CMRT09: Object Extraction for 3D City Models, Road Databases and Traffic Monitoring - Concepts, Algorithms, and Evaluation
160
• slice no. 3 for displaying intensities in elevation
direction
According to the defined slices, necessary data in slant-range,
azimuth and elevation are extracted out of the data pool
provided by the sampling process in POV Ray.
Slant Range |m|
Figure 5: Slice 1: elevation heights in slant-range direction
(slice 1 in Figure 4 corresponds to slant range interval
60 m to 140 m); blue: single bounce contributions,
green: double bounce contributions
-40 -30 -20 -10 O 10 20 30 40
Azimuth [mj
Figure 6: Slice 2: elevation information along azimuth direction
displayed in height over ground; blue: single bounce
contributions, green: double bounce contributions;
zero level = level of ground surrounding the step
Since the incidence angle used for sampling the 3D model
scene is known, slice no. 1 pointing in slant range direction can
be presented by two versions, either by displaying elevation
heights (Figure 5), i.e. elevation coordinates with respect to a
master height situated in the center of the image plane used for
sampling the scene, or by providing height information in
height over ground geometry, i.e. heights with respect to the
ground surrounding the box.
Following the slant-range direction from left to right, displaying
height data in elevation heights enables to distinguish between
range intervals containing one scatterer and areas containing
several scatterers resulting in layover effects, which can not be
separated in reflectivity maps such as shown in Fig. 3 (right). In
Figure 5, reflection caused by direct backscattering are colored
in blue color while double bounce contributions are indicated
by green spots. Due to the incidence angle of 45 degrees,
double bounce effects are focused at the same position in slant-
range and are overlaid by both single bounce contributions at
the ground and single bounce contributions reflected at the end
of the step up-side.
Following slice no. 2 along its way, elevation information is
shown along the azimuth direction in height over ground
(Figure 6). After passing an interval of contributions directly
backscattered at the ground, the layover region starts showing
the width of the double bounce areas in azimuth, which are
equal to the width of the step model. As expected, double
bounce contributions caused by the interaction between
perpendicular faces are concentrated at the corresponding
intersection lines and, hence, show a height value of 0 and 20
meters, respectively.
Figure 7: Slice 3: normalized intensities along elevation
direction; step width in elevation: 2 meters; blue:
single bounce contributions, green: double bounce
contribution
Slice no. 3 pointing in elevation direction is shown in Figure 7.
After the spatial sampling along elevation direction is chosen
by the operator, intensity contributions are assigned to elevation
intervals and summed up. Since the selected pixel is located
within the double bounce area of two dihedrals, slice no. 3
shows two strong double bounce contributions caused by the
interaction of step faces (colored in green) accompanied by
weak direct backscattering derived at the step faces (colored in
blue). Although the radiometric quality of detected intensity
contributions is moderate due to simplified reflection models
and the approximation of SAR signals by rays, proportions
between single and double bounce intensities within one
resolution cell are well represented.
In the following Section, simulation results will be compared to
real data derived by tomographic analysis.
3. COMPARISON: SIMULATION VS. REAL DATA
For demonstrating potential applications of SAR simulation in
elevation dimension, a practical example extracted from
tomographic analysis using TerraSAR-X high resolution
spotlight data is provided in this section and compared to
simulation results.
3.1 Object modelling
Fig.8 shows the 2D intensity map for the convention center of
Las Vegas acquired by TerraSAR-X. For the purpose of this
paper, an azimuth-range pixel marked by a green dot has been
taken as example. The complex valued measurement at this
pixel corresponds to the integration of the reflected radar signal
