occasions.