Environmental Engineering Reference
In-Depth Information
R
0
01110001
R
1
R
2
R
3
W
0
W
1
W
2
W
3
AB
B
1
B
2
B
3
B
4
110011
111100
101001
010110
001001
001100
000110
000011
B
1
R
0
R
0
R
1
R
2
R
3
W
0
W
1
W
2
W
3
W
0
R
1
R
2
R
3
W
0
W
1
W
2
W
3
0110100
0
R
2
B
4
01000
00010
0101
010
01
0
A
W
1
R
1
R
0
W
3
B
B
2
R
3
W
2
B
3
(a)
(b)
(c)
FIGURE 6.10
A hipped roof model.
a hipped roof. Figure 6.10(b) presents the adjacent matrix of the
roof surfaces and vertical walls in Fig. 6.10(a), where
R
0
∼
R
4
are the roof surfaces,
W
0
∼
W
4
are the vertical walls. Under
the adjacent matrix, the corner-surface matrix can be formed as
shown in Fig. 6.10(c), whose purpose is to get the coordinates of
the building corners. Typically, the coordinate of a spatial point
needs at least three surfaces to solve. Thus, in the corner-surface
matrix, each column represents a corner of the 3D building
model. Corners located within the boundary can be calculated by
the intersection of the roof surfaces. For example, corner A can be
calculated by the intersection of the adjacent roofs
R
0
,
R
1
,and
R
2
.
Corner
B
can be calculated by the roofs
R
0
,
R
1
,and
R
3
.Inreality,
the ridge line is horizontal, so the height value of the corner
A
is equal to that of the corner
B
. Similarly, corners
B
1
∼
6.3.1
Datasets
The lidar datasets used in this study were collected over 1000m
above ground level by an Optech ALTM 3100 system. The system
also carried a 4k
4k digital camera. The datasets include the
first and the last returns of the laser beam and the true color
aerial imagery. Figure 6.11 shows a residential scene, which is a
subset of a typical urban area in the City of Toronto, Ontario,
Canada. As shown in Fig. 6.11(a) presents the raster-based DSM,
containing a total of 105,298 points, which was interpolated with
both the first and last pulse returns by the bilinear interpolation
method. The width and height of grid equals to the ground sample
distance (GSD) of the aerial image (0.5m). The elevation of the
study area ranges from 150.00m to 178.11m. Besides buildings,
there are several clusters of trees. Figure 6.11(b) shows the true
color aerial image that was resampled to 0.5m ground pixel. The
majority of buildings appeared in the color image are with gable
roofs or hipped roofs. Figure 6.11(c) and 6.11(b) illustrate the
lidar range images of the first and last returns, respectively.
×
B
4
can
also be calculated by the intersection of roofs and vertical walls.
For instance, four adjacent surfaces (two roof surfaces
R
1
and
R
2
, two vertical walls
W
0
and
W
1
) are used to get the corner
B
1
.
More attention needs to be paid to the heights of the corners
B
1
∼
B
4
. They should be the same value because they generally
are horizontal in the real world. For a building with a hipped roof
(see Fig. 6.10a), the coordinates of six corners with 14 unknown
parameters are retrieved by the least-squares adjustment.
6.3.2
Results
6.3
Results and discussion
Figure 6.12(a) shows the filtered results overlaid on the color
aerial image. The results demonstrate that all the on-terrain
points have been removed quite well. The parameter of the
iterative distance in the filtering algorithmwhich is usually below
This section first describes the lidar point cloud data and color
aerial image data used in this chapter and then presents the results
of building extraction.
178.11 m
150.88 m
Low ------------------------ High
(a)
(b)
(c)
(d)
FIGURE 6.11
Lidar range data and colour aerial image of the study area: (a) DSM, (b) true colour aerial image, (c) first return
range image, and (d) last return range image.
Search WWH ::
Custom Search