International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B3. Istanbul 2004 
  
  
Wild RC10 Zeiss RMK A 30/23 
  
Zeiss RMKA-TOP30 
Zeiss RMKA-TOP3050 
Figure 2. Subimages of fiducial marks for several models. 
3. IMAGE MATCHING 
Matching can be described as the establishment of the 
correspondence between various data sets, such as images, 
maps or GIS data in photogrammetry. Moreover, the matching 
problem can be described as correspondence problem. A 
number of photogrammetric tasks is related to matching. Some 
of these are following: the reestablishment of the interior 
orientation, relative orientation and point transfer in aerial 
triangulation, absolute orientation, generation of digital terrain 
model (DTM) and interpretation step. The image of a fiducial 
mark is mátched with a two-dimensional model in the 
reestablishment of the interior orientation. Parts of one image 
are matched with parts of other images in order to generate tie 
points for relative orientation and point transfer in aerial 
triangulation. Parts of the image are matched with a description 
of control features in absolute orientation. Parts of an image are 
matched with parts of another image in order to generate a 
three-dimensional object description in the generation of DTM. 
Features extracted from the image are matched with object 
models in order to identify and localize the depicted scene 
objects in the interpretation step (Heipke® 1996). 
Matching algorithm is generally categorized as area based and 
feature based matching. The area based matching aims to shift 
and possibly warp one of the images such that its intensities 
best fit to the intensities of the other image. Area based 
matching consist of cross correlation and least squares 
matching. Cross correlation is a powerful technique to have the 
correspondence between digital images. It is based on two 
assumptions: 
1. the two images geometrically differ only due to 
translation. 
2. the two images radiometrically differ only due to 
brightness and contrast (Lang and Forstner 1998). 
In order to compute the cross correlation function of two 
windows, a reference window (Fig3) is shifted across a larger 
search window (Fig4). In each position the cross correlation 
coefficient between the reference window and the 
corresponding part of the search window is computed according 
to Equ.1 (Heipke® 1996). 
  
Figure 3. Reference Window 
  
Figure 4. Search Window 
m 
TEE) CE kt) 
n=1 
  
  
p= (1) 
> 2 (P(&.n) zu Y » (gb. m - uy 
-In- m 
usps 
where 
f(&n) = individual grey values of reference window 
Hi = average grey value of reference window 
g (En) = individual grey values of corresponding part 
of search window 
po — average grey value of corresponding part of 
search window 
m, n — number rows and columns of reference 
window 
Least squares matching is a generalization of cross correlation. 
It has the following fundamental features: 
e any parametric type of mapping function can be 
assumed, 
e any parametric type of radiometric relation between 
the two images may be dealt with, 
Inte 
Fe: 
im: 
as 
we 
cor 
rel: 
per 
wit 
gec 
Fi 
Ta 
cal 
  
de 
thi 
or 
di; 
fic 
th 
Bx