International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part Bl. Istanbul 2004
higher change in forest areas is due to mismatching in these
areas. Many falsely matched points are eliminated in the case of
the first filter and blunders are eliminated by the median
filtering.
Table 3: Comparison of the regularized SPOT-DEM and
the reference DEM for selected areas and two surface types
(SPOT DEM - Reference DEM)
Reference | Mean Height | Mean Height Height-
area Difference [m]| Difference [m] difference
non-forest/ | forest/ STDV | forest/non-
STDV forest
DEM-02 78 1 56 176 / 97 11.9
DEM-02 85 / 48 150 / 80 7:5
statist. filter
DEM-02 80 / 39 172 / 82 9.2
median filter
DEM-06 65 / 65 19.0 / 9.0 12.5
DEM-06 62 / 42 16.9 / 6.8 10.7
median filtering
Fig. 4 shows the difference image of the Laser reference DEM
to the SPOT DEM, forest areas can be seen clearly (brighter
grey values) because of the higher mean difference. Also some
blunders can be seen in the lower right part of the DEM, which
has not been filtered in this case.
2
Figure 4. Lower part: Difference DEM (SPOT DEM
— Reference DEM): Bright: forest areas and some
blunders, the black parts have no value in the
reference DEM. Upper part: Map of the same
Another possibility of comparison is to look at profiles of the
DEMS along a given line. In the profiles, the structure of the
DEM and its variability can be seen easily. Fig 5 shows the
profiles along the same line for the SPOT DEM and the
reference DEM. The rough structure of the profile is very
similar, but there is more variability in the SPOT DEM (Fig. 5
without filtering!).
1300
1040 === ===
1
i
1
! | /
+ 1 1
snas VSEVED mE MUT 4 130.0
: : M ;
So X i i
i 1 + i
py 1 1 1
oo 104.0 2080 312.0 4160 520.0
Figure 5. Profiles of SPOT-DEM (upper) and reference
DEM (lower)
7, COMPARISON AND FUSION OF SPOT-DEM WITH
SRTM-DEM FOR BARCELONA AREA
Additional to an area based comparison with the reference
DEM, a SRTM-DEM (derived from data of the Shuttle Radar
Topography Mission, Bamler et al. 2003) which is available for
most of the area has been investigated. Table 4 shows the mean
differences as well as standard deviation and min/max values of
the differences for four different areas in and around Barcelona.
Table 4: Area-wise comparison of height of SPOT-/SRTM-
DEM and reference DEM
Reference Size of Mean Height | STDV | Min/Max
area Area Difference [m] | [m]
SPOT-HRS-DEM
Barcelona 71 km? 11.2 4.4 47 / +37
City
Rural Area 161 km? 10.4 5.9 -59 / +53
Moderate 105 km? 11-1 6.5 -62 / +63
Mountain
Montserrat 84 km? 9.8 13.5 | -158 / *191
Whole area | 1882 km? 10.0 6.3 | -158/ *191
SRTM-DEM
Barcelona 74 km? 1.0 A7 -22/ *25
City
Rural Area 161 km? -1.5 4. 432 -98 / +62
Moderate 105 km? -1.4 8.7 | -215/ +135
Mountain
Montserrat 84 km? -1.8 25.2 | -484 / +394
Whole area 623 km? -1.2 8.5 | -484 / +394
Mean height differences are very similar in all four cases, while
the standard deviation for HRS-DEM gets higher with more
slopes and forest areas. In the case of Montserrat, with very
steep slopes (many above 45°) the standard deviation and the
min/max values become very high. Applying a coarse
classification of forest areas, the mean height difference in
those areas is only 2-3 meters higher. This can be due to less
dense forest cover than in the Bavarian case. The SRTM-DEM
shows a very high accuracy in absolute height values due to the
Intei
sea |
devi
mou
effe
diffi
mod
fusi
AD
laye
prod
dens
optic
stant
mat
DEN
and
com
abso
meas
imp
one
from
qual
Tal
anc
R
Bar
IRur
Mo
Mo
Mor
Wh
Fron
prov
in al
reme
conti
blocl
is us
modi
from
valu
table
2003
Cer
intro
the v
was
test s
Tal
