The Monitoring and Restoration Technology of Geolocation Accuracy of High Spatial Resolution Remote Sensing Data

The technology of on-ground data processing from the Russian EO systems allowing geocoding images of the Earth's surface with a specified accuracy in case of an abnormal mode of the satellite measuring systems is considered. This technology is based on two-stage refinement of the rigorous model of the image geocoding using the coordinates of characteristic objects recognized on the analyzed and reference images. At the first stage, significant errors in the geolocation using the reference image of the Earth's surface obtained from the Landsat-8 satellite images are eliminated. At the second stage, the parameters of survey model are refined using the coordinates of individual reference objects geocoded with an accuracy of 5-10 m. The features of building of the reference information databank are considered. The algorithm SURF implemented on GPU is proposed to use for rapid and reliable identification of the same name objects at the first stage of the model refinement. The filtering of falsely identified of the same objects is carried out using statistical analysis. The examples of practical implementation of the developed technology for processing of the data from the satellites Kanopus-V and Resurs-P are presented.