International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B3. Istanbul 2004
(average values) are the opposite direction pairs. The results are
significantly different between them. It seems that the centre
points of the buildings have moved about 5 cm to the flight
direction. There is an across track shift (<5 cm) between the
flight strips with the same flight direction. The standard
deviations of the differences for opposite direction flight strips
are larger than for the flight strips to the same direction. To
conclude, the results indicate that the flight direction was a
significant reason for planimetric errors.
In Figure 4 comparisons were made between the last pulse data
and the Espoo city map buildings. The coordinate
transformation from the Espoo system to the Finnish national
system kkj also brings some errors to the results. There is a
systematic shift of the map buildings to the east (northeast)
compared to the laser derived buildings.
Mean north and east differences between laser derived
buildings and maps
0,25 x
02 | Onorth
= 0,15 Bl east
$ 01
o
- = 0,05 ;
=
se 0 ; :
z ©
w -0,05 |]
9 2. 23 2 2
o o Uv v o Uu
= -0,1 2
T 0,15 :
& mn
© 0:2 i + ;
003- 005- 00Z- O0 006- 003- 005- 002- 004- 006-
-0,25 "map--map--map--map- map -map- map--map-map--map-
Compared flight strips with map
Figure 4. Comparison of different last pulse flight strips with
map buildings. 3 and 5 were flown to the southeast
and 2, 4 and 6 to the northwest.
The use of ridges of buildings
A ridge of a building roof can be obtained in a more accurate
way than the edges and corner points of the building since the
ridge is defined by two intersecting planes that are determined
using even hundreds of laser points. Thus, previous conclusions
were checked by calculating the shift of the ridges between
repeated acquisitions. Table 7 summarizes the comparison of
these two approaches.
Compared flight strips Distance between
Centre points Ridges
Same 5-3 14+12 S43
direction 2-4 1128 4+4
2-6 9+9 3+4
4-6 14x11 SS
Opposite 2-3 13:13 6:5
direction 4-3 17:12 846
6-3 18x15 6x6
2-5 16:8 7X4
4-5 2049 7X4
6-5 19:9 7X4
Table 7. Mean differences and standard deviations (cm)
calculated from centre points and ridges of ridge
roof buildings. Last pulse data.
The compared flight strips and the modelling method of the
buildings are the same. Again flight strips to the same direction
gave smaller biases than the flight strips to the opposite
directions. However, obtained errors are significantly smaller
using ridges than centre points of buildings.
3.3 Analysis of size of buildings
The areas of the roofs were measured along the slopes (i.e. to
the normal of the roof) Areas were calculated for every
building as a mean value of the five flight strips. The
differences between the first and the last pulse mode derived
roof areas in percentage were calculated for each building and
the mean values were obtained for each roof type, c.f. Table 8.
Internat
Directic
of edg:
Across
Alone
TU
Obliqu
Total
Table 9
The me
number
observa:
the roof
the roof
results |
expectec
cm) bec
quite sn
the first
lengths :
Directio
of edge
Across
Along
Oblique
Total
Roof First pulse | Last pulse | Mean
type flight flight difference
and # strips strips
Hipped | Mean area | 163.0 m“ 145.9 m” 11.6 96
roof, minimum
N=6 Mean area | 166.8 m? 151.3 m
maximum
Double | Mean area | 165.7 m“ 145.6 m* 14.0 %
slope minimum
pitch Mean area | 167.8 m? 146.8 m*
roof, maximum
N=3
Flat or | Mean area | 17.9 m* 16.3 m° 44%
pitch minimum
roof, Mean area | 171.6 m 159.7 m?
Nes maximum
Ridge Mean area | 94.3 m“ 58.8 m° 13.2 %
roof, minimum
N=20 Mean area | 234.8 m‘ | 211.5 m“
maximum
Connec | Mean area | 88.3 m° 85.8 m° 11.4%
ted minimum
ridge Mean area | 189.8 m 170.5 m?
roof, maximum
N=5
Table 8. Mean differences between the first and the last pulse
mode derived roof areas.
The buildings were one- or two-storied and relatively small
ones. Differences between the first and the last pulse measured
areas were 11-14 % except for the flat/pitch roofs (4.4 96) when
the trees obstructed the views of these roofs. This flat or pitch
roof building group included also car shelters that were lower
than the human settlements. The first pulse mode observations
tend to exaggerate the area and the last pulse data systematically
underestimate the results.
Roof lengths were calculated from the RTK measured roof
corner points and these lengths were compared with the laser
derived roof lengths. The laser derived roof lengths were
actually the side lengths of the extracted roof planes. Because
the Toposys laser scanner has a special scan pattern the lengths
were divided to three categories: lengths across the flight strip,
along the flight strip and oblique to the flight strip. The distance
between the scan lines across the fight strip direction is about
80 cm (footprints of neighbouring fibres) when the flying height
is 400 m. In the along track direction the pulses overlap on the
ground. Results are shown in Tables 9 and 10.
Table 1(
The last
directior
across aj
(-81 cm)
lengths.
about 2.
measure
The obt:
Also pl:
strips flo
The mea
and stan
it was +
results fc
Centre p:
test how
obtained
the accu
Rónnhol
internal
data is gi
The mea
derived |
than 20
+9-16 cn
building