Multi-camera calibration based on coplanar projection of points

摘要

To solve the problem of multi-camera calibration, an algorithm is brought forward based on coplanar projection of points. A coordinate system is unified between world coordinates and projective plane ones. After feature patterns projected onto a planar screen and corners extracted from images, each camera is calibrated respectively, from which it can obtain internal parameters and relative poses among cameras. Initial values of spatial points are derived from projective transformation. Then it finds out corresponding points of each image pair by epipolar geometry constraint and calculates spatial points in terms of triangle principle again. Root mean square error between initial values and estimation values are calculated. Furthermore, it eliminates points until errors of all spatial points within threshold. The experimental analyses indicate that the algorithm can reduce computational complexity and manual intervention. Calibration accuracy can meet the need.