International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B3. Istanbul 2004 
The topological structure of boundary part is unique when there 
is exactly one given boundary model if coplanar bases are 
merged into one so that the base is forced to have maximum 
extent. Since some campus buildings include combinations of 
boundaries with different heights, boundary models are not 
closed under set operations. The union of two different height 
level boundary models does not result in a new valid boundary 
model. The edges or the polygon of one solid boundary touches 
any element of the other. Under this circumstance, these 
Boolean set operations become very difficult dealing with 
boundary walls. However, the data structure is still useful for 
modeling a man-made object including visualization tasks, 
architectural reconstruction and geospatial query. Figure 4b 
shows the residential buildings around the campus. Most 
buildings are well reconstructed. However, roofs of residential 
buildings show much more diversity than campus building. 
Some complicated roofs have one rectangular base, but roof 
points are not included in our primitive models. The approach 
fails to reconstruct buildings in those cases. This can be 
improved by applying more complex roof models in the future. 
The other reason for those features that cannot be correctly 
reconstructed is the erroneous data or missing points occur 
during measurement. Such faultiness certainly can be 
minimized through more careful data acquisition. 
  
  
  
  
  
Figure 4a). School buildings 
  
  
  
  
Figure 4b). Residence houses 
Figure 4. Reconstructed Purdue University campus buildings 
4. CONCLUSION 
In this study, we have developed a methodology for 3-D 
building reconstruction from unstructured distinct points. The 
558 
underlying mechanism is that the buildings with right angle 
corners can be rectangulated on each height level. Based on this, 
we propose a novel rectangulation approach to regularly 
construct these unstructured points. 
The rectangulation process facilitates the model identification of 
roof structures according to the construction between the roof 
points and corresponding closest rectangular bases. Moreover, 
once the roof models are determined, merging the rectangular 
bases also can easily outline the building boundary and finally 
reconstruct the building by including the vertical walls. In this 
paper, such a rectangulation approach for 3-D building 
reconstruction is introduced through regularization procedures. 
Through this key issue, the unstructured data points initially 
without any sequence can be constructed step by step from 
model identification to building-boundary projection. 
To demonstrate our methodology, we present successfully 
reconstructed Purdue University campus building and their 
comparison with aerial images. Because we consider three type 
CAD roof models only, some complicated roof types require 
more primitive models included. Our experience also shows that 
a reliable rectangulation approach is necessary for model 
classification. Buildings with non-right angle corners need 
certain modification and adaptation of the reported hierarchical 
methodology. 
REFERENCES 
Fischer, A., Kolbe, T.H., Lang, F., Cremers, A.B., Fórstner, W., 
Pliimer, L.,. Steinhage, V. (1998) Extracting Buildings from 
Aerial Images Using Hierarchical Aggregation in 2D and 3D, 
Computer Vision and Image Understanding, Vol. 72, no. 2, pp 
185-203. 
Grün, A. and Dan, H. (1997) TOBAGO — A Topology builder 
for the automated generation of building models, Automatic 
Extraction of Man-Made Object from Aerial and Space Images 
(I), Birkhauser, Basel, pp.149-160. 
Griin, A. and Wang, X. (1998) CC-Modeler: A Topology 
Generator for 3-D Building Models, ZJPRS, Vol.53, pp.286-295. 
Gülch, E. Muller, H.& Labe, T. (1999) Integration of 
Automatic Processes Into Semi-Automatic Buildin g Extraction, 
Proc. of ISPRS Conference "Automatic Extraction Of GIS 
Objects From Digital Imagery", September 8-10, (INVITED). 
Lang, F. and Fórstner, W. (1996) 3D City modeling with a 
digital one-eye-stereo system, Proc. of ISPRS Comm. IV, 4, 
Vienna, pp.261-266 
Norbert, H. and Brenner, C. (1998) Interpretation of urban 
surface models using 2D building information, Computer Vision 
and Image Understanding, Vol. 72, No. 2. pp. 204-214. 
Wenidner, U. (1997) Digital surface models for building 
extraction, Automatic Extraction of Man-Made Object from 
Aerial and Space Images (II), Birkháuser, Basel, pp.193-200. 
Zlatanova, S., Paintsil, J., Tempfli, K. (1998) 3D object 
reconstruction from aerial stereo images, Proceeding of the 6th 
International Conference in Central Europe on Computer 
Graphics and Visualization'98, Vol. III 9-13 February, Plzen, 
Czech Republic, pp. 472-478 
"Oi 
For 
pres 
moc 
ther 
play 
com 
ber 
the 
geo, 
geol 
sup] 
shoi 
be s 
Zlat