TerraSAR-X Big Data: The DLR Solutions for Data Mining

Abstract

Earth Observation (EO) data has increased significantly over the last decades with satellite and airborne remote sensing sensors collecting and transmitting to Earth receiving stations several terabytes of data a day. This is also the case with TerraSAR-X and TanDEM-X missions. This data acquisition rate is a major challenge to the existing data exploitation, dissemination used approaches for extracting information from these images. And, with plans for more EO systems the challenge is increasingly going to be how to enlarge the usability of the millions of images being stored in archives to a broader group of users. The presentation goal is two folded: it overviews the latest results obtained at DLR for very high resolution SAR image content extraction and presents the latest formalism and solutions to solve the challenge of accessing Big Data repositories. The methods of information retrieval are introduced and discussed. Particularly, the latest concepts of Image Information Mining for enhancing the performance of the PDGS are explained and exemplified. The methods have to be common to all application domains without the weakening effect of specializing it for specific, particular applications fields. The objective is re-formulating the definition of the “relevant information” in relation to the notions of “data content” and “context”. The presentation is introducing: the latest models and algorithms for SAR and InSAR feature extraction, the alternative of using the methods of Algorithmic Information Theory for parameter free SAR data interpretation, kernel methods for data classification, methods for SAR Image Time Series analysis, Visual Data Mining modules, and the overall new architecture for the PDGS including KDD and Data Mining functions.