Figure. 1.The colors ynthetized map of three polarizations basis (HH, HV, VV) of the LingShui area. The three test areas represented
as red, blue and fuchsine color rectangles respectively.
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume XXXIX-B7, 2012
XXII ISPRS Congress, 25 August — 01 September 2012, Melbourne, Australia
bere
surface
crop forest building areca
lychee paper — mango
Figure.2.Ground measurements superimposed on the polarimetric color synthetized map.
During the period of data acquisition, the associated ground
survey campaign was also implemented in test area, composed
of bare surface, orchards, cropland, forests, and residential area,
as shown in figure.2. Figure.1. presents the study area as a color
composite where the backscattered powers corresponding to the
three standard basis’ polarizations (HH, HV, VV) are
respectively coded in red, green, and blue. Orchards and forests
appear in green color, buildings in red color, and cropland in
blue or green color. The typical test areas marked with
rectangles are selected as experimental areas of land
classification, man-made objects detection and DSM mapping
shown in figure. 2. The test areal in red color is selected as land
classification experiment, test area 2 in fuchsine color is
selected as man-made objects detection experiment, test area3
in blue color is selected as DSM mapping experiment.
3. POLARIMETRIC ANALYSIS
3.1 Polarimetric Parameters
The classical polarimetric parameters, the entropy, the alpha
angle, and the anisotropy describe the polarimetric properties of
different surfaces which have been proposed in (S.R.Cloude,
1997). As the alternative parameters, surface scattering fraction,
scattering diversity and etc. can be used by a fairly simple
algorithm (E.Colin, 2010). The alternative parameters will
provide the easier way of image interpretation, classification
and visualization. Therefore, as the new adjustment for
polarimetric parameters, the alpha angle is replaced by surface
scattering fraction, entropy is adjusted to the simpler formula
avoiding any eigenvalue calculation(J.Praks, 2009), anisotropy
retains the classical formula, scattering diversity is appended as
the new polarimetric parameter.
3.2 Polarimetric Analysis for Datasets
Surface scattering fraction, entropy, anisotropy and scattering
diversity of datasets are firstly computed over a 9 X 9 sliding
window. To proceed to a more quantitative analysis of
polarimetric parameters, some representative areas are analyzed
in surface scattering fraction /entropy, the surface scattering
fraction /anisotropy and surface scattering fraction /scattering
diversity planes with a 30 X30 pixel window. Entropy-Surface
fraction scatter plots presented in figure.3 show that surface
scattering fraction can discriminate between bare surface, forest,
building and cultivated lands (crops and orchards). However,
for different types orchards and crops, surface scattering
fraction can not discriminate them. Entropy represents the best
discrimination capability, as shown in figure.3, except for the
same capability as surface scattering fraction has, it still can
identify areca from other cultivated lands since it behaviours
similar property of higher entropy with forest covered with
more leaves. Figure.4. show that scattering diversity gives the
results close to the entropy, but it represents the lower
discrimination capability than Entropy. Figure.5. show that
anisotropy can identify buildings from the other types well as
the observations of eigenvalues are different. Therefore, surface
scattering fraction, entropy, anisotropy are selected for land
classification and objects detection as the following experiments.
08
gue
> 067 E
8 ® bung
= 044 i
ud ^
02 +
0 i 1 i : "m
15 25 35 45 55 65 75
Surface scattering fraction
Figure.3.Entropy-Surface fraction scatter plot of different types
of lands
