Petra Zimmermann 
  
Bounding Box Criteria within Blob i Preferences 
.Rloh of the blob of hue values Small 
E A of hue values Not n, blue 
X No of i Small 
. Segment C (Tree) No of 1 : x 
  
  
  
   
  
No of “holes” in on Few 
Magni of ients 
  
  
Segment B H ity in texture 
Segment A (Roof) direction and distance to ridge line Small 
(Roo) — Mum Wo Distance of centre of gravities from blob and Low 
Segment D Segment 
(Tree) No of pixels o blob 
ient peri to No of pixels Low 
Figure 17: Examples for several segments Table 2: Predefined preferences of the BuildingRecognitionAgent 
5 | CONCLUSIONS AND OUTLOOK 
We showed possible improvements for building detection through combining multiple cues, colour segmentation, edge 
detection, texture segmentation and blob detection. The resulting attributed image primitives may be used directly to 
further derive 3D information, e.g. to perform and constrain edge matching with colour edges. We have to test our 
algorithm for the whole area of the new Zürich Hóngg dataset including the old centre of the quarter. The results of our 
subset datasets showed good results, the detected blobs contained 95% of the buildings compared to a manually 
measured dataset and visual inspection, within these blobs about 4046 of the long strong ridgelines could be found by 
computing slope and aspect. At the moment blobs indicate the presence of buildings as a first guess, an additional check 
should be done to detect buildings that could not be detected by blobs. Merging similar colour regions may produce 
misclassifications and has to be improved, also the last step in combining the features by has to be improved. Our main 
goal is to refine this system for processing aerial imagery of dense settled areas and also to able to extract building 
details, this should be able with the same algorithms but adjusted parameters in an iterative processing. Results from 
point and edge matching (Zhang et al., 2000), (Park et al., 2000) will be integrated. 
ACKNOWLEDGEMENTS 
The city of Zürich collected the Dataset. Maria Pateraki contributed by measuring the DSM with the software Virtuozo, 
DEM and reference data were measured with software from Xinhua Wang through his wife Ms. Fu. 
REFERENCES 
Grün, A., E. Baltsavias, O. Henricsson (Eds.) 1997. Automatic Extraction of Man-Made Objects from Aerial and Space 
Images (II), Birkhäuser Verlag, Basel. 
Grün, A., Wang, X., 1999. CyberCity Modeler, a tool for interactive 3-D city model generation, Proceedings of German 
Photogrammetry Week, Stuttgart, Germany. 
Gülch, E., Müller, H., Lübe, T., Ragia L. 1998. On the Performance of Semi-automatic Building Extraction, IAPRS Vol 
32, Part 3/1, pp. 331-338. 
Haralick, R.M., Shanmugam, K. and Dinstein, I. 1973. Textural features for image classification. IEEE Transactions on 
Systems, Man, and Cybernetics, 3(6), pp.610-611. 
Henricsson, O. 1996. Analysis of image structures using color attributes and similarity relations, PhD thesis No. 11663, 
ETH Zurich. 
Henricsson, O. 1998. The Role of Color Attributes and Similarity Grouping in 3-D Buiding Reconstruction, Computer 
Vision and Image Understanding, Vol. 72, No. 2, November, pp. 163-184. 
Mason, S., Baltsavias, E., Stallmann, D., 1994. High Precison Photogrammetric Data Set for Building Reconstruction 
and Terrain Modelling, Internal Report, Institute of Photogrammetry and Geodesy. 
Niederóst, M., 2000. Reliable reconstruction of buildings for digital map revision, accepted for IAPRS, Vol. XXIII, 
Amsterdam 2000. 
Park, H.J, Zimmermann, P. , 2000. Colour Image Matching for DTM generation and house extraction, accepted for 
IAPRS, Vol. XXXIII, Amsterdam 2000. 
Scholze, S.,Moons, T., Ade, F.,Van Gool, L., 2000. Exploiting colour for edge extraction and line segment stereo 
matching in high resolution aerial imagery, accepted for IAPRS, Vol. XXXIII, Amsterdam 2000. 
Skidmore, A.K., 1989. A comparison of techniques for calculating gradient and aspect from a gridded digital elevation 
model, Int. J. Geographical Information Systems, 3 (4): 323-334; London, New York, Philadelphia. 
Zhang, C., Baltsavias, E. 2000. Knowledge-based image analysis for 3D edge extraction and roadreconstruction, 
accepted for IAPRS, Vol. XXXIII, Amsterdam 2000. 
Zimmermann, P. 2000. Description of the Dataset of Zürich Hóngg, A new high precision dataset for building 
reconstruction and terrain modeling, Internal Report, Institute of Geodesy and Photogrammetry, ETH Zürich. 
  
1070 International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B3. Amsterdam 2000.