FIL A * 
| SR. RawDataCenter U»/ 
    
  
Aerial ; ni 
Triangulation anf 
Framework i DragonFly 
infrastructure & Gisteibuted Processing | à interacaon 8 Visualization. 
  
  
  
  
  
  
Figure 2. UltraMap concept and modules. 
With the upcoming version of UltraMap 3.0 two new 
modules will be added to further extend the workflow. 
These two modules provide revolutionary new features, 
namely the automated generation of point clouds, digital 
surface models (DSM), digital terrain models (DTM), 
DSMOrtho images and DTM Ortho images, all derived 
from a set of overlapping UltraCam images. Results from 
the basic image processing and the aero triangulation are 
used by the new modules to generate a point cloud, then a 
DSM, then a DTM and then two different ortho images, the 
so called DSMOrtho (images rectified by a DSM) and 
DTMOrtho (images rectified by the DTM). 
   
   
UltraMap 
, MitraMap/AT 
  
DTM 
Figure 3. New UltraMap 3.0 modules 
The processing is being kicked-off automatically after the 
aero triangulation and fully supports the automated 
distributed processing and the full 16-bit workflow. The 
new modules support processing on GPU(s) if available in 
the system. Visual output and QC are smoothly integrated 
into the existing viewer. 
2. Dense Matching and 3D Point Clouds 
A significant change in photogrammetry has been achieved 
by Multi-Ray Photogrammetry which became possible high 
performance digital aerial camera such as UltraCam and a 
fully digital workflow by software systems such as 
UltraMap. The cameras enable a significantly increased 
forward overlap as well as the ability to collect more 
images but literally without increasing acquisition costs. 
These highly redundant image datasets enable to generate 
new products such as point clouds highly automated and 
robust. 
     
* Orthos j 
However, Multi-Ray Photogrammetry in a first step is not a 
new technology; it is a specific flight pattern with a very 
high forward overlap (8096, even 9096) and an increased 
sidelap (up to 6096). The result is this highly redundant 
dataset that allows automated “dense matching” to generate 
high resolution, highly accurate point clouds from the 
imagery by matching the pixels of the overlapping images 
automatically. 
  
For the point cloud generation, the dense matcher analyzes 
the images and calculates a range image for each pixel on 
the ground from each stereo pair covering the pixel on the 
ground. The location (x, y,) of the pixel is defined by its 
position in the geo-referenced image; the range image 
defines a z-value for each pixel. Due to the highly 
redundant data set (thanks to the high forward and sidelap), 
usually multiple stereo pairs exist which cover one pixel on 
the ground. Thus multiple range images can be processed 
by the dense matcher who leads to multiple z-vales per 
pixel. That makes the whole process very robust and 
increases accuracy of the derived z-value. 
The 3D point cloud generated by the dense matcher of 
UltraMap has a point density of several hundred points per 
square meter and thus is much denser than any airborne 
Lidar scanning point cloud. 
   
    
    
    
    
  
  
   
    
   
   
   
     
    
  
  
  
  
  
  
  
  
  
  
    
    
   
   
    
   
   
   
    
    
    
   
    
   
   
   
  
   
    
  
  
     
  
Figu 
Figure 
based 
densit 
visual 
The a 
than tl 
accurz 
of th 
signifi 
Geom 
have ¢ 
Once 
DSM 
ortho 
image 
pixel 
surfac 
ortho 
in soi 
object 
Figure 
gener: 
level 
white 
showı