Correlation Analysis of Camera Self-Calibration in Close Range Photogrammetry

While the Brown self-calibration model has been widely used in close range photogrammetry for over 40years, the negative effects of the well-known high correlations between parameters are not clear. This paper presents a novel view and study of these correlations, which are shown to be inherent in the Brown model due to its polynomial nature; they occur exactly as with additional parameters in aerial photogrammetry. Although it has high correlations with the decentring distortion parameters, the principal point can be located precisely (up to a few pixels) in self-calibration with an appropriate image configuration. An alternative model of the in-plane distortion is proposed, which helps to reduce the correlation with principal distance.