Interactive 3D landscape visualization: improved realism through use of remote sensing data and geoinformation

We describe strategies to generate interactive real-time landscape visualizations using remote sensing data and additional geoinformation. The goal is to achieve highly realistic and geoscientifically correct visualizations of real landscapes. To minimize costs, the modeling process should be automated as much as possible. The resulting data structure has to allow interactive real-time walk or fly-throughs. The terrain surface is represented by digital elevation models (DEM) and geospecific textures derived from remote sensing data of the Thematic Mapper (TM) sensor on the Landsat satellites and data from the high resolution visible sensor (HRV) on the French SPOT satellite. Additional remote sensing and geographic datasets are used to extract visually relevant features (e.g. distribution, height and structure of the vegetation) of existing landscapes. To automate the modeling process, we present a strategy which uses templates describing the transformation of geoinformation into computer graphics structures. As an illustrative example, a database covering the territory of Switzerland has been created.