International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B3. Istanbul 2004 
  
  
CE 
A Eu Deae 
ZE (11) 
— = 6, +2c,y=0 
Putting x =x, and y =y, in (59) we get 
€ C, 
Xo = — Ye 
26; 2c, 
It is obvious from equation (11) that the PSF 
fitting method, unlike moment analysis, can only 
be used when the star image is spread over more 
than four pixels. However the process can be 
easily extended to fitting many images 
simultaneously. This allows position 
determination in a much more crowded situation 
and complete automation of the whole process. It 
is said to give better accuracy than the moment 
analysis method (Schildknecht, 1994). Thus for 
accuracies better than 0.05 one has to use the PSF 
fitting method. Theoretical estimations by 
Eisfeller and Hein (Eisfeller and Hein, 1994) 
using the PSF fitting method gave accuracies of 
the order of 0.05 to 0.15 of a pixel. 
3. Image co-ordinate measurements 
In this experiment centroids of real star objects 
were determined using both the moment analysis 
and PSF fitting methods and analysed: 
A To find out the measuring accuracy of CCD 
co-ordinates of star images that can be 
achieved in practice. 
B To compare the two methods 
In this work images were taking using the /FEN 
CCD integrated telescope with exposure times 
ranging from | to 3 seconds and the telescope 
pointing to the zenith. These exposure times were 
found to be optimum, in terms of avoiding trails 
and obtaining good signal-to-noise ratios, after 
many trials. All the sets of images were multiple. 
One of the images used for the analysis is shown 
in figure 1 below. The common images in every 
multiple image were identified and their co- 
ordinates measured. 
Two sets of observations were used. The 
centroids in the first set, found in table 4, were 
determined with the moment analysis software, 
astrometrica, developed by Raab (Raab, 1996) 
while the second set made up of 17 multiple 
images taking under 45 minutes found in table 5 
were determined by both moment analysis and 
PSF fitting methods using astrometrica and the 
CCD software developed by Ploner (Ploner, 
1996) respectively. The PSF fitting method was 
found to be more adaptive to geodetic use. 
614 
  
Figure 1: CCD250 2 a 2.5 sec. exposure image after 
calibration and median filtering 
4. Analysis of results 
Before the analysis the image co-ordinates, which were 
given in pixels, were converted to sensor co-ordinate, 
also in pixels, using the following expressions. 
total colum. 
x—sensor co-ord-col measured- (13) 
total rows 
y-sensor co-ordz— ————- -—row measure 
It is pertinent to note that the images were taken 
independently and secondly the stars were in motion. In 
order to analyse these images they were registered or 
transformed to the same datum using the two- 
dimensional similarity transformation. The 
> . . 4t m m > 
transformation also gives mean positions [= >); | , of 
the common points in the new co-ordinate system. 
The standard errors of the estimated image co-ordinates 
were computed. An output of the computer program 
developed for the computations is shown in tables 1 to 3. 
The mean of the standard errors for every multiple image 
were calculated. 
Table 1: Registration of images from a multiple exposure 
  
  
(CCD250) 
Transformation coefficients 
Images 
a b e f 
  
CCD250 1| 0.99999 | 0.00000 | 09024 | 0.0008 
CCD250 2| 1.00076 | 0.00065 | 63.1416 | 12146 
CCD250 3| 100109 |. 0.00127 |. 1257476! 2 1138 
CCD250 4| 1.00115 | 0.00125 |. 1863867! 3,9443 
  
  
  
  
  
  
  
Fi 
ac