Estimation of Camera Projection Matrix Using Linear Matrix Inequalities
Yoshimichi Ito, Yuta Oda · 2016
This paper proposes some methods for estimating camera projection matrix from given 3D coordinate vectors of feature points and 2D coordinate vectors of the projected feature points on the image plane. It is well-known that the problem is formulated as the L2minimization problem of the sum of reprojection errors, which is very hard to solve because the sum of reprojection errors is non-convex with respect to camera projection matrix. We show that the original problem is reduced to a tractable problem with linear matrix inequality (LMI) constraints by utilizing a property of a trace minimization problem. The proposed methods are easily implemented by MATLAB using tools for LMI such as SeDuMi, SDPT3, and YALMIP. The proposed methods are evaluated through numerical examples.