Comperative analysis of different geometric correction methods for very high resolution pleiades images

Abstract

This research aims to analyze geometric correction accuracy of empirical Rational Function Model (RFM), empirical Rational Polynomial Coefficient (RPC) Refinement Model and physical Toutin Model applied to Very High Resolution (VHR) Pleiades satellite images. Two different pilot regions located in Turkey with different topographic characteristics were selected and analysis were conducted for these regions using ground control points obtained from 1:5000 scale aerial ortho-photos. Further analysis was conducted to analyze positional accuracy of resulting images using independent check points collected from aerial ortho-photos and Google Earth separately. The Advanced Spaceborne Thermal Emission and Reflectance Radiometer Global Digital Elevation Model (ASTER GDEM) was used for all different geometric correction processes for both regions. The accuracy of each geometric correction model was presented by root mean square error (RMSE) metric. Results showed that first and second order RPC refinement models together with Toutin Model provided highly accurate results for both regions with RMSE between 3–6 meters while rational function model with three and four RPC's provided comparatively lower horizontal accuracy. Applying the rational function model with the 20 RPC's, generated from ground control points using PCI Geomatica, reduced the RMSE comparatively to a better level but increased the number of ground control points required to perform geometric correction with this model. Moreover, results indicated a possible effect of terrain structure and land use /cover types on the accuracy of geometric correction.