CMRT09: Object Extraction for 3D City Models, Road Databases and Traffic Monitoring - Concepts, Algorithms, and Evaluation
126
Figure 11 shows in parallel, the seed DSM computed in ground
range along with the improved one obtained after 25 iterations.
Figure 11 : Seed and improved DSM
While the simulated SAR scene is clearly improved after 25
iterations, improvement is less evident observing the obtained
DSM.
Figure 12 represents a DSM sample line, in ground range,
before and after improvement. Globally, we observe that the
modified DSM appears less noisy and more structured. At this
stage, it is difficult to assert if the reached structure is a correct
representation of the observed scene and if it can be used in
man-made structure detection or identification. But, we can
conclude that the achieved structure, together with the proposed
model and the used parameter set, allows simulating a SAR
intensity image close to the really detected one.
Figure 11: DSM sample line before (green) and after (blue)
improvement
Obtaining a DSM representation closer to the observed one will
require testing the influence of all parameters as also improving
our simplistic model. But, the main point is that we performed a
proof of concept of the proposed principle: “Iterative DSM
improvement through SAR scene simulation and comparison
with observed one”.
Since the proposed method is global and does not require any a
priori knowledge on buildings shapes and orientation, it can be
envisioned as a first improvement of the DSM to be used in
more sophisticated and context-based man-made structure
detection techniques.
Nevertheless, if stable, the reached simulated SAR intensity
image stays, for the moment, still far from the really detected
SAR intensity image. We have well concentrated the energy
where it should, but still not with the degree of details offered
by the real data. One must thus keep in mind that the obtained
improved DSM is just one possible representation of the
observed scene. Other representations are possible provided
simulation model and set of parameters that are used are
optimized
6. CONCLUSIONS
We developed the tools required for simulating a SAR intensity
image in slant range geometry starting from a seed DSM given
in ground range and issued from InSAR processing.
Our objective was first to perform a proof of concept, showing
that in its principle, it is possible to perform an iterative
improvement of a seed DSM by simulation of SAR intensity
image in slant range - azimuth projection and comparison with
the corresponding detected one. Therefore, we developed a
simplistic model allowing to associate a backscattered energy to
ground range - azimuth resolution cells with respect to local
heights.
Effort was principally put on the reliability and accuracy of
back and forth referencing and projection processes.
Clearly, the proof of concept is performed: comparing simulated
and detected backscattered energy in slant range allows
correcting iteratively the underlying DSM.
The process converges monotonically toward a DSM structure
that is thus one possible representation of the observed scene.
Monotonic convergence shows that the obtained solution is
stable and is, in itself, the result that had to be obtained to
validate the proposed iterative process.
Complementary analysis must be performed to assess if the
derived DSM can efficiently be used for man-made structures
detection.
7. REFERENCES
Balz T, Haala N. (2003), SAR-based 3D reconstruction of
complex urban environments, The International Archives of the
Photogrammetry, Remote Sensing and Spatial Information
Sciences, Vol. 34, Part 3/W13, pp. 181-185
Balz, T (2006), Automated CAD model-based geo-referencing
for high-resolution SAR data in urban environments. Radar,
Sonar and Navigation, IEE Proceedings - Vol. 153(3), June
2006, pp. 289-293
Bamler. R and Schattier B. (1993), SAR data acquisition and
image formation. In: Schreier G. (ed.) SAR geocoding: data and
systems.. Wichmann, Karlsruhe, pp. 53-101.
Dupuis X., Dupas J., Oriot H. (2000) 3D extraction from
interferometric high resolution SAR images using the RAMSES
sensor, PROC. 3rd European Symposium on Synthetic Aperture
Radar, EUSAR'2000, München (Germany), VDE, pp. 505-507
Soerger U., Thoennessen U., Stilla U. (2003), Reconstruction of
buildings from inteiferometric SAR data of built-up areas. The
International Archives of the Photogrammetry, Remote Sensing
and Spatial Information Sciences, Vol. 34, Part 3AVI3, pp. 59-
64
Thiele, A.; Cadario, E.; Schulz, K.; Thonnessen, U.; Soergel, U
(2007), Building Recognition From Multi-Aspect High-
Resolution InSAR Data in Urban Areas, IEEE Transactions on
Geoscience and Remote Sensing, 45(11), pp. 3583 - 3593
Tison C., Tupin F., Maitre H. (2007), A fusion scheme for joint
retrieval of urban height map and classification from high-
resolution inteiferometric SAR images, IEEE Tranactions on
Geosciences and Remote Sensing,, vol. 45(2), pp. 496-50
