Institute for Photogrammetry, Stuttgart University
Keplerstrasse 11, 70174 Stuttgart, Germany
Tel.: +49-711-121 3383 Fax : +49-711 121 3297
The main steps of a process for the recognition and reconstruction of objects are object modelling, i.e. the definition and generation of suitable object models, segmentation, i.e. the extraction of primitives from the observed data and matching, I.e. the detection and definition of correspondences between the primitives extracted by the segmentation process and the primitives contained in the object model. A large number of buildings can be represented by a polyhedron since the boundaries of most buildings consist of a number of planar surfaces and straight lines. For this reason a segmentation algorithm should aim to extract planar surfaces and straight lines from the observed data in order to match these features against the corresponding object primitives in the consecutive step of the automatic building reconstruction. Since the use of height data is considered to be the most important topic of this paper, the segmentation of DSM data, which is a prerequisite using this type of data will be discussed in detail. Afterwards two approaches aiming on the combination of different types of information will be presented within this paper. First break-lines extracted from both DSM and image data are used to reconstruct a rather simple type of building. Therefore the high reliability of DSM break-lines are combined with the high geometric accuracy of grey value edges. The second approach aims on the extraction of planar surfaces likely to be roof planes from the DSM. To support this process given ground plans are utilized as a priori information. Within this approach the use of polyhedrons as building models allows to reconstruct very general building types.