Object-based Analysis of Remote Sensing Data for Landscape Monitoring: Recent Developments

Environmental remote sensing is dominated by land use and land cover classifications. Different land cover features reflect electromagnetic radiation in different ways and remotely sensed images provide a basic and synoptic representation of land cover variation on the Earth’s surface. The pixel based analysis has been ‘the holy grail’ for many years. Land-based features may be categorized according to land cover classes (e.g., grass, concrete, water, etc.) but not likely according to land use classes (urban park, golf course, etc.). Only contextual information allows to encountering spatial relationships, object shapes or neighbourhood information. Even where land cover information is not the ultimate goal of remote sensing studies, it is often a useful aid for further analysis. This Progress Report will focus on one recent development in land cover research, namely the advent of object-based image processing. We report on the rapid diffusion of this approach and critically discuss the possibilities and methodological shortcomings. We discuss that tangible objects are desirable in many applications and conclude that many environmental management applications necessitate the control and use of object-based land cover distributions – in the opposite to pixel-based pattern to exploit land resources while safeguarding environmental concerns.