Proposed Methods
Table 1 A Comparison Between RMS Errors for the Existing and the
Method Root Mean Square Error in Unit of Meter
The Existing Method with
Least Square Method
44. 8
The Proposed Method with
Simulated Annealing
28. 3
reflectance at the 0.7um is
assumed.
3.4 Model derived at sensor
radiance image
Mode! derived at sensor radiance
image can be generated throughout
from simulated annealing. Figure 3
shows an example for the image which
is corresponding to the Unzen,
Kyushu in Japan. Through a
comparison between real image and
the model derived at sensor radiance
image, it is clear that the both
images are very similar except the
detailed portions around steep areas.
3.5 RMS error
The root mean square errors on DEM
estimation for the existing method
and Ee proposed method are shown in
able 1.
The results show that the proposed
method is superior to the existing
method in terms of root mean square
error of DEM estimation accuracy.
One of the reasons for this is that
the existing method does not take
into account the physical model and
that there are too many unknown
variables in the minimization of the
real and model derived at sensor
radiance. On the other hand, the
proposed method take into account
the physical models of the geometric
relationship among the satellite,
the surface and the sun, atmospheric
transmittance and the surface
reflectance so that these models
generate constraints for the
solution space in the optimization
processes.
4. CONCLUDING REMARKS
Further study is required for
sensitivity analysis and the others.
Sensitivity of the DEM is not large
enough compared to the other factors
so that other constraints are
required for improvement of the DEM
estimation accuracy.
40
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B3. Vienna 1996
REFE
(DH
Heipl
matcl
recol
(2$
T. Ve
Nume
