rlin was chosen, consisting 
All images were taken at 
te of that, the automatic tie 
.]l. The achieved relative 
in with imagery of the 
r scenarios not included in 
e to the generally lower 
ckBird in compare to its 
3), as shown in Table 2, is 
pointing accuracy of the 
he absolute and relative 
a set of precisely measured 
ints not used for bundle 
ndependent reference. 
ell PowerEdge T610 with 
CPUs at 2.93GHz. For all 
could be performed in at 
> preparation time tpreperation 
tep requires relatively little 
| SGM processing step on 
The Dunedin scenario took 
arger height range, caused 
"took very little time as 
2 m because of the low 
4 processing per square 
GSDpsm, the number of 
able 2 the processing times 
matching and orthophoto 
'nt scenarios. 
tof Berlin, textured by 
    
  
For aerial image processing, it has been shown (Hirschmüller 
and Bucher, 2010) that the height error of SGM is around 
half of the GSD, additional to the registration error. In our 
experience, the same quality can be reached with satellite 
images. 
     
Figure : Reconstruction of a part of Cape Town 
  
5. CONCLUSIONS AND OUTLOOK 
It has been discussed that processing high resolution digital 
surface models from satellite images on a productive level 
requires a fully automatic and robust approach. This paper 
presented a solution to this problem. This results in a huge 
gain in productivity and cost-efficiency. From the large 
number of already processed scenes, examples were given 
with images from DigitalGlobe satellites. However, the 
developed method has also been tested with many datasets 
from an airborne line camera (Wohlfeil, 2010 and 2011). 
The processing times needed on the mentioned test machine 
can be reduced significantly to a very small fraction via 
parallelization since most time is consumed by SGM 
matching, which can run in parallel if more CPU cores or 
more computers are available. 
The remaining manual interaction can be reduced further. 
Suitable parameters (orientation, season, etc.) from an image 
database will be subject of further investigation in order to 
select images automatically. We also see good chances in 
refining the SRTM-based water masks by existing or new 
image processing algorithms in order to get optimal results 
even in regions where water covered areas change fast. 
Besides water masking, it is also important to mask clouds. 
As only in one of all processed scenarios clouds occurred this 
issue was not treated yet. But it is regarded to be solved 
easily as there are many different algorithms available that 
are capable of segmenting clouds automatically due to their 
high intensity values and homogenous structure. 
6. ACKNOWLEDGEMENTS 
We would like to thank DigitalGlobe for kindly allowing us 
to use the imagery for our research. 
7. REFERENCES 
Bay, H., Ess, A., Tuytelaars, T., and Gool, L. V. 2008. “Surf: 
Speeded up robust features,” Computer Vision and Image 
Understanding (CVIU) Vol. 110, No. 3, 346-359. 
Bouguet, J-Y. 2000. Pyramidal Implementation of the Lucas 
Kanade Feature Tracker. Description of the algorithm. Tech. 
Report, Intel Corporation Microprocessor Research Labs. 
79 
Bovolin, V., De Chiara, G., Migliaccio, M. and Villani, P., 2006. 
Remote Sensing Technique to Estimate the Water Surface of 
Artificial Reservoirs: Problems and Potential Solutions, IEEE 
GOLD Remote Sensing Conference Bari 
Ernst, I. and Hirschmiiller, H. (2008). Mutual Information based 
Semi-Global Stereo Matching on the GPU, International 
Symposium on Visual Computing, Las Vegas 
Gehrig, S., Eberli, F. and Meyer, T. (2009). A Real-Time Low- 
Power Stereo Vision Engine using Semi-Global Matching, 
International Conference on Computer Vision Systems (ICVS), 
LNCS 5815 
Gehrke, S., Morin, K., Downey, M., Boehrer, N. and Fuchs, T. 
2010. Semi-Global Matching: An Alternative to LIDAR for 
DSM Generation?, In International Archives of the Photo- 
grammetry, Remote Sensing and Spatial Information Sciences, 
Vol. XXXVIII 
Hirschmüller, H. 2005. Accurate and Efficient Stereo Processing 
by Semi-Global Matching and Mutual Information, Proc. of 
IEEE Conference on Computer Vision and Pattern Recognition, 
San Diego, Vol. 2, pp. 807-814 
Hirschmüller, H., Scholten, F. and Hirzinger, G. 2005. Stereo Vi- 
sion Based Reconstruction of Huge Urban Areas from an 
Airborne Pushbroom Camera, Lecture Notes in Computer 
Science: Pattern Recognition, Proceedings of the 27" DAGM 
Symposium, Vienna, Austria, Vol. 3663, pp. 58-66 
Hirschmiiller, H. 2008. Stereo Processing by Semi-Global 
Matching and Mutual Information, IEEE Transactions on 
Pattern Analysis and Machine Intelligence, 30(2), pp. 328-341 
Hirschmiiller, H. and Scharstein, D. 2009. Evaluation of Stereo 
Matching Costs on Images with Radiometric Differences, IEEE 
Transactions on Pattern Analysis and Machine Intelligence, 
31(9), pp. 1582-1599 
Hirschmiiller, H. and Bucher, T. 2010. Evaluation of Digital 
Surface Models by Semi-Global Matching, DGPF Vienna 
Hirschmiiller, H. 2011. Semi-Global Matching - Motivation, 
Developments and Applications, Photogrammetric Week, 
September 2011, pp. 173-184 
Lowe, D.G. 2004. Distinctive Image Features from Scale- 
Invariant Keypoints. International. Journal of Computer Vision, 
60(2), pp. 91-110 
Lourakis, M. and Argyros, A. 2004. The design and 
implementation of a generic sparse bundle adjustment software 
package based on the levenberg-marquardt algorithm. Technical 
report, ICS-FORTH, Heraklion, Greece. 
Shi, J and Tomasi, C. 1994. Good Features to Track. Proc. of 
IEEE Conference on Computer Vision and Pattern Recognition, 
pp. 593-600. 
Tomasi, C. and Kanade, T. 1991. Shape and Motion from Image 
Streams: a Factorization Method - Part 3: Detection and Tracking 
of Point Features. Tech. Rep. CMU-CS-91-132, Computer 
Science Dept., Carnegie Mellon University. 
Wohlfeil, J. 2010. Completely optical orientation determination 
for an unstabilized aerial three-line camera. Proceedings of SPIE 
Sensors, Systems, and Next-Generation Satellites XIV, 7826. 
Wohlfeil, J. 2011. Dissertation: Optical Orientation Determi- 
nation for Airborne and Spaceborne Line Cameras, Humboldt 
Universität zu Berlin, Naturwissenschaftliche Fakultät II 
Wohlfeil, J. 2012. Determining fast orientation changes of multi- 
spectral line cameras from the primary images. /SPRS Journal of 
Photogrammetry and Remote Sensing, 67, pp. 45-51 
Zhaohui, Z., Prinet, V. and Songde, M., 2003 *Water body 
extraction from multisource satellite images”. [EEE, 0-7803- 
7929-2/03