I
AE
S In
t
er
na
t
io
na
l J
o
urna
l o
f
Art
if
icia
l In
t
ellig
ence
(
I
J
-
AI
)
Vo
l.
14
,
No
.
2
,
A
p
r
il
2
0
2
5
,
p
p
.
1
2
1
1
~
1
2
1
9
I
SS
N:
2
2
5
2
-
8
9
3
8
,
DOI
: 1
0
.
1
1
5
9
1
/ijai.v
14
.i
2
.
p
p
1
2
1
1
-
1
2
1
9
1211
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
a
i
.
ia
esco
r
e.
co
m
Lea
rning
high
-
lev
el spect
ra
l
-
spa
tial
f
ea
tur
es for
hy
pe
rspectral
ima
g
e clas
sifica
tion with
i
nsuff
icie
nt
la
beled
sa
mple
s
Do
ug
la
s
O
m
weng
a
Ny
a
bu
g
a
,
G
o
dfr
ey
Ny
a
rik
i
D
e
p
a
r
t
me
n
t
o
f
I
n
f
o
r
ma
t
i
o
n
T
e
c
h
n
o
l
o
g
y
,
S
c
h
o
o
l
o
f
C
o
m
p
u
t
i
n
g
,
I
n
f
o
r
m
a
t
i
c
s
a
n
d
M
e
d
i
a
S
t
u
d
i
e
s
,
M
o
u
n
t
K
i
g
a
l
i
U
n
i
v
e
r
si
t
y
,
K
i
g
a
l
i
,
R
w
a
n
d
a
Art
icle
I
nfo
AB
S
T
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
Ap
r
2
4
,
2
0
2
4
R
ev
is
ed
No
v
7
,
2
0
2
4
Acc
ep
ted
No
v
1
4
,
2
0
2
4
Hy
p
e
rsp
e
c
tral
ima
g
e
(H
S
I)
c
las
si
fica
ti
o
n
re
se
a
rc
h
is
a
h
o
t
a
re
a
,
wi
th
a
m
a
ss
o
f
n
e
w
m
e
th
o
d
s
b
e
i
n
g
d
e
v
e
l
o
p
e
d
to
imp
r
o
v
e
p
e
rfo
rm
a
n
c
e
fo
r
sp
e
c
ifi
c
a
p
p
li
c
a
ti
o
n
s
th
a
t
u
se
sp
a
ti
a
l
a
n
d
sp
e
c
tral
ima
g
e
m
a
teria
l.
Ho
we
v
e
r,
th
e
m
a
in
o
b
sta
c
le
fo
r
sc
ien
ti
sts
is
d
e
term
in
in
g
h
o
w
to
id
e
n
ti
fy
HSIs
e
ffe
c
ti
v
e
ly
.
T
h
e
se
o
b
sta
c
les
in
c
l
u
d
e
a
n
i
n
c
re
a
se
d
p
re
se
n
c
e
o
f
re
d
u
n
d
a
n
t
s
p
e
c
tral
in
f
o
rm
a
ti
o
n
,
h
ig
h
d
ime
n
si
o
n
a
li
t
y
i
n
o
b
se
rv
e
d
d
a
ta,
a
n
d
li
m
it
e
d
s
p
a
ti
a
l
fe
a
tu
re
s
in
a
c
las
sifica
ti
o
n
m
o
d
e
l.
T
o
th
is
e
n
d
,
we
,
th
e
re
fo
re
,
p
ro
p
o
se
d
a
n
o
v
e
l
a
p
p
ro
a
c
h
fo
r
lea
rn
in
g
h
ig
h
-
lev
e
l
s
p
e
c
tral
-
sp
a
ti
a
l
fe
a
tu
re
s
fo
r
HSI
c
las
sifi
c
a
ti
o
n
with
in
su
fficie
n
t
lab
e
led
sa
m
p
les
.
F
irst
,
we
imp
lem
e
n
ted
t
h
e
p
rin
c
i
p
a
l
c
o
m
p
o
n
e
n
t
a
n
a
ly
sis
(P
CA)
tec
h
n
i
q
u
e
t
o
re
d
u
c
e
th
e
h
i
g
h
d
ime
n
sio
n
a
li
ti
e
s
e
x
p
e
rien
c
e
d
.
S
e
c
o
n
d
,
a
f
u
sio
n
o
f
2
D
a
n
d
3
D
c
o
n
v
o
lu
ti
o
n
s
a
n
d
De
n
se
Ne
t,
a
tran
sfe
r
lea
rn
in
g
n
e
two
rk
fo
r
fe
a
tu
re
lea
rn
in
g
o
f
b
o
t
h
s
p
a
ti
a
l
-
sp
e
c
tral
p
i
x
e
ls.
T
h
e
a
c
h
iev
e
d
e
x
p
e
rime
n
tal
re
su
lt
s
a
re
c
o
m
p
a
ra
ti
v
e
ly
sa
ti
sfa
c
to
ry
to
c
o
n
tras
te
d
a
p
p
ro
a
c
h
e
s
o
n
t
h
e
wid
e
ly
u
se
d
H
S
I
ima
g
e
s,
i.
e
.
,
t
h
e
Un
i
v
e
rsity
o
f
P
a
v
ia
a
n
d
In
d
ian
P
i
n
e
s,
with
a
n
o
v
e
ra
ll
c
las
sifica
ti
o
n
a
c
c
u
ra
c
y
o
f
9
7
.
8
0
%
a
n
d
9
7
.
6
0
%
,
re
sp
e
c
ti
v
e
ly
.
K
ey
w
o
r
d
s
:
C
las
s
if
icatio
n
Den
s
eNe
t
Hy
p
er
s
p
ec
tr
al
im
ag
e
Prin
cip
al
co
m
p
o
n
en
t a
n
al
y
s
is
Sp
ec
tr
al
-
s
p
atial
T
h
is i
s
a
n
o
p
e
n
a
c
c
e
ss
a
rticle
u
n
d
e
r th
e
CC B
Y
-
SA
li
c
e
n
se
.
C
o
r
r
e
s
p
o
nd
ing
A
uth
o
r
:
Do
u
g
las Om
wen
g
a
Ny
a
b
u
g
a
Dep
ar
tm
en
t o
f
I
n
f
o
r
m
atio
n
T
e
ch
n
o
lo
g
y
,
Sch
o
o
l o
f
C
o
m
p
u
ti
n
g
,
I
n
f
o
r
m
atics a
n
d
Me
d
ia
Stu
d
ies
Mo
u
n
t K
ig
ali
Un
iv
er
s
ity
P.O
B
o
x
5
6
2
8
,
Kig
ali,
R
wan
d
a
E
m
ail:
d
n
y
ab
u
g
a@
m
k
u
r
wan
d
a.
ac
.
r
w
1.
I
NT
RO
D
UCT
I
O
N
Hy
p
er
s
p
ec
tr
al
im
ag
e
r
y
(
HSI
)
m
ea
s
u
r
es
r
ef
lecta
n
ce
v
alu
es
o
f
th
e
elec
tr
o
m
ag
n
etic
s
p
ec
tr
a
in
o
v
e
r
a
h
u
n
d
r
ed
s
p
ec
tr
al
b
a
n
d
s
to
ev
er
y
s
p
atial
r
eg
io
n
in
t
h
e
im
ag
e.
W
h
ile
th
ese
v
alu
ab
le
s
p
ec
tr
al
d
etails
im
p
r
o
v
e
th
e
ca
p
ac
ity
to
d
is
tin
g
u
is
h
o
b
jects
,
HSI
an
aly
s
is
n
ee
d
s
m
o
r
e
co
m
p
lex
alg
o
r
ith
m
s
b
ec
a
u
s
e
o
f
th
e
h
ig
h
d
im
en
s
io
n
o
f
th
e
p
ix
els,
h
ig
h
n
o
n
lin
ea
r
ity
,
an
d
th
e
s
m
all
-
s
am
p
le
p
r
o
b
lem
o
f
HSI
d
ata
[
1
]
,
[
2
]
.
T
h
er
ef
o
r
e,
m
a
n
y
r
esear
ch
er
s
,
f
o
r
ex
a
m
p
le
in
[
3
]
,
[
4
]
h
a
v
e
ex
p
l
o
r
ed
th
ese
HSI
d
im
en
s
io
n
ality
r
ed
u
ctio
n
tec
h
n
iq
u
es.
Ho
wev
er
,
th
er
e
is
an
in
cr
ea
s
e
in
v
ar
iatio
n
in
s
p
atial
d
im
en
s
io
n
.
Hy
p
er
s
p
ec
tr
al
s
en
s
o
r
s
p
r
o
d
u
c
e
m
ass
iv
e
v
o
lu
m
es
o
f
d
ata,
r
e
s
u
ltin
g
in
a
lar
g
e
v
o
lu
m
e
o
f
b
an
d
s
in
th
e
d
ata,
m
ak
in
g
r
ea
l
-
tim
e
p
ar
am
eter
s
d
if
f
icu
lt
an
d
lab
o
r
i
o
u
s
to
ac
h
iev
e.
As
a
r
esu
lt,
it
is
a
d
esira
b
le
s
tr
ateg
y
to
r
ed
u
ce
th
e
d
ata
s
ize
b
ef
o
r
e
s
tar
tin
g
h
ig
h
-
lev
el
p
r
o
ce
s
s
in
g
.
T
h
er
e
f
o
r
e,
i
n
r
ec
en
t
y
ea
r
s
,
a
d
im
en
s
io
n
ality
r
ed
u
ctio
n
s
tag
e
h
as
b
ec
o
m
e
a
s
ig
n
if
ican
t
p
ar
t
o
f
m
ac
h
in
e
lear
n
in
g
(
ML
)
.
Fu
r
t
h
er
,
t
h
e
r
esear
ch
in
[
5
]
,
[
6
]
m
ain
ly
ca
r
r
ied
o
u
t
d
im
en
s
io
n
a
lity
r
ed
u
ctio
n
u
s
in
g
th
e
p
r
in
ci
p
al
co
m
p
o
n
en
ts
tr
an
s
f
o
r
m
atio
n
,
wh
ich
ch
o
s
e
an
d
p
r
eser
v
ed
t
h
e
m
o
s
t
r
ele
v
an
t
d
ata
f
o
r
class
if
icatio
n
.
As
a
r
es
u
lt,
class
if
ier
s
cr
ea
te
ef
f
icien
t
m
o
d
els
at
m
in
im
al
co
m
p
u
tatio
n
al
co
s
t
an
d
en
h
an
ce
p
ix
el
class
if
icatio
n
ac
cu
r
ac
y
in
HSI
s
.
Nev
er
th
eless
,
th
er
e
is
a
ch
allen
g
e
in
h
o
w
to
m
ak
e
f
u
ll u
s
e
o
f
th
e
s
p
atial
-
s
p
ec
tr
al
f
ea
tu
r
es c
o
n
tain
e
d
in
HSI
to
im
p
r
o
v
e
HSI
class
if
icatio
n
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
14
,
No
.
2
,
Ap
r
il
20
25
:
1
2
1
1
-
1
2
1
9
1212
Fo
r
th
is
r
ea
s
o
n
,
th
e
p
r
in
cip
al
co
m
p
o
n
en
t
an
aly
s
is
(
PC
A)
h
as
r
ec
en
tly
b
ee
n
u
s
ed
as
a
m
u
l
tiv
ar
iab
le
ap
p
r
o
ac
h
f
o
r
d
im
e
n
s
io
n
ality
r
ed
u
ctio
n
[
5
]
,
[
7
]
,
[
8
]
.
I
t
is
th
e
m
o
s
t
ad
o
p
ted
m
eth
o
d
o
l
o
g
y
in
r
em
o
te
s
en
s
in
g
ap
p
licatio
n
s
,
m
ain
ly
th
o
s
e
ap
p
ly
in
g
HSI
s
.
T
h
e
ad
jace
n
t
b
a
n
d
s
ar
e
h
ig
h
l
y
co
r
r
elate
d
i
n
th
is
ty
p
e
o
f
im
ag
e,
h
en
ce
g
ai
n
in
g
litt
le
ad
d
itio
n
a
l
in
f
o
r
m
atio
n
.
PC
A
m
in
im
ize
s
th
e
am
o
u
n
t
o
f
d
ata
b
y
r
ed
u
cin
g
d
e
p
en
d
e
n
cies
b
etwe
en
th
e
v
ar
i
o
u
s
b
a
n
d
s
.
An
eig
en
v
ec
to
r
d
ec
o
m
p
o
s
itio
n
o
f
t
h
e
o
r
ig
in
al
d
ata'
s
co
v
ar
ian
ce
m
atr
ix
i
s
co
m
p
u
ted
to
ac
h
iev
e
th
is
[
6
]
.
Ho
wev
er
,
PC
A
o
n
ly
s
ee
k
s
th
e
b
est
o
r
th
o
g
o
n
al
v
ec
to
r
s
,
o
m
itt
in
g
cr
u
cial
f
ea
t
u
r
es
ess
en
tial f
o
r
HSI
class
if
icatio
n
.
Sin
ce
its
co
n
ce
p
tio
n
,
th
e
cla
s
s
if
icatio
n
o
f
HSI
s
h
as
d
r
aw
n
wid
esp
r
ea
d
atten
tio
n
an
d
s
p
awn
ed
a
p
leth
o
r
a
o
f
ap
p
r
o
ac
h
es
aim
e
d
at
allo
ca
tin
g
a
p
ix
el
(
o
r
a
s
p
ec
tr
u
m
)
t
o
o
n
e
o
f
a
s
et
o
f
class
es
[
9
]
,
[
1
0
]
.
Sev
er
al
ap
p
r
o
ac
h
es
in
th
e
liter
atu
r
e
h
av
e
f
o
cu
s
ed
o
n
in
v
esti
g
atin
g
th
e
im
p
o
r
tan
ce
o
f
HSI
d
ata
s
p
ec
tr
al
s
ig
n
atu
r
es
in
class
if
icatio
n
,
u
s
in
g
o
n
ly
t
h
e
s
p
ec
tr
u
m
o
f
a
p
ix
el
to
e
s
tab
lis
h
its
cla
s
s
m
em
b
er
s
h
ip
.
Ho
wev
er
,
two
f
u
n
d
am
e
n
tal
d
if
f
icu
lties
f
o
r
s
u
ch
p
ix
el
-
wis
e
tech
n
iq
u
es
b
en
ef
it
f
r
o
m
r
elativ
e
co
n
ce
p
tu
al
s
im
p
licity
an
d
im
p
lem
en
tatio
n
ea
s
e:
i
)
th
e
lim
ited
tr
ain
in
g
s
et
co
m
p
ar
ed
t
o
th
e
h
ig
h
-
d
im
en
s
io
n
al
s
p
ec
tr
a
an
d
ii
)
th
e
s
p
ec
tr
al
v
ar
iatio
n
s
.
T
h
e
f
ir
s
t
is
s
u
e,
wh
ich
h
as
b
ee
n
ex
te
n
s
iv
ely
s
tu
d
ied
in
lig
h
t
o
f
t
h
e
w
ell
-
k
n
o
wn
Hu
g
h
es
p
h
en
o
m
en
o
n
[
1
1
]
,
ca
u
s
es
p
r
o
b
lem
s
in
two
way
s
.
First,
d
u
e
to
th
e
lim
ited
n
u
m
b
er
o
f
la
b
eled
s
am
p
les,
th
e
s
am
p
le
co
v
ar
ian
ce
m
atr
ix
is
lik
ely
to
b
e
s
in
g
u
lar
,
r
esu
ltin
g
in
ill
-
p
o
s
ed
d
if
f
icu
lties
f
o
r
s
e
v
er
al
class
if
icatio
n
alg
o
r
ith
m
s
.
Seco
n
d
,
h
ig
h
-
d
i
m
en
s
io
n
al
s
p
ec
tr
a
r
e
q
u
ir
e
n
u
m
er
o
u
s
f
r
ee
p
ar
am
eter
s
f
o
r
co
m
p
u
tatio
n
in
a
p
ar
am
etr
ic
ap
p
r
o
ac
h
,
w
h
ich
i
s
in
clin
ed
to
o
v
er
f
it
an
d
co
n
s
eq
u
en
tly
d
ec
r
ea
s
es
th
e
g
en
er
a
lizatio
n
ca
p
ac
ity
o
f
class
if
ier
s
.
R
eg
ar
d
in
g
s
p
ec
tr
al
v
ar
iatio
n
ca
u
s
ed
b
y
s
ev
er
al
f
ac
to
r
s
s
u
ch
as
i
n
cid
en
t
li
g
h
t,
a
tm
o
s
p
h
er
ic
ef
f
ec
ts
,
u
n
d
esira
b
le
s
h
a
d
e
an
d
s
h
ad
o
w,
n
atu
r
al
s
p
ec
tr
u
m
f
lu
ctu
ati
o
n
,
a
n
d
i
n
s
tr
u
m
en
t
n
o
is
es
[
9
]
,
[
1
2
]
,
two
s
er
io
u
s
ch
allen
g
es m
ig
h
t m
a
k
e
ca
teg
o
r
izatio
n
d
if
f
icu
lt.
On
th
e
o
n
e
h
an
d
,
s
u
b
s
tan
tial
in
tr
a
-
class
s
p
ec
tr
a
v
ar
iab
ilit
y
m
ak
es
it
ch
allen
g
in
g
to
id
en
tif
y
a
s
p
ec
if
ic
class
.
B
esid
es,
lo
w
in
ter
-
class
s
p
ec
tr
al
v
ar
iatio
n
m
ak
es
d
is
ti
n
g
u
is
h
in
g
d
is
tin
ct
class
es
d
if
f
icu
lt.
T
h
ese
is
s
u
e
s
m
ak
e
HSI
class
if
icatio
n
c
o
m
p
lex
,
r
esu
ltin
g
in
p
o
o
r
class
if
icatio
n
r
esu
lts
wh
en
u
s
in
g
p
ix
el
-
wis
e
ap
p
r
o
ac
h
es.
B
ec
au
s
e
HSI
s
ar
e
n
atu
r
ally
3
-
D
an
d
v
is
u
al,
s
p
atial
r
elia
n
ce
,
an
alo
g
o
u
s
t
o
s
p
ec
tr
al
b
eh
av
io
r
,
is
a
n
atu
r
al
co
m
p
lem
en
t
to
s
p
ec
tr
a.
As
a
r
esu
lt,
th
e
in
clu
s
io
n
o
f
s
p
atial
d
ep
en
d
e
n
cy
h
as
th
e
p
o
ten
tial
to
im
p
r
o
v
e
p
ix
el
-
wis
e
class
if
icatio
n
.
T
h
e
u
s
e
o
f
s
p
atial
in
f
o
r
m
atio
n
in
HSI
clas
s
if
icatio
n
d
ates
b
ac
k
o
v
er
a
d
ec
ad
e,
an
d
s
o
m
e
s
u
cc
ess
f
u
l
r
esear
ch
h
as
d
em
o
n
s
tr
ated
its
ab
ilit
y
to
e
n
h
an
ce
class
if
icatio
n
p
er
f
o
r
m
an
ce
[
1
3
]
.
Sin
ce
t
h
en
,
th
er
e
h
as
b
ee
n
a
s
ig
n
if
ican
t
in
cr
ea
s
e
in
in
ter
est
in
s
p
ec
t
r
al
-
s
p
atial
class
if
icatio
n
.
Sch
o
lar
s
h
av
e
u
tili
ze
d
m
u
ltil
ay
er
s
tr
ateg
ies to
s
o
lv
e
th
is
ch
allen
g
e.
Ho
wev
er
,
th
ese
m
eth
o
d
s
o
n
l
y
an
aly
ze
2
D
an
d
3
D
s
p
ec
tr
al
-
s
p
atial
p
r
o
p
er
ties
in
d
e
p
en
d
e
n
tly
an
d
t
ak
e
a
lo
n
g
tim
e
to
co
m
p
lete.
T
o
th
is
en
d
,
we
p
r
esen
t
a
s
p
ec
tr
al
-
s
p
atial
HSI
class
if
icat
io
n
ap
p
r
o
ac
h
b
ased
o
n
Den
s
eNe
t
[
1
4
]
,
wh
ich
we
em
p
lo
y
as
a
u
n
iq
u
e
s
tr
ateg
y
f
o
r
HSI
d
ataset
class
if
icatio
n
.
Fu
r
th
er
m
o
r
e,
we
in
co
r
p
o
r
ate
d
th
e
Den
s
eNe
t
f
r
am
ewo
r
k
f
o
r
its
s
tan
d
ar
d
izatio
n
tech
n
iq
u
e,
wh
i
ch
h
as
n
u
m
er
o
u
s
ad
v
an
tag
es:
i)
t
h
e
r
eu
s
ab
ilit
y
o
f
th
e
in
f
o
r
m
atio
n
(
f
ea
t
u
r
e)
co
v
e
r
ed
in
HSI
s
,
ii)
th
e
c
o
n
ca
ten
at
io
n
o
f
d
if
f
er
e
n
t
p
at
h
s
(
wh
ich
r
ed
u
ce
s
th
e
n
u
m
b
e
r
o
f
p
a
r
am
eter
s
)
,
iii)
ai
d
ed
t
o
o
v
er
co
m
in
g
p
r
o
b
lem
s
s
u
ch
as
o
v
er
f
itti
n
g
an
d
th
e
v
an
is
h
in
g
g
r
ad
ien
t
wh
en
f
ew
tr
ain
in
g
s
am
p
les
ar
e
av
aila
b
le,
an
d
iv
)
s
tr
en
g
th
en
f
ea
tu
r
e
p
r
o
p
ag
atio
n
.
T
h
e
s
u
m
m
ar
y
o
f
th
e
m
ain
co
n
tr
ib
u
tio
n
s
o
f
o
u
r
p
r
o
p
o
s
ed
ap
p
r
o
ac
h
is
:
‒
W
e
im
p
lem
en
ted
th
e
PC
A
tec
h
n
iq
u
e
o
n
th
e
HSI
f
o
r
d
im
en
s
io
n
ality
r
ed
u
ctio
n
d
u
e
to
th
e
h
ig
h
d
im
en
s
io
n
s
in
v
o
lv
ed
in
th
e
HSI
s
.
‒
Ou
r
p
r
o
p
o
s
ed
a
p
p
r
o
ac
h
i
n
tr
o
d
u
ce
d
th
e
Den
s
eNe
t
-
b
a
s
ed
m
ec
h
an
is
m
to
m
o
d
el
th
e
s
em
an
tic
in
ter
d
ep
en
d
en
cies
in
s
p
atial
an
d
ch
an
n
el
d
im
en
s
io
n
s
to
im
p
r
o
v
e
f
ea
tu
r
e
r
e
p
r
esen
tatio
n
f
o
r
class
if
icatio
n
ab
ilit
y
.
‒
T
h
e
ac
h
ie
v
ed
r
esu
lts
d
em
o
n
s
tr
ate
th
at
o
u
r
s
u
g
g
ested
ap
p
r
o
a
ch
ca
n
b
e
tr
ain
e
d
e
n
d
-
to
-
en
d
a
n
d
esti
m
ated
as
th
e
s
tate
-
of
-
th
e
-
ar
t
f
o
r
b
o
th
d
a
tasets
co
n
cu
r
r
en
tly
.
T
h
e
r
em
ain
i
n
g
p
ar
ts
o
f
th
is
s
tu
d
y
ar
e
e
x
am
in
ed
as
f
o
llo
ws:
s
ec
tio
n
2
d
is
cu
s
s
es
r
elate
d
wo
r
k
s
.
Sectio
n
3
d
is
cu
s
s
es
o
u
r
p
r
o
p
o
s
ed
ap
p
r
o
ac
h
.
Sectio
n
4
p
r
esen
ts
th
e
ex
p
er
im
en
ts
,
i.e
.
,
th
e
d
ataset
an
d
p
er
f
o
r
m
an
ce
an
aly
s
is
m
etr
ics,
an
d
co
m
p
a
r
es
b
aselin
e
m
eth
o
d
s
,
r
esu
lts
,
an
d
d
is
cu
s
s
io
n
s
.
Fin
ally
,
s
ec
tio
n
5
co
n
clu
d
es o
u
r
s
tu
d
y
b
y
g
iv
in
g
a
s
u
m
m
ar
y
o
f
th
e
c
o
n
ten
t a
n
d
f
u
tu
r
e
r
ec
o
m
m
e
n
d
atio
n
s
.
2.
RE
L
AT
E
D
WO
RK
L
ately
,
m
an
y
r
esear
ch
e
r
s
h
av
e
d
ep
lo
y
ed
co
n
v
o
l
u
tio
n
al
n
eu
r
al
n
etwo
r
k
s
(
C
NNs)
in
HSI
ca
teg
o
r
izatio
n
.
T
h
ey
ar
e
b
etter
s
u
ited
f
o
r
HSI
an
al
y
s
is
an
d
f
ea
tu
r
e
ex
tr
ac
tio
n
[
1
5
]
,
[
1
6
]
.
T
h
o
u
g
h
th
ese
ap
p
r
o
ac
h
es
tr
y
to
o
p
tim
ize
th
e
u
s
e
o
f
b
o
th
s
p
ec
tr
al
-
s
p
atial
f
ea
tu
r
es,
th
ey
u
s
u
ally
d
iv
id
e
th
e
jo
in
t
s
p
atial
-
s
p
ec
tr
al
f
ea
tu
r
es
in
to
two
in
d
ep
e
n
d
en
t
lear
n
in
g
c
o
m
p
o
n
e
n
ts
,
ig
n
o
r
in
g
th
e
r
elatio
n
u
n
d
er
l
y
in
g
th
e
s
p
ec
tr
al
-
s
p
atial
f
ea
tu
r
es.
T
h
e
r
esear
ch
in
[
1
7
]
,
[
1
8
]
d
es
cr
ib
ed
th
e
f
ew
-
s
h
o
t
lear
n
in
g
t
ec
h
n
iq
u
e,
in
wh
ich
th
e
m
o
d
el
ef
f
icien
tly
d
is
cr
im
in
ated
ca
teg
o
r
ies
in
a
n
ewly
ac
q
u
ir
ed
d
ata
s
et
u
s
in
g
o
n
ly
a
s
m
all
n
u
m
b
er
o
f
lab
eled
s
am
p
les.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
Lea
r
n
in
g
h
ig
h
-
leve
l sp
ec
tr
a
l
-
s
p
a
tia
l fe
a
tu
r
es fo
r
h
yp
ers
p
ec
tr
a
l ima
g
e
…
(
Do
u
g
la
s
Omwen
g
a
N
ya
b
u
g
a
)
1213
Ho
wev
er
,
all
o
f
th
ese
ap
p
r
o
a
ch
es
r
ely
o
n
ar
tific
ially
ca
lc
u
lated
m
ea
s
u
r
em
en
t
d
is
tan
ce
s
,
wh
ich
m
a
y
o
n
l
y
p
ar
tially
ap
p
ly
to
th
e
f
ea
tu
r
es
r
etr
iev
ed
b
y
th
e
n
e
u
r
al
n
etwo
r
k
wh
en
ca
te
g
o
r
izin
g
th
em
.
As
a
r
esu
lt,
J
ia
et
a
l.
[
1
9
]
p
r
o
p
o
s
ed
a
n
ef
f
ec
tiv
e
tr
an
s
f
er
lear
n
in
g
s
tr
ateg
y
to
ad
d
r
ess
in
ad
eq
u
ate
tr
a
in
in
g
HSI
s
am
p
les.
E
v
en
th
o
u
g
h
th
is
s
tr
ateg
y
h
as
ac
h
iev
ed
s
ig
n
if
ican
t
ad
v
an
ce
s
in
HSI
class
if
icatio
n
,
it
p
er
f
o
r
m
s
p
o
o
r
ly
wh
en
o
n
ly
a
f
ew
lab
eled
s
am
p
les
ar
e
av
ailab
le.
T
h
is
h
as
r
esu
lt
ed
in
a
s
ig
n
if
ican
t
is
s
u
e
f
o
r
d
ee
p
lear
n
in
g
(
DL
)
m
o
d
els,
as
a
d
d
r
ess
ed
in
th
is
wo
r
k
.
T
h
e
r
esear
c
h
in
[
2
0
]
,
[
2
1
]
s
tate
th
at
o
p
tical
r
em
o
te
s
en
s
in
g
co
llects
r
ad
iatio
n
r
ef
lecte
d
an
d
em
itted
f
r
o
m
th
e
s
u
r
f
ac
es
u
n
d
er
s
tu
d
y
,
f
o
cu
s
in
g
o
n
th
e
r
eg
io
n
o
f
th
e
elec
tr
o
m
ag
n
etic
s
p
ec
tr
u
m
with
wav
ele
n
g
th
s
s
p
an
n
in
g
f
r
o
m
v
is
ib
le
to
n
e
ar
-
in
f
r
ar
e
d
to
th
er
m
al
i
n
f
r
ar
e
d
.
W
ith
a
HSI
th
at
ca
p
tu
r
es
a
v
ar
iety
o
f
p
r
ec
is
ely
ca
lib
r
ated
tin
y
s
p
ec
tr
al
b
a
n
d
s
o
f
th
e
v
is
ib
le
an
d
in
f
r
ar
ed
s
p
ec
tr
u
m
s
.
T
h
e
en
o
r
m
o
u
s
am
o
u
n
t
o
f
s
p
ec
tr
al
d
ata
p
r
o
v
id
es
im
p
o
r
ta
n
t
lan
d
-
co
v
e
r
in
f
o
r
m
atio
n
th
at
h
elp
s
p
r
ec
is
ely
clas
s
if
y
s
u
r
f
ac
e
lan
d
u
s
e
a
n
d
lan
d
co
v
er
.
Nev
e
r
th
eless
,
lab
o
r
an
d
t
im
e
-
in
ten
s
iv
e
p
r
o
ce
d
u
r
es
ar
e
r
eq
u
ir
ed
to
ex
tr
ac
t
tag
g
ed
tr
ain
in
g
d
ata
f
r
o
m
HSI
.
C
o
n
s
eq
u
en
tly
,
b
ased
o
n
ac
tiv
e
lear
n
in
g
,
a
class
if
ier
d
esig
n
th
at
u
s
es th
e
f
ewe
s
t
lab
eled
ex
am
p
les as p
r
ac
tical
f
o
r
class
if
icatio
n
was p
r
o
p
o
s
ed
[
2
2
]
.
R
em
o
te
s
en
s
in
g
im
ag
e
r
y
(
R
SI)
o
b
jects a
n
d
f
ea
tu
r
es
f
r
e
q
u
en
tly
h
av
e
u
n
cl
ea
r
b
ac
k
g
r
o
u
n
d
s
a
n
d
ca
n
n
o
t
y
ield
h
elp
f
u
l
in
f
o
r
m
atio
n
.
I
d
e
n
tify
in
g
th
e
R
SI
is
m
o
r
e
d
if
f
icu
lt b
ec
au
s
e
o
f
th
e
n
o
tab
le
in
tr
ac
lass
v
ar
ian
ce
s
.
G
u
o
e
t
a
l
.
[
2
3
]
p
r
e
s
e
n
t
e
d
a
m
u
l
t
i
-
v
i
e
w
-
f
e
a
t
u
r
e
-
l
e
a
r
n
i
n
g
n
e
t
w
o
r
k
t
o
a
d
d
r
e
s
s
t
h
i
s
p
r
o
b
l
e
m
a
n
d
g
a
t
h
e
r
t
h
r
e
e
s
p
e
c
i
f
i
c
d
o
m
a
i
n
f
e
a
t
u
r
e
s
f
o
r
t
h
e
s
c
e
n
e
c
a
t
e
g
o
r
i
z
a
t
i
o
n
c
h
a
l
l
e
n
g
e
.
O
n
t
h
e
o
t
h
e
r
h
a
n
d
,
P
u
n
d
i
r
a
n
d
A
k
s
h
a
y
[
2
4
]
in
tr
o
d
u
ce
d
m
o
d
el
-
a
g
n
o
s
tic
m
e
ta
-
lear
n
in
g
an
d
th
e
en
s
em
b
le
o
f
p
r
o
to
ty
p
e
n
etwo
r
k
s
to
o
v
er
co
m
e
th
e
p
r
o
b
lem
s
ass
o
ciate
d
with
s
tan
d
ar
d
d
ee
p
-
lear
n
in
g
n
etwo
r
k
s
.
T
h
is
tech
n
iq
u
e
tack
led
th
e
R
SI
ca
teg
o
r
izatio
n
p
r
o
b
lem
b
y
ap
p
ly
in
g
m
eta
-
lea
r
n
in
g
.
Mu
l
tire
s
o
lu
tio
n
ca
teg
o
r
izatio
n
o
f
p
an
ch
r
o
m
atic
an
d
m
u
ltis
p
e
ctr
al
p
ictu
r
es
is
a
p
o
p
u
lar
a
r
ea
o
f
r
esear
ch
.
T
h
e
m
ain
ch
allen
g
e
in
th
is
f
ield
i
s
ass
e
s
s
in
g
d
ata
an
d
ex
tr
ac
tin
g
ch
ar
ac
ter
is
tics
to
im
p
r
o
v
e
class
if
icatio
n
ac
cu
r
ac
y
p
r
o
p
er
ly
.
An
ad
ap
tiv
e
h
y
b
r
i
d
f
u
s
io
n
n
etwo
r
k
th
at
in
co
r
p
o
r
ates
b
o
t
h
d
ata
f
u
s
io
n
an
d
f
ea
tu
r
e
f
u
s
io
n
was
p
r
esen
ted
in
[
2
5
]
to
class
if
y
m
u
ltire
s
o
lu
tio
n
R
SI.
Ho
wev
er
,
th
ese
m
eth
o
d
s
d
ep
en
d
o
n
a
lar
g
e
n
u
m
b
er
o
f
lab
eled
tr
ain
in
g
s
am
p
les
to
o
b
tain
an
e
x
ce
llen
t
class
if
icatio
n
p
er
f
o
r
m
an
ce
.
T
o
s
o
lv
e
th
e
class
if
icatio
n
p
r
o
b
lem
,
Sath
y
an
a
r
ay
an
a
an
d
Sin
g
h
[
2
6
]
d
esig
n
e
d
a
m
u
ltil
ay
er
f
ee
d
f
o
r
war
d
ar
tific
ial
n
eu
r
al
n
etwo
r
k
an
d
u
s
ed
a
h
is
to
g
r
am
tech
n
iq
u
e
t
o
ex
tr
ac
t
th
e
p
ix
el
d
en
s
ity
d
is
tr
ib
u
tio
n
an
d
n
o
r
m
aliza
tio
n
to
m
ak
e
th
e
r
esu
lt
in
d
ep
en
d
en
t
f
r
o
m
th
e
p
h
y
s
ical
p
r
o
p
er
ties
o
f
th
e
im
ag
e.
L
iu
et
a
l.
[
2
7
]
s
u
g
g
ested
a
tech
n
iq
u
e
th
at
co
m
b
in
es
a
lo
w
-
r
eso
lu
tio
n
HSI
with
a
h
ig
h
-
r
eso
lu
tio
n
(
HR
)
m
u
lti
-
s
p
ec
tr
al
im
ag
e
(
MSI
)
to
e
x
tr
a
ct
d
ee
p
m
u
ltis
ca
le
p
r
o
p
er
ties
f
r
o
m
an
HSI
s
ce
n
e.
T
r
ain
in
g
d
ata
was
o
p
tio
n
al
f
o
r
th
is
s
tr
ateg
y
.
T
h
is
r
esear
ch
aim
s
to
r
eliab
ly
class
if
y
HSI
in
to
a
class
o
r
ca
teg
o
r
y
r
eg
ar
d
less
o
f
s
o
u
r
ce
,
r
e
s
o
lu
tio
n
,
o
r
s
ize
b
y
d
ev
el
o
p
in
g
a
s
p
ec
tr
al
-
s
p
atial
ap
p
r
o
ac
h
b
ased
o
n
th
e
De
n
s
eNe
t n
etwo
r
k
.
T
h
u
s
,
it will e
x
p
ed
ite
th
e
p
r
o
ce
s
s
an
d
en
h
a
n
ce
s
p
ee
d
.
3.
M
E
T
H
O
D
T
h
is
s
ec
tio
n
co
v
er
s
th
e
d
etai
ls
o
f
o
u
r
p
r
o
p
o
s
ed
a
p
p
r
o
ac
h
f
o
r
s
p
ec
tr
al
-
s
p
atial
HSI
clas
s
if
icatio
n
.
Fig
u
r
e
1
d
em
o
n
s
tr
ates
th
e
g
en
er
al
f
r
am
ewo
r
k
o
f
th
e
p
r
o
p
o
s
ed
m
eth
o
d
.
T
h
e
in
p
u
t
o
f
o
u
r
m
o
d
el
is
h
y
p
er
s
p
ec
tr
al
d
ata
with
th
e
−
s
pe
c
tr
a
l
b
an
d
an
d
s
ize
o
f
×
.
T
h
u
s
,
we
co
n
s
id
er
it
as
th
e
m
atr
i
x
o
f
o
r
d
e
r
×
×
.
T
h
e
PC
A
is
ap
p
lied
to
th
e
H
SI
f
o
r
d
ata
d
im
e
n
s
io
n
ality
r
ed
u
ctio
n
.
T
h
e
3
D
s
p
ec
tr
al
-
s
p
atia
l
p
ix
els
ar
e
co
n
v
o
l
u
ted
o
r
co
n
ca
ten
ate
d
.
W
e
in
tr
o
d
u
ce
d
a
Den
s
eNe
t la
y
er
b
ef
o
r
e
th
e
f
u
ll c
o
n
v
o
lu
tio
n
al
(
FC
)
lay
er
s
f
o
r
3
D
s
p
ec
tr
al
-
s
p
atial
f
ea
tu
r
e
lea
r
n
in
g
.
Fig
u
r
e
1
.
Pro
p
o
s
ed
s
ch
em
e
f
o
r
s
p
atial
-
s
p
ec
tr
al
class
if
icatio
n
o
f
HSI
3
.
1
.
L
o
w
-
s
pa
ce
pro
j
ec
t
i
o
ns
Fo
r
th
e
HSI
d
im
en
s
io
n
ality
r
e
d
u
ctio
n
th
r
o
u
g
h
PC
A
tak
es th
e
m
ath
em
atica
l f
o
r
m
u
latio
n
o
f
(
1
)
.
=
×
×
(
1
)
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
14
,
No
.
2
,
Ap
r
il
20
25
:
1
2
1
1
-
1
2
1
9
1214
L
et
∈
ℝ
×
r
ep
r
esen
t
t
h
e
r
aw
HSI
,
with
,
,
an
d
in
d
icatin
g
th
e
r
o
w,
c
o
lu
m
n
,
an
d
b
a
n
d
n
u
m
b
er
s
,
r
esp
ec
tiv
ely
.
On
ly
th
e
f
ir
s
t
p
r
in
cip
al
co
m
p
o
n
en
ts
ar
e
r
eser
v
ed
wh
en
th
e
PC
A
i
s
ap
p
lied
to
th
e
r
aw
im
ag
e
to
m
in
im
ize
th
e
co
n
v
o
lu
tio
n
p
h
ase's
co
m
p
u
tatio
n
al
co
s
ts
.
T
h
e
d
im
en
s
io
n
-
r
ed
u
ce
d
im
ag
e
is
r
ep
r
esen
ted
b
y
∈
ℝ
×
.
A
n
eig
h
b
o
r
zo
n
e
with
th
e
d
im
en
s
io
n
s
×
×
is
ex
tr
ac
ted
ar
o
u
n
d
ea
ch
p
ix
el.
Su
p
p
o
s
e
we
tak
e
tr
ain
in
g
s
am
p
les;
th
en
∈
ℝ
×
×
×
d
en
o
tes th
e
tr
ain
in
g
s
et.
3
.
2
.
Sp
ec
t
ra
l pix
el
ex
t
ra
ct
io
n
HSI
s
h
av
e
a
lo
t
o
f
s
p
ec
tr
u
m
in
f
o
r
m
atio
n
an
d
s
p
ec
tr
al
r
eso
lu
tio
n
;
h
en
ce
,
th
eir
class
if
icatio
n
ap
p
r
o
a
ch
is
b
ased
o
n
s
p
ec
tr
al
f
ea
tu
r
es.
E
ac
h
p
ix
el
ca
n
ex
tr
ac
t
1
D
s
p
ec
tr
al
v
ec
to
r
s
to
class
if
y
o
b
jects.
W
e
d
ep
lo
y
ed
th
e
1D
-
C
NN
to
ex
tr
ac
t th
e
s
p
ec
tr
a
l f
ea
tu
r
es f
r
o
m
HSI
s
an
d
ca
teg
o
r
ize
th
em
,
as d
e
m
o
n
s
tr
ated
in
Fig
u
r
e
2
.
Fig
u
r
e
2
.
T
h
e
s
p
ec
tr
al
p
ix
el
e
x
tr
ac
tio
n
T
h
e
co
n
v
o
lu
tio
n
al
lay
er
is
in
tr
o
d
u
ce
d
f
ir
s
t.
T
h
e
v
alu
e
o
f
n
e
u
r
o
n
,
at
a
p
o
s
itio
n
o
f
th
e
ℎ
f
ea
tu
r
e
m
ap
in
th
e
ℎ
lay
er
is
d
ef
in
e
d
as
(
2
)
.
,
=
(
∑
∑
,
,
(
−
1
)
+
+
,
−
1
=
0
)
(
2
)
W
h
er
e
in
d
ex
es
th
e
f
ea
tu
r
e
m
ap
in
th
e
p
r
e
v
io
u
s
lay
er
(
(
−
1
)
ℎ
lay
er
)
co
n
n
ec
ted
to
th
e
c
u
r
r
e
n
t
f
ea
tu
r
e
m
ap
,
,
,
d
e
n
o
tes
th
e
weig
h
t
o
f
p
o
s
itio
n
co
n
n
ec
te
d
to
th
e
ℎ
f
ea
tu
r
e
m
a
p
,
d
e
n
o
tes
th
e
wid
th
o
f
th
e
k
er
n
el
to
war
d
th
e
s
p
ec
tr
al
d
im
en
s
io
n
,
an
d
,
d
en
o
tes th
e
b
ias o
f
th
e
ℎ
f
ea
tu
r
e
m
ap
i
n
th
e
ℎ
lay
er
.
3
.
3
.
Sp
a
t
ia
l pix
el
e
x
t
ra
ct
i
o
n
Sp
a
tial
p
ix
els
,
i.
e.
,
co
n
t
ex
t
i
n
f
o
r
m
a
ti
o
n
,
w
h
i
c
h
f
o
r
m
s
p
ar
t
o
f
t
h
e
HS
I
im
a
g
es
,
a
r
e
u
s
e
d
to
c
lass
i
f
y
th
e
HSI
s
.
T
h
e
s
p
ati
al
p
ix
els
e
x
t
r
a
c
ted
f
r
o
m
th
e
c
o
n
ti
g
u
o
u
s
o
f
t
h
e
p
ix
el
a
r
e
u
ti
liz
ed
i
n
s
t
ea
d
o
f
t
h
e
s
p
e
ct
r
al
f
ea
tu
r
es
d
e
r
i
v
e
d
f
r
o
m
a
s
p
ec
if
ic
p
ix
el
.
Du
e
t
o
t
h
e
wi
d
e
r
a
n
g
e
o
f
h
y
p
e
r
s
p
ec
tr
al
d
ata
,
th
e
s
ta
n
d
a
r
d
s
t
r
ate
g
y
f
o
r
o
b
tai
n
i
n
g
2
D
(
s
p
a
tia
l)
f
e
at
u
r
es
is
t
o
c
o
n
d
e
n
s
e
t
h
e
d
at
aset
f
i
r
s
t,
u
s
e
a
t
wo
-
d
i
m
e
n
s
i
o
n
al
C
NN
t
o
o
b
ta
in
m
o
r
e
d
e
tai
le
d
s
p
at
ial
f
ea
tu
r
es
,
a
n
d
c
lass
i
f
y
u
s
i
n
g
s
p
ati
al
d
etai
ls
.
Fi
g
u
r
e
3
d
e
p
ic
ts
t
h
e
s
p
e
ci
f
ic
p
r
o
c
e
d
u
r
e.
T
o
e
x
ec
u
t
e
a
co
n
v
o
lu
ti
o
n
o
p
e
r
at
io
n
o
n
2
D
d
ata
i
n
th
e
2
D
co
n
v
o
l
u
ti
o
n
p
r
o
c
ed
u
r
e
,
w
e
a
p
p
lie
d
a
2
D
c
o
n
v
o
l
u
ti
o
n
k
er
n
e
l.
I
n
t
h
e
2
D
c
o
n
v
o
l
u
t
io
n
o
p
er
ati
o
n
,
i
n
p
u
t
d
ata
is
c
o
n
v
o
lv
ed
w
it
h
2
D
k
er
n
e
ls
,
a
n
d
t
h
e
p
r
o
ce
s
s
c
an
b
e
f
o
r
m
u
lat
e
d
as
(
3
)
.
,
,
=
(
∑
∑
∑
,
,
ℎ
(
−
1
)
(
+
ℎ
)
(
+
)
+
,
−
1
=
0
−
1
ℎ
=
0
)
(
3
)
3
.
4
.
Sp
ec
t
ra
l
-
s
pa
t
ia
l pix
el
ex
t
ra
ct
io
n
So
le
s
p
ec
tr
al
in
f
o
r
m
atio
n
is
u
s
u
ally
em
p
lo
y
e
d
in
tr
a
d
itio
n
al
HSI
f
o
r
class
if
icatio
n
p
u
r
p
o
s
es.
B
ec
au
s
e
o
f
th
e
in
f
lu
en
ce
o
f
th
e
n
atu
r
al
atm
o
s
p
h
er
e,
id
en
tical
lan
d
f
e
atu
r
es
will
d
i
s
p
lay
d
if
f
er
en
t
s
p
ec
tr
al
cu
r
v
es.
T
h
e
so
-
k
n
o
wn
alik
e
-
o
b
ject
h
eter
o
-
s
p
ec
tr
u
m
an
d
u
n
r
elate
d
-
o
b
ject
b
u
t
with
s
im
ilar
s
p
ec
tr
u
m
p
h
en
o
m
en
a
ca
n
ca
u
s
e
d
if
f
er
en
t
g
r
o
u
n
d
o
b
jects
to
h
av
e
th
e
s
am
e
s
p
ec
tr
al
c
u
r
v
e.
Fo
r
e
x
am
p
le,
s
p
ec
if
ic
p
ix
els
(
elem
en
ts
)
in
ter
co
n
n
ec
ted
o
n
ea
r
th
ar
e
d
e
s
ig
n
ated
as p
ar
k
in
g
lo
ts
; th
er
ef
o
r
e,
p
ix
els with
s
p
ec
tr
al
f
ea
tu
r
es th
at
lo
o
k
h
ig
h
l
y
s
im
ilar
to
m
etal
s
p
ec
tr
al
f
ea
t
u
r
es
ar
e
m
o
s
t
lik
ely
to
r
ep
r
e
s
en
t
v
eh
icles.
I
f
a
p
ix
el
h
as
m
an
y
g
r
ass
p
ix
els
s
u
r
r
o
u
n
d
in
g
it,
t
h
e
p
ix
els
in
th
e
m
id
d
le
ar
e
m
o
s
t
lik
ely
g
r
ass
.
Hy
p
er
s
p
ec
tr
al
d
ata
h
as
a
3
D
s
tr
u
ctu
r
e
th
a
t
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
Lea
r
n
in
g
h
ig
h
-
leve
l sp
ec
tr
a
l
-
s
p
a
tia
l fe
a
tu
r
es fo
r
h
yp
ers
p
ec
tr
a
l ima
g
e
…
(
Do
u
g
la
s
Omwen
g
a
N
ya
b
u
g
a
)
1215
in
clu
d
es
1
D
(
s
p
ec
tr
al
f
ea
tu
r
es)
an
d
2
D
(
s
p
atial
f
ea
tu
r
es).
A
3
D
-
C
NN
m
ay
e
x
tr
ac
t
b
o
th
s
p
ec
tr
al
an
d
s
p
atial
in
f
o
r
m
atio
n
at
th
e
s
am
e
tim
e.
Fig
u
r
e
4
d
e
p
icts
th
e
p
r
o
ce
d
u
r
e
f
o
r
ex
t
r
ac
tin
g
th
e
s
p
ec
tr
al
-
s
p
atial
f
ea
tu
r
es.
Fig
u
r
e
3
.
T
h
e
s
p
atial
p
ix
el
ex
t
r
ac
tio
n
Fig
u
r
e
4
.
T
h
e
s
p
ec
tr
al
-
s
p
atial
p
ix
el
ex
tr
ac
tio
n
T
h
e
co
n
ca
ten
ati
o
n
o
f
th
e
2
D
a
n
d
3
D
co
n
v
o
l
u
tio
n
s
o
f
t
h
e
HSI
,
it'
s
d
ef
in
ed
th
r
o
u
g
h
in
(
4
)
.
=
(
∑
∑
∑
ℎ
(
−
1
)
(
+
ℎ
)
(
+
)
(
+
)
+
−
1
=
0
−
1
=
0
−
1
ℎ
=
0
)
(
4
)
wh
er
e
d
en
o
tes
th
e
s
p
ec
tr
al
d
ep
th
o
f
th
e
3
-
D
k
e
r
n
el,
r
ep
r
e
s
en
ts
th
e
n
u
m
b
e
r
o
f
f
ea
tu
r
e
c
u
b
es
in
th
e
p
r
io
r
lay
er
,
d
en
o
tes
th
e
n
u
m
b
e
r
o
f
k
er
n
els
in
th
is
lay
er
.
d
en
o
te
s
th
e
o
u
tp
u
t
at
p
o
s
itio
n
(
,
,
)
,
wh
ich
is
co
m
p
u
ted
b
y
co
n
v
o
lv
in
g
th
e
ℎ
f
ea
tu
r
e
cu
b
e
o
f
th
e
p
r
ev
io
u
s
lay
er
with
th
e
ℎ
k
er
n
el
o
f
th
e
ℎ
l
ay
er
,
an
d
ℎ
is
th
e
(
ℎ
,
,
)
ℎ
v
alu
e
o
f
th
e
k
e
r
n
el
c
o
n
n
ec
ted
t
o
th
e
ℎ
f
ea
tu
r
e
cu
b
e
in
th
e
p
r
ec
ed
in
g
la
y
er
.
As
s
u
ch
,
th
e
o
u
tp
u
t d
ata
o
f
th
e
ℎ
co
n
v
o
l
u
tio
n
lay
er
co
m
p
r
is
es
×
3
-
D
f
ea
t
u
r
e
cu
b
es.
I
n
o
u
r
m
o
d
el,
we
f
o
u
n
d
o
u
t th
e
ex
p
ec
ted
lab
els v
ia
th
e
FC
lay
er
s
an
d
a
So
f
tMa
x
lay
er
.
T
h
e
n
,
th
e
lo
s
s
f
u
n
ctio
n
o
f
th
e
en
tire
n
etwo
r
k
is
f
o
r
m
u
lated
u
s
in
g
(
5
)
.
=
1
∑
[
(
̂
)
+
(
1
−
)
(
1
−
̂
)
]
=
1
(
5
)
wh
er
e
an
d
̂
d
en
o
te
th
e
lab
el
an
d
p
r
ed
icted
o
f
th
e
ℎ
d
ata,
r
esp
ec
tiv
ely
.
r
ep
r
esen
ts
t
h
e
n
u
m
b
er
o
f
tr
ain
in
g
s
am
p
les.
Ou
r
ap
p
r
o
ac
h
ad
o
p
ted
th
e
r
ec
tifie
d
lin
ea
r
u
n
it
(
R
eL
U)
f
u
n
c
tio
n
(
)
=
(
0
,
)
.
R
eL
U
ac
tiv
atio
n
g
u
ar
an
tees
th
e
co
n
v
o
l
u
tio
n
al
f
ea
tu
r
e
ex
tr
a
cto
r
s
(
FE)
n
o
n
lin
e
ar
ity
an
d
aid
s
th
e
q
u
ick
er
tr
ain
in
g
o
f
th
e
n
etwo
r
k
.
I
n
ad
d
itio
n
,
we
d
ep
lo
y
ed
th
e
m
in
i
-
b
atch
s
to
ch
asti
c
g
r
ad
ien
t
d
escen
t
(
SGD)
ap
p
r
o
ac
h
t
o
o
p
tim
ize
th
e
n
etwo
r
k
ef
f
ec
tiv
ely
.
T
o
tr
ain
o
u
r
d
ataset
in
th
e
ex
p
er
im
en
ts
,
we
s
et
th
e
tr
ain
in
g
ep
o
ch
s
to
1
0
0
,
th
e
lear
n
in
g
r
ate
(
lr
)
t
o
0
.
0
0
1
,
a
n
d
th
e
p
atch
s
ize
t
o
2
5
.
I
n
a
d
d
itio
n
,
all
s
im
u
latio
n
s
wer
e
ex
ec
u
te
d
o
n
a
Ma
cBo
o
k
Pro
la
p
to
p
with
a
n
I
n
tel
i7
-
5
8
2
0
K
3
.
3
0
GHz
p
r
o
ce
s
s
o
r
,
8
GB
o
f
R
AM
,
an
d
an
NVI
DI
A
GT
X1
0
8
0
g
r
ap
h
ics
ca
r
d
an
d
GPU
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
14
,
No
.
2
,
Ap
r
il
20
25
:
1
2
1
1
-
1
2
1
9
1216
(
C
o
lab
with
2
5
GB
R
AM
)
,
al
l
r
u
n
n
in
g
o
n
Py
th
o
n
3
.
9
.
Fig
u
r
e
5
,
th
u
s
d
ep
icts
th
e
s
u
m
m
ar
y
o
f
o
u
r
p
r
o
p
o
s
ed
m
o
d
el
p
ar
a
m
eter
s
.
Fig
u
r
e
5
.
Su
m
m
ar
y
o
f
th
e
p
r
o
p
o
s
ed
m
o
d
el
p
ar
am
eter
s
4.
RE
SU
L
T
S AN
D
D
I
SCU
SS
I
O
N
W
e
co
n
d
u
cted
ex
p
e
r
im
en
ts
to
d
em
o
n
s
tr
ate
h
o
w
s
p
atial
in
f
o
r
m
atio
n
s
ig
n
if
ican
tly
in
f
lu
en
ce
s
HSI
class
if
icatio
n
.
T
wo
HSI
b
en
ch
m
ar
k
im
ag
es,
in
clu
d
in
g
th
e
I
n
d
ian
Pin
es
(
I
P)
an
d
th
e
Un
iv
er
s
ity
o
f
Pav
ia
(
Pav
iaU)
,
wer
e
s
tu
d
ied
to
e
v
alu
ate
th
e
ef
f
icien
cy
o
f
o
u
r
p
r
o
p
o
s
ed
m
o
d
el,
a
n
d
we
co
m
p
ar
ed
o
u
r
f
i
n
d
in
g
s
wit
h
v
ar
io
u
s
s
p
ec
tr
al
-
s
p
atial
HSI
b
aselin
e
ap
p
r
o
ac
h
es
to
d
eter
m
in
e
th
e
p
er
f
o
r
m
a
n
ce
o
f
t
h
e
p
r
o
p
o
s
ed
m
o
d
el.
T
ab
le
1
ex
p
lain
s
th
e
s
ig
n
if
ican
t
f
ea
tu
r
es
o
f
ea
ch
d
ataset
u
s
ed
in
o
u
r
ex
p
er
i
m
en
ts
,
in
clu
d
in
g
th
e
n
u
m
b
e
r
o
f
p
ix
els,
th
e
n
u
m
b
e
r
o
f
s
p
ec
tr
a
l
b
an
d
s
,
wav
elen
g
th
r
an
g
e
,
s
p
atial
r
eso
lu
tio
n
,
th
e
n
u
m
b
er
o
f
class
es,
an
d
th
e
s
en
s
o
r
,
r
esp
ec
tiv
ely
.
Fu
r
th
er
,
t
o
estab
lis
h
th
e
ef
f
icac
y
o
f
o
u
r
m
o
d
el,
we
r
an
d
o
m
ly
ch
o
s
e
N=
5
%
s
am
p
les
f
r
o
m
ea
ch
class
to
f
o
r
m
th
e
t
r
ain
in
g
d
ata
s
et
f
o
r
th
e
IP
an
d
N=
1
%
th
e
tr
ain
in
g
s
am
p
le
s
ize
f
o
r
th
e
Pav
iaU
d
ataset.
Fu
r
th
er
,
we
d
esig
n
ated
th
e
r
e
m
ain
in
g
r
ef
er
e
n
ce
s
am
p
les
as
th
e
test
in
g
d
ata
s
et.
Her
e,
x
i
d
en
o
tes
th
e
n
u
m
b
e
r
o
f
s
am
p
le
s
et
s
izes.
E
ac
h
ex
p
e
r
im
en
t w
as r
ep
ea
ted
f
i
v
e
tim
es with
r
an
d
o
m
ly
ch
o
s
en
tr
ain
in
g
s
am
p
les.
T
o
q
u
an
titativ
ely
test
th
e
ef
f
ec
tiv
en
ess
o
f
th
e
p
r
o
p
o
s
e
d
m
o
d
el
an
d
v
alid
ity
o
f
th
e
d
er
iv
ed
co
n
clu
s
io
n
s
,
th
r
ee
p
er
f
o
r
m
an
ce
an
aly
s
is
m
etr
ics,
i.e
.
,
o
v
er
all
ac
cu
r
ac
y
(
OA)
,
a
v
er
ag
e
a
cc
u
r
ac
y
(
AA)
,
an
d
k
ap
p
a
co
ef
f
icien
t,
wer
e
u
s
ed
.
AA
is
th
e
m
ea
n
p
er
ce
n
tag
e
o
f
co
r
r
ec
tly
class
if
ied
p
ix
els
f
o
r
ea
ch
class
.
T
h
e
k
ap
p
a
co
e
f
f
icien
t
g
iv
es
th
e
p
er
ce
n
tag
e
o
f
co
r
r
ec
tly
class
if
ied
p
ix
els
ab
o
u
t
class
if
icatio
n
f
in
d
in
g
s
ex
p
ec
ted
p
u
r
ely
b
y
ch
a
n
ce
.
T
ab
le
1
.
T
h
e
h
y
p
er
s
p
ec
tr
al
d
at
aset d
etails
[
2
8
]
H
S
I
c
l
a
ssi
f
i
c
a
t
i
o
n
#
P
i
x
e
l
s
B
a
n
d
s
W
a
v
e
l
e
n
g
t
h
r
a
n
g
e
(
µ
m)
S
p
a
t
i
a
l
r
e
s
o
l
u
t
i
o
n
#
C
l
a
ss
e
s
S
e
n
s
o
r
I
n
d
i
a
n
P
i
n
e
s
1
4
5
×
1
4
5
2
0
0
0
.
4
-
2
.
5
2
0
m
16
A
V
I
R
I
S
U
n
i
v
e
r
si
t
y
o
f
P
a
v
i
a
6
1
0
×
3
4
0
1
0
3
0
.
4
3
-
0
.
8
6
1
.
3
m
9
R
O
S
I
S
T
o
ass
ess
th
e
ac
h
iev
em
e
n
t
o
f
o
u
r
p
r
o
p
o
s
ed
ap
p
r
o
ac
h
,
we
c
o
m
p
ar
ed
th
e
class
if
icatio
n
r
es
u
lts
o
f
o
u
r
m
o
d
el
u
s
in
g
th
e
latest
p
u
b
lis
h
ed
b
aselin
e
m
eth
o
d
s
co
m
m
o
n
ly
u
s
ed
C
NN
-
b
ased
HSI
class
if
icatio
n
m
eth
o
d
s
,
in
clu
d
in
g
th
e
2
D
-
C
NN
[
2
9
]
(
i.e
.
,
s
p
atial
ap
p
r
o
ac
h
)
,
m
u
lti
-
s
ca
le
3
D
d
ee
p
co
n
v
o
lu
tio
n
al
n
eu
r
al
n
etwo
r
k
(
M3
D
-
DC
NN
)
[
3
0
]
,
an
d
r
esid
u
al
h
y
b
r
id
s
p
ec
tr
al
n
etwo
r
k
(
R
-
Hy
b
r
id
SN
)
[
3
1
]
(
i.e
.
,
s
p
ec
tr
al
-
s
p
atial
m
eth
o
d
s
)
.
T
ab
les
2
a
n
d
3
s
h
o
w
th
e
tr
ai
n
in
g
s
am
p
les
f
o
r
ea
ch
d
ataset
an
d
th
e
m
ea
n
v
alu
es,
c
o
n
f
i
d
en
ce
in
te
r
v
al,
a
n
d
class
if
icatio
n
p
er
f
o
r
m
an
ce
f
o
r
th
e
two
d
atasets
.
W
e
ca
n
atte
s
t
th
at
o
u
r
p
r
o
p
o
s
ed
m
o
d
el
o
u
t
p
er
f
o
r
m
ed
b
aselin
e
m
eth
o
d
s
f
r
o
m
th
e
ac
h
iev
e
d
ex
p
er
im
en
tal
r
esu
lts
.
T
h
u
s
,
we
attr
ib
u
te
th
is
to
th
e
in
tr
o
d
u
c
ed
Den
s
eNe
t
-
b
ased
m
ec
h
an
is
m
to
m
o
d
el
th
e
s
em
a
n
tic
in
ter
d
ep
en
d
en
cies in
s
p
ati
al
an
d
ch
a
n
n
el
d
im
e
n
s
io
n
s
.
Fig
u
r
e
6
illu
s
tr
ates
o
u
r
m
o
d
el'
s
OA
ag
ain
s
t
th
e
b
aselin
e
m
et
h
o
d
s
.
Ou
r
p
r
o
p
o
s
ed
m
o
d
el
h
a
s
th
e
b
est
OA
o
n
th
e
I
P
d
ata
s
et
co
m
p
ar
ed
to
M3
D
-
DC
NN
an
d
2
D
-
C
NN
m
eth
o
d
s
b
u
t
is
s
lig
h
tly
b
etter
th
an
th
e
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
Lea
r
n
in
g
h
ig
h
-
leve
l sp
ec
tr
a
l
-
s
p
a
tia
l fe
a
tu
r
es fo
r
h
yp
ers
p
ec
tr
a
l ima
g
e
…
(
Do
u
g
la
s
Omwen
g
a
N
ya
b
u
g
a
)
1217
R
-
Hy
b
r
id
SN
m
eth
o
d
with
a
d
i
f
f
er
en
ce
o
f
1
.
1
4
%
(
OA
=9
7
.
6
0
%).
C
o
m
p
ar
i
n
g
th
e
r
esu
lts
o
f
th
e
Pav
iaU
d
ataset
(
i.e
.
,
5
%
tr
ain
s
et)
in
co
n
t
r
a
s
t
to
th
e
b
aselin
e
m
et
h
o
d
s
,
o
u
r
m
o
d
el
ac
h
iev
e
d
e
x
ce
llen
t
OA,
i.e
.
,
9
7
.
8
0
%,
as
in
d
icate
d
in
T
ab
le
3
an
d
r
ep
r
esen
ted
in
Fig
u
r
e
5
.
Ad
o
p
ti
n
g
th
e
Den
s
eNe
t
lay
er
o
n
o
u
r
m
o
d
el
co
n
tr
ib
u
te
d
h
ig
h
ly
to
p
r
ec
is
e
lear
n
in
g
o
f
th
e
s
p
atial
-
s
p
ec
tr
al
class
if
icatio
n
o
f
HSI
s
.
Fu
r
th
e
r
,
as
ev
id
en
ce
d
i
n
T
ab
les
2
an
d
3
,
we
n
o
tice
m
o
s
t
class
es
p
r
ed
icted
with
h
ig
h
ac
cu
r
ac
y
.
T
a
b
le
4
lis
ts
th
e
co
m
p
u
tatio
n
al
co
m
p
lex
ity
r
esu
lts
f
o
r
v
a
r
io
u
s
b
aselin
e
m
eth
o
d
s
.
B
ec
au
s
e
th
is
im
ag
e
is
r
o
u
g
h
ly
1
4
5
×1
4
5
×
2
0
0
p
ix
els,
it
co
s
ts
2
6
.
9
s
ec
o
n
d
s
o
n
t
h
e
I
P d
ata
s
e
t,
o
u
tp
er
f
o
r
m
in
g
th
e
o
th
er
c
o
m
p
ar
ed
m
et
h
o
d
s
.
T
ab
le
2
.
T
h
e
class
es d
etail
in
f
o
r
m
atio
n
,
t
h
e
tr
ain
in
g
an
d
test
in
g
s
am
p
le
s
ize,
an
d
t
h
e
ac
cu
r
ac
y
o
f
d
if
f
er
e
n
t
b
aselin
e
m
eth
o
d
s
,
p
l
u
s
o
u
r
p
r
o
p
o
s
ed
ap
p
r
o
ac
h
o
n
th
e
I
P d
ata
s
et
#
N
ame
#
Sa
mp
l
e
s
T
ra
i
n
(
5
%
)
T
e
s
t
(
9
5
%
)
2D
-
C
N
N
M3
D
-
D
C
N
N
R
-
H
y
b
r
i
d
S
N
O
u
rs
1
A
l
fa
l
fa
46
2
44
7
.
9
5
2
7
.
5
45
7
3
.
8
1
2
Co
r
n
-
N
o
t
i
l
l
1
4
2
8
71
1
3
5
7
7
0
.
6
9
5
9
.
1
5
9
5
.
4
5
9
4
.
1
4
3
Co
r
n
-
Mi
n
t
i
l
l
8
3
0
42
7
8
9
5
2
.
8
4
4
5
.
0
7
9
7
.
3
6
9
9
.
7
4
4
Co
r
n
2
3
7
12
2
2
5
2
7
.
5
1
3
8
.
4
9
9
4
.
8
9
9
.
0
8
5
G
ras
s
-
Pa
s
t
u
re
4
8
3
24
4
5
9
9
0
.
4
4
7
0
.
3
3
9
8
.
8
5
9
7
.
3
6
G
ras
s
-
T
ree
s
7
3
0
37
6
9
4
9
8
.
5
9
9
7
.
2
9
9
.
3
2
9
9
.
6
6
7
G
ras
s
-
Pa
s
t
u
re
-
M
o
w
e
d
28
1
27
1
0
.
3
7
1
8
.
5
2
9
5
.
5
6
9
6
.
3
1
8
H
a
y
-
W
i
n
d
r
o
w
e
d
4
7
8
24
4
5
4
9
9
.
9
6
9
8
.
0
4
1
0
0
1
0
0
9
O
a
t
s
20
1
19
1
6
.
3
2
2
5
.
7
9
6
5
.
2
6
8
8
.
8
9
10
So
y
b
ea
n
-
N
o
t
i
l
l
9
7
2
49
9
2
3
6
7
.
8
4
5
5
.
8
5
9
5
.
9
9
7
.
3
2
11
So
y
b
ea
n
-
M
i
n
t
i
l
l
2
4
5
5
1
2
3
2
3
3
2
7
8
.
1
6
7
6
.
2
9
8
.
0
9
9
8
.
5
4
12
So
y
b
ea
n
-
C
l
ea
n
5
9
3
30
5
6
3
4
2
.
0
1
3
3
.
8
9
8
9
.
1
5
9
5
.
2
4
13
W
h
ea
t
2
0
5
10
1
9
5
9
8
.
9
7
9
1
.
2
3
9
9
.
7
4
9
9
.
4
1
14
W
o
o
d
s
1
2
6
5
63
1
2
0
2
9
7
.
6
5
9
4
.
6
8
9
9
.
2
6
9
9
.
2
3
15
Bu
i
l
d
i
n
g
s
-
G
ra
s
s
-
T
ree
s
-
D
r
i
v
es
3
8
6
19
3
6
7
6
2
.
6
2
4
2
.
3
7
8
7
.
6
6
9
8
.
0
3
16
St
o
n
e
-
S
t
ee
l
-
T
o
w
er
s
93
5
88
7
6
.
0
2
4
9
.
3
2
8
8
.
1
8
9
1
.
7
6
A
A
(
%
)
6
2
.
3
7
±
1
.
6
4
5
7
.
7
3
±
6
.
5
2
9
0
.
6
0
±
1
.
5
3
9
5
.
1
5
±
0
.
3
1
O
A
(
%
)
7
5
.
4
7
±
1
.
6
4
6
8
.
8
8
±
3
.
7
7
9
6
.
4
6
±
0
.
3
3
9
7
.
6
0
±
0
.
3
4
K
a
p
p
a
(
%
)
0
.
7
1
8
±
0
.
0
1
0
.
6
4
2
±
0
.
0
4
5
0
.
9
6
0
±
0
.
0
0
4
0
.
9
7
3
±
0
.
5
6
T
ab
le
3
.
T
h
e
class
es d
etail
in
f
o
r
m
atio
n
,
t
h
e
tr
ain
in
g
an
d
test
in
g
s
am
p
le
s
ize,
an
d
t
h
e
ac
cu
r
ac
y
o
f
d
if
f
er
e
n
t
b
aselin
e
m
eth
o
d
s
,
p
l
u
s
o
u
r
p
r
o
p
o
s
ed
ap
p
r
o
ac
h
o
n
th
e
Pav
iau
d
ataset
#
N
a
me
#
S
a
m
p
l
e
s
Tr
a
i
n
(
1
%)
Te
st
(
9
9
%)
2D
-
C
N
N
M3D
-
D
C
N
N
R
-
H
y
b
r
i
d
S
N
O
u
r
s
1
A
sp
h
a
l
t
6
6
3
1
66
6
5
6
5
9
6
.
8
8
9
0
.
5
6
9
6
.
9
4
9
7
.
4
2
M
e
a
d
o
w
s
1
8
6
4
9
1
8
6
1
8
4
6
3
9
9
.
0
1
8
9
.
4
7
9
9
.
6
9
9
9
.
1
8
3
G
r
a
v
e
l
2
0
9
9
21
2
0
7
8
7
5
.
0
8
5
9
.
1
1
8
7
.
1
7
8
8
.
3
4
4
Tr
e
e
s
3
0
6
4
31
3
0
3
3
8
7
.
7
4
9
3
.
2
5
8
9
.
1
5
8
9
.
8
8
5
P
a
i
n
t
e
d
m
e
t
a
l
sh
e
e
t
s
1
3
4
5
13
1
3
3
2
9
8
.
1
7
9
3
.
6
6
9
9
.
5
1
8
9
.
2
6
6
B
a
r
e
S
o
i
l
5
0
2
9
50
4
9
7
9
7
5
.
5
1
6
9
.
6
3
9
8
.
4
4
9
8
.
5
4
7
B
i
t
u
m
e
n
1
3
3
0
13
1
3
1
7
6
1
.
3
2
6
5
.
7
1
9
5
.
8
2
9
8
.
2
5
8
S
e
l
f
-
B
l
o
c
k
i
n
g
B
r
i
c
k
s
3
6
8
2
37
3
6
4
5
8
0
.
6
1
7
8
.
3
5
9
3
.
2
8
9
3
.
5
1
9
S
h
a
d
o
w
s
9
4
7
9
9
3
8
9
7
.
9
7
9
4
.
4
1
7
7
.
8
2
7
8
.
1
8
O
A
(
%)
9
1
.
1
3
±
0
.
5
5
8
4
.
6
3
±
1
.
2
1
9
6
.
5
9
±
0
.
5
0
9
7
.
8
0
±
0
.
0
9
A
A
(
%)
8
5
.
8
1
±
1
.
4
8
8
1
.
5
7
±
1
.
7
9
9
3
.
0
9
±
1
.
2
0
9
4
.
9
4
±
0
.
0
8
K
a
p
p
a
(
%)
0
.
8
8
1
±
0
.
0
0
8
0
.
7
9
8
±
0
.
0
1
6
0
.
9
5
5
±
0
.
0
0
7
0
.
9
6
4
±
0
.
3
1
Fig
u
r
e
6
.
T
h
e
OA
f
o
r
b
o
th
I
P (
5
% tr
ain
s
et)
an
d
Pav
iaU
(
1
% tr
ain
s
et)
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
14
,
No
.
2
,
Ap
r
il
20
25
:
1
2
1
1
-
1
2
1
9
1218
T
ab
le
4
.
T
h
e
tim
e
(
s
)
co
m
p
u
ta
tio
n
co
s
t o
f
v
ar
io
u
s
b
aselin
e
m
eth
o
d
s
in
IP
M
e
t
h
o
d
Tr
a
i
n
i
n
g
t
i
me
(
s)
Te
st
i
n
g
t
i
m
e
(
s)
2D
-
C
N
N
5
2
8
.
3
7
.
4
M3D
-
D
C
N
N
6
6
0
.
6
1
2
6
.
3
R
-
H
y
b
r
i
d
S
N
4
6
4
.
3
6
0
.
9
O
u
r
s
4
3
0
.
8
2
6
.
9
5.
CO
NCLU
SI
O
N
I
n
th
is
s
tu
d
y
,
we
p
r
o
p
o
s
ed
a
tech
n
iq
u
e
f
o
r
lear
n
in
g
h
ig
h
-
lev
el
s
p
ec
tr
al
-
s
p
atial
f
ea
tu
r
es
f
o
r
HSI
class
i
ficatio
n
with
in
s
u
f
f
icien
t
lab
eled
s
am
p
les
f
r
a
m
ewo
r
k
t
h
at
jo
in
tly
u
s
ed
d
im
en
s
io
n
r
e
d
u
ctio
n
a
n
d
tr
an
s
f
er
lear
n
in
g
tech
n
iq
u
es f
o
r
lear
n
in
g
th
e
3
D
s
p
atial
-
s
p
ec
tr
al
clas
s
i
f
icatio
n
o
f
HSI
s
.
Ou
r
p
r
o
p
o
s
e
d
f
r
am
ewo
r
k
f
u
s
ed
a
Den
s
eNe
t
tr
an
s
f
er
lear
n
in
g
n
etwo
r
k
f
o
r
s
p
ec
tr
al
-
s
p
atial
f
e
atu
r
e
lear
n
in
g
,
th
e
PC
A
tech
n
iq
u
e
f
o
r
e
x
tr
ac
tin
g
f
ea
tu
r
es
f
r
o
m
h
i
g
h
-
d
im
e
n
s
io
n
al
h
y
p
er
s
p
ec
t
r
al
d
ata
s
ets,
an
d
th
e
3
D
an
d
2
D
c
o
n
v
o
lu
tio
n
s
.
T
h
e
ex
p
er
im
en
ts
wer
e
co
n
d
u
cted
o
n
two
d
atase
ts
,
i.e
.
,
I
P
an
d
th
e
Pav
iaU.
T
h
e
r
esu
lts
r
ev
ea
led
o
u
r
m
o
d
el'
s
h
ig
h
er
p
e
r
f
o
r
m
an
ce
th
an
o
th
er
b
aselin
e
m
eth
o
d
s
t
h
at
we
co
m
p
ar
ed
o
u
r
m
o
d
el.
T
h
u
s
,
th
is
m
eth
o
d
h
as
g
o
o
d
p
o
ten
tial
f
o
r
p
r
ec
is
e
HSI
clas
s
if
icatio
n
,
o
f
ten
co
n
t
ain
in
g
h
ig
h
d
im
e
n
s
io
n
s
o
f
s
p
ec
tr
al
-
s
p
atial
f
ea
tu
r
es.
W
ith
p
o
ten
tial
b
en
ef
its
f
o
r
th
e
en
v
ir
o
n
m
e
n
tal
an
d
ag
r
icu
ltu
r
al
d
o
m
ai
n
s
,
th
is
s
u
g
g
ested
tech
n
iq
u
e
h
o
ld
s
s
ig
n
if
ican
t
v
alu
e
in
en
h
a
n
cin
g
p
r
ec
is
io
n
an
d
ef
f
icac
y
i
n
HSI
ca
teg
o
r
izatio
n
.
W
e
will
in
clu
d
e
s
p
atial
in
f
o
r
m
atio
n
i
n
th
e
m
o
d
el,
wh
ich
was
a
lim
itatio
n
in
th
e
cu
r
r
en
t
wo
r
k
in
th
e
f
u
tu
r
e
to
im
p
r
o
v
e
f
ea
tu
r
e
class
if
icatio
n
ca
p
ab
ilit
y
.
Fu
r
th
er
m
o
r
e,
b
ec
au
s
e
th
e
v
alu
e
o
f
s
p
ec
tr
al
a
n
d
s
p
a
tial
f
ea
tu
r
es
d
if
f
er
s
d
ep
en
d
in
g
o
n
th
e
m
ate
r
ial
an
d
s
ce
n
ar
io
,
it
is
also
wo
r
th
lo
o
k
in
g
at
h
o
w
to
au
to
m
atica
ll
y
lear
n
th
e
p
r
o
p
er
weig
h
tin
g
f
ac
to
r
.
RE
F
E
R
E
NC
E
S
[
1
]
Q
.
W
a
n
g
,
Z.
M
e
n
g
,
a
n
d
X
.
Li
,
“
L
o
c
a
l
i
t
y
a
d
a
p
t
i
v
e
d
i
s
c
r
i
m
i
n
a
n
t
a
n
a
l
y
s
i
s
f
o
r
s
p
e
c
t
r
a
l
–
sp
a
t
i
a
l
c
l
a
ssi
f
i
c
a
t
i
o
n
o
f
h
y
p
e
r
s
p
e
c
t
r
a
l
i
ma
g
e
s,
”
I
E
EE
G
e
o
s
c
i
e
n
c
e
a
n
d
Re
m
o
t
e
S
e
n
si
n
g
L
e
t
t
e
rs
,
v
o
l
.
1
4
,
n
o
.
1
1
,
p
p
.
2
0
7
7
–
2
0
8
1
,
2
0
1
7
,
d
o
i
:
1
0
.
1
1
0
9
/
LG
R
S
.
2
0
1
7
.
2
7
5
1
5
5
9
.
[
2
]
L.
H
e
,
J.
Li
,
A
.
P
l
a
z
a
,
a
n
d
Y
.
Li
,
“
D
i
s
c
r
i
mi
n
a
t
i
v
e
l
o
w
-
r
a
n
k
g
a
b
o
r
f
i
l
t
e
r
i
n
g
f
o
r
s
p
e
c
t
r
a
l
-
sp
a
t
i
a
l
h
y
p
e
r
s
p
e
c
t
r
a
l
i
ma
g
e
c
l
a
ss
i
f
i
c
a
t
i
o
n
,
”
I
EEE
T
r
a
n
s
a
c
t
i
o
n
s
o
n
G
e
o
sc
i
e
n
c
e
a
n
d
R
e
m
o
t
e
S
e
n
s
i
n
g
,
v
o
l
.
5
5
,
n
o
.
3
,
p
p
.
1
3
8
1
–
1
3
9
5
,
2
0
1
7
,
d
o
i
:
1
0
.
1
1
0
9
/
TG
R
S
.
2
0
1
6
.
2
6
2
3
7
4
2
.
[
3
]
P
.
C
h
e
n
,
L.
Ji
a
o
,
F
.
L
i
u
,
S
.
G
o
u
,
J.
Zh
a
o
,
a
n
d
Z
.
Z
h
a
o
,
“
D
i
me
n
si
o
n
a
l
i
t
y
r
e
d
u
c
t
i
o
n
o
f
h
y
p
e
r
sp
e
c
t
r
a
l
i
m
a
g
e
r
y
u
si
n
g
s
p
a
r
s
e
g
r
a
p
h
l
e
a
r
n
i
n
g
,
”
I
EEE
J
o
u
r
n
a
l
o
f
S
e
l
e
c
t
e
d
T
o
p
i
c
s
i
n
A
p
p
l
i
e
d
E
a
r
t
h
O
b
s
e
rv
a
t
i
o
n
s
a
n
d
R
e
m
o
t
e
S
e
n
s
i
n
g
,
v
o
l
.
1
0
,
n
o
.
3
,
p
p
.
1
1
6
5
–
1
1
8
1
,
2
0
1
7
,
d
o
i
:
1
0
.
1
1
0
9
/
JS
TA
R
S
.
2
0
1
6
.
2
6
0
6
5
7
8
.
[
4
]
A
.
K
i
a
n
i
s
a
r
k
a
l
e
h
a
n
d
H
.
G
h
a
sse
mi
a
n
,
“
M
a
r
g
i
n
a
l
d
i
s
c
r
i
m
i
n
a
n
t
a
n
a
l
y
si
s
u
si
n
g
s
u
p
p
o
r
t
v
e
c
t
o
r
s
f
o
r
d
i
me
n
s
i
o
n
a
l
i
t
y
r
e
d
u
c
t
i
o
n
o
f
h
y
p
e
r
s
p
e
c
t
r
a
l
d
a
t
a
,
”
Re
m
o
t
e
S
e
n
s
i
n
g
L
e
t
t
e
rs
,
v
o
l
.
7
,
n
o
.
1
2
,
p
p
.
1
1
6
0
–
1
1
6
9
,
2
0
1
6
,
d
o
i
:
1
0
.
1
0
8
0
/
2
1
5
0
7
0
4
X
.
2
0
1
6
.
1
2
2
2
0
9
9
.
[
5
]
F
.
P
a
l
sso
n
,
J.
R
.
S
v
e
i
n
ss
o
n
,
M
.
O
.
U
l
f
a
r
sso
n
,
a
n
d
J.
A
.
B
e
n
e
d
i
k
t
ss
o
n
,
“
M
o
d
e
l
-
b
a
se
d
f
u
si
o
n
o
f
m
u
l
t
i
-
a
n
d
h
y
p
e
r
s
p
e
c
t
r
a
l
i
ma
g
e
s
u
si
n
g
P
C
A
a
n
d
w
a
v
e
l
e
t
s,”
I
E
EE
T
ra
n
sa
c
t
i
o
n
s
o
n
G
e
o
sc
i
e
n
c
e
a
n
d
Re
m
o
t
e
S
e
n
si
n
g
,
v
o
l
.
5
3
,
n
o
.
5
,
p
p
.
2
6
5
2
–
2
6
6
3
,
2
0
1
5
,
d
o
i
:
1
0
.
1
1
0
9
/
TG
R
S
.
2
0
1
4
.
2
3
6
3
4
7
7
.
[
6
]
S
.
S
h
i
n
d
e
a
n
d
H
.
P
a
t
i
d
a
r
,
“
H
y
p
e
r
s
p
e
c
t
r
a
l
i
m
a
g
e
c
l
a
ss
i
f
i
c
a
t
i
o
n
u
s
i
n
g
p
r
i
n
c
i
p
l
e
c
o
m
p
o
n
e
n
t
a
n
a
l
y
s
i
s
a
n
d
d
e
e
p
c
o
n
v
o
l
u
t
i
o
n
a
l
n
e
u
r
a
l
n
e
t
w
o
r
k
,
”
J
o
u
rn
a
l
o
f
Am
b
i
e
n
t
I
n
t
e
l
l
i
g
e
n
c
e
a
n
d
H
u
m
a
n
i
ze
d
C
o
m
p
u
t
i
n
g
,
v
o
l
.
1
4
,
n
o
.
1
2
,
p
p
.
1
6
4
9
1
–
1
6
4
9
7
,
2
0
2
3
,
d
o
i
:
1
0
.
1
0
0
7
/
s
1
2
6
5
2
-
0
2
2
-
0
3
8
7
6
-
z.
[
7
]
G
.
Li
c
c
i
a
r
d
i
,
P
.
R
.
M
a
r
p
u
,
J.
C
h
a
n
u
sso
t
,
a
n
d
J.
A
.
B
e
n
e
d
i
k
t
ss
o
n
,
“
Li
n
e
a
r
v
e
r
s
u
s
n
o
n
l
i
n
e
a
r
P
C
A
f
o
r
t
h
e
c
l
a
ss
i
f
i
c
a
t
i
o
n
o
f
h
y
p
e
r
s
p
e
c
t
r
a
l
d
a
t
a
b
a
s
e
d
o
n
t
h
e
e
x
t
e
n
d
e
d
mo
r
p
h
o
l
o
g
i
c
a
l
p
r
o
f
i
l
e
s,
”
I
EE
E
G
e
o
sc
i
e
n
c
e
a
n
d
R
e
m
o
t
e
S
e
n
si
n
g
L
e
t
t
e
rs
,
v
o
l
.
9
,
n
o
.
3
,
p
p
.
4
4
7
–
4
5
1
,
2
0
1
2
,
d
o
i
:
1
0
.
1
1
0
9
/
LG
R
S
.
2
0
1
1
.
2
1
7
2
1
8
5
.
[
8
]
D
.
K
.
P
a
t
h
a
k
,
S
.
K
.
K
a
l
i
t
a
,
a
n
d
D
.
K
.
B
h
a
t
t
a
c
h
a
r
y
a
,
“
H
y
p
e
r
s
p
e
c
t
r
a
l
i
m
a
g
e
c
l
a
ss
i
f
i
c
a
t
i
o
n
u
s
i
n
g
s
u
p
p
o
r
t
v
e
c
t
o
r
m
a
c
h
i
n
e
:
a
s
p
e
c
t
r
a
l
s
p
a
t
i
a
l
f
e
a
t
u
r
e
b
a
se
d
a
p
p
r
o
a
c
h
,
”
E
v
o
l
u
t
i
o
n
a
ry
I
n
t
e
l
l
i
g
e
n
c
e
,
v
o
l
.
1
5
,
n
o
.
3
,
p
p
.
1
8
0
9
–
1
8
2
3
,
2
0
2
2
,
d
o
i
:
1
0
.
1
0
0
7
/
s
1
2
0
6
5
-
021
-
00591
-
0.
[
9
]
J.
M
.
B
i
o
u
c
a
s
-
D
i
a
s,
A
.
P
l
a
z
a
,
G
.
C
a
m
p
s
-
V
a
l
l
s,
P
.
S
c
h
e
u
n
d
e
r
s,
N
.
M
.
N
a
sr
a
b
a
d
i
,
a
n
d
J
.
C
h
a
n
u
s
so
t
,
“
H
y
p
e
r
sp
e
c
t
r
a
l
r
e
mo
t
e
s
e
n
s
i
n
g
d
a
t
a
a
n
a
l
y
s
i
s
a
n
d
f
u
t
u
r
e
c
h
a
l
l
e
n
g
e
s,
”
I
EE
E
G
e
o
s
c
i
e
n
c
e
a
n
d
Re
m
o
t
e
S
e
n
si
n
g
Ma
g
a
zi
n
e
,
v
o
l
.
1
,
n
o
.
2
,
p
p
.
6
–
3
6
,
2
0
1
3
,
d
o
i
:
1
0
.
1
1
0
9
/
M
G
R
S
.
2
0
1
3
.
2
2
4
4
6
7
2
.
[
1
0
]
G
.
C
a
m
p
s
-
V
a
l
l
s,
D
.
Tu
i
a
,
L
.
B
r
u
z
z
o
n
e
,
a
n
d
J
.
A
.
B
e
n
e
d
i
k
t
sso
n
,
“
A
d
v
a
n
c
e
s
i
n
h
y
p
e
r
s
p
e
c
t
r
a
l
i
ma
g
e
c
l
a
s
si
f
i
c
a
t
i
o
n
:
E
a
r
t
h
mo
n
i
t
o
r
i
n
g
w
i
t
h
st
a
t
i
s
t
i
c
a
l
l
e
a
r
n
i
n
g
me
t
h
o
d
s,”
I
EE
E
S
i
g
n
a
l
Pr
o
c
e
ssi
n
g
Ma
g
a
zi
n
e
,
v
o
l
.
3
1
,
n
o
.
1
,
p
p
.
4
5
–
5
4
,
2
0
1
4
,
d
o
i
:
1
0
.
1
1
0
9
/
M
S
P
.
2
0
1
3
.
2
2
7
9
1
7
9
.
[
1
1
]
D
.
A
.
La
n
d
g
r
e
b
e
,
S
i
g
n
a
l
t
h
e
o
ry
m
e
t
h
o
d
s
i
n
m
u
l
t
i
sp
e
c
t
ra
l
rem
o
t
e
se
n
s
i
n
g
.
H
o
b
o
k
e
n
,
U
S
A
:
J
o
h
n
W
i
l
e
y
&
S
o
n
s,
I
n
c
.
,
2
0
0
3
,
d
o
i
:
1
0
.
1
0
0
2
/
0
4
7
1
7
2
3
8
0
0
.
[
1
2
]
A
.
Za
r
e
a
n
d
K
.
C
.
H
o
,
“
En
d
me
m
b
e
r
v
a
r
i
a
b
i
l
i
t
y
i
n
h
y
p
e
r
s
p
e
c
t
r
a
l
a
n
a
l
y
s
i
s:
A
d
d
r
e
ssi
n
g
s
p
e
c
t
r
a
l
v
a
r
i
a
b
i
l
i
t
y
d
u
r
i
n
g
s
p
e
c
t
r
a
l
u
n
m
i
x
i
n
g
,
”
I
EEE
S
i
g
n
a
l
Pro
c
e
ss
i
n
g
Ma
g
a
zi
n
e
,
v
o
l
.
3
1
,
n
o
.
1
,
p
p
.
9
5
–
1
0
4
,
2
0
1
4
,
d
o
i
:
1
0
.
1
1
0
9
/
M
S
P
.
2
0
1
3
.
2
2
7
9
1
7
7
.
[
1
3
]
M
.
F
a
u
v
e
l
,
Y
.
Ta
r
a
b
a
l
k
a
,
J.
A
.
B
e
n
e
d
i
k
t
sso
n
,
J.
C
h
a
n
u
ss
o
t
,
a
n
d
J.
C
.
Ti
l
t
o
n
,
“
A
d
v
a
n
c
e
s
i
n
sp
e
c
t
r
a
l
-
s
p
a
t
i
a
l
c
l
a
ssi
f
i
c
a
t
i
o
n
o
f
h
y
p
e
r
s
p
e
c
t
r
a
l
i
ma
g
e
s
,
”
Pro
c
e
e
d
i
n
g
s
o
f
t
h
e
I
EEE
,
v
o
l
.
1
0
1
,
n
o
.
3
,
p
p
.
6
5
2
–
6
7
5
,
2
0
1
3
,
d
o
i
:
1
0
.
1
1
0
9
/
JP
R
O
C
.
2
0
1
2
.
2
1
9
7
5
8
9
.
[
1
4
]
G
.
H
u
a
n
g
,
Z
.
Li
u
,
L.
V
.
D
.
M
a
a
t
e
n
,
a
n
d
K
.
Q
.
W
e
i
n
b
e
r
g
e
r
,
“
D
e
n
sel
y
c
o
n
n
e
c
t
e
d
c
o
n
v
o
l
u
t
i
o
n
a
l
n
e
t
w
o
r
k
s,
”
i
n
2
0
1
7
I
EE
E
C
o
n
f
e
re
n
c
e
o
n
C
o
m
p
u
t
e
r
Vi
s
i
o
n
a
n
d
Pa
t
t
e
r
n
Re
c
o
g
n
i
t
i
o
n
(
C
VP
R)
,
2
0
1
7
,
p
p
.
2
2
6
1
–
2
2
6
9
,
d
o
i
:
1
0
.
1
1
0
9
/
C
V
P
R
.
2
0
1
7
.
2
4
3
.
[
1
5
]
M
.
H
a
m
o
u
d
a
,
K
.
S
.
Et
t
a
b
a
a
,
a
n
d
M
.
S
.
B
o
u
h
l
e
l
,
“
S
mart
f
e
a
t
u
r
e
e
x
t
r
a
c
t
i
o
n
a
n
d
c
l
a
ssi
f
i
c
a
t
i
o
n
o
f
h
y
p
e
r
s
p
e
c
t
r
a
l
i
ma
g
e
s
b
a
s
e
d
o
n
c
o
n
v
o
l
u
t
i
o
n
a
l
n
e
u
r
a
l
n
e
t
w
o
r
k
s,
”
I
ET I
m
a
g
e
Pr
o
c
e
ssi
n
g
,
v
o
l
.
1
4
,
n
o
.
1
0
,
p
p
.
1
9
9
9
–
2
0
0
5
,
2
0
2
0
,
d
o
i
:
1
0
.
1
0
4
9
/
i
e
t
-
i
p
r
.
2
0
1
9
.
1
2
8
2
.
[
1
6
]
L.
Y
a
n
g
e
t
a
l
.
,
“
F
u
si
o
n
N
e
t
:
A
c
o
n
v
o
l
u
t
i
o
n
–
t
r
a
n
sf
o
r
m
e
r
f
u
s
i
o
n
n
e
t
w
o
r
k
f
o
r
h
y
p
e
r
sp
e
c
t
r
a
l
i
ma
g
e
c
l
a
ssi
f
i
c
a
t
i
o
n
,
”
Re
m
o
t
e
S
e
n
s
i
n
g
,
v
o
l
.
1
4
,
n
o
.
1
6
,
2
0
2
2
,
d
o
i
:
1
0
.
3
3
9
0
/
r
s
1
4
1
6
4
0
6
6
.
[
1
7
]
B
.
L
i
u
,
K
.
G
a
o
,
A
.
Y
u
,
L
.
D
i
n
g
,
C
.
Q
i
u
,
a
n
d
J.
Li
,
“
ES
2
F
L:
E
n
s
e
m
b
l
e
sel
f
-
s
u
p
e
r
v
i
se
d
f
e
a
t
u
r
e
l
e
a
r
n
i
n
g
f
o
r
smal
l
sam
p
l
e
c
l
a
ss
i
f
i
c
a
t
i
o
n
o
f
h
y
p
e
r
sp
e
c
t
r
a
l
i
m
a
g
e
s,
”
Re
m
o
t
e
S
e
n
si
n
g
,
v
o
l
.
1
4
,
n
o
.
1
7
,
2
0
2
2
,
d
o
i
:
1
0
.
3
3
9
0
/
r
s1
4
1
7
4
2
3
6
.
[
1
8
]
X
.
L
i
u
e
t
a
l
.
,
“
H
-
R
N
e
t
:
H
y
b
r
i
d
r
e
l
a
t
i
o
n
n
e
t
w
o
r
k
f
o
r
f
e
w
-
s
h
o
t
l
e
a
r
n
i
n
g
-
b
a
se
d
h
y
p
e
r
sp
e
c
t
r
a
l
i
m
a
g
e
c
l
a
ssi
f
i
c
a
t
i
o
n
,
”
Re
m
o
t
e
S
e
n
s
i
n
g
,
v
o
l
.
1
5
,
n
o
.
1
0
,
2
0
2
3
,
d
o
i
:
1
0
.
3
3
9
0
/
r
s
1
5
1
0
2
4
9
7
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
Lea
r
n
in
g
h
ig
h
-
leve
l sp
ec
tr
a
l
-
s
p
a
tia
l fe
a
tu
r
es fo
r
h
yp
ers
p
ec
tr
a
l ima
g
e
…
(
Do
u
g
la
s
Omwen
g
a
N
ya
b
u
g
a
)
1219
[
1
9
]
S
.
Ji
a
,
S
.
J
i
a
n
g
,
Z
.
Li
n
,
N
.
L
i
,
M
.
X
u
,
a
n
d
S
.
Y
u
,
“
A
s
u
r
v
e
y
:
D
e
e
p
l
e
a
r
n
i
n
g
f
o
r
h
y
p
e
r
s
p
e
c
t
r
a
l
i
ma
g
e
c
l
a
ssi
f
i
c
a
t
i
o
n
w
i
t
h
f
e
w
l
a
b
e
l
e
d
samp
l
e
s
,
”
N
e
u
r
o
c
o
m
p
u
t
i
n
g
,
v
o
l
.
4
4
8
,
p
p
.
1
7
9
–
2
0
4
,
2
0
2
1
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
n
e
u
c
o
m
.
2
0
2
1
.
0
3
.
0
3
5
.
[
2
0
]
E.
R
.
K
o
n
d
a
l
a
n
d
S
.
S
.
B
a
r
p
a
n
d
a
,
“
H
y
p
e
r
sp
e
c
t
r
a
l
i
m
a
g
e
c
l
a
ssi
f
i
c
a
t
i
o
n
u
s
i
n
g
H
y
b
-
3
D
c
o
n
v
o
l
u
t
i
o
n
n
e
u
r
a
l
n
e
t
w
o
r
k
s
p
e
c
t
r
a
l
p
a
r
t
i
t
i
o
n
i
n
g
,
”
I
n
d
o
n
e
s
i
a
n
J
o
u
r
n
a
l
o
f
El
e
c
t
ri
c
a
l
En
g
i
n
e
e
ri
n
g
a
n
d
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
,
v
o
l
.
2
9
,
n
o
.
1
,
p
p
.
2
9
5
–
3
0
3
,
2
0
2
3
,
d
o
i
:
1
0
.
1
1
5
9
1
/
i
j
e
e
c
s.
v
2
9
.
i
1
.
p
p
2
9
5
-
3
0
3
.
[
2
1
]
Z.
W
u
,
Y
.
G
a
o
,
L.
L
i
,
J.
X
u
e
,
a
n
d
Y
.
Li
,
“
S
e
ma
n
t
i
c
s
e
g
me
n
t
a
t
i
o
n
o
f
h
i
g
h
-
r
e
so
l
u
t
i
o
n
r
e
m
o
t
e
s
e
n
s
i
n
g
i
m
a
g
e
s
u
si
n
g
f
u
l
l
y
c
o
n
v
o
l
u
t
i
o
n
a
l
n
e
t
w
o
r
k
w
i
t
h
a
d
a
p
t
i
v
e
t
h
r
e
s
h
o
l
d
,
”
C
o
n
n
e
c
t
i
o
n
S
c
i
e
n
c
e
,
v
o
l
.
3
1
,
n
o
.
2
,
p
p
.
1
6
9
–
1
8
4
,
2
0
1
9
,
d
o
i
:
1
0
.
1
0
8
0
/
0
9
5
4
0
0
9
1
.
2
0
1
8
.
1
5
1
0
9
0
2
.
[
2
2
]
V
.
K
.
S
h
r
i
v
a
s
t
a
v
a
a
n
d
M
.
K
.
P
r
a
d
h
a
n
,
“
H
y
p
e
r
sp
e
c
t
r
a
l
r
e
m
o
t
e
s
e
n
si
n
g
i
mag
e
c
l
a
s
si
f
i
c
a
t
i
o
n
u
si
n
g
a
c
t
i
v
e
l
e
a
r
n
i
n
g
,
”
S
t
u
d
i
e
s
i
n
C
o
m
p
u
t
a
t
i
o
n
a
l
I
n
t
e
l
l
i
g
e
n
c
e
,
v
o
l
.
9
0
7
,
p
p
.
1
3
3
–
1
5
2
,
2
0
2
1
,
d
o
i
:
1
0
.
1
0
0
7
/
9
7
8
-
3
-
0
3
0
-
5
0
6
4
1
-
4
_
8
.
[
2
3
]
Y
.
G
u
o
,
J
.
Ji
,
D
.
S
h
i
,
Q
.
Y
e
,
a
n
d
H
.
X
i
e
,
“
M
u
l
t
i
-
v
i
e
w
f
e
a
t
u
r
e
l
e
a
r
n
i
n
g
f
o
r
V
H
R
r
e
mo
t
e
se
n
si
n
g
i
m
a
g
e
c
l
a
ss
i
f
i
c
a
t
i
o
n
,
”
Mu
l
t
i
m
e
d
i
a
T
o
o
l
s
a
n
d
Ap
p
l
i
c
a
t
i
o
n
s
,
v
o
l
.
8
0
,
n
o
.
1
5
,
p
p
.
2
3
0
0
9
–
2
3
0
2
1
,
2
0
2
1
,
d
o
i
:
1
0
.
1
0
0
7
/
s
1
1
0
4
2
-
0
2
0
-
0
8
7
1
3
-
z.
[
2
4
]
S
.
P
u
n
d
i
r
a
n
d
J.
A
.
A
k
sh
a
y
,
“
EP
M
:
M
e
t
a
-
l
e
a
r
n
i
n
g
m
e
t
h
o
d
f
o
r
r
e
mo
t
e
se
n
si
n
g
i
m
a
g
e
c
l
a
ss
i
f
i
c
a
t
i
o
n
,
”
i
n
M
a
c
h
i
n
e
I
n
t
e
l
l
i
g
e
n
c
e
a
n
d
S
m
a
r
t
S
y
s
t
e
m
s
,
S
i
n
g
a
p
o
r
e
:
S
p
r
i
n
g
e
r
,
2
0
2
2
,
p
p
.
3
2
9
–
3
3
9
,
d
o
i
:
1
0
.
1
0
0
7
/
9
7
8
-
9
8
1
-
16
-
9
6
5
0
-
3
_
2
5
.
[
2
5
]
W
.
M
a
e
t
a
l
.
,
“
A
n
o
v
e
l
a
d
a
p
t
i
v
e
h
y
b
r
i
d
f
u
si
o
n
n
e
t
w
o
r
k
f
o
r
mu
l
t
i
r
e
s
o
l
u
t
i
o
n
r
e
mo
t
e
se
n
si
n
g
i
ma
g
e
s
c
l
a
s
si
f
i
c
a
t
i
o
n
,
”
I
EE
E
T
ra
n
s
a
c
t
i
o
n
s
o
n
G
e
o
sc
i
e
n
c
e
a
n
d
R
e
m
o
t
e
S
e
n
si
n
g
,
v
o
l
.
6
0
,
p
p
.
1
–
1
7
,
2
0
2
2
,
d
o
i
:
1
0
.
1
1
0
9
/
TG
R
S
.
2
0
2
1
.
3
0
6
2
1
4
2
.
[
2
6
]
N
.
S
a
t
h
y
a
n
a
r
a
y
a
n
a
a
n
d
S
.
S
i
n
g
h
,
“
H
y
b
r
i
d
a
d
a
p
t
i
v
e
n
e
u
r
a
l
n
e
t
w
o
r
k
f
o
r
r
e
mo
t
e
se
n
si
n
g
i
ma
g
e
c
l
a
s
si
f
i
c
a
t
i
o
n
,
”
I
AES
I
n
t
e
r
n
a
t
i
o
n
a
l
J
o
u
rn
a
l
o
f
Art
i
f
i
c
i
a
l
I
n
t
e
l
l
i
g
e
n
c
e
,
v
o
l
.
1
3
,
n
o
.
2
,
p
p
.
2
2
9
1
–
2
3
0
0
,
2
0
2
4
,
d
o
i
:
1
0
.
1
1
5
9
1
/
i
j
a
i
.
v
1
3
.
i
2
.
p
p
2
2
9
1
-
2
3
0
0
.
[
2
7
]
S
.
Li
u
,
S
.
M
i
a
o
,
J
.
S
u
,
B
.
L
i
,
W
.
H
u
,
a
n
d
Y
.
D
.
Z
h
a
n
g
,
“
U
M
A
G
-
N
e
t
:
A
n
e
w
u
n
s
u
p
e
r
v
i
s
e
d
m
u
l
t
i
a
t
t
e
n
t
i
o
n
-
g
u
i
d
e
d
n
e
t
w
o
r
k
f
o
r
h
y
p
e
r
s
p
e
c
t
r
a
l
a
n
d
mu
l
t
i
s
p
e
c
t
r
a
l
i
m
a
g
e
f
u
si
o
n
,
”
I
EE
E
J
o
u
rn
a
l
o
f
S
e
l
e
c
t
e
d
T
o
p
i
c
s
i
n
A
p
p
l
i
e
d
E
a
rt
h
O
b
s
e
rv
a
t
i
o
n
s
a
n
d
R
e
m
o
t
e
S
e
n
si
n
g
,
v
o
l
.
1
4
,
p
p
.
7
3
7
3
–
7
3
8
5
,
2
0
2
1
,
d
o
i
:
1
0
.
1
1
0
9
/
JS
TA
R
S
.
2
0
2
1
.
3
0
9
7
1
7
8
.
[
2
8
]
M
.
G
r
a
ñ
a
,
M
.
A
.
V
e
g
a
n
z
o
n
s
,
a
n
d
B
.
A
y
e
r
d
i
,
“
H
y
p
e
r
s
p
e
c
t
r
a
l
r
e
m
o
t
e
s
e
n
s
i
n
g
s
c
e
n
e
s
,
”
G
r
o
u
p
D
e
I
n
t
e
l
i
g
e
n
c
i
s
a
C
o
m
p
u
t
a
c
i
o
n
a
l
.
2
0
1
1
.
A
c
c
e
s
s
e
d
:
J
a
n
.
0
6
,
2
0
2
0
.
[
O
n
l
i
n
e
]
.
A
v
a
i
l
a
b
l
e
:
h
t
t
p
:
/
/
w
w
w
.
e
h
u
.
e
u
s
/
c
c
w
i
n
t
c
o
/
i
n
d
e
x
.
p
h
p
?
t
i
t
l
e
=
H
y
p
e
r
s
p
e
c
t
r
a
l
_
R
e
m
o
t
e
_
S
e
n
s
i
n
g
_
S
c
e
n
e
s
[
2
9
]
K
.
M
a
k
a
n
t
a
si
s
,
K
.
K
a
r
a
n
t
z
a
l
o
s,
A
.
D
o
u
l
a
m
i
s,
a
n
d
N
.
D
o
u
l
a
mi
s
,
“
D
e
e
p
s
u
p
e
r
v
i
se
d
l
e
a
r
n
i
n
g
f
o
r
h
y
p
e
r
sp
e
c
t
r
a
l
d
a
t
a
c
l
a
ss
i
f
i
c
a
t
i
o
n
t
h
r
o
u
g
h
c
o
n
v
o
l
u
t
i
o
n
a
l
n
e
u
r
a
l
n
e
t
w
o
r
k
s,”
i
n
2
0
1
5
I
EE
E
I
n
t
e
r
n
a
t
i
o
n
a
l
G
e
o
sc
i
e
n
c
e
a
n
d
R
e
m
o
t
e
S
e
n
s
i
n
g
S
y
m
p
o
si
u
m
(
I
G
AR
S
S
)
,
2
0
1
5
,
p
p
.
4
9
5
9
–
4
9
6
2
,
d
o
i
:
1
0
.
1
1
0
9
/
I
G
A
R
S
S
.
2
0
1
5
.
7
3
2
6
9
4
5
.
[
3
0
]
M
.
H
e
,
B
.
Li
,
a
n
d
H
.
C
h
e
n
,
“
M
u
l
t
i
-
s
c
a
l
e
3
D
d
e
e
p
c
o
n
v
o
l
u
t
i
o
n
a
l
n
e
u
r
a
l
n
e
t
w
o
r
k
f
o
r
h
y
p
e
r
sp
e
c
t
r
a
l
i
m
a
g
e
c
l
a
ssi
f
i
c
a
t
i
o
n
,
”
i
n
2
0
1
7
I
EEE
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
r
e
n
c
e
o
n
I
m
a
g
e
Pr
o
c
e
ssi
n
g
(
I
C
I
P)
,
2
0
1
7
,
p
p
.
3
9
0
4
–
3
9
0
8
,
d
o
i
:
1
0
.
1
1
0
9
/
I
C
I
P
.
2
0
1
7
.
8
2
9
7
0
1
4
.
[
3
1
]
F
.
F
e
n
g
,
S
.
W
a
n
g
,
C
.
W
a
n
g
,
a
n
d
J.
Z
h
a
n
g
,
“
Le
a
r
n
i
n
g
d
e
e
p
h
i
e
r
a
r
c
h
i
c
a
l
s
p
a
t
i
a
l
–
s
p
e
c
t
r
a
l
f
e
a
t
u
r
e
s
f
o
r
h
y
p
e
r
sp
e
c
t
r
a
l
i
ma
g
e
c
l
a
ss
i
f
i
c
a
t
i
o
n
b
a
se
d
o
n
r
e
s
i
d
u
a
l
3
D
-
2
D
C
N
N
,
”
S
e
n
s
o
rs
,
v
o
l
.
1
9
,
n
o
.
2
3
,
2
0
1
9
,
d
o
i
:
1
0
.
3
3
9
0
/
s
1
9
2
3
5
2
7
6
.
B
I
O
G
RAP
H
I
E
S O
F
AUTH
O
RS
Dr
.
Do
u
g
l
a
s
O
m
we
n
g
a
Ny
a
b
u
g
a
is
a
lec
tu
re
r,
p
o
stg
ra
d
u
a
te
a
n
d
u
n
d
e
rg
ra
d
u
a
te
lec
tu
re
r
in
th
e
u
n
i
v
e
rsity
lev
e
l.
He
h
o
l
d
s
a
P
h
.
D
.
in
e
n
terp
r
ise
in
fo
rm
a
ti
o
n
s
y
ste
m
s
a
n
d
e
n
g
in
e
e
rin
g
fr
o
m
Do
n
g
h
u
a
Un
i
v
e
rsity
,
C
h
i
n
a
,
M
a
ste
rs
in
Co
m
p
u
ter
S
c
ie
n
c
e
fro
m
P
e
riy
a
r
Un
iv
e
rsity
,
I
n
d
ia.
P
u
b
li
s
h
e
d
m
o
re
th
a
n
1
5
a
c
a
d
e
m
ic
p
a
p
e
rs
in
a
c
a
d
e
m
ic
jo
u
rn
a
ls
a
n
d
c
o
n
fe
re
n
c
e
s.
He
h
a
s b
e
e
n
i
n
c
lu
d
e
d
i
n
m
a
n
y
p
a
p
e
rs
b
y
S
CI
a
n
d
EI.
His
m
a
in
re
se
a
rc
h
d
irec
ti
o
n
s
a
re
;
h
u
m
a
n
-
c
o
m
p
u
ter
in
tera
c
ti
o
n
(m
a
in
ly
e
d
u
c
a
ti
o
n
sy
ste
m
s),
ima
g
e
p
ro
c
e
ss
in
g
,
re
m
o
t
e
se
n
sin
g
,
p
a
tt
e
r
n
re
c
o
g
n
it
io
n
,
a
n
d
m
a
c
h
in
e
lea
rn
i
n
g
.
He
is
a
d
i
stin
g
u
ish
e
d
re
v
iew
e
r
a
t
IET
Im
a
g
e
P
ro
c
e
ss
in
g
Jo
u
rn
a
l
,
IAES
In
tern
a
ti
o
n
a
l
Jo
u
rn
a
l
o
f
Artifi
c
ial
In
telli
g
e
n
c
e
(IJ
-
AI),
a
n
d
a
p
ro
g
ra
m
m
e
c
o
m
m
it
tee
m
e
m
b
e
r
fo
r
th
e
in
ter
n
a
ti
o
n
a
l
c
o
n
fe
re
n
c
e
o
n
a
rti
ficia
l
n
e
u
ra
l
n
e
tw
o
r
k
s.
His
re
se
a
rc
h
in
tere
st
fo
c
u
se
s
o
n
u
ti
li
z
i
n
g
c
o
m
p
u
ti
n
g
a
n
d
re
m
o
te
se
n
sin
g
tec
h
n
i
q
u
e
s
t
o
s
o
lv
e
p
ro
b
lem
s
re
late
d
to
n
a
tu
ra
l
r
e
so
u
rc
e
m
a
n
a
g
e
m
e
n
t,
e
c
o
lo
g
y
,
n
a
tu
ra
l
d
isa
ste
r
m
a
p
p
i
n
g
,
v
e
g
e
tatio
n
p
ro
p
e
rty
e
x
trac
ti
o
n
,
a
n
d
u
rb
a
n
re
m
o
te
se
n
sin
g
.
He
c
a
n
b
e
c
o
n
tac
ted
a
t
e
m
a
il
:
d
n
y
a
b
u
g
a
@m
k
u
rwa
n
d
a
.
a
c
.
rw.
G
o
d
fr
e
y
Ny
a
r
ik
i
re
c
e
iv
e
d
t
h
e
M
.
S
c
.
c
o
m
p
u
ter
sc
ien
c
e
d
e
g
re
e
f
ro
m
th
e
Ala
g
a
p
p
a
Un
iv
e
risty
,
Ka
ra
ik
u
d
i,
I
n
d
ia,
wit
h
th
e
d
isse
rtatio
n
“
In
c
e
n
ti
v
ize
d
p
e
e
r
-
a
ss
isted
stre
a
m
in
g
fo
r
o
n
d
e
m
a
n
d
se
rv
ice
s”
.
He
is
a
lec
tu
r
e
r
o
f
c
o
m
p
u
ter
sc
ien
c
e
in
th
e
De
p
a
rtme
n
t
o
f
In
f
o
rm
a
ti
o
n
Tec
h
n
o
l
o
g
y
,
M
o
u
n
t
Ki
g
a
li
Un
iv
e
risty
,
Ki
g
a
li
,
Rwa
n
d
a
sin
c
e
2
0
1
7
.
His
re
se
a
rc
h
in
tere
sts
a
re
in
h
u
m
a
n
-
c
o
m
p
u
ter
i
n
tera
c
ti
o
n
,
c
o
m
p
u
ter
s
y
ste
m
s
&
n
e
two
r
k
s,
n
e
u
ra
l
n
e
two
r
k
s,
p
a
tt
e
rn
re
c
o
g
n
it
i
o
n
,
a
n
d
ima
g
e
p
r
o
c
e
ss
in
g
.
He
c
a
n
b
e
c
o
n
tac
ted
a
t
e
m
a
il
:
g
n
y
a
rik
i@m
k
u
.
a
c
.
k
e
.
Evaluation Warning : The document was created with Spire.PDF for Python.