limited control. One of the chief advantages of satellite 
remote sensing is that it in short time can give access 
to information of remote areas. Maps for providing 
ground control may then be unavailable or unreliable, 
making it necessary to interpolate or extrapolate the 
geometry from surrounding better mapped areas. 
Results from investigations of planimetric accuracy in 
JERS-1 models were reported by Westin (1995). This 
paper is an extension of these investigations to 
include also altimetric accuracy. 
2. DATA SET 
The satellite data used for this study consists of six 
scenes acquired in a contiguous strip on June 30, 
1993. The strip covers an area in north-western 
Sweden (Figure 2) and is essentially cloudfree and of 
good quality. The images display moderate horizontal 
striping as reported by Nishidai (1993). The images 
were acquired in high gain mode, which reduces the 
potential problem with low dynamic range in the 
visible bands. This caused on the other hand that 
some bright areas in the IR band were oversaturated. 
   
  
FINLAND 
Figure 2. Location of the study area. 
The scenes were provided preprocessed to level 1. 
This means that the scenes were geometrically 
uncorrected, but that a radiometric correction of 
detector gain and bias was performed. No further 
radiometric correction was applied in this study. 
938 
The ground control points were extracted from the 
1:10,000 topographic orthophoto maps from the 
Swedish Land Survey. These maps allow the 
extraction of control point features with a positional 
accuracy of 3-4 m in three dimensions. 
A digital elevation model from the Swedish Land 
Survey was used for evaluation of the computed 
DEM. It has a height accuracy of about 2 m, with 50 m 
grid distance. 
3. EXTERIOR ORIENTATION 
3.1 Adjustment model 
The similarities between the JERS-1 VNIR instrument 
and the SPOT XS instrument made it feasible to 
implement the JERS-1 model in an already existing 
and well proven system for rectification of SPOT 
scenes. This model will here briefly be explained, 
while a more detailed description can be found in 
Westin (1990,1991). 
The system uses a satellite orbital model based the 
six Keplerian parameters. Investigations in Westin 
(1990) showed that it was possible to reduce these 
parameters to the following four without loss of 
accuracy: 
to time at the ascending node 
Q right ascension of ascending node 
| inclination 
fo orbit radius at t; 
The attitude modelling is utilizing the gyro rate 
measurements present in the telemetry. Integration of 
the gyro rates gives a detailed description of relative 
attitude changes within the scene. The unknown 
constant offsets remains to be estimated and are 
introduced as parameters in the model: 
Do roll 
Po pitch 
Ko yaw 
Each control point give rise to an observation vector 
With five parameters: 
o 
coordinate for the detector position 
time for control point imaging 
control point latitude 
control point longitude 
control point elevation 
Yee tx 
These are feed into a least squares adjustment. In 
this adjustment, a priori estimates of the parameters 
are used to resolve singularities and to allow solutions 
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B4. Vienna 1996 
also fc 
points : 
In the | 
series 
parame 
(Westir 
used fc 
parame 
importe 
3.2 An 
Before 
integra 
data. F 
over tl 
rates a 
the rol 
(NASD 
distorti 
thatar 
Attitude angle [deg] 
Figure 
A clos 
reveals 
of abo 
distorti 
necess 
informe