Hiroshi Masaharu
The 3-
mitigat
a build
In this
homog
[2 UE IIT
A 100m
Figure 1. Laser Scanner Data of Minokamo City Figure 2. Bird's-eye View Representation
(A Portion) of Laser Scanner Data
The data were taken in February, 1998 and shown as Figure 1 in layer
tint image. The data used here is the height data in regular grid
spacing of 0.5 m produced by the software of the sensor system.
Bird's-eye view of this original data is shown in Figure 2. An area
were cut from the original data for testing the method. The size of the
area is 400 x 800 pixels, which corresponds to 200m x 400m. The
threshold of region segmentation was determined as 0.4 m after try and
error. Figure 3 shows the obtained polygon data after generalization.
3-D model is obtained by determining heights and roof shapes of
buildings using the laser scanner data and the polygon data. Figure 4
shows the obtained 3-D model as bird's-eye view.
3.3 An Example of Tsukuba City
Tsukuba city was newly developed around 1980 according to the
government policy to concentrate national research institutes. The
population is about 160,000. Many of the buildings in the city central
area are high-rise buildings with 4 to 13 stories and detached houses are
few. As the size of each building is large, one-meter resolution height
data, which was made by averaging 2 x 2 pixels of the original data, was
used in the processing. The size of the area is 1200 x 600 pixels and is
1200m x 600m. The threshold for region segmentation was 0.8 m.
Obtained polygon data is shown in Figure 5 and the bird’s-eye view of
the 3-D model is shown in Figure 6.
4 DISCUSSION
4.1 Evaluation of the Results
\
The polygons extracted by region segmentation (Figure 3 and Figure 5) Figure 3. Extracted Polygons by Region Wi
are rather complicated and somewhat distorted. Actual shapes of Segmentation, Minokamo City ins
buildings and houses are very complicated when objects on the roofs are e
taken into account. Furthermore many buildings have parts with different stories. There are also many trees along the n 3
street in both cities. The complicated structure reflects these circumstances. We consider the distortions of the eR
resulting polygon shapes are mainly due to the position and attitude measuring error of the sensor system. We expect
that more accurate shapes can be derived from laser scanner data in the near future by improved sensor systems.
558 International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B3. Amsterdam 2000.