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.
1
4
,
No
.
5
,
Octo
b
er
2
0
2
5
,
p
p
.
4
3
1
9
~
4
3
3
1
I
SS
N:
2
2
5
2
-
8
9
3
8
,
DOI
: 1
0
.
1
1
5
9
1
/ijai.v
14
.i
5
.
p
p
4
3
1
9
-
4
3
3
1
4319
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
a
i
.
ia
esco
r
e.
co
m
Build
ing
chang
e
detec
tion v
ia
clas
sifica
tion in
high
-
r
eso
lution
a
eria
l ima
g
ery
H
a
y
der
M
o
s
a
M
er
za
,
I
ha
b S
beit
y
,
M
o
ha
m
ed
Dbo
uk
,
Z
ein A
l
-
Abid
i
n Ib
ra
h
im
D
e
p
a
r
t
me
n
t
o
f
C
o
mp
u
t
e
r
S
c
i
e
n
c
e
,
F
a
c
u
l
t
y
o
f
S
c
i
e
n
c
e
s (I
)
,
Le
b
a
n
e
se
U
n
i
v
e
r
si
t
y
,
B
e
i
r
u
t
,
L
e
b
a
n
o
n
Art
icle
I
nfo
AB
S
T
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
Oct
2
2
,
2
0
2
4
R
ev
is
ed
J
u
l 1
4
,
2
0
2
5
Acc
ep
ted
Au
g
6
,
2
0
2
5
Th
is
re
se
a
rc
h
i
n
v
e
sti
g
a
tes
t
h
e
d
e
tec
ti
o
n
o
f
c
h
a
n
g
e
s
in
b
u
il
d
i
n
g
stru
c
tu
re
s
with
in
h
i
g
h
-
re
so
l
u
ti
o
n
a
e
rial
ima
g
e
s
o
f
Ba
g
h
d
a
d
,
Ira
q
,
o
v
e
r
two
y
e
a
rs,
2
0
0
7
a
n
d
2
0
2
4
.
Emp
l
o
y
i
n
g
a
d
v
a
n
c
e
d
r
e
m
o
te
se
n
sin
g
tec
h
n
i
q
u
e
s
a
n
d
s
o
p
h
isti
c
a
ted
ima
g
e
p
ro
c
e
ss
in
g
a
lg
o
rit
h
m
s,
t
h
is
stu
d
y
a
ims
to
id
e
n
ti
fy
a
n
d
q
u
a
n
ti
fy
a
lt
e
ra
ti
o
n
s
i
n
th
e
u
rb
a
n
lan
d
s
c
a
p
e
a
c
c
u
ra
tely
b
y
a
d
d
re
ss
in
g
th
e
k
e
y
c
h
a
ll
e
n
g
e
s
in
h
e
re
n
t
in
th
e
ima
g
e
re
g
istratio
n
p
r
o
c
e
ss
,
a
s
we
l
l
a
s
th
e
a
v
a
il
a
b
il
it
y
a
ss
o
c
iate
d
with
c
h
a
n
g
e
d
e
tec
ti
o
n
(CD)
tec
h
n
i
q
u
e
s.
We ex
a
m
in
e
d
th
e
d
a
ta
c
o
ll
e
c
ti
o
n
stra
teg
ies
,
e
v
a
lu
a
ted
m
a
tch
in
g
m
e
th
o
d
s,
a
n
d
c
o
m
p
a
re
d
CD
a
p
p
ro
a
c
h
e
s.
Ae
rial
ima
g
e
s
we
re
a
c
c
u
ra
tely
a
n
a
ly
z
e
d
to
d
e
tec
t
c
h
a
n
g
e
s
i
n
b
u
il
d
in
g
f
o
o
t
p
ri
n
ts,
c
o
n
stru
c
ti
o
n
a
c
ti
v
it
ies
,
a
n
d
d
e
stru
c
ti
o
n
.
We
d
e
v
e
lo
p
e
d
a
c
o
m
p
re
h
e
n
siv
e
a
n
n
o
tat
io
n
m
e
th
o
d
o
lo
g
y
tail
o
re
d
to
t
h
e
c
o
m
p
l
e
x
u
rb
a
n
e
n
v
iro
n
m
e
n
t
o
f
Ba
g
h
d
a
d
.
Th
e
se
fin
d
i
n
g
s
e
m
p
h
a
siz
e
th
e
ra
p
id
l
y
e
v
o
lv
i
n
g
n
a
tu
re
o
f
Ba
g
h
d
a
d
’s
u
rb
a
n
fa
b
ric
a
n
d
th
e
c
rit
ica
l
n
e
e
d
f
o
r
o
n
g
o
i
n
g
m
o
n
it
o
r
in
g
to
i
n
fo
rm
u
r
b
a
n
p
la
n
n
in
g
a
n
d
m
a
n
a
g
e
m
e
n
t
stra
teg
ies
.
Th
e
re
su
lt
s
d
e
m
o
n
stra
te
t
h
e
e
ffica
c
y
o
f
u
ti
l
izin
g
h
i
g
h
-
re
so
l
u
ti
o
n
a
e
rial
ima
g
e
ry
wit
h
o
b
jec
t
-
b
a
se
d
CD
tec
h
n
i
q
u
e
s
f
o
r
d
e
tailed
u
rb
a
n
a
n
a
ly
sis.
Th
is
re
se
a
rc
h
a
d
v
a
n
c
e
s
th
e
e
x
isti
n
g
k
n
o
wle
d
g
e
b
y
p
ro
v
id
i
n
g
a
ro
b
u
st
fra
m
e
wo
rk
fo
r
u
r
b
a
n
CD
,
with
imp
li
c
a
ti
o
n
s
fo
r
e
n
h
a
n
c
in
g
u
rb
a
n
p
lan
n
i
n
g
a
n
d
p
o
li
c
y
-
m
a
k
in
g
p
ro
c
e
ss
e
s.
F
u
tu
re
re
se
a
rc
h
will
f
o
c
u
s
o
n
re
fin
i
n
g
th
e
a
n
n
o
tatio
n
p
ro
c
e
ss
e
s
a
n
d
i
n
c
o
rp
o
ra
ti
n
g
a
d
d
it
io
n
a
l
d
a
ta
so
u
rc
e
s
to
e
n
h
a
n
c
e
th
e
a
c
c
u
ra
c
y
a
n
d
c
o
m
p
re
h
e
n
siv
e
n
e
ss
o
f
u
r
b
a
n
CD
m
e
th
o
d
o
lo
g
ies
.
K
ey
w
o
r
d
s
:
Aer
ial
h
ig
h
-
r
eso
lu
tio
n
im
ag
es
C
h
an
g
e
d
etec
tio
n
C
las
s
if
icatio
n
alg
o
r
ith
m
s
L
an
d
u
s
e/lan
d
c
o
v
er
R
em
o
te
s
en
s
in
g
im
ag
es
Sift
m
atch
in
g
alg
o
r
ith
m
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
:
Hay
d
er
Mo
s
a
Me
r
za
Dep
ar
tm
en
t o
f
C
o
m
p
u
ter
Scie
n
ce
,
Facu
lty
o
f
Scien
ce
s
(
I
)
,
L
eb
an
ese
Un
iv
er
s
ity
B
eir
u
t,
L
eb
an
o
n
E
m
ail:
h
ay
d
er
.
m
er
za
@
u
l.e
d
u
.
l
b
1.
I
NT
RO
D
UCT
I
O
N
C
h
an
g
e
d
etec
tio
n
(
CD
)
in
ae
r
ial
im
ag
es
p
lay
s
a
p
iv
o
tal
r
o
l
e
in
an
aly
zin
g
an
d
m
o
n
ito
r
i
n
g
d
y
n
a
m
ic
en
v
ir
o
n
m
en
ts
.
Aer
ial
im
ag
es
ca
p
tu
r
ed
f
r
o
m
v
a
r
io
u
s
p
latf
o
r
m
s
,
s
u
ch
as
s
atell
ites
,
d
r
o
n
es,
o
r
air
cr
af
t,
p
r
o
v
id
e
a
v
alu
ab
le
s
o
u
r
ce
o
f
in
f
o
r
m
atio
n
f
o
r
d
etec
tin
g
an
d
u
n
d
er
s
tan
d
in
g
ch
a
n
g
es
o
cc
u
r
r
i
n
g
o
n
th
e
E
ar
th
’
s
s
u
r
f
ac
e
.
CD
in
v
o
lv
es
th
e
id
e
n
tific
atio
n
a
n
d
c
h
ar
ac
ter
izatio
n
o
f
alter
at
io
n
s
b
etwe
en
two
o
r
m
o
r
e
i
m
ag
es
ac
q
u
ir
e
d
at
d
if
f
er
en
t
p
er
io
d
s
.
T
h
ese
ch
a
n
g
es
ca
n
e
n
co
m
p
ass
a
wid
e
r
an
g
e
o
f
p
h
e
n
o
m
en
a
,
in
c
lu
d
in
g
la
n
d
c
o
v
e
r
tr
an
s
f
o
r
m
atio
n
s
,
u
r
b
an
g
r
o
wt
h
,
n
at
u
r
al
d
is
aster
s
,
v
eg
etatio
n
d
y
n
am
ics,
an
d
i
n
f
r
astru
ctu
r
e
d
ev
elo
p
m
en
t.
T
h
e
ab
ilit
y
to
d
etec
t
an
d
q
u
an
tif
y
ch
an
g
es
in
ae
r
ial
im
ag
es
h
as
s
ig
n
if
ican
t
im
p
licatio
n
s
in
m
u
ltip
le
f
ield
s
,
in
clu
d
in
g
en
v
ir
o
n
m
e
n
tal
m
o
n
ito
r
in
g
,
s
p
atial
p
la
n
n
in
g
,
em
er
g
en
cy
r
esp
o
n
s
e
m
an
ag
em
e
n
t,
ag
r
icu
lt
u
r
e,
a
n
d
r
eso
u
r
ce
m
a
n
ag
em
en
t.
I
t
en
ab
les
d
ec
is
io
n
-
m
ak
er
s
,
r
esear
ch
er
s
,
an
d
p
o
licy
m
ak
er
s
to
g
ain
in
s
ig
h
ts
in
t
o
tem
p
o
r
al
a
n
d
s
p
atial
p
atter
n
s
o
f
ch
a
n
g
e,
ass
ess
th
e
im
p
ac
ts
o
f
h
u
m
an
ac
tiv
ities
,
an
d
m
ak
e
in
f
o
r
m
ed
d
ec
is
io
n
s
f
o
r
s
u
s
tain
ab
le
d
e
v
elo
p
m
e
n
t.
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.
1
4
,
No
.
5
,
Octo
b
er
2
0
2
5
:
4
3
1
9
-
4
3
3
1
4320
−
W
h
en
g
eo
g
r
ap
h
ic
m
etad
ata
is
u
n
a
v
ailab
le
o
r
u
n
r
eliab
l
e,
wh
at
a
r
e
th
e
in
h
e
r
en
t
c
h
allen
g
es
an
d
p
er
f
o
r
m
an
ce
lim
itatio
n
s
o
f
co
n
v
en
tio
n
al
r
eg
is
tr
atio
n
tech
n
iq
u
es
(
s
u
ch
as
s
ca
le
in
v
ar
ian
t
f
ea
tu
r
e
tr
an
s
f
o
r
m
(
SIFT
)
c
o
m
b
in
e
d
w
ith
B
r
u
te
Fo
r
ce
Ma
tch
er
)
?
−
I
n
s
ce
n
ar
io
s
lack
in
g
s
p
atial
r
ef
er
en
cin
g
in
f
o
r
m
atio
n
,
wh
at
s
tr
ateg
ies
ca
n
b
e
em
p
lo
y
ed
t
o
im
p
r
o
v
e
t
h
e
r
o
b
u
s
tn
ess
an
d
ac
cu
r
ac
y
o
f
k
e
y
p
o
in
ts
d
etec
tio
n
an
d
m
atch
in
g
?
−
Ho
w
d
o
v
ar
io
u
s
CD
m
eth
o
d
o
lo
g
ies,
s
u
ch
as
alg
eb
r
aic
,
clas
s
if
icatio
n
-
b
ased
,
an
d
tr
an
s
f
o
r
m
atio
n
-
b
ased
ap
p
r
o
ac
h
es,
co
m
p
a
r
e
in
ter
m
s
o
f
th
eir
ac
c
u
r
ac
y
,
r
eliab
ilit
y
,
a
n
d
co
m
p
u
tatio
n
al
ef
f
icien
cy
?
−
W
h
en
ap
p
lied
to
r
ea
l
-
wo
r
l
d
a
er
ial
d
atasets
(
e.
g
.
,
im
ag
er
y
o
f
B
ag
h
d
ad
)
,
wh
ic
h
o
f
th
ese
CD
tech
n
iq
u
es
d
em
o
n
s
tr
ates th
e
h
ig
h
est lev
el
o
f
p
er
f
o
r
m
an
ce
an
d
p
r
ac
tical
ap
p
licab
ilit
y
?
CD
tech
n
iq
u
es
p
r
o
v
id
e
r
esear
ch
er
s
an
d
a
n
aly
s
ts
with
ess
en
tial
in
s
ig
h
ts
in
to
th
e
d
y
n
am
ic
ch
ar
ac
ter
is
tics
o
f
th
e
ea
r
th
’
s
s
u
r
f
ac
e.
I
n
th
e
f
ield
o
f
CD
,
n
u
m
er
o
u
s
tech
n
iq
u
es
ar
e
av
ailab
le;
h
o
wev
er
,
id
en
tify
in
g
an
o
p
tim
al
an
d
d
e
f
in
itiv
e
m
eth
o
d
r
em
ain
s
a
ch
all
en
g
e.
Du
e
to
th
e
co
m
p
lex
ity
o
f
CD
,
d
ata
an
aly
s
t
s
em
p
lo
y
a
v
ar
iety
o
f
m
eth
o
d
s
,
u
tili
zin
g
th
eir
ex
p
er
tis
e
to
ef
f
ec
tiv
ely
d
etec
t
ch
an
g
es.
No
n
eth
eless
,
th
e
p
r
o
ce
s
s
in
g
o
f
h
ete
r
o
g
e
n
eo
u
s
d
ata
is
wid
ely
r
ec
o
g
n
ized
as
o
n
e
o
f
th
e
m
o
s
t
s
ig
n
if
ican
t
ch
allen
g
es
in
CD
.
Dete
ctio
n
o
f
ch
an
g
es
in
r
ea
l
-
tim
e
ap
p
licatio
n
s
p
o
s
es
s
ig
n
if
ican
t
ch
allen
g
es
d
u
e
to
th
e
n
ee
d
f
o
r
m
u
ltip
le
p
r
o
ce
s
s
in
g
s
tep
s
.
T
h
ese
s
tep
s
in
clu
d
e
id
en
tify
in
g
is
s
u
es
r
elate
d
to
CD
,
p
r
ep
r
o
ce
s
s
in
g
th
e
im
ag
es,
an
d
ass
es
s
in
g
th
e
p
er
f
o
r
m
an
ce
o
f
t
h
e
ap
p
licatio
n
-
s
p
ec
if
ic
alg
o
r
it
h
m
.
T
h
e
ch
allen
g
e
is
f
u
r
th
er
c
o
m
p
o
u
n
d
ed
b
y
th
e
n
ee
d
to
d
ev
elo
p
an
ap
p
r
o
p
r
iat
e
m
eth
o
d
o
lo
g
y
f
o
r
d
etec
tin
g
c
h
an
g
es in
v
e
r
y
h
i
g
h
-
r
eso
lu
tio
n
im
ag
er
y
.
Ad
v
an
ce
m
en
ts
in
r
em
o
te
s
en
s
in
g
im
ag
in
g
tech
n
o
lo
g
ies
an
d
th
e
g
r
o
win
g
av
ailab
ilit
y
o
f
h
ig
h
-
r
eso
lu
tio
n
s
atellite
d
ata
h
av
e
g
r
ea
tly
f
ac
ilit
ated
im
ag
e
a
n
al
y
s
is
tech
n
iq
u
es
tailo
r
ed
f
o
r
r
e
m
o
te
s
en
s
in
g
-
b
ased
ap
p
licatio
n
s
,
wh
ich
wer
e
p
r
ev
io
u
s
ly
r
elian
t
o
n
lab
o
r
io
u
s
f
ield
s
u
r
v
e
y
s
.
C
o
n
d
u
ctin
g
an
in
d
i
v
id
u
al
ar
ea
s
u
r
v
e
y
was
tim
e
-
co
n
s
u
m
in
g
an
d
ar
d
u
o
u
s
,
b
u
t
t
o
d
ay
,
th
e
u
tili
za
tio
n
d
er
iv
e
d
f
r
o
m
s
atellite
o
b
s
er
v
atio
n
s
co
llected
in
r
ea
l
tim
e
h
as
s
im
p
lifie
d
th
is
p
r
o
ce
s
s
.
R
ea
l
-
tim
e
s
atellite
d
at
a
p
lay
s
a
cr
u
cial
r
o
le
in
v
ar
io
u
s
ap
p
licatio
n
s
,
with
r
em
o
te
s
en
s
in
g
b
ein
g
a
p
r
i
m
ar
y
b
e
n
ef
iciar
y
.
I
t
en
ab
les
ac
cu
r
ate
d
etec
tio
n
o
f
en
v
ir
o
n
m
en
tal
ch
a
n
g
es,
co
n
tr
ib
u
tin
g
to
a
b
etter
u
n
d
er
s
tan
d
in
g
o
f
h
u
m
an
-
n
atu
r
e
in
ter
ac
tio
n
s
.
T
h
is
u
n
d
er
s
tan
d
in
g
,
in
tu
r
n
,
ai
d
s
d
ec
is
io
n
-
m
ak
in
g
,
p
ar
ticu
la
r
ly
in
th
e
co
n
te
x
t o
f
u
r
b
a
n
d
ev
el
o
p
m
en
t.
2.
RE
L
AT
E
D
WO
RK
S
I
n
th
e
f
ield
o
f
ag
r
ic
u
ltu
r
al
s
tu
d
ies,
CD
tech
n
iq
u
es
ar
e
e
m
p
lo
y
ed
f
o
r
m
o
n
ito
r
in
g
d
ef
o
r
estatio
n
,
ev
alu
atin
g
th
e
im
p
ac
ts
o
f
n
at
u
r
al
d
is
aster
s
,
an
d
an
al
y
zin
g
p
atter
n
s
o
f
s
h
if
ti
n
g
cu
ltiv
ati
o
n
.
I
n
th
e
m
ilit
ar
y
d
o
m
ain
,
th
ese
m
eth
o
d
s
p
lay
a
cr
itical
r
o
le
in
in
tellig
en
ce
g
ath
er
in
g
.
T
h
is
in
clu
d
es
th
e
id
e
n
tific
atio
n
o
f
n
ewly
estab
lis
h
ed
m
ilit
ar
y
in
s
talla
ti
o
n
s
,
th
e
m
o
n
ito
r
in
g
o
f
en
em
y
tr
o
o
p
m
o
v
em
en
ts
,
b
attlef
ie
ld
ass
es
s
m
en
t,
an
d
d
am
ag
e
ev
al
u
atio
n
[
1
]
.
I
n
th
e
civ
il
co
n
tex
t,
it
f
u
n
ctio
n
s
as
a
r
eg
u
lato
r
y
m
ec
h
an
is
m
f
o
r
m
a
n
ag
in
g
u
r
b
a
n
d
e
v
el
o
p
m
en
t
an
d
g
u
id
in
g
th
e
s
p
atial
ex
p
an
s
io
n
o
f
cities
[
2
]
.
Alth
o
u
g
h
CD
alg
o
r
ith
m
s
p
r
o
v
i
d
e
s
u
b
s
tan
tial b
en
ef
its
ac
r
o
s
s
a
wid
e
r
an
g
e
o
f
ap
p
licatio
n
s
,
th
ey
ar
e
also
ass
o
ciate
d
with
n
o
tab
le
ch
allen
g
es.
Fo
r
ex
am
p
le
,
v
e
g
etatio
n
g
r
o
wth
an
d
ch
an
g
es
in
s
u
r
f
ac
e
r
ef
lecta
n
c
e
ch
ar
ac
ter
is
tics
,
s
u
ch
as
th
o
s
e
ca
u
s
ed
b
y
s
o
il
co
n
d
itio
n
s
b
ef
o
r
e
an
d
af
ter
r
ain
f
all,
ca
n
s
ig
n
if
ica
n
tly
im
p
ac
t
th
e
r
eliab
ilit
y
an
d
v
alid
i
ty
o
f
th
e
d
etec
ted
c
h
an
g
es
[
3
]
.
Acc
u
r
ate
im
a
g
e
r
eg
is
tr
atio
n
is
cr
u
cial
f
o
r
r
eliab
le
CD
,
p
ar
ticu
lar
ly
wh
en
a
n
aly
zin
g
m
u
ltit
em
p
o
r
al
o
r
m
u
lti
-
s
o
u
r
ce
im
ag
er
y
.
T
h
e
p
r
o
ce
s
s
in
v
o
l
v
es
ex
tr
ac
ti
n
g
k
ey
f
ea
tu
r
es
a
n
d
esti
m
atin
g
a
s
p
atial
tr
an
s
f
o
r
m
atio
n
t
o
alig
n
th
e
m
o
v
i
n
g
im
ag
e
with
th
e
f
ix
e
d
im
ag
e,
e
n
ab
lin
g
c
o
n
s
is
ten
t c
o
m
p
ar
is
o
n
ac
r
o
s
s
tim
e
an
d
s
p
ac
e
[
4
]
.
Featu
r
e
-
b
ased
im
ag
e
r
eg
is
tr
at
io
n
m
atch
es
d
is
tin
ct
f
ea
tu
r
es,
wh
ile
h
y
b
r
id
m
eth
o
d
s
co
m
b
i
n
e
r
eg
i
o
n
-
b
ased
an
d
attr
ib
u
te
-
o
r
ien
ted
m
eth
o
d
s
.
T
r
ad
itio
n
al
tech
n
iq
u
es
o
p
er
ate
in
th
e
s
p
atial
d
o
m
ain
an
d
ar
e
class
if
ied
as
m
an
u
al,
s
em
i
-
a
u
to
m
atic,
o
r
au
to
m
atic
[
5
]
.
Ad
v
a
n
ce
m
en
t
s
in
r
em
o
te
s
en
s
in
g
tec
h
n
o
lo
g
ies,
s
u
ch
as
h
ig
h
er
s
p
atial
an
d
s
p
ec
tr
al
lev
els
o
f
d
etail,
im
p
r
o
v
ed
im
a
g
e
ac
q
u
is
itio
n
tech
n
iq
u
es,
an
d
ad
v
an
c
ed
d
ata
p
r
o
ce
s
s
in
g
alg
o
r
ith
m
s
,
h
av
e
r
e
v
o
lu
tio
n
iz
ed
th
e
f
ield
o
f
CD
in
ae
r
ial
i
m
ag
er
y
.
T
h
ese
ad
v
an
ce
m
en
ts
h
av
e
en
h
an
ce
d
o
u
r
ab
ilit
y
to
ex
tr
ac
t
v
alu
ab
le
in
f
o
r
m
atio
n
f
r
o
m
im
ag
es
an
d
d
ete
ct
s
u
b
tle
o
r
c
o
m
p
lex
ch
a
n
g
es
t
h
at
m
ay
h
a
v
e
b
ee
n
ch
allen
g
in
g
to
id
e
n
tify
u
s
in
g
t
r
ad
itio
n
al
s
u
r
v
ey
m
et
h
o
d
s
alo
n
e.
I
n
th
is
co
n
tex
t,
th
is
ar
ticle
aim
s
to
ex
p
lo
r
e
th
e
f
u
n
d
am
e
n
tal
co
n
ce
p
t,
m
eth
o
d
o
lo
g
ies,
an
d
a
p
p
licatio
n
s
o
f
CD
in
ae
r
ial
im
ag
es.
Ma
ch
in
e
lear
n
i
n
g
al
g
o
r
ith
m
s
ar
e
co
m
m
o
n
ly
class
if
ied
in
to
s
u
p
er
v
is
ed
,
u
n
s
u
p
er
v
is
ed
,
an
d
r
ein
f
o
r
ce
m
e
n
t
lear
n
in
g
.
Su
p
e
r
v
is
ed
class
if
ier
s
p
er
f
o
r
m
b
e
s
t
with
wid
e
lab
ele
d
d
ata
a
n
d
ca
n
b
e
f
u
r
th
e
r
ca
teg
o
r
ized
as
p
ar
am
etr
ic
o
r
n
o
n
p
a
r
am
etr
ic
b
ased
o
n
d
ata
d
is
tr
ib
u
tio
n
ass
u
m
p
tio
n
s
[
6
]
.
E
x
ec
u
tio
n
tim
e
is
a
k
ey
f
ac
to
r
in
im
p
lem
en
tin
g
a
n
d
o
p
er
atio
n
alizin
g
m
ac
h
in
e
l
ea
r
n
in
g
m
o
d
els,
r
ef
e
r
r
in
g
to
t
h
e
d
u
r
atio
n
n
ee
d
ed
f
o
r
a
s
in
g
le
in
f
er
en
ce
.
Op
tim
i
za
tio
n
tech
n
iq
u
es
aim
to
r
ed
u
ce
th
is
tim
e
wh
ile
p
r
eser
v
in
g
ac
cu
r
ac
y
,
en
h
an
cin
g
th
e
m
o
d
el’
s
p
r
ac
tical
u
tili
ty
.
I
n
th
is
co
n
te
x
t,
r
a
n
d
o
m
f
o
r
est
(
R
F)
is
wid
ely
u
s
ed
b
o
th
f
o
r
b
u
ild
in
g
ac
cu
r
ate
p
r
ed
ictiv
e
m
o
d
els
an
d
f
o
r
ev
alu
atin
g
th
e
r
elativ
e
im
p
o
r
ta
n
ce
o
f
in
p
u
t
v
ar
iab
les
[
7
]
.
CD
in
ea
r
th
s
u
r
f
ac
e
im
ag
er
y
h
as lo
n
g
b
ee
n
r
ec
o
g
n
i
ze
d
as a
f
u
n
d
a
m
en
tal
ch
allen
g
e
in
th
e
f
ield
o
f
r
em
o
te
s
en
s
in
g
[
8
]
.
E
f
f
ec
tiv
e
CD
in
r
em
o
te
s
en
s
in
g
r
eq
u
ir
es c
le
ar
ly
d
ef
in
e
d
r
esear
ch
o
b
jectiv
es a
n
d
a
well
-
s
p
ec
if
ied
s
tu
d
y
a
r
ea
.
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
B
u
ild
in
g
ch
a
n
g
e
d
etec
tio
n
vi
a
cla
s
s
ifica
tio
n
in
h
ig
h
-
r
eso
lu
tio
n
a
eria
l ima
g
ery
(
Ha
yd
er Mo
s
a
Mer
z
a
)
4321
CD
en
tails
th
e
id
en
tific
atio
n
o
f
alter
atio
n
s
i
n
s
u
r
f
ac
e
f
e
atu
r
es
o
v
e
r
tim
e
u
s
in
g
m
u
lt
i
-
tem
p
o
r
al
im
ag
er
y
,
o
f
f
er
in
g
v
alu
a
b
le
in
s
ig
h
ts
in
to
th
e
tem
p
o
r
al
d
y
n
a
m
ics
o
f
n
atu
r
al
an
d
an
th
r
o
p
o
g
en
ic
p
r
o
ce
s
s
es
[
9
]
.
R
eliab
le
CD
r
eq
u
ir
es
im
ag
e
p
air
s
to
b
e
s
p
ec
tr
ally
,
s
p
atially
,
an
d
tem
p
o
r
ally
alig
n
ed
.
T
h
is
is
ac
h
iev
ed
t
h
r
o
u
g
h
p
r
ep
r
o
ce
s
s
in
g
s
tep
s
s
u
ch
as
co
-
r
eg
is
tr
atio
n
o
f
m
u
lti
-
tem
p
o
r
al
im
ag
es
o
v
er
th
e
s
am
e
lo
ca
tio
n
[
1
0
]
.
Kar
k
e
r
[
1
1
]
d
escr
ib
es
im
ag
e
alig
n
m
en
t
as
a
p
r
o
ce
s
s
th
at
u
s
e
s
im
ag
e
tie
p
o
in
ts
(
I
T
Ps
)
to
c
o
m
p
u
te
g
eo
m
etr
ic
tr
an
s
f
o
r
m
atio
n
s
,
en
ab
lin
g
o
n
e
im
ag
e
to
b
e
s
p
atially
alig
n
e
d
with
an
o
t
h
er
.
B
y
u
n
d
er
s
tan
d
i
n
g
an
d
h
ar
n
ess
in
g
th
e
p
o
wer
o
f
CD
in
ae
r
ial
i
m
ag
es,
we
ca
n
g
ain
d
ee
p
er
in
s
ig
h
ts
in
to
th
e
ev
o
lv
i
n
g
e
ar
th
’
s
s
u
r
f
ac
e
an
d
co
n
tr
ib
u
te
to
m
o
r
e
in
f
o
r
m
ed
d
ec
is
io
n
-
m
ak
in
g
a
n
d
s
u
s
tain
ab
l
e
d
ev
elo
p
m
en
t p
r
ac
tices.
C
h
en
et
a
l.
[
1
2
]
i
n
tr
o
d
u
ce
s
an
s
y
n
th
etic
ap
er
tu
r
e
r
ad
ar
(
SAR
)
r
em
o
te
s
en
s
in
g
alg
o
r
ith
m
f
o
r
d
etec
tin
g
ch
an
g
es
in
im
ag
e
r
y
th
at
u
s
es
ad
ap
tiv
e
tech
n
i
q
u
es
f
o
r
r
ea
l
-
ti
m
e
p
ar
am
eter
esti
m
atio
n
a
n
d
a
s
p
ar
s
e
au
to
m
atic
en
co
d
er
t
o
d
etec
t
s
ig
n
if
ican
t
r
eg
io
n
s
an
d
r
ed
u
ce
s
p
ec
k
le
n
o
is
e.
Prin
cip
al
co
m
p
o
n
en
t
a
n
aly
s
is
an
d
K
-
m
ea
n
s
clu
s
ter
in
g
en
h
an
ce
CD
b
y
m
in
im
izin
g
is
o
lated
p
ix
el
im
p
ac
t.
E
x
p
er
im
e
n
tal
r
esu
lts
s
h
o
w
h
ig
h
d
etec
tio
n
ac
cu
r
ac
y
,
ef
f
ec
tiv
ely
h
a
n
d
lin
g
en
v
ir
o
n
m
en
tal
in
ter
f
er
en
ce
s
lik
e
s
ea
wate
r
f
lu
ct
u
atio
n
s
an
d
s
h
ip
p
r
esen
ce
.
Acc
o
r
d
in
g
to
B
ao
et
a
l.
[
1
3
]
,
p
atch
an
d
p
i
x
el
ch
an
g
e
n
etw
o
r
k
(
PP
C
NE
T
)
f
o
r
d
etec
tin
g
i
n
b
item
p
o
r
al
h
ig
h
-
r
eso
lu
tio
n
im
ag
es.
T
h
is
d
ee
p
lear
n
in
g
ap
p
r
o
ac
h
in
teg
r
ates
p
atch
an
d
p
ix
el
-
lev
el
CD
to
ac
h
iev
e
p
r
ec
is
e
b
o
u
n
d
ar
ies
o
f
ch
a
n
g
e
ar
ea
s
wh
ile
also
s
u
r
p
ass
in
g
th
e
s
p
ee
d
o
f
p
i
x
el
-
lev
el
-
b
ased
d
ee
p
lear
n
in
g
m
eth
o
d
s
.
E
x
ten
s
iv
e
ex
p
er
im
en
ts
co
m
p
ar
in
g
PP
C
NE
T
,
tr
ad
itio
n
al
m
eth
o
d
s
,
an
d
o
th
er
d
ee
p
n
etw
o
r
k
s
u
s
in
g
s
atellites
an
d
ae
r
ial
im
a
g
es
d
em
o
n
s
tr
ated
th
e
ef
f
ec
tiv
en
ess
an
d
f
ea
s
ib
ilit
y
o
f
d
etec
tin
g
ch
an
g
es
in
h
ig
h
-
r
eso
lu
tio
n
r
em
o
te
s
en
s
in
g
im
ag
es.
Pen
g
et
a
l.
[
1
4
]
claim
ed
a
n
e
wly
p
r
o
p
o
s
ed
e
n
d
-
to
-
en
d
f
r
am
ewo
r
k
f
o
r
CD
m
eth
o
d
u
s
in
g
t
h
e
UNe
t++
ar
ch
itectu
r
e,
wh
ich
lear
n
s
ch
a
n
g
e
m
ap
s
d
ir
ec
tly
f
r
o
m
an
n
o
t
ated
d
atasets
.
T
h
is
ap
p
r
o
ac
h
ad
d
r
ess
es
lim
itatio
n
s
o
f
ex
is
tin
g
CD
m
eth
o
d
s
b
y
r
ed
u
cin
g
er
r
o
r
ac
cu
m
u
latio
n
a
n
d
in
ter
m
ed
iate
p
r
o
ce
s
s
in
g
s
tep
s
.
I
t
o
u
tp
er
f
o
r
m
s
o
th
er
m
eth
o
d
s
in
v
is
u
al
an
d
q
u
an
titativ
e
ev
alu
atio
n
s
b
u
t
r
el
ies
o
n
a
s
u
b
s
tan
tial
n
u
m
b
er
o
f
tr
u
e
ch
an
g
e
m
a
p
s
,
p
o
ten
tially
lim
itin
g
its
b
r
o
ad
er
ap
p
licatio
n
.
Hu
an
g
et
a
l.
[
1
5
]
d
ev
elo
p
ed
a
n
au
to
m
atic
CD
m
eth
o
d
u
s
in
g
p
lan
ar
-
v
e
r
tical
f
ea
tu
r
es,
o
b
je
ct
-
b
ased
tem
p
o
r
al
c
o
r
r
ec
tio
n
,
an
d
a
m
u
lti
-
tem
p
o
r
al
CD
m
o
d
el
to
id
en
tif
y
non
-
c
o
v
er
e
d
b
u
ild
in
g
ar
ea
s
(
NC
B
As).
T
esti
n
g
in
B
e
ijin
g
an
d
Sh
an
g
h
ai
s
h
o
wed
s
atis
f
ac
to
r
y
r
esu
lts
,
b
u
t
lim
itatio
n
s
in
clu
d
ed
er
r
o
r
s
in
d
etec
tin
g
r
eb
u
ilt
ar
ea
s
an
d
co
n
s
tr
ain
ts
f
r
o
m
th
e
Z
Y
-
3
s
atellite,
wh
ich
lim
ited
tim
e
-
s
er
ies im
ag
e
av
ailab
ilit
y
.
Vian
a
et
a
l.
[
1
6
]
ex
am
in
e
d
la
n
d
u
s
e
an
d
lan
d
c
o
v
er
(
L
UL
C
)
ch
an
g
es
in
a
r
u
r
al
r
eg
io
n
o
v
er
2
1
y
ea
r
s
(
1
9
9
5
-
2
0
1
5
)
u
s
in
g
L
an
d
s
at
i
m
ag
er
y
.
T
h
ey
u
s
ed
o
p
e
n
-
s
o
u
r
ce
L
UL
C
d
ata
an
d
K
-
m
ea
n
s
clu
s
ter
in
g
to
r
ef
in
e
s
p
ec
tr
al
s
ig
n
atu
r
es
f
o
r
ea
ch
cl
ass
.
B
y
in
teg
r
atin
g
d
ata
f
r
o
m
th
e
o
f
f
icial
Po
r
tu
g
u
ese
L
UL
C
m
ap
,
C
ar
ta
d
e
Uso
e
Ocu
p
aç
ã
o
d
o
s
o
l
o
(
C
OS)
,
f
o
r
1
9
9
5
,
2
0
0
7
,
2
0
1
0
,
an
d
2
0
1
5
,
th
e
y
g
en
er
ated
r
e
p
r
esen
tativ
e
tr
ain
in
g
s
am
p
les.
T
h
e
m
eth
o
d
ac
h
iev
e
d
an
o
v
e
r
all
ac
cu
r
ac
y
o
f
7
6
%,
d
em
o
n
s
tr
atin
g
its
ef
f
ec
tiv
en
ess
an
d
p
r
o
v
id
i
n
g
v
alu
ab
le
in
s
ig
h
ts
in
to
s
ig
n
if
ican
t
L
UL
C
ch
an
g
es
d
u
r
in
g
th
e
p
er
i
o
d
.
Viv
ek
an
a
n
d
a
et
a
l.
[
1
7
]
f
o
c
u
s
ed
o
n
class
if
y
in
g
L
UL
C
ch
an
g
es
b
etwe
en
th
e
y
ea
r
s
1
9
9
9
an
d
2
0
1
9
.
R
esear
ch
er
s
em
p
lo
y
e
d
a
co
m
b
in
atio
n
o
f
I
n
d
ia
’
s
to
p
o
g
r
a
p
h
ic
m
a
p
s
u
r
v
e
y
an
d
tem
p
o
r
al
s
atellite
im
ag
er
y
to
g
ath
er
d
ata.
R
em
o
te
s
en
s
in
g
an
d
g
eo
g
r
a
p
h
ic
in
f
o
r
m
atio
n
s
y
s
tem
(
GI
S)
tech
n
iq
u
es
wer
e
in
teg
r
ated
t
o
q
u
an
tify
an
d
co
m
p
r
e
h
en
d
th
e
L
UL
C
c
h
an
g
es
in
An
an
th
a
r
am
an
s
p
a
n
n
in
g
a
p
e
r
io
d
o
f
4
0
y
ea
r
s
,
f
r
o
m
1
9
7
8
to
2
0
1
8
.
T
h
e
co
n
f
u
s
io
n
m
atr
ix
was
em
p
lo
y
ed
to
ass
ess
th
e
c
lass
if
icatio
n
ac
cu
r
ac
y
,
wh
ich
was
s
ati
s
f
ac
to
r
y
.
T
ewa
b
e
an
d
Fen
tah
u
n
[
1
8
]
an
aly
ze
d
L
UL
C
ch
an
g
es
in
th
e
T
an
a
b
asin
u
s
in
g
L
an
d
s
at
th
em
atic
m
ap
p
er
(
TM
)
im
ag
es
f
r
o
m
1
9
8
6
,
2
0
0
2
,
an
d
2
0
1
8
.
T
h
ey
class
if
ied
s
ix
lan
d
co
v
e
r
ty
p
es
an
d
ass
ess
ed
ac
cu
r
ac
y
u
s
in
g
t
h
e
Kap
p
a
co
e
f
f
icien
t.
T
h
e
f
in
d
in
g
s
r
ev
ea
led
o
v
e
r
all
ac
cu
r
ac
ie
s
o
f
8
4
.
2
1
%,
8
3
.
3
2
%,
an
d
9
1
.
4
0
%
f
o
r
th
e
y
ea
r
s
1
9
8
6
,
2
0
0
2
,
an
d
2
0
1
8
,
r
esp
ec
t
iv
ely
,
with
in
th
e
b
asin
.
T
h
e
c
o
r
r
esp
o
n
d
in
g
k
a
p
p
a
co
e
f
f
icien
ts
wer
e
d
eter
m
in
ed
as 7
9
.
0
2
%,
8
3
.
3
2
%,
an
d
8
9
.
6
6
%.
3.
M
AT
E
R
I
AL
S AN
D
M
E
T
H
O
D
T
h
e
r
esear
ch
em
p
lo
y
ed
a
q
u
an
titativ
e
ap
p
r
o
ac
h
k
n
o
wn
as
CD
.
T
h
is
m
eth
o
d
in
v
o
lv
ed
class
if
y
in
g
ea
ch
s
atellite
im
ag
e
an
d
c
o
m
p
ar
in
g
it
b
ased
o
n
a
p
i
x
el
-
by
-
p
ix
el
m
eth
o
d
with
t
h
e
r
esu
lt
in
g
L
UL
C
m
ap
s
in
2
0
2
4
th
at
wer
e
p
r
o
v
id
e
d
b
y
th
e
r
e
m
o
te
s
en
s
in
g
d
e
p
ar
t
m
en
t
o
f
th
e
Un
iv
er
s
ity
o
f
B
ag
h
d
ad
.
O
u
r
f
ir
s
t
m
eth
o
d
o
l
o
g
y
in
em
p
lo
y
e
d
in
th
e
s
tu
d
y
co
n
s
is
ted
o
f
th
e
s
tep
s
as
illu
s
tr
ated
in
Fig
u
r
e
1
:
i
)
c
o
llectin
g
th
e
d
ata,
ii
)
im
ag
e
p
r
e
-
p
r
o
ce
s
s
in
g
,
iii
)
im
ag
e
r
eg
is
tr
atio
n
(
I
R
)
,
iv
)
d
ata
class
if
icatio
n
,
v
)
o
u
tlie
r
r
em
o
v
i
n
g
m
o
d
el,
vi
)
tim
e
p
er
f
o
r
m
a
n
ce
m
o
d
el
a
s
in
o
u
r
cited
[
1
9
]
,
an
d
f
in
ally
,
th
e
v
ii)
CD
s
tep
will
b
e
p
r
es
en
ted
in
th
is
s
tu
d
y
.
T
h
e
s
ec
o
n
d
m
eth
o
d
o
lo
g
y
is
p
r
esen
ted
in
s
ec
tio
n
4
.
1
.
3
.
1
.
Cha
ng
e
det
ec
t
i
o
n m
et
h
o
ds
B
ased
o
n
th
e
CD
p
r
o
ce
s
s
in
g
as
ca
n
b
e
s
ee
n
in
Fig
u
r
e
2
,
th
er
e
ar
e
a
m
u
lti
v
ar
io
u
s
ap
p
r
o
ac
h
es
th
at
h
av
e
b
ee
n
d
ev
elo
p
ed
t
o
b
e
u
s
ed
f
o
r
,
an
d
ea
ch
tailo
r
ed
to
its
s
p
ec
if
ic
ap
p
licatio
n
.
I
n
ca
s
es
wh
er
e
th
e
CD
im
ag
es
ar
e
ac
q
u
ir
ed
th
r
o
u
g
h
v
ar
io
u
s
s
en
s
o
r
m
o
d
alities
,
it
b
ec
o
m
es
cr
u
cial
to
p
er
f
o
r
m
im
ag
e
r
eg
is
tr
atio
n
b
ef
o
r
e
a
p
p
ly
in
g
th
e
alg
o
r
it
h
m
d
esig
n
e
d
f
o
r
CD
.
I
m
a
g
e
r
eg
is
tr
atio
n
e
n
s
u
r
es
p
r
o
p
er
alig
n
m
e
n
t
an
d
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.
1
4
,
No
.
5
,
Octo
b
er
2
0
2
5
:
4
3
1
9
-
4
3
3
1
4322
s
y
n
ch
r
o
n
izatio
n
o
f
th
e
im
a
g
e
s
,
en
ab
lin
g
ac
cu
r
ate
an
d
m
ea
n
in
g
f
u
l
c
h
an
g
e
a
n
aly
s
is
[
2
0
]
.
B
y
em
p
lo
y
in
g
CD
tech
n
iq
u
es,
r
esear
ch
e
r
s
an
d
a
n
aly
s
ts
ca
n
g
ain
v
alu
ab
le
i
n
s
ig
h
ts
in
to
th
e
d
y
n
a
m
ic
n
atu
r
e
o
f
t
h
e
E
ar
th
’
s
s
u
r
f
ac
e
.
Fig
u
r
e
1.
Ph
ase
o
n
e
o
f
th
e
p
r
o
p
o
s
al
m
eth
o
d
o
lo
g
y
Fig
u
r
e
2.
CD
p
r
o
ce
s
s
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
B
u
ild
in
g
ch
a
n
g
e
d
etec
tio
n
vi
a
cla
s
s
ifica
tio
n
in
h
ig
h
-
r
eso
lu
tio
n
a
eria
l ima
g
ery
(
Ha
yd
er Mo
s
a
Mer
z
a
)
4323
3
.
1
.
1
.
Cha
ng
e
det
ec
t
io
n
-
ba
s
ed
a
lg
ebra
Alg
eb
r
a
-
b
ased
is
o
n
e
o
f
th
e
CD
ap
p
r
o
ac
h
es,
wh
ich
in
v
o
l
v
es
em
p
lo
y
i
n
g
m
at
h
em
atica
l
o
p
e
r
atio
n
s
o
n
in
d
iv
id
u
al
im
ag
e
p
ix
els
to
g
en
er
ate
a
d
if
f
er
en
t
im
ag
e.
T
h
is
m
eth
o
d
ca
lcu
lates
th
e
d
is
cr
ep
an
cies
b
etwe
en
co
r
r
esp
o
n
d
in
g
p
ix
els
in
two
o
r
m
o
r
e
im
ag
es,
h
ig
h
lig
h
tin
g
ar
ea
s
wh
er
e
s
ig
n
if
ican
t
c
h
an
g
es
h
av
e
o
cc
u
r
r
ed
.
On
e
wid
ely
u
s
ed
alg
eb
r
a
-
b
as
ed
CD
tech
n
iq
u
e
is
im
ag
e
d
if
f
er
en
cin
g
[
2
1
]
.
T
y
p
ically
,
th
is
ap
p
r
o
ac
h
i
n
clu
d
es
s
ev
er
al
m
eth
o
d
s
lik
e
im
ag
e
r
at
io
in
g
,
im
ag
e
d
i
f
f
er
en
ci
n
g
,
an
d
ch
an
g
e
v
ec
to
r
an
aly
s
is
(
C
VA
)
,
wh
ich
in
v
o
l
v
es a
m
ath
em
atica
l
tech
n
iq
u
e
f
o
r
c
h
an
g
e
id
e
n
tific
atio
n
.
T
h
is
ty
p
e
o
f
CD
h
as
a
d
is
ad
v
an
tag
e,
s
p
ec
if
ically
with
n
o
s
in
g
,
wh
ich
o
cc
u
r
r
ed
d
u
r
in
g
p
r
ep
r
o
ce
s
s
in
g
.
3
.
1
.
2
.
Cha
ng
e
det
ec
t
io
n
-
ba
s
ed
t
ra
ns
f
o
rm
T
h
is
ca
teg
o
r
y
in
v
o
l
v
es
u
s
in
g
th
e
tr
an
s
f
o
r
m
atio
n
-
b
ased
p
ix
e
l.
T
h
e
p
r
im
ar
y
ad
v
a
n
tag
e
o
f
t
h
is
ty
p
e
is
its
ab
ilit
y
to
m
in
im
ize
r
ed
u
n
d
an
t
in
f
o
r
m
atio
n
ac
r
o
s
s
th
e
b
an
d
s
.
T
h
e
illu
s
tr
atio
n
in
Fig
u
r
e
3
p
r
esen
ts
th
e
s
tr
u
ctu
r
e
o
f
CD
b
ased
o
n
tr
an
s
f
o
r
m
atio
n
.
Ma
n
y
s
tu
d
ies
h
av
e
in
v
esti
g
ated
m
u
lti
-
tem
p
o
r
al
CD
b
y
u
s
in
g
s
atellite
im
ag
e
f
u
s
io
n
tech
n
iq
u
es
lik
e
d
is
cr
ete
wav
elet
tr
an
s
f
o
r
m
(
DW
T
)
an
d
h
o
m
o
g
e
n
eo
u
s
p
ix
el
tr
an
s
f
o
r
m
atio
n
(
HPT)
in
[
2
2
]
,
[
2
3
]
,
r
esp
ec
tiv
ely
.
Ho
wev
e
r
,
tr
an
s
f
o
r
m
-
b
ased
m
eth
o
d
s
s
tr
u
g
g
le
to
p
r
ec
is
ely
lab
el
ch
an
g
e
a
r
ea
s
in
th
e
tr
a
n
s
f
o
r
m
ed
im
ag
e
[
2
0
]
.
Fig
u
r
e
3.
CD
-
b
ased
tr
a
n
s
f
o
r
m
atio
n
3
.
1
.
3
.
Cha
ng
e
det
ec
t
io
n
-
ba
s
ed
cla
s
s
if
ica
t
io
n
T
h
e
p
r
im
a
r
y
b
e
n
ef
it
o
f
th
is
a
p
p
r
o
ac
h
,
as
in
Fig
u
r
e
4
,
lies
in
its
ab
ilit
y
to
o
f
f
e
r
p
r
ec
is
e
alter
atio
n
d
etails
th
at
r
em
ain
lar
g
ely
u
n
i
n
f
lu
en
ce
d
b
y
ex
ter
n
al
elem
en
t
s
s
u
ch
as
atm
o
s
p
h
er
ic
d
is
tu
r
b
a
n
ce
s
.
I
n
p
u
t
im
ag
es
f
r
o
m
d
if
f
e
r
en
t
tim
e
p
o
in
ts
u
n
d
er
g
o
f
ea
tu
r
e
ex
tr
ac
tio
n
,
a
n
d
th
e
r
esu
ltin
g
f
ea
tu
r
e
m
a
p
s
ar
e
f
ilter
ed
an
d
co
n
ca
ten
ated
.
T
h
ese
ar
e
th
en
p
r
o
ce
s
s
ed
b
y
a
CD
n
etwo
r
k
,
t
r
ain
ed
o
n
s
im
u
lated
s
am
p
les
to
g
en
e
r
ate
th
e
f
in
al
ch
an
g
e
m
ap
.
I
t
en
co
m
p
ass
es
p
o
s
t
-
class
if
icatio
n
co
m
p
r
ess
io
n
,
CD
co
n
d
u
cted
th
r
o
u
g
h
u
n
s
u
p
er
v
is
ed
m
eth
o
d
o
l
o
g
ies
an
d
tech
n
iq
u
es,
an
d
a
p
p
r
o
ac
h
es
b
ased
o
n
a
r
tific
ial
n
eu
r
al
n
etwo
r
k
s
.
T
h
e
e
f
f
ec
tiv
en
ess
o
f
th
is
ca
teg
o
r
y
is
co
n
tin
g
en
t u
p
o
n
th
e
ca
r
ef
u
l c
h
o
ice
o
f
tr
ai
n
in
g
d
ata.
T
ab
le
1
p
r
o
v
i
d
es a
n
ex
ten
s
i
v
e
o
v
er
v
iew
o
f
CD
tech
n
iq
u
es c
en
ter
e
d
ar
o
u
n
d
cl
ass
if
icatio
n
.
Fig
u
r
e
4.
CD
-
b
ased
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.
1
4
,
No
.
5
,
Octo
b
er
2
0
2
5
:
4
3
1
9
-
4
3
3
1
4324
T
ab
le
1.
C
o
m
p
r
eh
en
s
iv
e
a
n
aly
s
is
o
f
CD
s
tr
ateg
ies em
p
lo
y
in
g
a
class
if
icatio
n
f
r
am
ewo
r
k
A
u
t
h
o
r
F
e
a
t
u
r
e
u
s
e
d
C
l
a
s
si
f
i
c
a
t
i
o
n
a
l
g
o
r
i
t
h
m
R
a
h
b
a
n
i
a
n
d
P
a
k
h
i
r
e
h
z
a
n
[
2
4
]
P
o
si
t
i
o
n
o
f
t
h
e
c
l
o
ses
t
c
e
n
t
r
o
i
d
S
h
e
p
a
r
d
c
l
a
ss
i
f
i
c
a
t
i
o
n
Li
u
e
t
a
l
.
[
2
3
]
I
mag
e
d
e
e
p
f
e
a
t
u
r
e
s
b
y
p
r
e
d
e
f
i
n
e
d
sc
a
l
e
D
C
N
N
S
i
n
g
h
a
n
d
S
i
n
g
h
[
2
5
]
C
l
u
st
e
r
c
e
n
t
e
r
s f
i
n
d
i
n
g
R
a
d
i
a
l
b
a
s
i
s
t
r
a
i
n
i
n
g
b
y
g
e
n
e
t
i
c
a
l
g
o
r
i
t
h
m
A
z
z
o
u
z
i
e
t
a
l
.
[
2
6
]
Te
x
t
u
r
e
a
t
t
r
i
b
u
t
e
s,
sp
a
t
i
a
l
,
a
n
d
s
p
e
c
t
r
a
l
F
u
n
c
t
i
o
n
o
f
G
a
u
ss
i
a
n
r
a
d
i
a
l
G
o
n
g
e
t
a
l
.
[
2
7
]
I
n
t
r
a
-
c
l
u
st
e
r
si
m
i
l
a
r
i
t
y
C
N
N
M
o
n
t
e
si
n
o
s
e
t
a
l
.
[
2
8
]
A
t
t
r
i
b
u
t
e
s
o
f
d
a
t
a
c
o
l
l
e
c
t
i
o
n
C
l
a
s
si
f
i
e
r
s
o
f
B
a
y
e
si
a
n
n
e
t
w
o
r
k
s
3
.
1
.
4
.
Cha
ng
e
det
ec
t
io
n ba
s
ed
a
dv
a
nced
m
o
del
T
h
e
ad
v
a
n
ce
d
m
et
h
o
d
o
lo
g
y
f
o
r
d
etec
tin
g
ch
a
n
g
es
en
co
m
p
ass
es
v
ar
io
u
s
r
ef
lectio
n
an
d
s
p
ec
tr
al
m
ix
in
g
m
o
d
els.
T
h
ese
a
p
p
r
o
a
ch
es
in
v
o
lv
e
tr
an
s
f
o
r
m
in
g
im
a
g
e
v
alu
es
in
t
o
s
ig
n
if
ican
t
v
ar
i
ab
les
b
y
em
p
lo
y
in
g
th
e
p
r
in
ci
p
les
o
f
lin
ea
r
p
atter
n
an
aly
s
is
[
2
9
]
.
T
h
e
au
t
h
o
r
s
ar
e
p
r
esen
ted
with
th
e
H
o
p
f
i
eld
n
eu
r
al
n
etwo
r
k
(
HNN)
[
3
0
]
.
An
alter
n
ativ
e
CD
ap
p
r
o
ac
h
,
k
n
o
wn
as
th
e
te
m
p
o
r
al
u
n
-
m
ix
in
g
m
eth
o
d
,
h
a
s
b
ee
n
in
tr
o
d
u
ce
d
b
y
[
3
1
]
.
T
h
e
an
aly
s
es
o
f
lan
d
s
c
ap
e
im
ag
es
aim
to
id
e
n
tify
c
h
an
g
es
in
th
e
c
o
v
er
a
g
e
ar
ea
o
v
er
tim
e.
A
n
o
v
el
ap
p
r
o
ac
h
f
o
r
CD
is
in
v
esti
g
ated
,
u
tili
zin
g
a
h
y
b
r
id
s
p
ec
tr
al
ch
an
g
e
-
b
ased
m
eth
o
d
o
lo
g
y
b
y
[
3
2
]
.
T
h
is
ap
p
r
o
ac
h
d
elin
ea
tes
d
if
f
er
e
n
c
es
in
s
p
ec
tr
al
v
alu
es
an
d
s
h
ap
e,
r
ely
in
g
s
o
lely
o
n
s
p
ec
tr
al
f
ea
tu
r
es
to
id
en
tify
m
o
d
if
icatio
n
s
th
at
ar
e
n
o
t e
asi
ly
d
etec
tab
le.
3
.
2
.
Co
m
pa
riso
n o
f
i
m
a
g
e
ma
t
ching
a
lg
o
rit
hm
s
C
o
m
p
ar
ativ
e
s
tu
d
ies
h
av
e
ev
alu
ated
th
e
p
er
f
o
r
m
an
ce
o
f
im
a
g
e
m
atch
in
g
alg
o
r
ith
m
s
,
f
o
cu
s
in
g
o
n
th
e
ch
allen
g
e
o
f
ac
h
iev
in
g
in
v
ar
i
an
t
f
ea
tu
r
e
d
etec
tio
n
ac
r
o
s
s
d
iv
er
s
e
tr
an
s
f
o
r
m
atio
n
s
.
T
h
e
c
h
o
ice
o
f
alg
o
r
ith
m
d
ep
en
d
s
o
n
th
e
im
ag
e
t
y
p
e
a
n
d
v
ar
iatio
n
s
lik
e
s
ca
le
an
d
o
r
ien
tatio
n
[
3
3
]
.
T
h
e
cr
iter
ia
f
o
r
ac
h
ie
v
in
g
tr
u
e
in
v
ar
ian
t f
ea
tu
r
e
d
etec
tio
n
u
n
d
er
tr
an
s
f
o
r
m
atio
n
s
ar
e
as f
o
llo
ws
,
−
C
o
n
s
is
ten
cy
:
th
e
d
etec
ted
p
o
s
i
tio
n
m
u
s
t r
em
ain
i
n
v
ar
ian
t t
o
v
ar
iatio
n
s
in
s
ca
le
an
d
o
r
ien
tat
io
n
.
−
Acc
u
r
ac
y
:
f
ea
tu
r
es
m
u
s
t b
e
id
en
tifie
d
as a
cc
u
r
ately
as p
o
s
s
ib
le
ab
o
u
t t
h
eir
tr
u
e
l
o
ca
tio
n
s
.
−
Sp
ee
d
:
th
e
alg
o
r
ith
m
m
u
s
t p
o
s
s
ess
s
u
f
f
icien
t e
f
f
icien
cy
t
o
p
r
o
ce
s
s
th
e
im
ag
e
s
wif
tly
.
3
.
2
.
1
.
Sca
le
inv
a
ria
nt
f
e
a
t
ure
t
ra
ns
f
o
rm
T
h
e
m
eth
o
d
was
o
r
ig
in
ally
p
r
o
p
o
s
ed
b
y
L
o
we
in
2
0
0
4
,
th
e
SIFT
alg
o
r
ith
m
h
as
b
ec
o
m
e
a
s
tan
d
ar
d
in
co
m
p
u
ter
v
is
io
n
an
d
p
h
o
t
o
g
r
am
m
etr
y
.
T
h
is
alg
o
r
ith
m
ex
t
r
ac
ts
d
is
tin
ctiv
e
f
ea
tu
r
es
f
r
o
m
im
ag
es,
en
ab
lin
g
r
o
b
u
s
t
m
atc
h
in
g
ac
r
o
s
s
d
iv
e
r
s
e
lan
d
s
ca
p
e
s
ce
n
es.
A
d
d
itio
n
ally
,
it
c
o
m
p
u
tes
d
escr
ip
to
r
s
f
o
r
th
ese
f
ea
tu
r
es,
f
ac
ilit
atin
g
m
o
r
e
ac
cu
r
ate
an
d
ef
f
icien
t im
ag
e
an
al
y
s
is
[
3
4
]
.
3
.
2
.
2
.
Sp
ee
ded
up
ro
bu
s
t
f
ea
t
ure
T
h
e
s
p
ee
d
ed
u
p
r
o
b
u
s
t
f
ea
tu
r
e
(
SUR
F
)
i
s
an
ac
ce
ler
ated
v
er
s
io
n
o
f
th
e
SIFT
alg
o
r
ith
m
.
I
t
is
b
o
th
a
lo
ca
l
k
ey
-
p
o
in
t
d
etec
to
r
a
n
d
d
escr
ip
to
r
.
I
t
g
en
e
r
ates
d
escr
i
p
to
r
s
with
eith
er
6
4
o
r
1
2
8
d
im
en
s
io
n
s
.
I
n
th
e
p
h
ase
o
f
f
ea
tu
r
e
d
etec
tio
n
,
SUR
F
u
tili
ze
s
th
e
L
ap
lacia
n
o
f
Gau
s
s
ian
(
L
o
G)
.
Ad
d
itio
n
ally
,
f
o
r
f
ea
tu
r
e
d
escr
ip
tio
n
,
SUR
F
ap
p
lies
wav
elet
r
esp
o
n
s
es
in
b
o
th
v
er
tic
al
an
d
h
o
r
izo
n
tal.
W
h
ile
SUR
F
o
f
f
er
s
im
p
r
o
v
ed
s
p
ee
d
o
v
er
SIFT
,
it is
s
till
n
o
t
id
ea
l f
o
r
r
ea
l
-
tim
e
ap
p
licatio
n
s
.
T
h
e
SUR
F wa
s
in
tr
o
d
u
ce
d
in
2
0
0
6
[
3
5
]
.
3
.
2
.
3
.
Acc
eler
a
t
io
n o
f
K
AZ
E
a
lg
o
rit
hm
Acc
eler
atio
n
o
f
KAZ
E
al
g
o
r
i
th
m
(
AKAZ
E
)
,
in
tr
o
d
u
ce
d
b
y
Sh
ar
m
a
an
d
J
ain
[
3
6
]
,
is
an
im
p
r
o
v
e
d
v
er
s
io
n
o
f
th
e
KAZ
E
al
g
o
r
it
h
m
.
AKAZ
E
d
etec
ts
f
ea
tu
r
es
b
y
f
i
n
d
in
g
ex
tr
em
a
o
f
s
ec
o
n
d
-
o
r
d
e
r
d
e
r
iv
ativ
es
with
in
a
n
o
n
lin
ea
r
m
u
lti
-
s
ca
l
e
p
y
r
am
id
b
ased
o
n
im
ag
e
d
if
f
u
s
io
n
.
AKAZ
E
u
tili
ze
s
f
as
t
ex
p
licit
d
if
f
u
s
io
n
(
FED)
with
in
a
p
y
r
am
id
al
f
r
a
m
ewo
r
k
,
o
p
tim
izin
g
th
e
s
p
ee
d
o
f
f
ea
tu
r
e
d
etec
tio
n
i
n
a
n
o
n
lin
ea
r
s
ca
le
s
p
ac
e.
T
h
is
alg
o
r
ith
m
s
ig
n
if
ican
tl
y
e
n
h
an
ce
s
th
e
ef
f
icien
cy
an
d
p
e
r
f
o
r
m
an
ce
o
f
th
e
f
ea
tu
r
e
d
etec
tio
n
p
r
o
ce
s
s
.
3
.
2
.
4
.
O
rient
ed
F
AST
a
nd
ro
t
a
t
ed
B
RIE
F
T
h
e
o
r
i
en
te
d
FAS
T
a
n
d
r
o
t
at
ed
B
R
I
E
F
(
OR
B
)
f
e
at
u
r
e
e
x
t
r
a
cti
o
n
m
e
th
o
d
w
as
c
h
o
s
e
n
f
o
r
f
e
at
u
r
e
ex
t
r
ac
ti
o
n
i
n
a
er
ial
i
m
a
g
es,
w
h
ic
h
w
as
i
n
t
r
o
d
u
c
ed
i
n
2
0
1
1
[
3
7
]
.
I
t
is
a
c
o
m
p
u
t
ati
o
n
al
l
y
ef
f
ic
ie
n
t
a
lte
r
n
at
iv
e
t
o
SIFT
a
n
d
S
UR
F,
co
m
b
i
n
i
n
g
FAST
k
e
y
-
p
o
i
n
t
d
ete
cti
o
n
wi
t
h
t
h
e
B
R
I
E
F
d
es
c
r
i
p
t
o
r
s
.
D
es
p
it
e
FAS
T
'
s
l
ac
k
o
f
o
r
ie
n
t
ati
o
n
ca
l
c
u
lat
io
n
a
n
d
B
R
I
E
F'
s
li
m
it
ati
o
n
s
wit
h
r
o
t
ati
o
n
,
m
o
d
i
f
i
ca
t
io
n
s
h
a
v
e
e
n
h
a
n
c
e
d
OR
B
'
s
p
e
r
f
o
r
m
a
n
c
e.
OR
B
e
x
h
ib
ite
d
an
is
s
u
e
wit
h
u
n
e
v
e
n
an
d
s
p
a
r
s
e
d
is
tr
ib
u
t
io
n
o
f
f
e
at
u
r
e
p
o
i
n
ts
.
Des
p
i
te
t
h
is
lim
i
tat
io
n
,
OR
B
d
e
m
o
n
s
tr
at
es
s
ig
n
i
f
ic
a
n
t
ad
v
a
n
ta
g
es
in
te
r
m
s
o
f
co
m
p
u
tat
io
n
a
l
ef
f
i
cie
n
c
y
.
W
e
c
h
o
s
e
t
h
e
S
I
FT
alg
o
r
it
h
m
f
o
r
o
u
r
s
t
u
d
y
b
ase
d
o
n
i
ts
s
u
p
e
r
i
o
r
p
e
r
f
o
r
m
a
n
c
e
i
n
o
u
r
c
o
m
p
ar
is
o
n
,
as
s
h
o
w
n
in
F
ig
u
r
e
s
5
a
n
d
6,
an
d
T
ab
le
2
.
E
ac
h
p
air
in
Fig
u
r
e
5
s
h
o
ws
th
e
o
r
ig
in
al
an
d
p
r
o
ce
s
s
ed
s
atell
ite
im
ag
es,
with
r
ed
lin
es
in
d
icatin
g
co
r
r
esp
o
n
d
in
g
k
ey
p
o
in
ts
d
et
ec
ted
an
d
m
atch
ed
b
y
th
e
alg
o
r
ith
m
s
.
T
h
is
co
m
p
ar
is
o
n
h
ig
h
lig
h
ts
th
e
p
er
f
o
r
m
an
ce
a
n
d
r
o
b
u
s
tn
ess
o
f
SIFT
an
d
SUR
F
in
id
en
tify
i
n
g
s
p
atially
co
n
s
is
ten
t
f
ea
tu
r
es.
T
h
e
b
ar
ch
ar
t
in
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
B
u
ild
in
g
ch
a
n
g
e
d
etec
tio
n
vi
a
cla
s
s
ifica
tio
n
in
h
ig
h
-
r
eso
lu
tio
n
a
eria
l ima
g
ery
(
Ha
yd
er Mo
s
a
Mer
z
a
)
4325
Fig
u
r
e
6
s
h
o
ws
th
at
SIFT
ac
h
iev
ed
th
e
h
ig
h
est
n
u
m
b
er
o
f
m
atch
es
(
5
1
)
,
f
o
llo
wed
b
y
SUR
F
(
3
0
)
,
wh
ile
AKAZ
E
an
d
OR
B
p
r
o
d
u
ce
d
f
ewe
r
m
atch
es.
T
h
e
d
ash
ed
lin
e
in
d
icate
s
th
e
m
in
im
u
m
m
at
ch
co
u
n
t
th
r
esh
o
ld
r
eq
u
ir
ed
f
o
r
ac
ce
p
tab
le
p
e
r
f
o
r
m
an
ce
.
Fig
u
r
e
5.
Vis
u
al
co
m
p
ar
is
o
n
o
f
f
ea
tu
r
e
m
atc
h
in
g
u
s
in
g
SIFT
an
d
SUR
F a
lg
o
r
ith
m
s
Fig
u
r
e
6.
C
o
m
p
a
r
is
o
n
o
f
th
e
n
u
m
b
er
o
f
f
ea
tu
r
e
m
atch
es d
ete
cted
b
y
f
o
u
r
alg
o
r
ith
m
s
T
ab
le
2.
R
esu
lts
o
f
f
o
u
r
m
atch
in
g
alg
o
r
ith
m
s
A
l
g
o
r
i
t
h
m
M
i
n
i
m
u
m
ma
t
c
h
e
s
c
o
u
n
t
R
e
s
u
l
t
N
o
t
e
s
S
I
F
T
10
S
I
F
T
mat
c
h
e
s
a
r
e
f
o
u
n
d
-
5
1
/
1
0
B
e
t
t
e
r
mat
c
h
e
s
S
U
R
F
10
S
U
R
F
ma
t
c
h
e
s
a
r
e
f
o
u
n
d
-
3
0
/
1
0
G
o
o
d
m
a
t
c
h
e
s
A
K
A
ZE
10
A
K
A
ZE
mat
c
h
e
s
a
r
e
f
o
u
n
d
-
9
/
1
0
N
o
t
e
n
o
u
g
h
mat
c
h
e
s
a
r
e
f
o
u
n
d
O
R
B
10
O
R
B
m
a
t
c
h
e
s
a
r
e
f
o
u
n
d
-
0
/
1
0
N
o
t
e
n
o
u
g
h
mat
c
h
e
s
a
r
e
f
o
u
n
d
4.
RE
SU
L
T
S AN
D
D
I
SCU
SS
I
O
N
Firstl
y
,
we
co
n
d
u
ct
a
co
m
p
r
e
h
en
s
iv
e
an
aly
s
is
an
d
co
m
p
ar
i
s
o
n
o
f
r
esu
lts
.
W
e
v
alid
ate
o
u
r
p
r
o
p
o
s
ed
m
eth
o
d
t
h
r
o
u
g
h
e
x
p
er
im
e
n
ts
th
r
o
u
g
h
e
m
p
lo
y
e
d
f
i
v
e
ev
al
u
atio
n
m
etr
ics:
p
r
ec
is
io
n
,
r
e
ca
ll
,
F1
-
s
co
r
e,
an
d
ac
cu
r
ac
y
,
s
ee
(
1
)
to
(
4
)
,
in
s
tead
o
f
th
e
ar
ea
u
n
d
e
r
th
e
r
ec
eiv
er
o
p
er
atin
g
c
h
ar
ac
ter
is
tics
cu
r
v
e
(
AURO
C
)
wh
ich
is
s
er
v
es
as
a
q
u
an
titativ
e
m
ea
s
u
r
e
em
p
l
o
y
ed
to
ass
ess
th
e
ef
f
ec
tiv
en
ess
th
e
ef
f
icac
y
o
f
class
if
icatio
n
m
o
d
els.
F1
-
s
co
r
e
s
er
v
es
as
a
m
ea
s
u
r
e
o
f
th
e
b
in
ar
y
class
if
icatio
n
m
o
d
el’
s
ac
c
u
r
ac
y
.
T
h
e
f
iv
e
m
etr
ics
a
r
e
em
p
lo
y
ed
o
v
er
f
iv
e
class
if
icatio
n
alg
o
r
ith
m
s
:
n
aï
v
e
Ba
y
es
(
NB
)
,
d
ec
is
io
n
tr
ee
(
DT
)
,
g
r
a
d
i
en
t
b
o
o
s
tin
g
(
GB
)
,
RF
,
an
d
lo
g
is
tic
r
eg
r
ess
io
n
(
L
R
)
,
as sh
o
wn
in
T
ab
le
s
3
an
d
4
,
an
d
Fig
u
r
e
7
.
=
(
+
)
(
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.
1
4
,
No
.
5
,
Octo
b
er
2
0
2
5
:
4
3
1
9
-
4
3
3
1
4326
=
(
+
)
(
2
)
1
−
=
2
×
(
×
)
(
+
)
(
3
)
=
(
4
)
T
ab
le
3.
C
lass
if
icatio
n
m
etr
ics b
ased
o
n
f
iv
e
alg
o
r
ith
m
s
C
l
a
s
si
f
i
c
a
t
i
o
n
a
l
g
o
r
i
t
h
m
D
a
t
a
s
e
t
P
r
e
c
i
s
i
o
n
R
e
c
a
l
l
F1
-
sc
o
r
e
Tr
u
e
(
1
)
F
a
l
se
(
0
)
Tr
u
e
(
1
)
F
a
l
se
(
0
)
Tr
u
e
(
1
)
F
a
l
se
(
0
)
Tr
u
e
(
1
)
F
a
l
se
(
0
)
NB
8
6
5
7
3
K
8
6
5
7
3
K
0
.
5
2
0
.
5
1
0
.
4
8
0
.
5
5
0
.
5
0
0
.
5
3
DT
8
6
5
7
3
K
8
6
5
7
3
K
0
.
5
2
0
.
5
1
0
.
4
8
0
.
5
4
0
.
5
0
0
.
5
3
GB
8
6
5
7
3
K
8
6
5
7
3
K
0
.
5
2
0
.
5
2
0
.
4
7
0
.
5
8
0
.
4
9
0
.
5
5
RF
8
6
5
7
3
K
8
6
5
7
3
K
0
.
5
7
0
.
5
6
0
.
5
3
0
.
6
1
0
.
5
5
0
.
5
8
LR
8
6
5
7
3
K
8
6
5
7
3
K
0
.
5
2
0
.
5
2
0
.
4
8
0
.
5
5
0
.
5
0
0
.
5
3
T
ab
le
4.
Acc
u
r
ac
y
an
d
AUROC
b
ased
o
n
f
iv
e
al
g
o
r
ith
m
s
C
l
a
s
si
f
i
c
a
t
i
o
n
a
l
g
o
r
i
t
h
m
D
a
t
a
s
e
t
r
e
c
o
r
d
s
S
u
p
p
o
r
t
A
c
c
u
r
a
c
y
A
U
R
O
C
Tr
u
e
(
1
)
F
a
l
se
(
0
)
Tr
u
e
(
1
)
F
a
l
se
(
0
)
NB
8
6
5
7
3
K
8
6
5
7
3
K
4
3
3
1
6
4
3
2
5
6
0
.
5
2
0
.
5
1
9
7
9
9
DT
8
6
5
7
3
K
8
6
5
7
3
K
4
3
4
4
7
4
3
1
2
5
0
.
5
1
0
.
5
1
7
7
1
7
GB
8
6
5
7
3
K
8
6
5
7
3
K
4
3
2
5
3
4
3
3
1
9
0
.
5
2
0
.
5
3
0
9
6
5
RF
8
6
5
7
3
K
8
6
5
7
3
K
2
1
6
7
1
2
1
6
1
5
0
.
5
7
0
.
5
5
2
9
2
8
LR
8
6
5
7
3
K
8
6
5
7
3
K
4
3
3
1
6
4
3
2
5
6
0
.
5
2
0
.
5
2
4
9
6
8
Fig
u
r
e
7.
R
esu
lts
o
f
m
etr
ics f
o
r
f
iv
e
class
if
icatio
n
alg
o
r
ith
m
s
4
.
1
.
Cha
ng
e
det
ec
t
i
o
n pro
ce
s
s
Fo
r
th
e
s
ec
o
n
d
p
ar
t
o
f
o
u
r
CD
m
eth
o
d
o
lo
g
y
,
as
illu
s
tr
ated
in
Fig
u
r
e
8
,
a
k
ey
ch
allen
g
e
in
o
u
r
s
tu
d
y
is
th
e
u
n
ce
r
tain
ty
o
f
th
e
im
a
g
e'
s
ex
ac
t
lo
ca
tio
n
with
in
th
e
g
lo
b
al
co
o
r
d
in
ate
s
y
s
tem
,
s
o
ex
tr
ac
tin
g
b
u
ild
i
n
g
f
o
o
tp
r
i
n
ts
f
r
o
m
ae
r
ial
im
ag
es
i
s
a
v
ital
p
r
ep
r
o
ce
s
s
in
g
s
tep
in
im
ag
e
an
aly
s
is
,
en
h
an
cin
g
ac
c
u
r
ac
y
in
CD
task
s
.
T
h
is
p
r
ep
r
o
ce
s
s
in
g
s
tep
en
h
a
n
ce
s
im
ag
e
m
atch
in
g
b
y
p
r
ep
ar
in
g
d
atasets
to
id
en
tify
alter
ed
f
ea
tu
r
es
in
v
er
y
h
ig
h
-
r
eso
lu
tio
n
im
a
g
es,
s
u
ch
a
s
b
u
ild
in
g
s
,
r
o
ad
s
,
o
r
v
eh
icles
[
3
8
]
.
Up
o
n
id
e
n
tify
in
g
th
e
c
o
r
r
esp
o
n
d
in
g
m
ask
s
,
th
e
SIFT
m
atch
in
g
alg
o
r
ith
m
d
em
o
n
s
tr
ates h
ig
h
e
f
f
icac
y
in
f
ea
tu
r
e
d
etec
tio
n
with
in
th
e
n
e
w
d
ataset.
Ou
r
co
n
tr
ib
u
tio
n
in
clu
d
es
th
e
p
r
ep
a
r
atio
n
o
f
v
er
y
h
ig
h
-
r
eso
lu
tio
n
ae
r
ial
im
a
g
er
y
at
a
co
n
s
is
ten
t
r
eso
lu
tio
n
o
f
6
cm
.
E
ac
h
im
a
g
e
p
air
is
ass
o
ciate
d
with
a
co
r
r
esp
o
n
d
i
n
g
m
ask
,
th
e
r
eb
y
m
i
n
im
izin
g
n
o
is
e
an
d
en
s
u
r
in
g
clar
ity
.
Fo
r
t
h
e
m
ask
in
g
p
r
o
ce
s
s
,
we
u
tili
ze
d
a
s
in
g
le
m
ask
f
o
r
th
e
im
a
g
e
p
air
o
f
B
ag
h
d
ad
2
0
0
7
an
d
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
B
u
ild
in
g
ch
a
n
g
e
d
etec
tio
n
vi
a
cla
s
s
ifica
tio
n
in
h
ig
h
-
r
eso
lu
tio
n
a
eria
l ima
g
ery
(
Ha
yd
er Mo
s
a
Mer
z
a
)
4327
2
0
2
4
,
with
a
to
tal
o
f
2
5
im
a
g
e
p
air
s
in
T
I
F
f
o
r
m
at,
w
h
ich
g
en
er
ated
(
8
6
,
5
7
3
K)
r
ec
o
r
d
s
,
co
n
s
is
tin
g
o
f
b
o
th
tr
u
e
an
d
f
alse
lab
eled
d
ata.
Alth
o
u
g
h
th
e
r
e
ar
e
b
en
ch
m
a
r
k
d
atasets
av
ailab
le
f
o
r
CD
,
as
s
e
s
s
m
en
t
o
f
ch
an
g
es
in
th
e
L
UL
C
was c
o
n
d
u
cted
b
y
u
tili
zin
g
o
u
r
n
ewly
c
o
llected
d
ataset.
Fig
u
r
e
8.
Ph
ase
two
o
f
th
e
p
r
o
p
o
s
al
m
eth
o
d
o
lo
g
y
4
.
2
.
Cha
ng
e
det
ec
t
i
o
n ba
s
ed
o
n t
hree
a
pp
ro
a
ches
Fo
r
th
e
p
r
o
b
lem
o
f
CD
,
we
u
s
ed
two
p
air
s
o
f
im
a
g
es
as
a
s
am
p
le,
d
r
awn
f
r
o
m
o
u
r
d
atasets
o
f
2
0
0
7
an
d
2
0
2
4
.
Ou
r
p
r
o
p
o
s
ed
m
et
h
o
d
was
im
p
lem
en
ted
b
y
u
s
in
g
Py
th
o
n
an
d
th
e
J
u
p
y
ter
p
l
atf
o
r
m
in
a
6
4
-
b
it
W
in
d
o
ws
s
y
s
tem
,
eq
u
ip
p
ed
w
ith
an
I
n
tel
I
r
is
C
PU
an
d
8
G
B
o
f
R
AM
.
Her
e,
th
e
CD
is
tr
ain
ed
an
d
test
ed
b
y
u
s
in
g
th
r
ee
CD
m
eth
o
d
s
as
s
h
o
wn
in
Fig
u
r
e
9
,
Fig
u
r
e
9
(
a
)
s
h
o
ws
th
e
CD
-
b
ased
class
if
icatio
n
,
Fig
u
r
e
9
(
b
)
s
h
o
ws th
e
CD
-
b
ased
tr
an
s
f
o
r
m
atio
n
,
an
d
Fig
u
r
e
9
(
c
)
s
h
o
ws th
e
CD
-
b
ased
alg
eb
r
a
.
T
h
e
f
ir
s
t
m
eth
o
d
p
r
o
d
u
ce
d
m
o
r
e
ac
cu
r
ate
r
esu
lts
th
an
o
th
er
m
eth
o
d
s
,
an
d
th
e
a
r
ea
s
o
f
c
h
an
g
e
d
em
o
n
s
tr
ate
a
h
i
g
h
d
eg
r
ee
o
f
co
n
s
is
ten
cy
with
th
e
g
r
o
u
n
d
tr
u
th
d
ata.
As
ca
n
b
e
s
ee
n
clea
r
l
y
th
at
th
e
c
h
an
g
in
g
ar
ea
h
ig
h
lig
h
ted
in
y
ello
w
co
l
o
r
was
p
er
f
ec
t.
Desp
ite
o
u
r
b
u
ild
in
g
'
s
co
n
ce
r
n
,
we
c
o
m
p
u
ted
m
u
ltip
le
class
es
(
u
r
b
an
,
wate
r
,
v
eg
etatio
n
,
n
o
n
-
v
eg
etatio
n
,
an
d
b
a
r
)
in
p
air
s
o
f
im
ag
es
as
s
h
o
wn
in
T
a
b
le
5
.
T
h
e
r
esu
lts
o
f
d
etec
tin
g
ch
an
g
es
in
th
e
s
ec
o
n
d
a
CD
ap
p
r
o
ac
h
g
r
o
u
n
d
ed
in
o
u
r
d
ataset,
wer
e
t
h
e
wo
r
s
t,
s
in
ce
in
th
is
m
eth
o
d
,
r
ed
u
n
d
an
t
s
p
ec
tr
al
b
an
d
s
ar
e
m
in
im
ized
b
y
d
ec
o
m
p
o
s
in
g
t
h
e
o
b
jects,
s
o
it
co
u
ld
n
’
t
co
v
er
b
o
th
s
m
all
an
d
lar
g
e
ar
ea
s
.
R
eg
ar
d
i
n
g
t
o
th
e
t
h
ir
d
CD
m
eth
o
d
,
th
e
b
lu
e
lin
e
s
in
d
icate
d
to
t
h
e
f
ea
tu
r
es
d
ete
ctio
n
th
at
d
i
f
f
er
in
g
b
etwe
en
two
im
ag
es,
an
d
th
e
g
r
ee
n
lin
es
r
ep
r
esen
t
th
e
b
o
u
n
d
ar
ies
o
f
c
h
an
g
es
ar
ea
s
in
p
air
s
o
f
im
ag
es
.
T
h
e
CD
r
esu
lts
was
g
o
o
d
,
b
u
t
as
a
cr
u
cial
asp
ec
t
o
f
tr
ad
itio
n
al
a
lg
eb
r
a
m
eth
o
d
is
to
d
eter
m
in
e
th
r
esh
o
ld
v
alu
e
,
as
it
d
ir
ec
tly
in
f
lu
e
n
ce
s
th
e
ab
ili
ty
to
id
en
tif
y
s
p
ec
if
ic
ar
ea
s
o
f
an
im
a
g
e
f
o
r
ev
al
u
atin
g
th
e
ex
ten
t
o
f
ch
a
n
g
e.
Alth
o
u
g
h
th
is
tech
n
iq
u
e
is
r
elativ
ely
s
im
p
le
to
im
p
lem
e
n
t,
s
elec
tin
g
a
n
a
p
p
r
o
p
r
iate
th
r
esh
o
ld
is
o
f
ten
d
if
f
icu
lt.
A
p
o
o
r
ly
ch
o
s
en
th
r
esh
o
ld
ca
n
r
esu
lt
in
i
n
ac
cu
r
at
e
esti
m
atio
n
s
o
f
th
e
d
eg
r
ee
o
f
ch
an
g
e.
Fin
ally
,
r
eg
ar
d
in
g
th
e
m
u
lti
-
class
o
f
L
UL
C
,
we
al
s
o
co
m
p
u
ted
th
e
ch
an
g
es
o
v
er
th
e
p
er
io
d
in
2
0
0
7
-
2
0
2
4
as
s
h
o
wn
in
T
ab
le
6
,
in
ad
d
itio
n
to
c
o
m
p
u
tin
g
th
e
p
e
r
ce
n
tag
e
o
f
b
u
ilt
-
u
p
an
d
n
o
n
-
b
u
ilt
-
u
p
th
r
o
u
g
h
u
s
in
g
th
e
n
o
r
m
alize
d
d
if
f
er
en
ce
b
u
ilt
-
u
p
in
d
e
x
(
ND
B
I
)
as r
ep
r
esen
ted
in
T
a
b
le
7
.
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.
1
4
,
No
.
5
,
Octo
b
er
2
0
2
5
:
4
3
1
9
-
4
3
3
1
4328
(
a)
(
b
)
(
c)
Fig
u
r
e
9.
Me
th
o
d
s
o
f
CD
b
ase
d
o
n
(
a)
c
lass
if
icatio
n
,
(
b
)
t
r
an
s
f
o
r
m
atio
n
,
a
n
d
(
c
)
alg
eb
r
a
T
ab
le
5.
Ar
ea
s
CD
in
2
0
0
7
-
2
0
2
4
Ty
p
e
2
0
0
7
A
r
e
a
(
%)
2
0
2
4
A
r
e
a
(
%)
U
r
b
a
n
1
1
0
7
.
4
9
8
1
2
1
.
7
3
1
2
4
3
.
8
4
5
1
2
4
.
4
1
W
a
t
e
r
1
8
4
.
7
5
3
7
3
.
6
2
1
7
0
.
2
1
3
2
3
.
3
5
V
e
g
e
t
a
t
i
o
n
8
9
9
.
1
2
9
6
1
7
.
6
4
1
0
3
3
.
6
6
5
2
0
.
2
8
N
o
n
-
v
e
g
e
t
a
t
i
o
n
8
7
8
.
0
7
7
2
1
7
.
2
4
8
0
1
.
5
1
5
2
1
5
.
7
2
B
a
r
e
2
0
2
6
.
9
3
5
4
3
9
.
7
7
1
8
4
7
.
1
5
4
7
3
6
.
2
4
T
ab
le
6.
C
h
an
g
es o
f
class
es in
2
0
0
7
-
2
0
2
4
Ty
p
e
2
0
0
7
2
0
2
4
C
h
a
n
g
e
2
0
0
7
-
2
0
2
4
U
r
b
a
n
1
1
0
7
.
4
9
8
1
1
2
4
3
.
8
4
5
1
1
3
6
.
3
4
7
W
a
t
e
r
1
8
4
.
7
5
3
7
1
7
0
.
2
1
3
2
-
1
4
.
5
4
0
5
V
e
g
e
t
a
t
i
o
n
8
9
9
.
1
2
9
6
1
0
3
3
.
6
6
5
1
3
4
.
5
3
6
N
o
n
-
v
e
g
e
t
a
t
i
o
n
8
7
8
.
0
7
7
2
8
0
1
.
5
1
5
2
-
7
6
.
5
6
2
B
a
r
e
2
0
2
6
.
9
3
5
4
1
8
4
7
.
1
5
4
7
-
1
7
9
.
7
8
0
7
T
ab
le
7.
B
u
ilt
-
u
p
a
n
d
n
o
n
-
b
u
il
t
-
u
p
in
2
0
0
7
-
2
0
2
4
Ty
p
e
2
0
0
7
A
r
e
a
(
%)
2
0
2
4
A
r
e
a
(
%)
B
u
i
l
t
-
up
9
9
8
.
1
2
7
1
9
.
5
8
1
1
8
1
.
5
8
4
2
3
.
1
8
N
o
n
-
b
u
i
l
t
-
up
4
0
9
8
.
2
6
7
8
0
.
4
2
3
9
1
4
.
8
1
7
6
.
8
2
5.
CO
NCLU
SI
O
N
T
h
is
p
ap
er
p
r
esen
ted
th
e
CD
ap
p
r
o
ac
h
in
m
u
lti
-
h
ig
h
-
r
eso
lu
tio
n
ae
r
ial
im
ag
es
in
2
0
0
7
an
d
2
0
2
4
at
th
e
s
tu
d
y
a
r
ea
in
th
e
ca
p
ital
o
f
I
r
aq
(
B
ag
h
d
a
d
)
.
T
o
h
ig
h
lig
h
t
s
ig
n
if
ican
t
ch
an
g
e
f
ea
tu
r
es
an
d
s
u
p
p
r
ess
Evaluation Warning : The document was created with Spire.PDF for Python.