es were captured 
construction made of 
the boards were not 
the camera from the 
were long. To enable 
ld was photographed 
fabout 1:8to 1: 10 
of about 30 to 40 cm. 
' exposure axes are 
ym the camera to the 
j) images cannot be 
bed below. The base 
1 as the final stereo 
every stereo pair can 
»d image maps of the 
axes can easily be 
ould be used in any 
as photo equipment 
camera over a small 
o about 15 cm, which 
is to align two of the 
r another line that is 
"sition of one leg and 
the required distance. 
lel axes are usually 
' when generating the 
on the illumination, 
. For a good stereo 
o images should be as 
ishes connected to the 
ase the position of the 
ges which deteriorates 
ation or sunlight can 
| very high differences 
This means that darker 
ved any more. Even if 
cht or brilliant, it is 
yder to get an image 
he whole scene. This 
tions and the options 
high quality lenses is 
covered with a big roof 
re rather dark. One of 
, for the other location 
images were taken in a 
h simplifies and speeds 
or the photogrammetric 
ei 6006 metric middle 
he image configuration 
; was similar to the one 
the images for stereo- 
ages with a 50 mm lens 
'ove the accuracy of the 
rked points in a bundle 
The exposure time was 
> the film material with 
ie small lens aperture of 
ufficient depth of focus 
06. 
  
   
  
  
it Making Referencing Project Display Window Help 
   
  
  
    
= Ee BEB BQ RAO 
RNS |] de Q * A 4 cn & X 
  
| Source Data 
i Photograph # 8 
| Number of dangles: n/a 
    
| Mapping plane points. [95,9293 Reorder | 
"C  Ssae/lnage EB —— | d 
(| C Imagesie  |257 [195 piéels 
| ir Image Scale | pinels per I 4 | 
: | C PintScale 1. [^ al uu d 
: (Sewer — e 
S Srl. 0 qu es us Number of border pixels fo 
low resolution preview 50x38 
  
  
    
   
  
  
  
7 = _ = . 
Figure 2. Generation of a rectified image using PhotoModeler® 
DATA PROCESSING DURING THE FIELD CAMPAIGN 
Image Rectification. The rectification of the vertical images is 
probably not necessary in many cases. If no height information 
of the object is available (just like in this case) and the object is 
not really flat, i.e. has some extent in the 3" dimension, the 
result of this process can only be a simply rectified image, i.e. 
not an orthophoto. Nevertheless, there are reasons to do this 
type of processing. Generating rectified images using the 
marked points leads to images that are orientated in a unique 
reference system and which have an approximately identical 
scale when printed. The rectification can be done using software 
like PhotoModeler®, which is an easy to use digital 
photogrammetry software. Otherwise, standard image 
processing or graphics software can be used to perform a coarse 
rectification and a resampling to a certain resolution. In this 
case the influence of the central projection will not be corrected, 
but dependent on the lens used (if not a fisheye) this can be 
accepted in many cases. It is important to realize and make 
understood to all persons involved, that these images are not 
metric image maps, but only image representations of the 
situation suitable for obtaining a general view or for numbering 
etc. 
In this project, the images were rectified using 
PhotoModeler® software (cf. fig. 2) and resampled for printing 
in a scale of 1:10 on A4 sized paper. The processing of the 
images was performed in the evenings, so that results on paper 
and the laptop screen could be presented to the partners the next 
day. The images are rectified in the plane of the marked control 
points, displacements remain where height differences occur. 
Anaglyph images. An anaglyph image is an image that 
contains the image information of two single images taken from 
different positions. If the geometric conditions (parallel axes of 
exposure, reasonable base to height ratio) are complied with and 
each eye is provided with the image information of one image, 
the observer can see the situation in stereo. When the geometric 
conditions are fulfilled during the image capture, the separation 
of the image information can be done using different 
techniques. A simple method is the separation of the images 
using different colors for the two images. These colors must be 
independent from each other. Two color combinations are quite 
popular for this procedure: red-green and red-cyan images. The 
red-cyan combination is better for color images, red-green 
rather for black and white images. With both methods, the color 
information of the images is strongly reduced. If the color 
information of the objects is essential even in 3D, these methods 
are not suitable, others should be taken into account like 
separation using polarization or shutter glasses. The latter have 
the disadvantage, that they cannot be used as easily. If the color 
information is not essential, or black and white images are used, 
the generation of the images is quite easy. In principle, the 
intensities of the single images are mapped to the intensity of 
the appropriate channels in the stereo image. The positions of 
the different color layers must be adjusted to let the parallaxes 
disappear for a chosen plane (here defined by the marked 
points). For objects in front of or behind this plane, the 
horizontal parallaxes represent height differences of the object. 
Thus, this simple technique leads to reasonable results only if 
the height differences from this plane are not too large. After the 
—381—