nd (c) 
of the 
iter at 
ror % 
-9.8 
0.0 
  
0.0 
quency 
zravıty 
Own In 
g scat- 
Signa- 
or var- 
ror be- 
values 
ict. the 
isation 
c SAR 
H, VV 
oposed 
polari- 
matrix 
y (18) 
en the 
ted po- 
ises, af 
| cross- 
Normalised o 
Normalised c 
  
  
Normalised c 
  
Figure 4: Co-polarisation signatures of ocean water at (a) P-, (b) L- and (c) C-bands measured by NASA/JPL 
AirSAR system. 
polarisation signatures plotted using the measured and 
simulated Mueller matrices for farmland, residential 
areas and the ocean water at P-band. The absolute 
differences between the measured and simulated sig- 
natures are also shown in the figures. 
5 CONCLUSIONS 
A method of optimal decomposition for radar polar- 
isation signatures has been developed. In the model 
the backscattering consists of single, double, Bragg 
and cross backscattering components and the Mueller 
matrix 1s the sum of the Mueller matrices of these 
four scattering mechanisms. The technique of the 
Weighted Least Squares is then used to find the op- 
timal combination of these four components. The 
method has been tested using NASA/JPL AirSAR 
data. The results of decomposition generally agree 
with the accepted understanding of radar backscat- 
ter, and in most cases, the accuracy of the decompo- 
sition is more than 9596 for linear polarisations and 
more than 8596 for any other polarisations. It should 
be noted that the decomposition is not unique and 
is dependent on the basic model assumptions. How- 
ever, the optimal decomposition method itself seems 
robust. It is expected that if better model assumptions 
are proposed, a better decomposition can be obtained. 
The distinct polarisation signatures of the land surface 
tested and the ability to understand and predict them 
from a limited number of scattering mechanisms sug- 
gested that these signatures can be used as a basis for 
land use classification. 
6 REFERENCES 
Cloude, S. R., 1991. Optimisation methods in radar 
polarimetry. Proceedings of ICAP '91 Conference, pp. 
392-395. London: The Institution. 
Dong, Y., and Richards, J. A., 1995a. Studies of the 
cylinder-ground double bounce scattering mechanism 
in forest backscatter models. IEEE Trans on Geo- 
science and Remote Sensing, 33(1), pp. 229-231. 
Dong, Y., and Richards, J. A., 1995b. Forest discrim- 
ination using SAR multifrequency and multipolarisa- 
tion data. Proceedings of IGARSS '95 Symposium. 
Noordwijk: ESA Scientific & Tech. 
Dubois, P. C., and Norikane, L.. 1987. Data volume 
reduction for imaging radar polarimetry. Proceedings 
of IGARSS ’87 Symposium, pp. 691-696. Noordwijk: 
ESA Scientific & Tech. 
Durden, S. L., Klein, J.. and Zebker, H. A., 1991. Po- 
larimetric radar measurements of a forested area near 
Mt. Shasta. IEEE Trans on Geoscience and Remote 
Sensing, 29(3). pp. 444-450. 
Flachi. C.. 1987. Introduction to the Physics and 
Techniques of Remote Sensing. New York: John Wi- 
ley and Sons Inc.. 
Evans, D. A., Farr, T. G., van Zyl’). J., and Zebker, 
H. A., 1988. Radar polarimetry: Analysis tools and 
application. IEEE Trans on Geoscience and Remote 
Sensing, 26(6), pp. 774-798. 
Freeman, A., and Durden, S.. 1992. A three- 
component scattering model to describe polarimetric 
SAR data. Proceedings of Radar Polarimetry: 23-24 
July 1992, San Diego, California, pp. 213-224. Wash- 
ington: SPIE. 
JPL, 1995. AirSAR integrated processor documen- 
tation: Data formats, version 0.01. California: Jet 
Propulsion Laboratory. 
Kwok, R., Rignot, E. J. M., Way, J., Freeman, A., and 
Holt, J., 1994. polarisation signatures of frozen and 
thawed forests of varying environmental state. IEEE 
Trans on Geoscience and Remote Sensing, 32(2), pp. 
371-381. 
Pierce, L. E., Ulaby, F. T., Sarabandi, k., and Dob- 
son, M. C., 1994. Knowledge-based classification of 
polarimetric SAR. images. IEEE Trans on Geoscience 
and Remote Sensing, 32(5), pp. 1081-1086. 
201 
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B7. Vienna 1996