6. RESULTS
The UltraMap v3 ortho processing algorithms already show up
a long testing and production phase. In the last years, Microsoft
produced a full US coverage and almost a complete European
coverage of 30cm ortho imagery based on UltraCam
technology. The joined project “Clear 30" between Microsoft
and DigitalGlobe was a big commitment in generating high
quality maps for most parts of the world.
In addition, we already applied our software on customer data
which have been kindly provided for testing. Figure 6 shows an
example of a DSM and DSMOrtho provided by City of Graz.
The presented dense matching approach allows for
reconstructing the fine crane structures. The conveyors are
shown with precise and sharp edges exploiting the redundancy
of the imagery (flown with a UC-Xp at 3cm GSD and with an
overlap of 60/60).
Figure 7 shows an example of a DTMOrtho which was also
generated in an automated way without any user interaction.
This imagery was provided by Ordnance Survey, UK. This
dataset was also flown with a UC-Xp at 10 cm ground
resolution. The DTMOrtho was projected on an internally
generated DTM using the presented constrained filter approach.
Figure 6 DSM and DSMOrtho screenshot showing a gravel pit
(data courtesy of City of Graz).
6.1 Processing Time
We have processed a demo dataset on our reference system. The
system consists of 4 machines each of which has 4 NVIDIA
Tesla cards built-in. Each machine has an i7 6 core processor
with 3.3 Ghz and 32 GB RAM. Our test dataset has 408
UltraCam Xp input images (each of which has an image
resolution of 17310x11310 pixels) and was processed on all 4
machines in parallel. The total processing time for generating
the DSM, DSMOrtho, and the DTMOrtho (without calculating
the time for Level-2 processing and aerial triangulation) was
less than 24 hours.
Figure 7 DTMOrtho showing a main junction without any
distortions (data courtesy of Ordnance Survey, UK).
Processing type Time [min]
DSM 190
DSMOrtho 546
DTM 39
DTMOrtho 661
7. REFERENCES
Amhar, F., & Ecker, R. (1996). An integrated solution for the
problems of 3d man-made objects in digital
orthophotos. International Archives of
Photogrammetry and Remote Sensing, (pp. 84-89).
Chen, N., Rau, J, & Chen, L. (2002). True orthophoto
generation built-up areas using multi-view images.
Photogrammetric Engineering Remote Sensing, 6, pp.
68-74.
Gruber, M., & Reitinger, B. (2008). UltraCamX and a new way
of photogrammetric processing. ASPRS 2008 Annual
Convention. Portland, Oregon.
Hirschmueller, H. (2011). Semi-Global Matching - Motivation,
Developments and Applications. Photogrammetric
Week '11, (pp. 173-184). Stuttgart.
Klaus, A., Sormann, M., & Karner, K. (2006). Segment-Based
Stereo Matching Using Belief Propagation and a Self-
Adapting Dissimilarity Measure. /CPR, (pp. 15-18).
Kolmogorov, V., & Zabih, R. (2004). What energy functions
can be minimized via graph cuts? IEEE PAMI, 26(2),
147-159.
Korytnik, P., Kuzmin, P., & Long, O. (2004). Polygon-based
true orthophoto generation. /SPRS:CI, (pp. 529-531).
Istanbul, Turkey.
Reitinger, B., Hoefler, M., Lengauer, A. Tomasi, R,
Lamperter, M., & Gruber, M. (2008). Dragonfly -
Interactive Visualization of Huge Aerial Image
Datasets. /nternational Archives of Photogrammetry,
Remote Sensing And Spatial Information Science, 31st
Congress in Beijing, (pp. 491-494). Beijing.
Uyttendaele, A., Eden, A., & Szeliski, R. (2006). Seamless
image stitching of scenes with large motions and
exposure differences. CVPR, 2, pp. 2498-2505.
Uyttendaele, M., Szeliski, R., & Steedly, D. (2011). Fast
poisson blending using multi-splines. /nternational
Conference on Computational Photography.
Wiechert, A., Gruber, M., & Ponticelli, M. (2011). UltraCam
Eagle, the new Super-Large Format Digital Aerial
Camera. Proceedings of the American Society for
Photogrammetry & Remote Sensing. Milwaukee, WI.
Zhou, G. (2004). Urban large-scale orthoimages standard for
national orthophoto program. [EEE International
Geoscience and Remote Sensing Symposium, 1, pp.
25-29.
144
