Michele Crosetto
Coherence Number of Percentage Mean Error Standard Deviation
Range Pixels [%] [m] [m]
0.8 + 1.0 49183 7.8 0.34 5.01
0.7 — 0.8 129600 20.7 0.34 6.13
0.6 = 0.7 136767 21.8 0.66 8.06
0.5 = 0.6 105632 16.8 0.93 10.37
0.4 + 0.5 74602 11.9 0.77 12.95
0.3 - 0.4 51289 8.2 - 0.36 15.22
0270.3 34206 5.5 - 1.47 17.11
0.1 +02 19663 3 - 2.30 17.76
0.0 — 0.1 7874 1.3 - 3.01 17.96
Interpolated 18262 2.0 - 4.28 23.48
eration Table 3: Ascending InSAR DEM. Statistics of the height differences (InSAR versus reference DEM) computed for
tlic phase different coherence classes.
E dus The extraction of height information is problematic especially in the terrain slopes facing the SAR antenna (see the
he-heicht slopes with positive values in Figure 2) where foreshortening and layover occur. As a consequence, InSAR DEMs are
stan E d characterised by a peculiar feature: in the slopes facing the SAR antenna the DEM quality is severely degraded, while in
those bent away from the SAR look direction the quality is higher (if the slopes are not in shadow). The decrease of
DEM quality due to terrain topography is illustrated in Table 4, where the error statistics for different classes of slopes
cherence are reported. The quality decrease concerns both positive (i.e. facing the SAR antenna) and negative slopes, but
line ve comparing classes having the same absolute slope values one may notice an important difference in their standard
$ very deviations. For instance, the 33-36 40 class and the corresponding negative one have standard deviation of 22.5 m and
15.5 m respectively.
and the
ence (e.g.
. Dealing
'over and 0 1.5 km
grids. N m
Figure 2: Ascending InSAR DEM. Map of the terrain slopes (left) and the corresponding map of the height differences
— InSAR versus reference DEM (right). The slopes are calculated along the ground range direction, i.e. perpendicular to
the satellite track. Slopes facing the SAR antenna have positive values.
International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part Bl. Amsterdam 2000. 49