A Generic Bundle Adjustment Methodology for Indirect RPC Model Refinement of Satellite Imagery

This article describes a generic bundle adjustment methodology for multi-view stereo pipelines 3D reconstruction from high-resolution optical satellite imagery. The Rational Polynomial Camera (RPC) model of each input view is refined by composing it with rotation that compensates errors related to the attitude angles encoding orientation. A set tie points, derived feature tracks detected across images, used find optimal corrective rotations minimization reprojection error. We evaluate performance our method using time series SkySat acquisitions over two different areas interest. As an internal evaluation metric we compute standard deviation between corresponding height estimations pairs, before and after correction. Lastly, results are also compared alternative solution eludes any explicit correction camera models directly registering independent dense surface models.