Wide Area Multiple Camera Calibration and Estimation of Radial Distortion

The calibration of cameras distributed in a wide area is a challenging task because it is impossible to use reference objects visible to all cameras and because wide field-of-view cameras suffer under radial distortion. The present work proposes the first algorithm in the literature for radial distortion estimation from multiple views without involving non-linear minimization. The correspondences between views are obtained by deliberately moving an LED in thousands of unknown positions in front of the cameras. Then both projection matrices and radial distortion parameters are simultaneously computed using a factorization approach. The algorithm is based on the application of two subspace approximation steps. At these steps, the estimated approximate solution for a matrix can be projected to the manifold of the parameter space by adjusting the singular values. It is remarkable, that our system does not involve a single non-linear minimization or outlier treatment and still produces accurate results which have been tested in a multi-camera reconstruction algorithm. In addition to real imagery results, we have analyzed the behavior of the algorithm in simulations.