rtian)
face
(4)
ar zenith
Latitude,
ime angle,
.ance
is assumed
allows us
tmospheric
ction of
zle, solar
so on so
ble for
tmospher ic
radiance
d be take
tmospher ic
r is
condition,
Lambertian
s study so
on angles
tivity are
ization of
Figure 2 Real DEM data of Unzen, Kyushu in Japan
Contour line was derived from 50 x 50 m of meshed data
2.4 Simulated Anneal ing
The simulated annealing is one of
the optimization methods which
allows us to determine a parameter
with minimizing the energy, E which
is corresponding to the difference
between real and model derived at
sensor radiance.
Let me assume that parameter set, S
which is corresponding to the DEM.
3. Exper iments
The actual DEM data with 50 meter of
mesh size of Unzen area in Kyushu
Japan, which is provided by
Geographical Survey of Japan is used
to show the effectiveness of the
proposed method through a comparison
between the existing surface
reconstruction method with
non-linear least square method
(Levenberg Markard) and the proposed
method. Figure 2 shows the DEM data
which consists of 512 x 512 pixels
corresponding to the 25 x 25 Km?.
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B3. Vienna 1996
3.1 Geometric Relationship Among
Ie Satellite, the Surface and the
un
The 256 line is corresponding to the
solar azimuth angle with the
elevation angle of 45 degree that
implies the sun direction is middle
of the DEM data. This Solar azimuth
and elevation angles are variables
and can be designated. On the other
hand, satellite altitude is
designated at 800 Km. The earth is
assumed as the sphere with the
diameter of 6, 378 Km.
3.2 Atmospheric Transmittance
Assuming that wavelength region of
the mission instrument is 0.7um,
atmospheric transmittance for nadir
viewing is calculated as 0.86.
Meanwhile atmospheric transmittance
for the arbitrary viewing angle is
calculated with MODTRAN-3 in the
same manner.
3.3 Surface Conditions
The target area is mostly covered
with forest so that Lambertian
surface with 0.55 of spectral
