Hier finden Sie wissenschaftliche Publikationen aus den Fraunhofer-Instituten.

Enhancement of generic building models by recognition and enforcement of geometric constraints

: Meidow, Jochen; Hammer, Horst; Pohl, Melanie; Bulatov, Dimitri

Volltext urn:nbn:de:0011-n-4047476 (1.5 MByte PDF)
MD5 Fingerprint: 90a37387788c9a8690d7d493447a6a83
Erstellt am: 26.7.2016

Halounova, L. ; International Society for Photogrammetry and Remote Sensing -ISPRS-:
XXIII ISPRS Congress. Vol.III, Pt.B3 : 12-19 July 2016, Prague, Czech Republic
Istanbul: ISPRS, 2016 (ISPRS Annals III-3)
International Society for Photogrammetry and Remote Sensing (ISPRS Congress) <23, 2016, Prague>
Konferenzbeitrag, Elektronische Publikation
Fraunhofer IOSB ()
building model; reasoning; adjustment; boundary representation; 3D city models

Many buildings in 3D city models can be represented by generic models, e.g. boundary representations or polyhedrons, without expressing building-specific knowledge explicitly. Without additional constraints, the bounding faces of these building reconstructions do not feature expected structures such as orthogonality or parallelism. The recognition and enforcement of man-made structures within model instances is one way to enhance 3D city models. Since the reconstructions are derived from uncertain and imprecise data, crisp relations such as orthogonality or parallelism are rarely satisfied exactly. Furthermore, the uncertainty of geometric entities is usually not specified in 3D city models. Therefore, we propose a point sampling which simulates the initial point cloud acquisition by airborne laser scanning and provides estimates for the uncertainties. We present a complete workflow for recognition and enforcement of man-made structures in a given boundary representation. The recognition is performed by hypothesis testing and the enforcement of the detected constraints by a global adjustment of all bounding faces. Since the adjustment changes not only the geometry but also the topology of faces, we obtain improved building models which feature regular structures and a potentially reduced complexity. The feasibility and the usability of the approach are demonstrated with a real data set.