IAPS 
(1) bad 
ur line 
Raster- 
ysteme 
id local 
rithms 
f small 
orming 
dis and 
adeoff. 
contour 
sraphic 
gle are 
rs: the 
(F-data 
In this 
contour 
le sens- 
scribed 
lygons 
smaller 
cerning 
r, while 
ess the 
üc map 
g chain 
ommon 
s when 
id SAR 
waves. 
:ntional 
  
  
Anselm Schmieder 
SAR (see Figure 1 (a)) it is not possible to discriminate between points P; and P; due to the height difference H, in 
InSAR the height difference causes a difference in the range AR. This AR is measured as a phase difference between the 
signals received by antennas 5, and S5. The separation between S, and S» is called the baseline. 
  
  
  
  
  
S B 
€. S 1 &——— — A 5 
Ra \\ 
\ N 
N R N Rı 
R x R4 N 
UN. 
i LH 
P, 
  
(a) (b) 
Figure 1: Distance Measurement using SAR: (a) Conventional SAR, (b) Interferometric SAR 
The phase difference measurement A is given by: 
Aj = BT AR (1) 
A 
Geometrical considerations lead to a phase difference measurement for each pixel. Phase measurement is restricted to the 
[0. .. 277) range, thus resulting in ambiguity. An interferogram visualizes those fringes in the phase. A range dependent 
phase component is removed from the interferogram through removal of the so called flat earth phase. The process 
of phase unwrapping is employed to resolve ambiguities caused by the [0...27) phase range. Unwrapped phase is 
related to terrain height by phase-to-height conversion. Figure 2 summarizes the necessary steps to derive a DEM from 
interferometric SAR processing. 
The InSAR for DEM generation principle was for instance reported in (Graham, 1974), (Zebker and Goldstein, 1986) and 
(Schwäbisch, 1998). Successful application of the mentioned technology using the the AeS-1 instrument was described 
in (Moreira, 1996). 
3 COUNTOUR LINE EXTRACTION AND APPROXIMATION 
In contour line approximation or generalization three requirements can be defined: (1) The contour lines must be simple, 
which means the number of points is reduced, (2) the contours must appear smooth, and (3) the characteristic of the 
contour must be retained, e.g. parallel lines, specific shapes, etc, (Li et al., 1999). 
Different measures of error between the approximation and the original contour points are possible. The two most promi- 
nent error measure are the maximum error E,,,,, and the summed square error E,,. (Pavlidis, 1982): 
Emas — max|p;—pi ^^ Esse — oi - ny, Q) 
i 
where p; is the original tie point and p; an approximated point (i.e. the closed point to p; on the approximation line). 
The most simple algorithm for line simplification, the takeM procedure, does not take into account the stated requirements, 
as expressed by error measures. Despite from this, it is computationally most efficient. It proceeds as follows (where M 
is a small integer): 
l. Let a polygon P be an ordered list of N points p;,i = 1,..., N. 
2. Construct a new polygon P' containing all p; for which (i — 1)/M is an integer. 
The algorithm by Pavlidis uses the E,,,, criterion. It employs a split-and-merge technique based on approximative 
collinearity and angle criterias for groups of lines: 
  
International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part Bl. Amsterdam 2000. 273