The straight-line detection process is performed for all line 
segments of the rough building footprints. The intersection 
points between the adjoining the straight line are calculated. 
And then, the each intersection points of all line segments are 
exported as the corrected building footprints. The results are 
shown in Figure 16. In this Figure, the rough building footprints 
are shown by the dotted line, and the corrected building 
footprints are shown by the solid line. 
Ean 
> 
     
720 740 760 780 500 
X {m} 
   
wi 
Figure 16 Straight-Lines Detection from Edge Image 
6. 3D CITY MODEL CREATION 
6.1 Building Label Image Creation 
The creation process of the building label image is as follows. 
And the results are shown in Figure 17. 
(1) The building footprints are sorted due to the area of 
polygons. The polygon’s labels are obtained as its order. 
(2) The label of buildings is over written onto label image in 
descending order of building footprint’s area. 
1000 
   
Y [m] 
ü BE ; 
B 200 B 1000 
X Im] 
Figure 17 Label Image of Buildings 
6.2 Determination Process of Building Height 
The determination process of the building height is as follows. 
(1) The laser point clouds are extracted based on the label 
image that is created in Section 6.1 from DSM and DEM. 
(2) The building height of upper level is determined based on 
the median (or mode) of height of the extracted points. 
(3) The building height of lower level is determined based on 
the minimum of height of the extracted points. 
6.3 Display of 3D Building Models 
The building models are displayed in the 3D Coordinates. The 
example is shown in Figure 18. In this section, the building 
polygons are risen up based on the height of building and the 
building’s walls are created vertically. 
  
Figure 18 3D City Models 
7. CONCLUSION 
In this study, the buildings footprints are tried to extract form 
the laser point clouds and the ortho images. In the case of the 
only using laser point clouds, the accuracy of vertexes 
coordinates is not enough, because the sampling interval of the 
laser point clouds is measured roughly as approximately 2m. 
And then, the buildings footprints are corrected with the exact 
edge lines that are detected from the ortho images In the case of 
the laser point clouds, it is used only to extract the rough 
building footprints. The real line segments are limited due to the 
rough building footprints, and then the straight lines can detect 
stably. The 3D City Models are created based on the results 
with the simplified method. 
In the present, we find the miss judgments into the straight lines 
detection from the ortho images, because it detects the road 
edges of contiguous to the buildings. And the building edges are 
seeing rarely doubled in the overlap area of ortho images. It is 
necessary to solve together with the problems of the high 
buildings are lying on the ortho image. And the correction 
process is not enough in the curve sections, because the 
correction process is assumed the straight lines as line segment. 
Therefore, we will try to use the method of Generalized Hough 
Transform for the correction process in the curve sections and 
to extract more exact building footprints in future. 
REFERENCES 
Zhao B. and Trinder J. (2000), Integrated Approach Based 
Automatic Building Extraction, Proceeding of the 19 th ISPRS 
Congress, Book 3B, pp: 1026-1032, Amsterdam. 
Abdullatif Alharthy and James Bethel (2002), Buildiing 
Extraction And Reconstruction From Lidar Data, XXII FIG 
International Congress ACSM-ASPRS Conference and 
Technology Exhibition, April, 2002 * Washington, DC 
Michel Morgan and Ayman Habib (2002), Interpolation Of 
Lidar Data And Automatic Building Extraction, XXII FIG 
International Congress ACSM-ASPRS Conference and 
Technology Exhibition, April, 2002 * Washington, DC 
—226—