Estimation of camera calibration uncertainty using LIDAR data

This paper presents an analysis of uncertainty in the calibration of a network of cameras. A global 3D scene model, acquired with a LIDAR scanner, allows calibrating cameras with non overlapping fields of view by means of the DLT-Lines algorithm. Once the projection matrix is computed for each camera, error sources are propagated to compute estimates of each camera position uncertainty. We validate the consistency of the uncertainty analysis with Monte-Carlo simulations, and apply the technique in a real camera network. This allows to evaluate the accuracy of DLT-Lines in real settings.