GeoGraph

The proposed project aims at methods for the use of graphs in geoinformatics. In particular, a landscape is characterized by hierarchies of plane attributed graphs. Thus the similarity of landscapes can be determined by hierarchical graph matching. We also aim at a comparison of the terminology in various research areas as computer vision, graph theory, and geoinformatics. Spatial information from remotely sensed data is processed for three purposes: acquisition, analysis, and visualization. We will develop practical processing methods for each of the purposes with special emphasis on applications in landscape ecology and nature protection. Land cover maps provide the basis for ecological landscape assessment and habitat modeling. In the proposed project the process of generating land cover maps from satellite images will be made explicit within the framework of graph pyramids and dual graph contraction. Thus a unified method for sub-pixel segmentation, pixel segmentation and meta-segmentation is provided. Visualization of geographic information will facilitate interaction with the user (geographers, landscape ecologists). This refers to visualization of retrieved information, but more important to visualization of typical landscapes by means of exemplary maps. The generation of exemplary maps is based on explicit definitions and rules or it is based on graph pyramids. Exemplary maps are used to verify the definitions and the rules. Furthermore, exemplary maps suggest landscape structures for renaturation tasks (e.g. renaturation of quarries or construction sites of power plants). Thus exemplary maps will support landscape planners.