Robust detection of buildings in digital surface models

There is growing interest in the automatic interpretation of range data acquired by airborne laser scanners. Huge volumes of data involved in land use analysis and building reconstruction applications necessitate a preliminary step to focus attention on interesting areas. We present a technique to detect and discriminate buildings and vegetation in dense digital surface models (DSM) by combining curvature based features and edge information. Our results demonstrate the effectiveness of differential geometric properties in the analysis of range data.