In: Wagner W., Székely, B. (eds.): ISPRS TC VII Symposium - 100 Years ISPRS, Vienna, Austria, July 5-7, 2010, IAPRS, Vol. XXXVIII, Part 7B 
STEREO RADARGRAMMETRY IN SOUTH-EAST ASIA USING TERRASAR-X 
STRIPMAP DATA 
Xueyan He, Timo Balz*, Lu Zhang, Mingsheng Liao 
State Key Laboratory of Information Engineering in Surveying, Mapping and Remote Sensing, Wuhan University, China 
Luoyu Road 129, 430079 Wuhan 
balz@lmars.whu.edu.cn 
Commission VII 
KEY WORDS: Radar, SAR, Stereo, Radargrammetry, TerraSAR-X 
ABSTRACT: 
Stereo radargrammetry is a mature technique for deriving height information from SAR image pairs. Because stereo radargrammetry 
is less sensible to temporal decorrelation, it can provide better results than interferometric SAR in certain situations. Using 
TerraSAR-X stripmap stereo pairs, digital surface models (DSM) with good height accuracy can be generated. We use SRTM data as 
our initial DEM and a pyramid layer based approach for our radargrammetric processing. Our results are relative noisy with a high 
error standard deviation. Nevertheless, with a mean overall error of less than 3 m in height the final DSM is quite precise. 
1. INTRODUCTION 
There are various ways for creating digital elevation models 
(DEM) from synthetic aperture radar (SAR) data. The most 
common technique is SAR interferometry (InSAR). InSAR uses 
the phase difference between the backscattered signals received 
at slightly different positions. From the phase differences the 
height of the backscattering object can be determined. InSAR is 
a very precise technique, but especially spacebome InSAR suf 
fers from errors caused by the atmosphere. The repeat pass in 
terferometry, which is commonly used with today’s sensors, as 
we are waiting for the TanDEM-X data, is very sensitive to 
temporal decorrelation. In repeat pass InSAR the time between 
two data acquisitions is rather large, for example 11 days for 
TerraSAR-X. In this time the situation on the ground can 
change, especially over vegetated areas, and the data decorre 
lates. 
This massively influences the usability of InSAR over strongly 
vegetated areas. Our test area around Kuala Kangsar in Malay 
sia is a strongly vegetated and mountainous area. The nearby 
mountains reach up to 1500 m, while Kuala Kangsar resides at 
around 40 m height. Under these conditions, repeat pass InSAR 
with TerraSAR-X does not provide good results. Most of the 
scene is strongly decorrelated. 
But InSAR is not the only method of creating DEMs from SAR 
data. Stereo radargrammetry (Stereo SAR) is far less affected by 
the atmosphere and by temporal decorrelation. StereoSAR is 
therefore, although in general less precise than InSAR, more 
suitable for this area. 
StereoSAR with TerraSAR-X can deliver quite precise results 
(Raggam et al. 2010). In our approach we use the rational poly 
nomial coefficient (RPC) model (Tao & Hu 2001) for the calcu 
lation of the 3D geo-coordinates of detected homologous points. 
In Section 2, the approach for StereoSAR with high-resolution 
TerraSAR-X data is explained. In Section 3, we describe the 
usage of the RPC model for stereo radargrammetry. Our expe 
rimental results are shown in Section 4 and finally conclusions 
are drawn. 
* corresponding author 
2. STEREO RADARGRAMMETRY WITH TERRASAR- 
X 
Comparable to InSAR, StereoSAR needs at least two images. 
But the images need to have a rather large acquisition angle dif 
ference. To use StereoSAR with spacebome systems, the sys 
tems must be able to acquire data under different incidence an 
gles. Before the launch of the new generation of high-resolution 
SAR satellites, Radarsat data was used for StereoSAR (e.g. 
Toutin 2000). The new satellite generation is also capable of 
acquiring data to be used for StereoSAR and in our experiments 
we used a pair of TerraSAR-X stripmap data. StereoSAR can be 
acquired in a same-side or opposite-side configuration. In the 
following we always assume a same-side configuration. In a 
same-side configuration the images are acquired from parallel 
tracks under different incident angles. 
The geometry of a SAR image is different in range and in azi 
muth direction. In azimuth direction the location of an object in 
a SAR image depends on the Doppler Effect and is typically 
precise if the object is not moving. In range direction the loca 
tion of an object in a SAR image depends on the distance be 
tween the object and the sensor. When an object is seen from 
different angles, this, so called, relief displacements will be dif 
ferent on each SAR image. The resulting difference between the 
image coordinates of homologous points in a stereo pair is 
called parallax (Leberl 1990).