aditional 
d on the 
ssing of 
> version 
ping has 
‘is based 
y of the 
urface 
ı/pixel) 
' Airport) 
  
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B2. Istanbul 2004 
  
Based on these geo-referenced standard products traditional 
data analyses within common software packages or new 
developments, e.g. automated extraction of object features is 
performed. Functionality for manual or semi-automated real- 
time editing of multi-line scanner data products (e.g. DSM) will 
probably soon be available as a part of commercial software 
solutions. 
Since 1997, a multitude of successful airborne HRSC 
application projects have been carried out. After the first 
extensive camera experiments which had a focus on geoscience 
(Gwinner et al, 1999; Gwinner et al, 20002), different 
applications in environmental science, civil engineering and 
cartography followed soon after. Some examples include 
topographic mapping and map-updating (Hoffmann et al., 2000; 
Gwinner et al., 2000b), high mountain cartography (Hauber et 
al, 2000), geologic mapping and natural hazards (Gwinner, 
2001; Baldi et al., 2002), hydrologic (Martin & O'Kane, 2000) 
and ecologic investigations (Leser, 2002) as well as telecom 
network planning (Renouard & Lehmann, 1999) and digital city 
models (Móller, 2000). 
2. HRSC DATA FLOW 
Figure 2 shows the entire HRSC data flow from digital data 
acquisition of image and orientation data to ground data 
processing of the photogrammetric final products. In addition to 
data of the HRSC cameras, the processing system can handle 
any other multi-line scanner data, provided that the necessary 
specifications for data integration as well as sensor-specific 
calibration data are provided. For test purposes, a data set of 
LH SYSTEMS ADS40 has already been integrated and was pro- 
cessed successfully to DSM and ortho-images (see chapter 6). 
the DGPS/INS systems adapted to the HRSC cameras is suf- 
ficiently high (Scholten et al., 2001; Scholten et al., 2003a). 
Camera/IMU boresight alignment or camera/orientation time 
line offsets are determined by means of HRSC's internal multi- 
stereo capability. 
    
  
  
  
Geometric Correction 
of Image Data 
   
Image Matching 
DEM Generation i 
    
  
Orthoimage Generation 
Generation of 
Orthoimage Mosaics 
| Color Orthomosaics I 
  
  
  
  
| DEM Follow-Up Products I 
  
Figure 3. HRSC photogrammetric data processing 
4. PARALLEL PROCESSING ON PC CLUSTER 
  
| Image Data Geom. Calibration Exter. Orientation Inertial Boresight Offsets | 
  
1 
  
   
  
| Inertial exterior orientation for each image line | 
  
  
| Inertial correction of flight movement | 
5 Generation of coarse DTM | 
————— — 
| Correction of flight movement and coarse DTM | Determination of final 
  
  
| Coarse multi-Image matching 
  
+ v v Y v + + 
  
| 
| 
  
  
  
Data Acquisition Data Processing 
| boresight offsets 
  
  
  
  
  
  
  
  
radiom. calibration data 
geom. calibration data 
  
  
  
  
  
   
  
  
m 
i 
   
   
Camera Command Unit 
  
se 
ZEISS platform m] 
  
  
  
  
  
position and 
attitude data 
  
    
  
   
  
  
  
Produots: 
Surface models 
Orthoimag es 
Follow-up produots 
  
  
  
  
  
  
  
Full-resolution multi-image matching 
| Generation of final DTM | 
  
  
  
  
  
+ + + + v v + LLLÁ———————————— 
| final exterior orientation for each image line ] 
| Ortho-image generation [ Generation of 
  
color ortho-image mosaics 
and 3D follow-up products 
  
  
  
Figure 4. Parallel processing structure 
The efficiency of the previously described photogrammetric 
processing line could be enhanced significantly by 
implementing parallel processing capabilities on a cluster of 
Linux-PCs in contrast to traditional processing of aerial images 
based on photogrammetric single-workstation concepts. 
  
  
  
Figure 2. HRSC data flow 
3. PHOTOGRAMMETRIC DATA PROCESSING LINE 
All the above mentioned applications have exploited many 
TBytes of airborne HRSC data processed with the photo- 
grammetric data processing system (Figure 3), which includes 
radiometric adjustments such as corrections for illumination 
effects and relative adjustment of adjacent image strips for 
mosaicking. Tbe development of a high extent of automation is 
a precondition to handle these data volumes. No ground control 
information is necessary for the generation of ortho-image 
mosaics and DSM since the orientation data quality provided by 
The DLR cluster currently consists of 24 clients and 2 servers, 
each of them equipped with double-CPUs. The servers perform 
the project management and conduct all project-wide processes 
(e.g. mosaicking and DSM interpolation for all image strips), 
while processing steps which can be performed independently 
for each strip are distributed to the PC-clients. Each successful 
step signals its completion, and thus enables the automatic 
initialization of subsequent steps until the final products are 
derived. Client processes, continuously checking for. auto- 
matically requested operations, ensure that the load of the 
available hardware is maximized. 
409