Full text: CMRT09

CMRT09: Object Extraction for 3D City Models, Road Databases and Traffic Monitoring - Concepts, Algorithms, and Evaluation 
122 
limited the data to a sub set of approximately 2000x2000 pixels 
at full slant range-azimuth resolution. This subset contains both 
man-made structures and open vegetated areas. 
2.2 Seed DSM generation 
Seed DSM is first generated in Slant range projection using 
interferometric processing. Working at Very High Resolution 
may induce some local problems mainly in the phase 
unwrapping process. 
Man-made structures and more generally all features observed 
at VHR induce rapid height variation with respect to the 
resolution cell dimension. Since working at full resolution, 
these rapid height variations combined with phase noise induce 
in turn high spatial frequencies in the interferometric phase, 
making the phase unwrapping process potentially difficult even 
if the ambiguity of altitude is high compared to buildings 
heights. The generated InSAR DSM contains some small holes 
made of local DSM areas unwrapped independently. Figure 1 
shows the amplitude image of the sample data set in slant range 
with the derived DSM. 
Figure 1: Data set and corresponding seed DSM in slant range 
After the phase unwrapping process, the seed DSM is still in 
slant range azimuth geometry. Before being considered as the 
seed DSM to be iteratively improved, it must be geo-referenced 
and projected in a convenient geometry. 
A convenient geometry is a projection within which further 
processing for man-made structure detection, localisation and 
identification will be feasible but also a projection geometry 
within which SAR scene simulation will stay easy to model. 
Considering first that man-made structures have no preferential 
orientation within an observed scene, there is no peculiar 
advantage of using a specific geographic or cartographic 
projection rather than another. Therefore, with respect to man 
made structure detection, the important point is to work on geo- 
projected data to get rid of geometrical aspects linked to the 
slant range geometry. Consequently, working within a given 
geographic or cartographic projection is of no peculiar 
importance. 
Considering SAR scene simulation, we need a projection 
geometry allowing to easily model radar wave interaction with 
the observed scene. Interactions taken into account here are 
purely geometrical (ray tracing). At the present time, we do not 
intent to take a local backscattering coefficient into account, 
even if possibilities to integrate it in the model will be 
envisioned at each implementation steps. 
Based on these considerations, the ground range projection was 
chosen. This geometry is certainly the simplest to be considered 
for SAR scene simulation, while, with respect to man-made 
structure localization, it is not necessarily the most convenient. 
Therefore, when performing ground range projection, geo- 
referencing of each point in terms of longitude and latitude will 
be saved to allow further projection in any geographic or 
cartographic reference system. 
2.3 Structure definition 
Once geo-projected, the seed DSM must be used to define a 
structure that in turn will be used to model the backscattered 
SAR signal and simulate the detected SAR scene. Therefore, 
structure definition depends mainly on the way the simulation 
process is envisioned. The basic idea is to associate to each 
point of the DSM, a value that is proportional to the 
backscattered energy, giving then a peculiar weight to each 
point. Next, this map of backscattered energy will simply be 
back-projected in slant range to generate a simulated image. 
In a first approach, we simply aimed at considering non 
coherent dihedral reflection as the main backscattering process 
to be taken into account. 
2.3.1 Dihedral structures: Once more, for the sake of 
simplicity and in order to allow us to first perform a proof of 
concept, we choose to use directly the DSM as the structure 
itself. Simply, two consecutive heights are used to define a 
dihedral. The DSM is considered sequentially, azimuth lines by 
azimuth lines, and within a line, heights are considered 
sequentially with increasing ground range. If a given height is 
greater than the preceding one, a dihedral structure can be 
defined (fig 2). 
• Phase center 
Ground range Ground range 
Figure 2: DSM height interpretation 
The part of the incident beam intercepted by a dihedral structure 
will be fully backscattered toward the beam source. Therefore, 
the backscattered energy will be proportional to the square of 
the aperture of the considered dihedral structure; the aperture 
being the hypotenuse of the illuminated part of the dihedral. 
Any entering beam in the dihedral follows an optical path of the 
same length. Therefore, all entering beams will be imaged as 
localized at the phase centre of the dihedral. Since we are 
working azimuth lines by azimuth lines, our basis structure is 
defined in 2D and the phase centre is localized at the 
intersection of the local horizontal and the local vertical of the 
considered point. 
If we consider two consecutive points of our DSM along a 
ground range line having respectively heights hj., and hj, a 
dihedral structure will basically be defined if hj > hj_j; its phase 
centre will be localized at ground range coordinate of hj with 
local height hj.! and have a weight proportional to its aperture. 
2.3.2 Overestimation: Normally, the aperture of a dihedral 
should be computed taking into account shadowing of preceding 
dihedrals, if any, and be computed with respect to the height 
difference or with respect to the base, whatever the one is 
limiting the aperture the first.
	        
Waiting...

Note to user

Dear user,

In response to current developments in the web technology used by the Goobi viewer, the software no longer supports your browser.

Please use one of the following browsers to display this page correctly.

Thank you.