THE
Zribi" |
udy site
] roads.
I. The
ow that
at these
rameter
/ -angle
ands. In
hniques
nd. The
rimetric
lifferent
rive the
[TON
rimetric
nvector
mentary
in order
> matrix
sing the
nd non-
unitary
using 4
'attering
pendent
cattered
>ters are
entropy
defined
J^ and
t
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B1. Istanbul 2004
HS pog) , OSHEl
1
LM
l
where P; are the normalised eigenvalues :
À;
a Ci
P =
Aj
TM
1
The entropy H is a measure of randomness of scattering
mechanisms. Low entropy (H~0) indicates a single scattering
mechanism (isotropic scattering) while high entropy (H~I)
indicates a totally random mixture of scattering mechanisms
with equal probability and hence a depolarizing target.
The parameter a represents the mean dominant scattering
mechanism and is calculated from the eigenvectors and
eigenvalues of «[T]»:
g=3 af
i=]
where a; are the scattering mechanisms represented by the three
eigenvectors. @ =0° indicates a surface scattering, a =45° a
volume scattering, and @ =90° a double bounce scattering from
metallic surfaces (dihedral scatter).
uo
The anisotropy A indicates the distribution of the two less
significant eigenvalues:
das
A =
where À, and À, are the two lowest eigenvalues.
It becomes 0 if both scattering mechanisms are of an equal
proportion while values of A70 indicates increasing amounts of
anisotropic scattering.
3. DATA SET
Fully polarimetric X-band radar data were acquired by the
airborne RAMSES SAR of the French Aerospace Research
Center (ONERA) on March 20, 2002 over a study site near
Avignon (south of France). A single swath of about 1.8 km
width and 5.2 km long was acquired with incidence angles
ranging from 13° to 36°. Image quality was excellent except for
the range between 13° and 16° where blurring and banding
were apparent. The SAR data were processed and calibrated at
ONERA. The slant-range resolution and the azimuth resolution
were 0.66 and 0.64 m, respectively. A segment of the SAR
image used in this study is shown in Figure 1, which shows the
polarization colour composite image. HH is displayed as red,
HV as green, and VV as blue.
The study site consist mainly of agricultural landscape. It
includes agricultural areas, forest stands, houses, buildings, and
roads. The agricultural areas are composed mainly of bare soils
and wheat fields. But include orchards of various fruit trees,
among them peach, pear and apricot. The forest area present in
the image is mainly covered with pine and herbaceous. The
buildings can be considered to be of three distinct types:
houses, low building (1 - 2 floors) and high building (3 floors or
more). In general, the buildings are characterised by their flat
roofs (mostly composed of gravel or tar). During the SAR
survey, four trihedrals and one dihedral were deployed.
During the SAR survey, ground photographs and field surveys
were conducted in order to facilitate the identification. of
different surface types and to measure the characteristics of the
agricultural areas. The ground truth measurements were carried
out on wheat fields and on bare soils. Measurements of soil
roughness were carried out on five bare soil fields (R1 to R5)
using 2-m long needle profilometers of with 1-cm sampling
intervals. Ten roughness profiles were established for each
training field. From these measurements, the standard deviation
of surface height (rms) were calculated. The soil moisture at
field scale was assumed to be equal to the mean value estimated
from 15 samples (per field) collected from the top 1 cm of soil
using the gravimetric method. The surface roughness (rms) and
soil moisture (mv) fall within the ranges: 0.80 cm<rms<2.40 cm
and 8.6%<mv<22.2% (Table 1). At the time of the image
acquisition, the fields R1 to R4 were freshly tilled whereas RS
had been tilled several months before. The soil is composed of
about 53.0% loam, 31.6% clay and 15.4% sand.
Field | Incidence | Soil moisture | rms surface | Correlation
ID angle (°) 0-1 cm (%) height (cm) | length (cm)
RI 26.1 18.7 2.40+0.20 5.514165
R2 26.3 20.0 1.64+0.20 4.43+1.21
R3 26.5 21.0 1.28+0.15 3.34+0.64
R4 26.6 22.2 0.85+0.15 3.04+0.98
R5 31.7 8.6 0.80+0.22 6.42+1.99
Table 1. Ground measurements study site Avignon.
4. DATA ANALYSIS
4.1 Image Interpretation
A photo-interpretation of the composite image presented in
Figure 1 shows that the difference in backscatter is pronounced
between certain surface types. The bright signature is attributed
to multiple bounce scattering (presence of buildings, houses,
dihedral and trihedrals). Roads, on the other hand, are very
smooth surfaces, and thus appear in a very dark tone on the
image. Forests, wheat and fruit trees would seem to have very
similar signatures on the imagery. Similarly, it is impossible to
distinguish the various roughness states of the areas of bare soil
which have a rms surface height between 0.85 and 2.4 cm
(fields R1, R2, R3 and R4).
Houses
High buildings
Forest
Wheat Field
Lawn
Dihedral Low buildings
Bare soil Apricot
Bare soil (R5)
"Trihedral
Pear
Figure 1. Segment of the X-band RAMSES image. HH, HV and
VV are respectively coded in red, green and blue. The incidence
angles range from 25? on the left side of the image to 36? at the
right side.