Application of Bayesian Statistics in Photogrammetric Bundle Adjustment

Abstract

The basic mathematical principle for bundle adjustment (BA) in photogrammetry is the Gauss-Markov Theorem within the framework of classical statistical inference. In the present article we try to show how Bayesian statistics can be applied in this field, leading to a so-called Bayesian bundle adjustment. The rigorous implementation of the Bayesian approach is derived and a comparison with the traditional BA both in theory and practice is accomplished. The empirical test results show that the Bayesian approach achieves almost the same accuracy as the conventional one, but with advantages as well as some difficulties to be discussed at the end of this paper. © 2011 Published by Elsevier Ltd.