Comparison of two physical sensor models for satellite images: Position–Rotation model and Orbit–Attitude model

The main objective of this paper is to compare two types of physical sensor models of linear pushbroom satellite images: one that uses position and rotation angles as model parameters and one that uses orbit and attitude angles as model parameters. Comparison is carried out by two accuracy measures: the accuracy of bundle adjustments and the accuracy of estimating exterior orientation parameters. The first measure has been used widely to indicate the mapping accuracy of sensor models. It is argued that the second measure is also important for certain applications. The two types were implemented with different sets of unknown parameters and tested with two KOMPSAT‐1 Earth Observing Camera (EOC) scenes and GPS‐derived control points. In terms of the first measure the two models produced similar results whereas in terms of the second measure the one based on orbit and attitude outperformed the other. It seems better to use this model if one wishes to retrieve satellite orbit or attitude through bundle adjustments.