Robust Building Detection in Aerial Images

The robust detection of buildings in aerial images is an important part of the automated interpretation of these data. Applications are e.g. quality control and automatic updating of GIS data, automatic land use analysis, measurement of sealed areas for public authority uses, etc. As additional data like laser scan data is expensive and often simply not available, the presented approach is based only on aerial images. It starts with a seeded region growing algorithm to segment the entire image. Then, photometric and geometric features are calculated for each region. Especially, robust form features increase the reliability of the approach. A numerical classification is performed to differentiate the classes building and non-building. The approach is applied to a test site and the classification results are compared with manually interpreted images.