Application of object-oriented analysis to study landscape structure

Development of remote sensing has contributed to new approaches in landscape ecology, based on GIS technology. However, per-pixel classification, which has been in use since the 1970s, has been criticized as it does not take into account the contextual information, fundamental in geographic analysis. An alternative is objectoriented image analysis, which seems to approximate human perception in digital processing. Its basic processing units are segments (homogeneous groups of pixels) which may represent landscape entities. Contrary to pixels they allow to study e.g. topological and hierarchical relations (providing contextual information). Object-oriented analysis was used to work out a methodology of environmental analysis. It refers to traditional methods of Polish landscape ecology but take advantage of remote sensing data and GIS tools. As a case study area, Gorce Mts. and surroundings were selected. For the study area, hierarchically structured system of spatial units in three scales was built, which allowed to characterize the study area in respect to morphology and land use. The method enabled an effective landscape analysis through combining an approach specific for human perception with automatic methods, accelerating the research and increasing the level of objectivity.