Theoretical Uncertainty Evaluation of Stereo Reconstruction

This paper deals with the problem of evaluating the uncertainties associated with linear camera model based stereo reconstruction using the law of propagation of uncertainty. The procedure of stereo reconstruction involves two consecutive stages: calibrating camera models and reconstructing a 3D point from its image projections. The output quantities of the first stage, the parameters of camera models, constitute a part of the input quantities of the second stage. The analytical expressions for uncertainty propagation during reconstructing a 3D point are proposed, and the results of an experiment with synthetic data are also presented. The experiment results show that there is at least one significant digit in the evaluated uncertainties associated with the reconstructed coordinates of a 3D point.