nts, 
oint 
the 
two 
nce 
le is 
| as 
ied, 
ime 
the 
; on 
t is 
dge 
nd, 
ige, 
ace 
ot 
ere 
hio 
ical 
ich 
The 
lare 
hes 
the 
2b, 
S is 
two 
The 
ph's 
the 
the 
| are 
ring 
vitl 
ngs 
Th. 
contour interval used was 4 meters. Figure 3c depicts all 
detected humps. At this stage the number of humps, the 
locations and boundaries of the humps become known. 
Additionally, the elevations and shapes of the humps are 
determined as well. 
After the hump detection, all edges are associated 
to humps or topographic surface based on their 
geometrical locations. To test the results of this grouping 
process, a DEM was generated for every hump using 
only the edges belong to the hump. Figure 3d and 3e are 
two samples of them. One of the two humps is OSU 
library, and the other one is University Hall. The two 
humps have the same shape as they are in the DEM 
surface in Figure 2c, which indicates the result of 
grouping is correct. Finally the Figure 3f shows the 
DEM of the topographic surface after all the humps have 
been removed, with the exception of two incomplete 
humps(the contours of these two humps are not closed). 
Figure 4 shows the results of classification. Based 
on the derived information of edge properties, we 
generated the top of OSU library in Figure 4a by using 
all horizontal edges which are above the topographic 
surface in the hump "library". Figure 4b is a combination 
of vertical edges and horizontal edges which are above 
the topographic surface. 
The derived hump information and edge 
properties are made available to the matching anc 
interpolation processes. With this information, th: 
matching improved considerably[Zong, 1992]. The 
improvement of the interpolation part is shown in Figure 
5. Here we show the DEM after a new interpolation took 
place with hump information. The result in Figure 5 
demonstrates that the building walls in Figure 5 are more 
vertical than those in Figure 2c. 
5. CONCLUSION 
Surface reconstruction of urban areas is a very 
important step towards the automation of mapping 
processes. À complete surface is essential in order to 
recognize man-made objects and interpret images. 
Surface analysis is a key part of the OSU surface 
reconstruction system. 
The experimental results demonstrate that the 
surface analysis can substantially improve the matching 
and interpolation of the surface of urban area. 
Additionally, the results od the hump detection can be 
used to recognize buildings. 
ACKNOWLEDGMENTS 
Funding for this paper was provided in part by the 
NASA Center for the Commercial Development of Space 
Component of the Center for Mapping at The Ohio State 
University. The authors would like to thank Ms. Jia Zong 
for providing 2D edge matching results for the research, 
and Mr. W. Cho for providing some assistance. 
REFERENCE 
Al-Tahir, R., 1992. On the Interpolation Problem of 
Automated Surface Reconstruction. Proceedings of 
ISPRS. 
Attneave, F., 1954. Some Informational Aspects of 
Visual Perception. Psychological Review. vol. 61, No 3, 
pp183-193. 
Besl, P., 1988. Surfaces in Range Image Understanding. 
Springer-Verlag. 
Bian, Z, 1988. Pattern Recognition. Qinghua 
Dave Publishing Press. Beijing, China pp272- 
74. 
Fan, T., 1990. Describing and Recognizing 3-D Objects 
Using Surface Properties. Springer-Verlag. 
Fisher, R., 1989. From Surfaces to Objects. J. Wiley. 
Chichester, NY. 
Marr, D., 1982. Vision: A computational Investigation 
into The Human Representation and Processing of Visual 
Information. W.H. Freeman. San Francisco. 
McCafferty, J., 1990. Human and Machine Vision. Ellis 
Horwood Limited. Chichester, England. 
Schenk, A. and Toth, C., 1991. Knowledge-Based 
Systems for Digital Photogarmmetric Workstations. In 
Ebner H., Fritsch, D., and Heipke, C.(Ed.): Digital 
Photogrammetric Systems. Wichmann. pp123-134. 
Zong, J. Li, J., and Schenk, T., 1992. Aerial Image 
matching Based on Zero-Crossings. Proceedings of 
ISPRS. 
723