A new method for automatic calibration of an active vision system in dynamic applications

In this paper, we talk about a technique for dynamic calibration of the relative pose in our structured light system with possibly varying intrinsic parameters. Assuming that there is an arbitrary planar structure in the scene, we suggest a two-step method for dynamic calibrating the relative pose between the camera and the projector. First, the translation vector is computed geometrically from the plane-induced parallax. Then the rotation matrix can be obtained using the plane-based Homography. We also show that some of the intrinsic parameters of the camera, e.g. the focal length and aspect ratio, can be calibrated in this system without any further constraints. Finally, some experiments have been carried out to validate this method. ©2006 IEEE.