CMRT09: Object Extraction for 3D City Models, Road Databases and Traffic Monitoring - Concepts, Algorithms, and Evaluation
a b c
Figure 5. Results of building detection based on optical data (a), detected corner lines in the InSAR data (b), and of building
detection based on InSAR and optical data (c)
buildings. Such comer lines also appear in single SAR images
and hence this approach is not limited to InSAR data.
Further developed, this approach may be the basis for a change
detection method after natural hazards like flooding and
hurricanes. An optical image acquired before the hazard and
SAR data acquired afterwards can be analyzed using the
presented approach. A human interpreter would only have to
check those buildings for damages that were not detected from
both data sources. Hence, all buildings recognized from the
combination of optical and SAR features, shown in red in Fig.
5c, would be classified as undamaged. Only buildings in the
optical image that where not detected would have to be checked
speeding up the entire damage assessment step significantly.
Although first results are encouraging, further improvements
have to be made. One main disadvantage of the presented
classification approach is that its quality measures are not
interpretable as probabilities in a Bayesian sense. Although
many parameters have been learned from training data, parts of
the approach are still ad-hoc. A next step will thus be the
integration of the presented approach into a Bayesian
framework.
Furthermore, the differences of the sensor geometries should be
used for further building recognition enhancement. Since the
roofs of high buildings are displaced away from the sensor and
parts of the façade appear in the image, roof regions have to be
shifted towards the sensor in order to delineate building
footprints. Such displacement also bears height information
which may be used as an additional feature for building
recognition. More height information may also be derived
directly from the InSAR data.
Finally, three-dimensional modelling of the scene could be
accomplished based on the building footprints, a height
hypothesis and maybe even the estimation of the roof type. An
iterative joint classification and three-dimensional modelling in
a Bayesian framework, including context information, will be
the final goal of this project.
7. REFERENCES
Inglada, J., and Giros, A. 2004. On the possibility of Automatic
Multisensor Image Registration. IEEE Transactions on
Geoscience and Remote Sensing, Vol. 42, No. 10, pp. 2104-
2120.
Klein, L. A. 2004. Sensor and Data Fusion-A Tool for
Information Assessment and Decision Making, 3rd ed.,
Bellingham, WA: SPIE Press, pp. 127-181.
Mueller, S., and Zaum, D. W. 2005. Robust Building Detection
in Aerial Images. IntArchPhRS, Vol. XXXVI, Part B2AV24,
pp. 143-148.
Soergel, U., Thiele, A., Cadario, E., Thoennessen, U. 2007.
Fusion of High-Resolution InSAR Data and optical Imagery in
Scenes with Bridges over water for 3D Visualization and
Interpretation. In: Proceedings of Urban Remote Sensing Joint
Event 2007 (URBAN2007), 6 pages.
Thiele, A., Schulz, K., Thoennessen, U., Cadario, E. 2006.
Orthorectification as Preliminary Step for the Fusion of Data
from Active and Passive Sensor Systems. In: Proceedings of the
IEEE International Conference on Multisensor Fusion and
Integration for Intelligent Systems 2006 (MFI2006), 6 pages.
Thiele, A., Cadario, E., Schulz, K., Thoennessen, U., Soergel,
U. 2007. Building Recognition From Multi-Aspect High-
resolution InSAR Data in Urban Areas. IEEE Transactions on
Geoscience and Remote Sensing, Vol. 45, No. 11, pp. 3583-
3593.
Thiele, A., Cadario, E., Schulz, K., Thoennessen, U., Soergel,
U. 2008. Reconstruction of residential buildings from multi
aspect InSAR data. In: Proceedings of ESA-EUSC Workshop,
Frascati, Italy, available http://earth.esa.int/rtd/Events/ESA-
EUSC_2008/, 6p.
Tupin, F., Maitre, H., Mangin, J.-F., Nicolas, J-M., Pechersky,
E. 1998. Detection of Linear Features in SAR Images:
Application to Road Network Extraction. IEEE Transactions on
Geoscience and Remote Sensing, Vol. 36, No. 2, pp. 434-453.
Tupin, F., and Roux, M. 2003. Detection of building outlines
based on the fusion of SAR and optical features. ISPRS Journal
of Photogrammetry and Remote Sensing, Vol. 58, pp. 71-82.
Tupin, F., and Roux, M. 2005. Markov Random Field on
Region Adjacency Graph for the Fusion of SAR and Opitcal
Data in Radargrammetric Applications. IEEE Transactions on
Geoscience and Remote Sensing, Vol. 42, No. 8, pp. 1920-
1928.
Wegner J.D., and Soergel, U. 2008. Bridge height estimation
from combined high-resolution optical and SAR imagery.
IntArchPhRS, Vol. XXXVII, Part B7-3, pp. 1071-1076.
Xu, F., Jin, Y.-Q. 2007. Automatic Reconstruction of Building
Objects From Multiaspect Meter-Resolution SAR Images. IEEE
Transactions on Geoscience and Remote Sensing, Vol. 45,
No. 7, pp. 2336-2353.
174