ts related to 
South-Haiti. 
on both data 
ortant on the 
are easier to 
Symposium on Remote Sensing for Resources Development and Environmental Management / Enschede / August 1986 
Digital elevation modeling with stereo SIR-B image data 
obtained by 
in 1984 were 
f Haiti. Two 
of La Gonave 
ries between 
basalts were 
2) ancients 
ults oriented 
nticline and 
rated in the 
plate. 
générale et 
[aiti. Trav. 
Hautes Etudes 
r I, 194 p., 
mt senestre 
conséquences 
i région de 
prequ'île du 
>1. Nord, t C 
Holt and M. 
.maging radar 
>r: the SIR-B 
16-20. 
tebillard, J. 
1 J.M. Vila 
irale de l'île 
.ti) à partir 
ISS et Seasat 
iraibes symp., 
)83. Analysis 
it and SIR-A 
)es. Geophys. 
Calmus, B. 
m Den Berghe 
Gi rault ed.), 
R.Simard 
Canada Centre for Remote Sensing, Ottawa 
F.Plourde & T.Toutin 
DIGIM, Montréal Canada 
ABSTRACT: A stereo SIR-B image data set has been used to produce a digital elevation model (DEM) for a test 
site located in the Mount Shasta area in California. The stereo pair was formed with images acquired at 53 
and 29 degree incidence angles, both on the same side. A digital method has been developed for the extraction 
of differential parallaxes which relies mainly on a hierarchical nested zoomed correlation procedure. The 
efficiency of the proposed method for automatic production of DEM is investigated in relation to terrain and 
SAR parameters. 
1 INTRODUCTION 
The possibility of deriving elevation data from 
stereoscopic radar images has been investigated in 
the past mainly in relation to geometry 
information necessary for adequate terrain 
modeling (Leberl et al, 1985; Derenyi and Stuart, 
1984) . It has been shown that radargrammetry 
offers possibilities for accurate terrain mapping 
under certain conditions, but it is generally more 
difficult to achieve, compared to traditional 
photogrammetry, due primarly to the inherent 
geometric and radiometric characteristics of radar 
images especially those derived from aircraft 
systems. To date the question has been addressed 
using analog or hybrid analog/digital plotter 
equipment requiring film products and operator 
visual interaction. 
The second Shuttle Imaging Radar (SIR-B) 
experiment offered for the first time the 
possibility to investigate the use of digital 
techniques for terrain modeling with spaceborne 
multi-incidence angle and digital L-band 
(synthetic aperture radar) SAR images (the SIR-B 
Science...,1984). The present research paper 
concerns results on the evaluation of a method 
developed at the Canada Centre for Remote Sensing 
for production of digital elevation models from 
SIR-B data. The method had been primarily 
developed for processing visible/infrared (VIS/IR) 
imagery such as that from LANDSAT or SPOT 
satellites (Simard and Krishna, 1983; Simard, 
1982) but has been refined in order to effectively 
account for geometic and radiometric differences 
between VIS/IR and SAR images. A test site was 
selected in the rugged topographic area of Mt. 
Shasta in California (lat/long » 41° 53'N/122° 
52'W) where elevations range approximately from 
500 to 2200 metres above sea level. The Mt. 
Shasta site is also very well known from the 
production of a SIR-B perspective view movie 
presented at the JPL SIR-B Workshop (Kobrlck, 
1985) . 
SIR-B image data acquired in October 1984 at the 
two incidence angles of 29 and 53 degrees were 
investigated for automatic DEM production. 
Basically, the overall process was divided into 
four main steps: geometric and radiometric 
preprocessing, automatic stereo image matching, 
three-dimensional modeling (ray intersection) and 
orthographic DEM/SIR-B image production. Precise 
image correction and geometric modeling have been 
made possible by using few ground control points, 
(GCPs) digitized from 1:62500 scale topographical 
maps, together with post flight ephemeris data 
(OPS onorbit postflight...,1983) sampled every 10 
seconds. 
2 GEOMETRIC AND RADIOMETRIC PREPROCESSING 
The SIR-B image data were digitally processed at 
JPL (Curlander, 1984; the SIR-B science...,1984) 
with a four look 12.5 metre pixel size (in ground 
range representation). However in order to smooth 
the image statistics, the input images were 
averaged to square 25 metre pixels using a 2 x 2 
adjacent cell. 
The geometric rectification of raw imagery has 
been done by using external parameters including 
GCPs and shuttle orbital parameters defined by 
position and velocity vectors. The rectification 
procedure consisted of moulding the imagery to a 
quasi-cartographic projection leaving uncorrected 
the parallax effect only. By this mean, both 29 
and 53 degree incidence angle images were 
s 
\ I 
o 
Figure 1. SIR-B range viewing geometry.