A new approach of geo-rectification for time series satellite data based on a graph-network

Abstract

Earth observation is indisputably one of the most important data sources for current synoptic geo-data in diverse environmental applications. Nevertheless, obtained thematic information by means of Earth observation is only as good as the quality of the pre-processed data. Important pre-processing steps are e.g. data usability assessment, geo-referencing and atmospheric correction. While data usability assessment or atmospheric correction expects radiometric corrected data for a thematic interpretation, the accuracy of geo-location of interpreted data is ensured by geo-referencing. This applies especially to multi-temporal analyses of environmental processes. In this case, precise spatial allocation of data represents a prerequisite for a correct interpretation of process dynamics. The paper is dealed with a new geo-referencing algorithm. Corresponding graph-networks in reference data as well as in remote sensing data which are based on virtual object points (e.g. centroids) will be used for geo-rectification.