Comparison of positioning accuracy of a rigorous sensor model and two rational function models for weak stereo geometry

So far, two sensor models for the geometric modeling of satellite imagery have been studied and compared: a rigorous sensor model (RSM) and a rational function model (RFM). Even though it was concluded that the RFM could replace the RSM, this paper points out that the previous conclusions were drawn only for a strong geometry because of the conventional use of single-sensor stereo and that they may not apply to the weak geometry of dual-sensor stereo pairs. This work highlights that dual-sensor stereo often creates a weak geometry and that for such weak geometry, accuracy differences may occur between the RSM and the RFM, and also between the RFMs with different bias correction methods. The positioning accuracy of the three sensor models, RSM, RFM using second-order polynomials model, and RFM using an affine model, were compared on various geometries, using pairs from every conceivable combination of two QuickBird and IKONOS as well as four KOMPSAT-2 images covering the same area. Our results showed that the three sensor models differed slightly owing to the strong geometry. However, for the weak geometry, the RSM or second-order RFM performed better than RFM with an affine model, resulting in increase in the difference between the accuracies of the sensor models. This implies that the physically weak geometry of a satellite stereo may require a rigorous or high-order model for a more accurate geo-positioning.