An effective camera calibration method

Proposes a method for camera calibration that could be used in stereo systems subject to slight change in any camera parameter, as well as in stereo head navigation. A pinhole camera model plus linear and radial distortion is considered. An extended Kalman filter (EKF) is used to recover pinhole parameters and a Newton Raphson (NR) algorithm is used to obtain distortion terms. With the pinhole model the position and orientation of the camera is immediately recovered because the extrinsic and intrinsic parameters have been explicitly calculated, this is necessary for the modeling of the camera in a 3D environment. The only strong requirement for the proper convergence of the calibration algorithm is a good initial approximation to the solution. The method allows us to recalibrate the system easily if a small change has taken place. In that case, the initial approximation required for the EKF is supplied by the old parameters of the camera, after which the convergence is very quickly achieved. The on-going work is related to the application of this algorithm to the calibration of the cameras for stereo head navigation, allowing a 3D description of the environment surrounding the camera movements. The algorithm has been tested in the 3D reconstruction of real objects through a path described by a stereo camera system. Numerical results have been compared with a classical least square error algorithm. Reported errors by EKF are similar of those by LSE