XIX-B3, 2012 
  
d planes to serve 
t clouds derived 
rary, and one ad- 
, taken from the 
On the one side, 
ly shaped build- 
ult objects as L- 
rated facades or 
, and stairs. We 
our algorithm in 
lves during all 
1000 iterations. 
iaset to another, 
a known metric 
ute value of this 
major four verti- 
rts, such as bal- 
g models, so far. 
onding to minor 
ien constructing 
with this refine- 
t level-of-detail. 
21s by additional 
OD 2 models. 
  
our major walls 
loud consists of 
sfully detected, 
£g cuboid-based 
1 from multiple 
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume XXXIX-B3, 2012 
XXII ISPRS Congress, 25 August — 01 September 2012, Melbourne, Australia 
  
Figure & Left: Image of building consisting of four major walls 
standing on a slope. Right: The derived point cloud consists of 
273 000 points, the four major walls were successfully detected, 
segmented and connected. 
  
Figure 9: Left: Image of building consisting of four major walls 
having a dominant entrance and stairs at one side. Right: The 
derived point cloud consists of 271 000 points, the four major 
walls were successfully detected, segmented and connected. 
images. It detects vertical planar surfaces assuming the dom- 
inance of orthogonal intersections. The approach was tested 
with point clouds of buildings from image matching with vary- 
ing scale, point density and amount of noise and has proven its 
robustness. 
In future, we want to produce a closed building footprint harmo- 
nizing the height of each wall and its neighboring walls. Then we 
are able to estimate the top and the bottom borders of the build- 
ing, yielding a final closed polyhedral for the building. i.e., its 
LOD | model. Afterwards, we will refine the model to models 
with more details, LOD 2 and higher. Also, we anticipate as a 
future work, the integration of many more geometric primitives 
such as spheres, cones, and cylinders. 
References 
Agarwal, S., Furukawa, Y., Snavely, N., Curless, B., Seitz, 5. M. 
and Szeliski, R., 2010. Reconstructing Rome. IEEE Computer 
Almansa, A. Desoineux, A. and Vamech, S., 2003. Vanish- 
ing Point Detection without Any A Priori Information. IEEE 
Transactions on Pattern Analysis and Machine Intelligence 
25(4), pp. 502-507. 
  
Figure 10: Left Image of building consisting of four major walls 
showing balconies or oriels. Right: The derived point cloud con- 
sists of 64 000 points, the four major walls were successfully de- 
tected, segmented and connected. 
    
149 
Figure 11: Left: Image of building consisting of four major walls 
showing balconies or oriels. Right: The derived point cloud con- 
sists of 1.6 million points, the four major walls were successfully 
detected, segmented and connected. 
  
Figure 12: Left: Image of the inner yard of a building consist- 
ing of four major walls with many ornaments and two cylindrical 
towers in the yard's corners. Right: The derived point cloud con- 
sists of 1 million points, the four major walls were successfully 
detected, segmented and connected. 
Bartelsen, J. and Mayer, H., 2010. Orientation of Image Se- 
quences Acquired from UAVs and with GPS Cameras. Land 
and Information Science 70(3), pp. [51-139. 
Becker, S. and Haala, N., 2008. Integrated LIDAR 
and Image Processing for the Modelling of Building Fa- 
cades. Photogramme trie-Femerkundung—Geoinformation 
2008(2), pp. 65-81. 
Berkman, J. and Caelli, T., 1994. Computation of Surface Ge- 
ometry and Segmentation Using Covariance Techniques. IEEE 
Transactions on Pattern Analysis and Machine Intelligence 
16(11), pp. 1114-1116. 
Fischier, M. A. and Bolles, R. C., 1981. Random Sample Consen- 
sus: A Paradigm for Mode! Fitting with Applications to Image 
Analysis and Automated Cartography. Communications of the 
ACM 24(61, pp. 381—395. 
Fárstner, W., 2010. Optimal Vanishing Point Detection and Ro- 
tation Estimation of Single Images from a Legoland Scene. In: 
The International Archives of the Photogrammetry, Remote 
Sensing and Spatial Information Sciences XXXVII (3A), 
pp. 157-162. 
Frahm, J.-M., Fite-Georgel, P, Gallup, D., Johnson, T., Raguram, 
R., Wu, C., Jen, Y.-H., Dunn, E., Clipp, B., Lazebnik, S. and 
Pollefeys, M., 2010. Building Rome on a Cloudiess Day. In: 
{ fth European Conference on Computer Vision, pp. 368-381. 
Hansen, W.. 2007. Automatic Detection of Zenith Direction in 
3D Point Clouds of Built-Up Areas. In: Photogrammetric Im- 
age Analysis (PIA07), IAPRS 36 (3/W49A), pp. 93-07, 
Hart, P. E., 2009. How the Hough Transform was Invented. IEEE 
Signal Processing Magazine 26(6), pp. 18-22.