In: Stilla U, Rottensteiner F, Paparoditis N (Eds) CMRT09. IAPRS, Vol. XXXVIII, Part 3/W4 — Paris, France, 3-4 September, 2009
RAY TRACING AND SAR-TOMOGRAPHY FOR 3D ANALYSIS OF MICROWAVE
SCATTERING AT MAN-MADE OBJECTS
S. Auer 3 , X. Zhu a , S. Hinz b , R. Bamler 3C
3 Remote Sensing Technology, Technische Universität München, Arcisstrasse 21, 80333 München - (Stefan.Auer,
Xiaoxiang.Zhu)@bv.tum.de
b Institute for Photogrammetry and Remote Sensing, Universität Karlsruhe, Kaiserstrasse 12, 76128 Karlsruhe
- stefan.hinz@ipf.uni-karlsruhe.de
c Remote Sensing Technology Institute, German Aerospace Center (DLR), Münchner Strasse 20, 82234
Oberpfaffenhofen-Wessling - Richard.Bamler@dlr.de
Commission VI, WG VI/4
KEY WORDS: SAR Simulation, Ray Tracing, POV Ray, SAR Tomography, TerraSAR-X
ABSTRACT:
An inherent drawback of SAR imaging of complex 3D structures is the potential layover of more than one scatterer in one resolution
cell. Such scatterers can be separated by tomographic processing of multiple SAR images acquired with different across-track
baselines. Simulation tools may further support interpretation of such layover effects appearing in multi-body urban scenes. In this
paper, an existing 2D simulation approach, developed for separating different kinds of reflection effects in the azimuth-range plane,
is enhanced by including the elevation direction as third dimension and thus enabling the comparison of the SAR simulation results
with 3D imaging techniques such as tomography. After introducing the simulation concept, tools for three-dimensional analysis of
scattering effects are presented. Finally, simulated data are compared with real elevation data extracted from TerraSAR-X images
for showing potential fields of application.
1. INTRODUCTION
High resolution SAR sensors like TerraSAR-X or Cosmo-
SkyMed provide SAR images having a resolution of below one
meter in spotlight mode. While in SAR images of coarse
resolution several dominant scatterers from man-made objects
at slightly different ranges may be condensed into a single
pixel, these will be separable in high resolution images. Hence,
more image features can be distinguished due to an increased
number of deterministic effects and due to an increased signal
to clutter ratio for dominant scatterers (Adam et al., 2008).
However, visual interpretation of image features in high
resolution SAR images remains challenging due to range
dependent geometrical effects. SAR maps the 3D world
basically into a cylindrical coordinate system, where range and
azimuth are the image coordinates and elevation is the
coordinate, along which all scattering contributions are
integrated, i.e. all scatterers are mapped into the same resolution
cell in the azimuth-range plane if they have the same spatial
distance with respect to the SAR sensor. Access to the third
coordinate, elevation, is achieved by multi-baseline methods,
like Persistent Scatterer Interferometry (Ferretti et ah, 2001;
Kampes, 2006) or SAR tomography (Reigber & Moreira, 2000;
Fomaro et al, 2003; Zhu et ah, 2008).
Simulation of scattering effects for urban areas may support
visual interpretation of high resolution SAR images. In this
context, Franceschetti and co-workers (Franceschetti et ah,
1995) distinguish between two different kinds of SAR
simulators: image simulators and SAR raw data simulators. In
the past, different concepts have been presented for simulating
artificial SAR images for urban areas (Balz, 2006; Mametsa et
ah, 2001) and for simulating SAR raw data by illuminating
simplified building models (Franceschetti et ah, 2003).
The simulator presented in this paper applies ray tracing
algorithms and has been developed for simulating artificial
SAR reflectivity maps (Auer et ah, 2008). The approach is
focused on geometrical correctness while physical effects and
speckle effects are neglected. In addition to a reflectivity map
containing all backscattered intensities, reflection effects are
assigned to different image layers based on available bounce
level infonnation, i.e. separate layers for single bounce, double
bounce, etc. Hence, interpretation of deterministic reflection
phenomena appearing at man-made objects is simplified.
So far, for providing image data in azimuth and range, two out
of three dimensions of the imaging system have been exploited.
The novelty of the presented approach compared to other
simulation concepts relates to the fact that the complete 3D
geometry of the SAR imaging process is simulated and stored.
This enables one, based on the simulated 2D SAR, to retrieve
information about the existence of multiple scatterers in one
resolution cell in the SAR image. We show that simulation of
the distribution of point scatterers in elevation direction may
support the interpretation of estimated elevation coordinates
derived by SAR Tomography.
The structure of the paper is organized as follows. Firstly, the
basic simulation concept is introduced in Section 2 where four
major parts of the simulation concept are explained including
necessary developments for extraction and analysis of elevation
data. Simulation results displaying elevation data are compared
with real data extracted from a TerraSAR-X image in Section 3.
Finally, in Section 4, a short summary is given and future work
is addressed.
