Relations between Multi Scale Imagery and Gis Aggregation Levels for the Automatic Extraction of Vegetation Areas

This study deals with the automatic interpretation of digital imagery and focuses on natural topographic objects, in particular vegetation. As a first step for developing a sound strategy combining pixel based and model based image analysis, we investigate the possibilities and limitations of pixel based methods in large scale applications. Throughout the paper we assume multispectral imagery as the basis of our discussion. Connections between different aggregation levels in GIS and the information content of imagery in different scales are discussed, and various properties of linear and morphological scale space in order to generate the individual image resolution levels are described. The paper continues with a case study dealing with the differentiation between sealed and non-sealed regions in urban areas. While this task proves feasible using pixel based methods, a further model based image analysis step is necessary to extract individual GIS landscape features.