53. Istanbul 2004 
n restitution of 
mp P3 analytical 
progressive and 
obtained from 
in 25 m by 25 m 
1500 points were 
ressive sampling 
ured in selective 
>re obtained from 
1996). 
AOD software by 
a is gridded on 
were measured in 
. More than 10 
lated by bilinear- 
ten for the four 
at which the Z 
parameters 
int in object 
eps for generation 
low. (Fig.3). 
rdinates and grid 
Ye. coordinates (X, 
Er 4 (Z- Ze) 
Fra S (Z- 2.) 
(4) 
£n, *(Z-Z)) 
(7-2) 
| before in bundle 
  
where 
Xo. Yo, Zo — coordinates of perspective center in 
object system 
C — principal distance 
X,Y = image coordinate 
Xo, Yo — coordinate of principal point 
Li, O12, 3, 
I21, 129, 23, = parameters of transformation 
I31, F32, 733 
Xy — coordinates of point in object system 
     
      
  
  
  
     
  
    
eComputer Imaging : LEE 
  
  
    
ll: Digital Elevation Model. : 
Project Property ——-—— — — 00^ 
File Name : [Kampus pg a 
| DEM File: grid_interval_10.dem > | 
| Grid Space (mi: fi 0 
x: [4210470 
Y: [456715 
2 pete 
Dag o 9 6 m e 
| [D-\proje_kampus\2145.b zl D'\proje_kampus\2146.b 7] 
      
  
  
Figure 3. DEM window of eComputer Imaging software 
IN. Image coordinates are transformed to pixel coordinates. For 
this transformation, affine transformation parameters calculated 
before in interior orientation module are used. In steps II. and 
III, coordinates both left and right image of point known plane 
coordinates (X, Y) are determined. 
ay; e Xi 7g 
= 
by; Yi — be; 
lj 
X-24]*]l (5) 
| 
Il 
1j 
where 
ag, 31, az, bo, bi, b» affine transformation parameters 
Xi, yj = image coordinates 
J = image number 
X — vector of pixel coordinates 
IV. The windows, whose centers are already calculated at the 
step 111, are selected for matching. 
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B3. Istanbul 2004 
V. The Selected windows both at left and at right image are 
matched. Besides, Cross-correlation coefficient is calculated. 
VI. X, Y plane coordinates are constant and height (Z) is 
altered by AZ. These heights (Z;) are calculated by Z;7Z; ,*AZ. 
VIL. II, Ii IV., V., VL, steps are repeated for X^ Y; and 
altered Z coordinates. If the process ends for +AZ, VIII, step 
would be performed. 
VIII. The best fit of cross-correlation coefficients is selected 
and Height value (Z) which derives the best fit of cross- 
correlation coefficients between reference and research window 
is determined as height (Z) value of (X, Y) plane coordinates 
IX. X coordinate is increased by grid interval (New point is 
generated used). 
X. Height of the nearest point is used as approximately Z 
coordinate for the new point. 
XI. IL, HI, IV, V., VI, VII, VIL, IX, and X "steps are 
repeated for the new point. 
XII. If increasing of X coordinate was ended, Y coordinate is 
increased by grid interval and the overall steps are repeated. 
3.4 Results of the Experimental Tests 
Height error and root mean square error were computed by 
means of Equ.6 and Equ.7. 
£47 ZGROUND-ZDEM (6) 
  
m, = (7) 
where 
£j = actual error 
m, = height root mean square error 
n = number of point 
For this study, the extreme values of actual error, the mean of 
actual error and root mean square error have been derived for 
both 10 000 photograph scale(Table.3) and 16 000 photograph 
scale (Table.4). The descriptions of the terms used in Table.3 
and Table.4 are as follows: 
m, = height root mean square error 
mean - average value of actual error (£;) 
= the maximum value of actual error 
= the minimum value of actual error 
Emax 
“Emax