ormation 
révu pour 
nthétique 
leviendra 
créer des 
plines, et 
nnement 
xture, la 
sification 
nentation 
férencier 
ro-ondes 
; densités 
- à partir 
qu’il sera 
IX ont été 
t dans un 
de traiter 
eront pas 
raiter les 
rariations 
t rapport 
ace. Ces 
bre. Les 
nique, de 
d’image 
1esure de 
a ensuite 
produits 
PROCESSING ERS-1 SAR AND JERS-1 SAR FOR MAPPING 
Ian Dowman and Mark Upton 
University College London 
Gower Street 
London WC1E 6BT 
United Kingdom 
Tel: (44) 71 380 7226 
Fax: (44) 71 380 0453 
ABSTRACT 
Synthetic Aperture Radar data from ERS-1 has been proved to be very stable and well calibrated and is of 
considerable interest for a number of geoscience applications. The production of JERS-1 data has been subject 
to a number of problems but is nevertheless of considerable interest. University College London (UCL) has 
used data from both sensors for geocoding and for the determination of heights using stereomatching. This 
paper reports on the characteristics of the data from both sensors and the processing required for geocoding 
using software developed at UCL and by a consortium in which UCL was involved. The derivation of the 
parameters needed for geocoding depended very much on the processor used for creating the images and was 
also dependent on the accuracy of the orbit data. 
The determination of digital elevation models from both ERS-1 and JERS-1 also requires orbit data but 
depended on the radiometric characteristics of the data. The software derived for filtering and stereomatching 
the data is described and the results given. 
The paper describes the problems encountered and the steps which have to be taken to achieve geocoding and 
stereomatching. 
Keywords: SAR, ERS-1, JERS-1, geocoding, stereoscopy, speckle reduction filters, Digital Elevation Models. 
I. INTRODUCTION 
ERS-1 data has been available for some time now and 
the data has been used in the area of topographic 
mapping for geocoding and work has been done on 
the extraction of features. Results were presented at 
the first and second ERS-1 conferences, for example 
Dowman et al [1993b], Dowman et al [1994], Winter 
et al, [1993], Renouard and Perlant, [1994]. Work 
done at University College London (UCL) will be 
summarised and particularly the work on geocoding 
and stereomatching will be discussed. 
The JERS-1 satellite with the L band SAR and an 
optical sensor (OPS) giving along track stereoscopic 
coverage adds important additional data to that 
already available for mapping. The SAR provides 
data at a different wave length and incidence angle to 
ERS-1 The data from JERS should therefore 
complement existing data from ERS-1 SAR and 
SPOT by providing SAR data which will highlight 
437 
different surface features with the L band frequency, 
and allow two incidence angles with a potential for 
stereoscopic modelling for height derivation. 
The paper starts with a report on geocoding of ERS-1 
and JERS-1 and then gives a brief brief description of 
the method used for determination of height. The use 
of speckle reduction filters to improve the 
performance of stereo matching is then investigated. 
Some results are given which give an indication of the 
potential of the method and finally conclusions are 
drawn and the remaining problems are discussed. 
2. PRINCIPLES OF GEOCODING SAR DATA 
The principle of the system developed by the 
consortium in the UK (Dowman et al, 1993a ) is the 
transformation of a SAR image in a co-ordinate 
system reflecting time and range data into a standard 
map projection. An image processed to zero Doppler 
is assumed. The transformation is carried out by