Evaluation of 3D Reconstruction Accuracy in the Case of Stereo Camera-Pose Configuration

This paper presents an empirical methodology to assess the quality of camera calibration and to check out the effects of image geometry in 3D photogrammetric reconstruction. The proposed methodology is applied to the simple stereo-pair configuration, but it is also suitable to be extended to larger image block geometries. The basic concept consists in the empirical evaluation of potential effects of biases in camera calibration within the use of a 3D test field. The multiple repetition of the standard camera calibration procedure in order to provide a sample of parameter values to be averaged, is demonstrated to guarantee a better determination of the whole set of calibration parameters. It is also demonstrated that, as the background theory supports, a longer baseline and a small convergence angle between camera stations may furnish the best results in term of accuracy.