ic Lune 
eee en era en ne a ann de AA NCAA 
DES 
  
   
   
    
   
  
  
   
  
  
   
   
  
  
   
  
  
  
  
    
   
  
  
  
  
   
  
  
   
   
  
   
   
    
  
  
  
   
  
   
    
    
   
    
   
  
  
  
  
  
  
  
  
  
    
  
   
   
  
  
  
  
  
   
    
SURFACE 
SURFACE MESH 
=> SURFACE 
£=/) ELEMENT 
| 
| Y 
E x 
Figure 2-1 : Description of the surface with 
surface meshes and surface elements 
(from /SCHNEIDER 1989/) 
coordinate system into the coordinate system of 
the digital image, and thus creating the image 
points, one can assign certain gray values to 
these image points. 
In view of the reflective qualities of the 
surface, the imaging of the gray value of a 
surface element is identical with the gray value 
of the surface. Therefore the imaging can be 
regarded as an observation. 
While up to now, the knowledge of the surface 
has only been assumed, the surface is now 
described by approximate coordinates of the grid 
points. When a surface element of an image is 
transferred to the other images of the same 
object with the help of the above-mentioned 
approximate coordinates, the coordinates and the 
gray value of the element change. 
The discrepancy between this new gray value and 
the corresponding gray value of the object is 
the residual. The square sum of the residuals 
can be minimized through adjustment according to 
the method of the least squares, with the image 
gray values being the observations and the 
object gray values being the unknowns. 
The functional model of the adjustment includes 
the following observation equation for each 
image gray value : 
Vi,s = Gi - gi,s (2,0) (2-1) 
residuals 
number of the surface elements 
number of the images 
object gray value 
image gray value 
orientation parameters 
Z1,...,Z24, grid coordinates 
with : 
t Ow «d 
n nm ug ug nn 
After a Taylor linearization in O and Z, the 
equation can be solved according to the known 
adjustment algorithm /EBNER, HEIPKE 1988/. 
By using the approximate unknowns, which are 
defined through iteration processes, new image 
gray values can be found and applied as 
observations in the following adjustments. 
For the creation of digital multi image 
orthophotos it is necessary to simultaneously 
define the radiometrical corrections. This can 
be achieved by using the gray value spectrum of 
one image as a constant and adjusting the other 
images and their gray values to it (2.2). The 
corrections of the surface meshes remain 
constant in this procedure. 
Qi, 1 = 111 +72) 0'153 (2.2) 
    
+ = image gray value 
gi.j - corrected gray value 
ri;,r2; = radiometrical corrections 
3. PREPARATORY PROCEDURES 
Digital images are the fundamental data for 
digital multi image matching. Basically, there 
are two ways of producing these digital images. 
First, CCD-sensors can be used for the creation 
of images. They produce digital images right 
away, which is an advantage for further 
procedures. These sensors were examined and 
tested with regard to their photogrammetric 
applicability and their geometric and electronic 
precision e.g. by /LUHMANN, WESTER-EBBINGHAUS 
1986/, /BEHR 1989/, /BOSEMANN ET. AL 1990/. 
Second, if only analog images are available, or 
if the resolution of a common film can not be 
done without, the images have to be digitized by 
scanners which operate on different kind of 
sensors. Hard- and software which are able to 
support this technique were described by 
/SINNREICH 1989/ and /FAUST 1989/ e.g.. 
For the computation of digital multi image 
orthophotos the orientation parameters of the 
digital images have to be known. Therefore, it 
is necessary to measure homologous points in the 
images first. The orientation parameters can 
then be defined by means of photogrammetric 
bundle triangulation and adjustment according to 
the method of least squares /WESTER-EBBINGHAUS 
1985/. 
Bundle triangulation can help define the 
parameters of the interior orientation as well 
(simultaneous camera calibration) /PEIPE 1985/, 
/WESTER-EBBINGHAUS 1986/. 
4. PERFORMING MULTI IMAGE MATCHING 
As mentioned above, the method for the 
computation of digital multi image orthophoto is 
based on the program system PROSurf. 
Before the actual computation can be started, 
size and number of the surface elements must be 
defined. These control parameters depend on 
pixel size, imaging scale and the structure of 
the object; these features fundamentally 
determine the quality of the orthophoto. 
The starting data, i.e. the starting coordinates 
for the multi image matching, can be yielded by 
defining points manually or by using points from 
the previous adjustment. 
It is indispensable for the creation of 
orthophotos that the surface meshes are 
contiguous, if the object is to be described in 
its entirety. For this purpose, a measuring 
strategy has been developed which processes the 
images in a meander-shaped way. Furthermore, the 
approximate figures of the surrounding heights 
serve as heights for the grid points of the 
surface meshes. 
It is inevitable to test the convergence rate 
and to set up stopping rules for the 
computation, as the orthophotos are computed