SYNERGY OF IMAGE AND DIGITAL ELEVATION MODELS (DEMS) INFORMATIONFOR VIRTUAL REALITY

摘要

In the framework of 3D visualization and real-timernrendering of large remote sensing image databases,rnseveral signal processing techniques are presented andrnevaluated to filter/enhance SAR Digital ElevationrnModels (DEMs). Through the SRTM DEM, the interestrnof InSAR data for such applications is illustrated. A nonrnstationary bayesian filter is presented to remove noisernand small artefacts which pervade the SAR DEM whilernpreserving structures and information content. Resultsrnobtained are very good, nevertheless large artefactsrncannot be filtered and some artefacts remain.rnTherefore, image information have to be inserted tornproduce more realistic views. This second step is donernby using a segmentation algorithm on the image data.rnBy a topology analysis, the extracted objects arernclassified/stored in a tree structure to describe therntopologic relations between the objects and reflect theirrninterdependencies. An interactive learning procedure isrndone through a Graphical User Interface to link thernsignal classes to the semantic ones, I.e. to includernhuman knowledge in the system. The selectedrninformation in form of objets are merged/fused in thernDEM by assigning regularisation constraints.