Zheng Wang
USGS digital ortho of quarter quad (DOQQO) and is displayed in figure 4. In the aerial photo, buildings can be recognized
and they match the rectangular humps in the elevation image.
uses
The
ition.
| TIN
Iding
oints
le in
have
lares
oints
rated
ibea
nake
y not
rtical
oints
; still
1 can
y on
ring
that
d by
roof
Figure 3. An elevation image of the test LIDAR data set. Figure 4. An aerial photo of the test area.
To obtain the best results, both edges and contours were generated from the test LIDAR data set. A comparison
between edges and contours clearly indicated that the contours were in much better shape than the edges in terms of
sharp right corners and straight segments. Because of this, contours were used for the rest of the test, see figure 5.
1 by
nine
data
rged
F the
with
m a
Figure 5. Contours generated from the elevation image. Figure 6. Contour classification results. Contours marked
with * were the final building contours.
There were total forty-one qualified contours and they were classified into building contours and non-building
contours. Table 1 gives a statistics of all forty-one contours in six categories after symmetry and circularity
classification. Of the total forty-one contours, sixteen of them were classified as building contours, see figure 6. The rest
International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B3. Amsterdam 2000. 961