Fast vision-based pose estimation iterative algorithm

Determining the relative pose between the object and the camera using 2D to 3D point correspondences is one of key problems in photogrammetry and the accuracy and speed of its computation is very important. A fast iterative algorithm to solve the problem is proposed in this article. The algorithm formulates objective function to minimize the object-space collinearity error. Then the optimal translation vector is computed in terms of rotation matrix. Finally the optimal rotation matrix is estimated by applying a new absolute orientation method which is also proposed in this article. The new absolute orientation method uses the Fobenius norm, determinant and adjoint of matrix to formulate the closed-form optimal estimation of the rotation matrix, instead of the popular SVD-based method. Empirical results suggest the proposed algorithm performs the same accuracy and noise-resistance with extremely fast speed compared with the current employed algorithms.