rice and other class because of heterogeneity in the scattering
mechanisms. The rice crop showed high magnitude of double
bounce scattering than odd and volume bounce scattering (Fig-
ure 4a and 4b) while in double versus odd scattering plot lot of
overlap was observed among the classes (Figure 4c). The planta-
tion class showed high volume scattering than odd and double
bounce scattering mechanisms because of multiple reflections
from tree canopy. The rough fallow showed high magnitude of
double or volume scattering than odd bounce scattering.
The cluster diagrams of
DN
values in
LISS
-IV imagery were
shown in Figure 5. All the land-cover classes showed good
separarability with
NIR
band combination. In the
NIR
band,
rice and plantation crops have shown high
DN
value than oth-
er land-use and land-cover classes (fallow, settlement, water)
because of high reflection in
NIR
than red and green (Tucker
et
al
., 1986). In the
LISS
-
IV
imagery, a few pixels of the settlement
class were mixing up with fallow class. Low
DN
values were
observed for water bodies in all bands due to absorption.
Classification
A non-parametric parallelepiped minimum distance classifier
was chosen for classification of decomposition parameters be-
cause this classification algorithm can be applied on any data
set irrespective of its distribution (Jensen, 2005). The RGB of
the Raney decomposition image and the classified image is de-
picted in the Figure 6. The common area of
LISS
-IV and
RISAT
-1
along with ground samples is also shown in Figure 6.
The accuracy of the
RISAT
1 hybrid polarimetric data clas-
sification was assessed using the ground observations, and
the results were compared with the
LISS
-IV maximum likeli-
hood classification results. The spatial agreement of the rice
class from
RISAT
1 and
LISS
IV data were also compared. In
addition, the statistics of rice cropped area from
RISAT
1 and
LISS
-IV were compared with the Government revenue records
(Anonymous, 2014).
The classified images from
LISS
IV and decomposition
parameters from
RISAT
-1 are shown in Figure 7 depicting five
classes, i.e., rice, plantation, fallow, settlement, and water. Due
to the presence of clouds in the optical image, a cloud mask
has been applied on both of the images to enable the com-
parison of the area statistics. However, in the optical image,
misclassification was observed in the areas of cloud shadows.
From Figure 7, it was observed that field bounds, canals,
and road network covered with vegetation such as coconut
trees were also classified as rice class in the optical data but
were found to be separated in the
RISAT
-1 classified image.
Even the peripheries of settlements resembled the rice crop in
Figure 3. The degree of polarization, degree of circularity parameters of the Raney decomposition.
Figure 4. Double, Odd and volume bounce scattering mechanisms of the Raney decomposition.
560
July 2015
PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING
