期刊名称：ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences
印刷版ISSN：2194-9042
电子版ISSN：2194-9050
出版年度：2006
卷号：XXXVI-4/C42
出版社：Copernicus Publications
摘要：This paper presents a study on automated habitat delineation and classification using very high spatial resolution (VHSR) optical remote sensing data in an alpine environment. Two overlapping study sites, which are situated in the Berchtesgaden National Park (SE Germany) represent a mountainous area, characterized by high habitat diversity. Habitat classification and delineation used to be accomplished manually on aerial photography by experienced interpreters on large areas. Driven by monitoring obligations in the framework of Natura 2000 as well as for reasons of time and economy, there is an increasing demand on regularly updated image data. In parallel, promising advances in automated image interpretation have been made over the last years. The used approach of object-based image analysis (OBIA) resembles the human performance of interpretation, though still with some limitations. Pan- sharpened data from a QuickBird scene were analyzed using two strategies for dealing with high spectral and spatial variability, namely one-level representation (OLR, Lang & Langanke, 2006) and multi-scale segmentation/object relationship modelling (MSS/ORM, Burnett & Blaschke, 2003). The study compares the potential of both strategies in specific settings, (1) boundary delineation of a given set of habitat types via OLR and (2) classification of patches of mountain pine (Pinus mugo) using MSS/ORM
