Improvement of the PeruSAT-1 Satellite Image Orthorectification Using the Optimization Algorithm

The rational polynomial coefficients (RPCs) are constants available in the metadata of satellite images and are used for orthorectification. However, the accuracy of an orthorectified image can be improved by modifying the original RPCs through an optimization algorithm (OA) based on the cost function (CF) and gradient descent method (GD). The OA is applied to the rational function model (RFM) to modify the original RPCs to obtain new RPCs and improve image orthorectification. The analysis of results showed that the difference between the coordinates performed by differential GPS and the initial coordinates obtained from the original RPCs has an average error of 19.45 m. In comparison, the coordinates obtained from the modified RPCs have an error of around 2.25 m. Therefore, this methodology reduces error by 11.6% and improves orthorectification. Applying this new proposed methodology will allow us to improve the geometric accuracy of the satellite images and reduce the errors generated by the original RPCs.