Corresponding author. 
RESEARCH ON THE LINEAR AND NONLINEAR METHODS OF CORRECTING 
BASELINE ERRORS ON SAR INTERFEROGRAMS 
Shiyu ZHANG 3 ’ *, Tao LI a , Jingnan LIU 3 , Ye XIA b 
3 GNSS Research Center, Wuhan University, 129 Luoyu Road, Wuhan, P.R.China-sheperdshiyu@163.com 
taoli@whu.edu.cn 
b GeoForschungsZentrumPotsdam, Telegrafenberg A17,14473 Potsdam, Germany-xia@gfz-potsdam.de 
Commission VII, WG VII/2 
KEY WORDS: Satellite Remote Sensing, SAR Image, SAR Interferometry, Accuracy Analysis, Ground Deformation 
ABSTRACT: 
In this paper, the precise orbit and the coarse orbit are both used to compare the characteristics of the baseline errros on 
interferograms. To eliminate the fringes caused by baseline errors, a linear algorithm which is done on interferograms and a 
Repeat-pass interferometric synthetic aperture (InSAR) requires radial directions. The relationship between baseline vector 
orbital satellite state vectors to estimate baseline vector, reduce errors and across-track/radial errors can be given by: 
flat earth effects, and geocode. However, the orbits are difficult 
to determine with enough precision to satisfy the need of 
InSAR. The orbit errors will propagate directly into the 
absolute orbit accuracy would need to be on the order of 1mm, 
which is far below the current satellites’s precision (about 5- 
10cm, ESA ENVISAT satellite). 
Where ^radial = ra dial orbit error 
Hanssen (2001) has explained in detail how orbit errors affect 
the baseline and how baseline errors influence the reference O' x track ~ across-track error 
phase. The baseline vector can be described conveniently in 
three representations:(Dparallel/perpendicular, (2)horizontal/vert j n paper, the characteristics of baseline errors are studied, 
-ical, and ©baseline length/angle. and linear and nonlinear methods are introduced to reduce these 
nonlinear algorithm which is performed on phase-unwrapped maps are both used. The unwrapped phase values are decreased greatly 
after baseline errors are eliminated using bilinear method. The results show that the standard deviation (STD) is decreased from 10.9 
radians to 3.1 radians. In nonlinear method, a quadratic function is used to fit the surface of the phase-unwrapped map, then the 
fitted surface is subtracted. STD is decreased from 10.9 radians to 2.8 radians. The results show that the nonlinear method is a little 
effective than the linear one. 
1. INTRODUCTION 
The phase errors in interferogram come only from the 
propagation of the orbit errors in the satellite’s across-track and 
topography height and deformation maps. Because the ultimate 
products of InSAR are relative such as height or deformation, to 
fully correct the residual interferometric fringes, the required 
(2) 
errors. The precise orbit (from Delft University) and the coarse 
orbit in SAR SLC product are both used to compare the 
baseline errros on interferograms. 
2. ALGORITHMS 
where 
B v = perpendicular baseline 
B h = horizontal baseline 
B i = parallel baseline 
=baseline vector 
Massonnet and Feig (1998) proposed a method similar to the 
tie-point method proposed to diminish those baseline errors for 
topography height estimation. However, this method can only 
reduce the parallel component error vector, and cannot correct 
the perpendicular one, still leaving residual errors. Another 
problem that constrains the tie-point method is the atmospheric 
effect, which causes additional phase gradients. 
B v = vertical baseline 
CC = orientation angle 
In this paper a bilinear algorithm and a nonlinear algorithm are 
used to eliminate the baseline errors on interferograms and 
phase-unwraped maps. The linear algorithm is done on