Detection of buildings through multivariate analysis of spectral, textural, and shape based features

In recent publications, numerous methods for automated reconstruction/registration of buildings from stereo imagery have appeared (cf. e.g. Niederöst, 2003; Süweg, 2003). A common characteristic of these methods is the requirement of an existing approximate position (AP) of the building to be registered. In the present work, we develop a method for producing such APs based on sets of true colour (RGB) or colourinfrared (CIR) aerial imagery in combination with prior registrations from a GIS database. The primary intention with the work is to support the automated update of the GIS database. At the current stage of the project, the main goal is to produce a dataset where as much as possible of the “non building” area is masked out.