ntation 
  
  
ila | 
ion 
  
ints in 
'refore 
ntation 
curacy 
nanual 
ng the 
sed on 
DTM 
image 
tional 
o that 
  
Jeffrey Shan 
  
Z-F(X,Y) 2) 
where F denotes the elevation interpolation function, which is chosen as bilinear form through out the computation in 
this article. 
Linearizing above two equations will yield following observation equations 
zdX 4 -XxdZ -2z(Xc- X)-X(Zc-Z) 
MY —3dZ -:0€-Y)-9(c—2Z) 
F F 3 
sue dYar am -Z-F(X,Y) © 
4x Y 
where 
X X 
SIR y (4) 
ef 
Given the DTM as expressed in general in Eq.(2), for each image point, its 3-D ground coordinates can be calculated 
with Eq.(3). The calculation is iterative. Initially, the Z coordinate of that image point is given as the average elevation 
of the area, while in the subsequent iterations it is determined with Eq.(2) with approximate (X,Y) coordinates. The 
iteration stops when the coordinate corrections dX, dY and dZ are smaller than the given tolerances. Comparing thus 
obtained ground coordinates with their known coordinates provides an independent evaluation on the exterior image 
orientation. 
The entire approach is implemented with Java as a stand-alone program, therefore software developed for this article is 
platform independent and can be easily implemented in the Internet environment. 
3 TEST RESULTS AND ANALYSES 
3.1 Feature extraction and matching 
Tab.1 lists the number of image points obtained in automatic feature extraction and image matching steps. As is shown, 
each patch has average more than one thousand point features extracted. In total, more than ten thousands of feature 
points are extracted in the nine patches from aerial image and orthoimage respectively (denoted as a-image and o-image 
in the table, respectively). Through feature correspondence step, about one tenth of the feature points are left for the 
final orientation calculation. 
Tab.1 Amount of feature points,correspondence points and matched points 
  
  
  
  
  
  
  
  
  
  
  
  
Feature extraction Correspondence & matching 
Patch No. | a-image o-image correspondence | LSM matching 
l 957 1037 340 206 
2 1039 1237 285 195 
3 1072 2287 360 240 
4 1992 1542 506 320 
5 1335 1468 502 312 
6 3749 4022 218 156 
7 1252 1460 260 160 
8 966 2043 157 95 
9 924 1093 223 155 
Total 13286 16189 2851 1840 
  
  
  
  
  
  
  
  
International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B3. Amsterdam 2000. 833