CA, 9-11 Nov. 1999
n the building polygons and the
R DSM.
Vv.)
the formula used is as shown in
ion available within ARC/INFO
using the building polygon as a
potential locational mismatch
id the LIDAR DSM could also
d. Dev. because the heights of
roneously be placed within the
z) [2]
ithin a building polygon , Z is
olygon and z is the height value
polygon.
on of Std. Dev. on the building
erid resolutions can be found in
D DISCUSSION
from LIDAR DSM and mean
the RMSE is not related to the
| rather is used to detect relative
vs the vertical roof heights for
gs for the derived 3D model at
both residential and industrial
he grid resolution of the LIDAR
lowever, there is a difference
its of residential and industrial
eight of residential buildings
the LIDAR DSM grid resolution
re 12). On the other hand, even
though the vertical roof height of industrial buildings decreases as
the grid resolution increases, it consistently shows a higher
vertical roof height compared to residential buildings. The values
for the computed vertical roof height are between 4.1m and 2.4m
as the grid resolution increases. It appears, therefore, from Figure
12, that industrial buildings have a higher vertical roof height
compared to residential buildings. Differentiation of the two
building types by analysing the vertical roof height at various grid
resolutions between 2m to 20m seems possible. It is also found
that, the maximum difference between the computed vertical roof
height (1.70m) is experienced at 2m-grid resolution. It is
suggested that differentiating the building type at a smaller grid
resolution (less than 2m) will improve the result. This is due to
the fact that more height values will be available from the LIDAR
DSM and the possibilities of detecting the ‘true’ maximum height
of both building types is therefore greater. Further studies in
relation to this effect will be carried out. In general, it is shown
that the vertical roof height that corresponds to industrial
buildings is higher than residential buildings.
Height (m)
63
4 4 son] pio gy ame si vio |
27 li a le ge
0 T Grid Resolution
2m 4m 6m 8m | 10m | 12m | 14m | 16m | 18m | 20m
—0— Resident] 24 | 17 | 13 | 13 | 12 | 16 | 18 | 16 | 16 | 13
~o—induswial | 41 | 32 | 27 | 29 | 28 | 29. 27 | 28 | 25 | 24
Figure 12: Vertical roof heights for the residential and industrial
buildings (Maximum height derived from the LIDAR DSM to
construct the 3D model and for control height, 3D model
constructed using the mean height from 2m-grid resolution
LIDAR DSM).
4.2 Mean height derived from LIDAR DSM and mean
height as control
As noted in Section 3.1 (Case IT), apart from using the maximum
height from the LIDAR DSM to construct the 3D model, the use
of mean height is also analysed. In this case, apart from
understanding the roof height differences between the two
building types, accuracy estimates for the derived model using the
mean height from the LIDAR DSM are known. It is shown that
grid resolution in the LIDAR DSM plays an important role in
constructing the 3D model. The accuracy estimates for both
building types constructed using the mean height from the
LIDAR DSM increases as the grid resolution increases (Figure
13).
RMSE (m)
2:
1l
| ; ;
0 J Grid Resolution
- T T
| 2m 4m 6m | 8m 10m | 12m 14m 16m 18m 20m
Fo rl |
|—@— Residential 0 0.5 0.9 12 1.0 1.6 1.9 16 16 16
| | |
---.-t : :
—e— Industrial | © 05 1.0 15 [15 | 18 17 17 19 | 17
—— 1 : | | |
Figure 13: RMSE for the residential and industrial building (mean
height derived from the LIDAR DSM to construct the 3D model
International Archives of Photogrammetry and Remote Sensing, Vol. 32, Part 3W14, La Jolla, CA, 9-11 Nov. 1999
and for control height, 3D model constructed using the mean
height from 2m-grid resolution LIDAR DSM).
Figure 13 shows that both building types exhibit an increase in
RMSE as the 3D model is constructed with an increase in LIDAR
DSM grid resolution. This indicates that the roof structure of both
of the building types is not flat and shows some variation in
height. The industrial buildings, in general, have a slightly higher
RMSE compared to residential buildings.
4.3 Maximum height derived from LIDAR DSM and
maximum height as control
Figure 14 shows the RMSE plot for the residential and industrial
buildings when the maximum height is used to construct the 3D
models at various grid resolutions. The control height adopted in
Case III is the height of a 3D model constructed using the
maximum height from the 2m-grid LIDAR DSM.
RMSE (m)
|
{
:
> Grid Resolution
e-—-
2m 4m 6m 8m | 10m [12m | 14m [16m | tem | 20m
—@-— Residential | O 0.9 17 22 2.2 2.9 17 34 | si 27
i 0 1.1 2.0 2.9 2.6 3.2 1.8 28122 25
—@E— Industrial | |
Figure 14: RMSE for the residential and industrial building
(maximum height derived from the LIDAR DSM to construct the
3D model and for control height, 3D model constructed using the
maximum height from 2m-grid resolution LIDAR DSM).
Figure 14 shows that the RMSE for residential and industrial
buildings increases as the grid resolution increases. In these plots,
there are no distinct differences between the pattern of RMSE for
residential and industrial buildings. However, industrial buildings
appear to exhibit higher RMSE values compared to residential
buildings. One of the advantages observed from Figures 13 and
14 is that RMSE values computed using the mean height derived
from the LIDAR DSM (Figure 13) are much smaller compared to
the equivalent values derived using the maximum height (Figure
14). It appears that 3D models constructed using the mean height
from the LIDAR DSM are more reliable than using the maximum
height. This is due to the fact that the mean height reduces the
risk that derived building heights are influenced by localised high
points such as tall chimneys or by the inclusion of adjacent
ground heights resulting from mismatch between the LIDAR data
and the polygon data. However, for certain applications, such as
the computation of inter-visibility between two points or
telecommunications applications, the use of maximum height
from the LIDAR DSM to construct the 3D model would be more
appropriate.
4.4 Maximum height derived from LIDAR DSM and mean
height as control for individual buildings
Referring to Figures 12 to 14, it can be concluded that there is
some sort of typical variation in vertical roof height for both
building type. Industrial buildings appear to have a higher
