2004 
  
:sults 
lings, 
same 
  
ferent 
N and 
Some 
dings, 
m the 
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B4. Istanbul 2004 
  
Simplification of buildings includes deleting too small building 
polygons and eliminating short segments and unimportant 
vertexes of a building polygon. Figure 8 shows the simplified 
buildings of Figure 7(b). 
Figure 8. Simplifying buildings 
3. EXPERIMENTSAL RESULTS 
We implemented some experiments to confirm the principle and 
the effectiveness of the proposed framework. Figure 9(a) shows 
the original dataset, including road edges and building polygons. 
The generalized result is shown in figure 9(b). Here, the 
modeled road polygons are filled with different colors, and the 
road networks are represented by black lines inside all road 
polygons. Also, all near building polygons are aggregated to a 
single one. And, both aggregated buildings and left original 
building polygons are simplified and rectified properly. 
4. DISCUSSION AND CONCLUSIONS 
This paper has presented a framework to generalize road edges 
and building polygons. In the proposed framework, the 
generalization of road edges and building polygons were carried 
out with three modules: Road Modeler, Network Generator and 
Building Generalization. Road Modeler worked to create road 
polygons from original road edges; Network Generator is used 
to collapse road polygons to road networks; and Building 
Generalization addressed to cluster building polygons, to 
aggregate building clusters and to simplify both aggregated 
buildings and left original building polygons. With the 
algorithms in the framework, the created road networks are 
topologically connected well and suitable for GIS, such as car 
navigation systems. Using the created road networks as the 
constraints for generalizing buildings leads to the gencralized 
results without contradiction. Some experiments have also been 
implemented and the experimental results showed that the 
present framework runs well. 
The performance of an automatic process usually varies with the 
quality of the input data. The proposed framework and 
algorithms here are also partially limited by the input road 
edges and building polygons. For example, one-side or crossed 
road edges will influence the processing results of Road 
Modeler. One-side road edges mean that there are only right or 
left curbs for some road region. Crossed road edges may exist 
for overpasses or underpasses, such as a bridge over the other 
road. For such cases, the input data should be revised before 
automatic processing. These problems are going to be settled in 
the future works. 
  
Ny 
SS AM. NS 
(a) Original Roads and Buildings