摘要

Many 3D reconstruction tasks need implicit camera parameters with high accuracy. To this aim, we propose a novel calibration method using a translational planar object, which is cheaper and flexible. The calibration instruments are installed to collect both the calibration image and the correction image. The world coordinates of control points are assigned by its' spatial distribution in the designated world coordinate. We detect control points in the calibration image and then the initial camera parameters are calculated. According to the reprojection errors obtained by the last camera parameters, the location of control points are adjusted and refined until convergence. The skew bias, which caused by the finite precision of assembly of the planar object, are estimated with the information of the correction image. Then the world coordinates of control points are corrected with the skew factors and utilized to optimize the final camera parameters. Many experiments based on real images denote that the proposed method is more accurate than Heikkila's or Zhang's method. The camera parameters obtained by our method can be applied to 3D reconstruction directly and effectively.

  • 出版日期2012

全文