The preprocessing of remotely sensed data includes any 
process that has to take place before remote sensing 
data can be processed and analyzed by the scientific 
users for their own application objects. Data from remote 
sensing missions are not always optimal in quality. Each 
remote sensing system has its own characteristic 
capability depending on such things as the instrument 
design, altitude of the platform, and performance of the 
recording equipment. Other less predictable, often 
extraneous, influences may adversely affect the quality of 
the retrieved data. Preprocessing functions are necessary 
to reduce or eliminate such inadequacies. It consists of 
mainly two correction; radiometric correction and 
geometric — correction.  (Battett and Curtis, 1982) 
(Curran, 1985) 
This paper describes the systematic correction 
mechanism of heavy geometric distortions inherent in 
images from micro-satellites with poor pointing ability. It 
also shows the result in case that this mechanism is 
applied to the KITSAT-1 images. 
2. Background 
Most of the earth images from a satellite are heavily 
distorted geometrically. These distortions are mainly from 
the Earth, the satellite, the orbit and the image projection. 
The Earth contributes to the image deformation by its 
rotation, oblateness, and curvature. The contribution of 
the satellite comes from its variation in speed, attitude, 
and altitude. The projection of a spherical surface on a 
flat image also gives rise to the geometric distortions. 
These deformations, if not properly accounted for, will 
prevent meaningful result in further image processing 
such as mosaic, classification, comparison or fusion with 
other images and so on.(Mather,1987) 
Various geometric distortions inherent in remotely sensed 
images have been usually corrected by two methods. The 
first one is to use the sufficiently accurate geometric 
model based on the geometry of the sensor, the satellite, 
the orbit and the Earth. The second one is to use a simple 
polynomial derived from the relation in a set of ground 
control points (GCP) that correspond to those in real 
maps.(Muller,1988) 
Many microsatellites like the KITSAT-1 carry experimental 
Earth imaging payloads. However, since they usually 
provide very limited information about their positions and 
attitudes due to insufficient and rather inaccurate sensors, 
this makes it difficult to perform accurate geometric 
corrections of their images using the first method. The 
images from those satellites have been mostly 
geometrically corrected using the second method, which 
may cause additional distortions in the corrected images 
because the derived polynomial used in the correction is 
too coarse to represent accurately the whole process 
causing geometric distortion. 
The Attitude Determination and Control System(ADCS) of 
the KITSAT-1 stabilizes its attitude using gravity-gradient 
boom and uses several attitude sensors with coarse and 
indeterminate accuracy for attitude determination. The 
ADCS can provide just pointing error that explains how 
much the satellite pointing is deviated from the line 
between the satellite and the center of the Earth but 
cannot give directional information of the deviation. Since 
430 
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B3. Vienna 1996 
this makes unable to establish an accurate model for the 
KITSAT-1 image geometric distortion process, the second 
method was applied to carry out geometric correction but 
get just the results with poor accuracy. 
The geometric model in the KITSAT-1 can be represented 
by mainly six uncertain parameters: the pitch, yaw and roll 
angles of the satellite attitude and the latitude, longitude 
and altitude of the satellite position. In particular, the 
parameters of the attitude information are indeterminate 
as the ADCS provides just pointing errors on its attitude. 
The uncertain parameter values of the KITSAT-1 
geometric model are estimated by using a least mean 
squares -algorithm to utilize the relationships of GCP 
selected between satellite images and real maps. Fig. 1 
shows a flow chart of the systematic correction 
mechanism described here. 
  
1. Construct models with uncertain parameters 
  
  
  
2. Find GCPs in images and maps 
  
  
  
3. Estimate the parameters based on the GCPs 
using LMS algorithm 
  
  
  
4. Enhance the models by substituting 
the estimated parameters 
  
  
  
5. Correct the images by using the models 
  
  
  
Figure 1. Systematic Correction Mechanism 
3. Establishment of Geometric Model 
Fig.2 shows the fundamental principle to establish the 
geometric model of the KITSAT-1 CCD cameras. Three 
points are defined on CCD sensor(A), on the center of the 
lens(B), and on the Earth surface(C). Let L and M be the 
vector from the point A to the point B and the vector from 
the point B to the point C, respectively. If the coordinates 
of the points A and B are known, it is possible to derive 
the coordinates of the point C from the condition that the 
vector M is parallel to L so that the point C may be 
imaged on the point A on the CCD sensor. 
Lens 
    
   
CCD Sensor 
The Earth 
Figure 2. Fundamental Principle to Derive 
a Geometric Model 
Four coordinate systems are defined on the CCD sensor, 
the satellite, the orbit and the earth respectively as shown 
in Fig. 3. 
    
    
   
   
    
   
   
   
    
   
   
     
   
    
    
    
   
    
    
  
    
  
    
   
  
   
   
    
   
    
   
   
      
   
  
   
   
   
   
    
   
   
  
Satel 
CCD 
Sensor ' 
Figure 
Let the 
in C1 
values 
functio 
the rel 
3.1 Nc 
The fc 
Since 
systen 
Eq.(4) 
If the 
denote 
and C 
syster 
where 
S3=sin 
The p 
locatic 
satelli 
An ar 
and le 
The v 
The c