Methods for Real-time Visualization and Interaction with Landforms

Abstract

This thesis presents methods to enrich data modeling and analysis in the geoscience domain with a particular focus on geomorphological applications. First, a short overview of the relevant characteristics of the used remote sensing data and basics of its processing and visualization are provided. Then, two new methods for the visualization of vector-based maps on digital elevation models (DEMs) are presented. The first method uses a texture-based approach that generates a texture from the input maps at runtime taking into account the current viewpoint. In contrast to that, the second method utilizes the stencil buffer to create a mask in image space that is then used to render the map on top of the DEM. A particular challenge in this context is posed by the view-dependent level-of-detail representation of the terrain geometry.<br /> After suitable visualization methods for vector-based maps have been investigated, two landform mapping tools for the interactive generation of such maps are presented. The user can carry out the mapping directly on the textured digital elevation model and thus benefit from the 3D visualization of the relief. Additionally, semi-automatic image segmentation techniques are applied in order to reduce the amount of user interaction required and thus make the mapping process more efficient and convenient. The challenge in the adaption of the methods lies in the transfer of the algorithms to the quadtree representation of the data and in the application of out-of-core and hierarchical methods to ensure interactive performance. <br /> Although high-resolution remote sensing data are often available today, their effective resolution at steep slopes is rather low due to the oblique acquisition angle. For this reason, remote sensing data are suitable to only a limited extent for visualization as well as landform mapping purposes. To provide an easy way to supply additional imagery, an algorithm for registering uncalibrated photos to a textured digital elevation model is presented. A particular challenge in registering the images is posed by large variations in the photos concerning resolution, lighting conditions, seasonal changes, etc. <br /> The registered photos can be used to increase the visual quality of the textured DEM, in particular at steep slopes. To this end, a method is presented that combines several georegistered photos to textures for the DEM. The difficulty in this compositing process is to create a consistent appearance and avoid visible seams between the photos. In addition to that, the photos also provide valuable means to improve landform mapping. To this end, an extension of the landform mapping methods is presented that allows the utilization of the registered photos during mapping. This way, a detailed and exact mapping becomes feasible even at steep slopes.