Advanced classification of urban form - Integrated analysis of land cover and plot boundaries for building type identification

Land cover data represents an important basis for the fulfillment of numerous analysis. In particular for questions in the context of spatial planning, urban settlement monitoring is essential. The objective of this study is to develop a method which permits identification of different building types. Seven categories are chosen for the classification. The methodology relies on integrating location, area, and height parameters from the land cover data with plot boundaries from the cadaster. Combination of these data sets enables identification of building structures, from open to closed coverage types. The differentiation of detached, semi-detached, and terraced houses can only be achieved by the integration of plot boundaries. Result of the classification approach is a model, implemented in ArcGIS 9.2 and tested on LISA data. LISA t t d t i b d i l i i b l i d t i d t Th t t t f K. Kalian , K. Steinnocher , C. Aubrecht INPUT DATA PRE-CLASSIFICATION prototype data is based on aerial imagery, airborne laser scanning, and remote sensing data. The test area covers part of Klagenfurt, Carinthia. Realization in GIS starts with an intersection of land cover data with the plot boundaries and proceeds with a spatial join in order to get information form both data sets. The final classification is based on predefined rules referring to a set of selected parameters for each building type. Applying this rule set to the study area results in a total coverage of over 98% leaving less than 2% of the buildings unclassified. Accuracy assessment is based on a random sample of 5% of all buildings, indicating that 83% of the building objects are classified correctly.