Camera Calibration Technique Based On Planar Mirror

With the broad application of high-precision computer vision systems and the acquisition of 3D features information requested by gradually increasing precision, camera calibration has become the basic work and the important issues to first try to resolve. Camera calibration is a major field in computer vision and an indispensable step to obtain 3D geometric information from 2D images.Camera model must first be established before camera is calibrated. In this paper, mathematical models of four parameters, five parameters and non-linear are established on the basis of projective theory.In this paper, a new method of camera calibration for intrinsic parameters is proposed. The method is based on a planar mirror and does not need any other calibration object. The intrinsic camera parameters can be linearly calibrated by using the reflection property of planar mirror and the theories such as vanishing points and vanishing lines and by controlling the camera or the mirror to undergo not less than 2 translations —there should be rotation of camera between any 2 translations. On basis of obtaining a vanishing point of normal mirror and a vanishing line of direction of mirror, two solution of intrinsic parameters are introduced, one of which is only used to 4 parameters model of camera, but the other can be used to 4 or 5 parameters model. Both computer simulation and real data have been used to test the proposed technique and the results show that the proposed method can quickly and conveniently calibrate the camera and is correct, and workable.In addition, a method of calibration of camera image center and scale factor is proposed on basis of introducing these intrinsic parameters of camera in detail. This method is based on a planar mirror and is realized by rotating camera around its optical axis. According to the nature of plane mirror, the translation of camera will not affect the position of optical center in the images, so the rotation of the camera around optical axis is easier to achieve. The virtue images of optical center in themirror make up of a circle. With only radial distortions, a circle remains a circle in the pixel coordinates system. Experiments with simulated data as well as with real images show that the proposed method has high precision and good stability.