Fie.7 The image of Ea.(19). 
First, we analyze the case that the satellite 
has attitude fluctuation around the roll axis. 
Here we approximate that it is 4.0 [4 rad / 7 »]. 
This value is about ten times larger than the 
actual attitude fluctuation value. Fig.8(a) shows 
an CCD output image fluctuated around the roll 
axis. From this figure, the influence of attitude 
fluctuation and disturbance by noise obviously 
observed. Here we apply the TDI method at 
eight. Fig.8(b) and 8(c) show TDI output images 
with and without correction respectively. 
Then we analyze the second case that the 
satellite has both the attitude fluctuation 
around roll and pitch axis, and show simulation 
results in Fig.9. From this figure, it is obvious 
that the proposed correction method can 
improve the output image S/N under roll and 
pitch axis fluctuation condition. 
In TDI method, the influence of the satellite's 
attitude fluctuation increases with the number 
of integration (i.e. the number of CCD array). 
Fig.9 shows the relationships between S/N 
improvement and the number of integration. 
From this figure, the TDI without correction 
method can t acquire S/N improvement ratio 
when the integration times exceeds 10. 
As compared with this TDI method using 
correction of attitude fluctuation can acquire 
the same S/N improvement as it when satellite 
has no attitude fluctuation. Fig.10 and 11 show 
TDI at thirty-two and sixty-four times. 
Fig.12(a) shows the level of an output of CCD. 
Fig.12(b) and (c) show levels of output of TDI 
without correction the attitude fluctuation and 
TDI with correction the attitude fluctuation at 
thirty-two times. Fig.12(d) and (e) show the 
profile of outputs of TDI without correction and 
with correction the attitude fluctuation at 
sixty-four times. Fig.13 shows the profile of the 
simulated images 
  
c» Image level £5 
— X 255 
(a) output of an CCD 
  
c» Image level £3 
— X 955 
(b) TDI without correction at thirty-two times 
25 
  
c» Image leve 
‘ - X 955 
(c)TDI with correction at thirty-two times 
25 
  
c» Image leve 
— X 255 
(d)TDI without correction at sixty-four times 
25 
  
c» Image leve 
— X 255 
(e)TDI with correction at sixty-four times 
Fig.12 The level of image. 
From Fig.12, it is clear that high-frequency 
components of the image are improved by 
correcting the attitude fluctuation. 
The relationship between S/N improvement and 
integration times when attitude fluctuation of 
roll axis is 4.0[ 4 rad / * »] is shown in Fig.13(a) 
20 Intemational Archives of Photogrammetry and Remote Sensing. Vol. XXXII, Part 7, Budapest, 1998 
#22 P RO ino * "RO "VERA. ON 
(2 
Fig