—— 
wavelet 
changes 
Slightly 
cts (e.g., 
' imaged 
'ested in 
ll size of 
al area in 
lution of 
to reduce 
] area, is 
eometric 
f a target 
airborne 
ed to the 
he UTM 
ly a few 
1ich can 
ich more 
y houses 
images 
iS of the 
ge of an 
topic of 
weather 
able the 
updating 
now has 
might be 
"national 
n, S.N., 
6(6), pp. 
Remote 
tions on 
itures of 
R, IEEE 
9-360. 
  
Dirk Stallmann 
RIGOROUS PHOTOGRAMMETRIC PROCESSING OF 
HIGH RESOLUTION SATELLITE IMAGERY 
Dieter FRITSCH, Dirk STALLMANN 
Institute for Photogrammetry (ifp), Stuttgart University, Germany 
Geschwister-Scholl-Strasse 24D, D—70174 Stuttgart 
Dieter.Fritsch@ifp.uni-stuttgart.de 
Dirk.Stallmann@ifp.uni-stuttgart.de 
Working Group IV/4 
KEY WORDS: remote sensing, orientation, calibration, accuracy 
ABSTRACT 
This paper deals with investigations on the geometric reconstruction and sensor calibration of satellite-based optical 
imaging systems using linear arrays in pushbroom mode, like SPOT, IRS-1C and MOMS-2P/PRIRODA. The geometric 
model of the sensor is based on an extension of a SPOT model developed by V. Kratky. This geometric solution combines 
the principle of rigorous photogrammetric bundle formulation with additional constraints derived from known relations 
assuming an elliptic orbit. The attitude parameters are modelled by a simple polynomial model being linear or quadratic. 
Ephemeris data (orbital position and attitude data) are not necessary but are optional. The parameters of the interior 
orientation, e.g. focal length and principle point coordinates, are determined by self—calibration. The sensor model can be 
easily extended to process images from other high resolution imaging systems as they become available. The flexibility 
and very good accuracy of the solution will be demonstrated with MOMS—2P/PRIRODA imagery, since multiple scenes 
with overlapping images are available with a relatively high image resolution. 
1 INTRODUCTION 
Past and current optical sensors like SPOT, IRS-1C/ID and MOMS-Q? use linear arrays in pushbroom mode. These 
systems provide panchromatic and multispectral image acquisition with a geometric resolution between 5 and 20m ground 
sampling distance. Stereographic coverage is provided by along-track (MOMS-02) or across-track (SPOT) image data 
collection. New and future systems like IKONOS-2 and QuickBird will have improved features, especially an higher 
geometric resolution, up to 1m in panchromatic and 4m in multispectral mode with a better dynamic radiometric range. 
In addition high precision orbital position and attitude data will be provided by the on-board Global Positioning System 
(GPS) receivers, Inertial Measurement Units (IMU) and star trackers. This additional information allows for reducing 
the number of ground control points. Furthermore, this information enables direct georeferencing of the imagery without 
geometric reconstruction of the imaging process (photogrammetric triangulation) and ground control. These developments 
offer new posibilies for the derivation of follow-ups like digital surface/terrain models, orthoimages and classification 
maps. Nevertheless to provide accurate and reliable products the calibration and verification of the whole sensor system, 
consisting of the imaging (camera), position and attitude sensors, is necessary. For this tasks we suggest an extended 
bundle adjustment for reconstruction of the exterior orientation and point determination with self-calibration. 
In our work the used geometric model is based on an extension of a SPOT model developed by V. Kratky (Kratky, 
1989). This geometric solution combines the principle of rigorous photogrammetric bundle formulation with additional 
constraints assuming an elliptic orbit. The sensor position is derived from known nominal orbit relations, while the 
attitude variations are modelled by a simple polynomial model (linear or quadratic). For self-calibration two additional 
parameters are added: the focal length (camera constant) and the principle point correction. The exterior orientation and 
the additional parameters of the sensor model are determined in a general formulation of the least-squares adjustment 
(Gauss-Helmert model). The use of additional information, e. g. from supplemented data files is not mandatory, but if this 
information is available it can be used to approximate or preset some of the unknown parameters. 
The current implementation has two main drawbacks. First, the number of images that can be processed simultanously is 
restricted to single — resection — or stereo images, multi-image processing (block adjustment) is not feasible. And sec- 
ond, for self-calibration only two additional parameters are available. In the new implementation of the geometric sensor 
model the method is capable to handle any number of images, using orbital position and/or attitude data as observations 
in the adjustment and a refined camera model for a improved self-calibration. 
  
International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B1. Amsterdam 2000. 313