TELKOM
NIKA Indonesia
n
Journal of
Electrical En
gineering
Vol. 13, No. 3, March 2
015,
pp. 441 ~ 44
8
DOI: 10.115
9
1
/telkomni
ka.
v
13i3.721
0
441
Re
cei
v
ed
No
vem
ber 2
0
, 2014; Re
vi
sed
De
cem
ber 2
9
,
2014; Accep
t
ed Jan
uary 1
5
, 2015
PDF Ba
sed Icing Image Recognition Applied to Online
Early Warning System for Transmissio
n
Lines
Xin Yin
1
, X
u
Chen*
2
, Erta
o Lei
2
, Jinrui Tang
2
, Hong
Wang
1
, Minghao Wen
2
1
School of Elec
trical an
d Elect
r
onic En
gin
eer
i
ng, T
he Univer
sit
y
of Manc
he
ster,
Oxford R
oad,
Manch
e
ster M13 9PL, U.K.
2
State Ke
y
La
b
o
rator
y
of Adva
nced El
ectr
om
agn
etic Eng
i
ne
erin
g and T
e
ch
nol
og
y,
Huaz
hon
g Un
i
v
ersit
y
of Sci
e
n
c
e and T
e
chno
log
y
, W
uha
n 4
300
74, Ch
ina
*Corres
p
o
ndi
n
g
author, e-ma
i
l
: chen
xu
_h
ust@qq.com
A
b
st
r
a
ct
T
h
is pap
er pro
poses
an o
n
li
n
e
early w
a
rn
in
g te
chn
i
qu
e a
nd the pr
ob
abi
lity distrib
u
tio
n
functio
n
(PDF
) base
d
ic
ing
i
m
a
ge r
e
co
gniti
on for
over
hea
d p
o
w
e
r tra
n
smissio
n
l
i
n
e
s
. T
he
mai
n
fu
nction
ality
of th
e
online ear
ly warning system for over
head
transmission li
nes firstly s
u
ggested in t
h
is
paper is the early
w
a
rning of ic
in
g, forest fire, lightin
g, insu
lato
r
flashover, co
nductor
gal
lop
i
ng an
d inv
a
si
o
n
w
h
ich is bas
e
d
on a schem
e
of the online inspection
system
for the transmission lines wi
th optical fibre comm
unic
ation
techno
lo
gy. As
a c
a
se stu
d
y,
the
early
w
a
rnin
g of
ic
in
g is
disc
ussed
in
this pap
er and
a
co
mpreh
ens
i
v
e
icing
e
a
rly w
a
r
n
in
g sch
e
m
e
h
a
s be
en
pro
p
o
s
ed: the
lo
c
a
l mete
oro
l
og
ical
cond
itio
ns
a
n
d
mech
an
ical
data
are
applied for
the
initial criteria
of the icing
ear
ly
warning
system
,
and the PD
F
bas
ed im
age recognition
techno
lo
gy is then
used to fi
nally
deci
de th
e icin
g
con
d
iti
on of trans
mis
s
ion l
i
nes, w
h
i
c
h can b
e
mo
r
e
effectively de
al
t w
i
th
than co
mplex p
i
ctures a
c
quir
ed at tow
e
rs.
Ke
y
w
ords
:
overh
ead
pow
er trans
missi
on
lines, o
n
l
i
ne
e
a
rly w
a
rnin
g, p
r
oba
bil
i
ty distri
butio
n functi
on,
icing
imag
e recog
n
iti
o
n
Copy
right
©
2015 In
stitu
t
e o
f
Ad
van
ced
En
g
i
n
eerin
g and
Scien
ce. All
rig
h
t
s reser
ve
d
.
1. Introduc
tion
The po
wer
system is one
of the most compl
e
x indu
strial sy
stem
s in the wo
rld, who
s
e
function
ality is to tran
sform the prima
r
y energy
into elect
r
ical energy by powe
r
ge
neratin
g
device
s
. Th
e
electri
c
ity is
distrib
u
ted to
the ce
ntre
of load by p
o
wer tr
a
n
smissi
on, tran
sform
i
ng
as
well
a
s
distribution
sy
stem, an
d the
n
co
nverte
d to
othe
r
kind
s o
f
ene
rgy a
m
o
ng
whi
c
h
po
wer
transmission system
takes a
responsi
b
ility for power energy
transmi
s
sion. F
aults of
transmissio
n lines
will not only pose a threat to
the safe and sta
b
l
e
operation o
f
power
syste
m
s
itself, but also cau
s
e p
o
wer u
s
ers’ hu
g
e
loss. As the
high pa
ce of
mode
rn po
we
r grid a
nd
sm
art
grid con
s
tru
c
tions, the hi
gh-voltag
e-le
vel tr
ansmi
ssion line
s
ha
ve becom
e more a
nd m
o
re
popul
ar, whi
c
h requi
re
s hi
gh relia
bility for power
su
pply of the transmi
ssion li
nes. Howeve
r
,
transmissio
n
lines m
a
y easily broke d
o
wn a
nd
ev
en be
com
e
a disa
ste
r
b
e
ca
use they are
exposed in th
e air
all yea
r
l
ong a
nd influ
enced by
the
adverse
clim
ate. Some tra
n
smi
ssi
on lin
es
are eve
n
set
across the b
ogs, de
se
rts
and mo
untai
n
s
. Therefore, we ne
ed to condu
ct in-dep
th
study on the
early wa
rni
ng tech
nolog
y of trans
mi
ssi
on line so
that we ca
n find faults o
f
transmissio
n line in time, whi
c
h gu
ara
n
t
ees the
safe
ty operation
of transmi
ssi
on line an
d its
device
s
.
To ove
r
com
e
the difficu
lty of transmissi
on li
ne
insp
ecto
r
a
nd ea
rly wa
rning,
an
insp
ectio
n
sy
stem for the
transmissio
n
line is
pr
opo
sed in thi
s
p
a
per,
whi
c
h
co
nsi
s
ts of to
wer
terminal
s,
fib
r
e-opti
c
com
m
unication systems and
the
ma
ster stati
on. Due t
o
la
rge
amo
unt
of
real
-time dat
a, such as i
m
age
s and p
i
cture
s
at ea
ch towe
r alo
ng with long
-distan
c
e line
s
, a
novel fibre-o
p
tic commu
ni
cation
syste
m
is p
r
e
s
ent
ed, whi
c
h i
s
based o
n
t
he fusi
on
spl
i
cing
techni
que
s o
f
the optical
fibre comp
osite
ove
r
he
ad groun
d
wire
s
(OPG
Ws) an
d Ethern
e
t
passive o
p
tical networks (EPONs). The
informatio
n
of the ea
rly warning i
s
o
b
tained f
r
om
the
online in
sp
ect
i
on syste
m
of overhe
ad tra
n
smi
ssi
on lin
es. Pape
rs [1
] have analyzed the re
al time
perfo
rman
ce
of the EPON system for b
e
tter monitorin
g
.
Based
on th
e inform
ation
acq
u
ire
d
by
the tran
smi
ssi
on line
ea
rly wa
rning
system,
influen
ce
s of
all
kind
s of
disa
sters
on
the tran
smi
ssion line
an
d
corre
s
p
ondin
g
ea
rly warni
n
g
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 13, No. 3, March 2
015 : 441 – 4
4
8
442
method
s can
be studie
d
. The
m
a
in
fu
nction
alit
ies of
the
onli
n
e
early wa
rnin
g
for overh
e
ad
transmissio
n lines that a
r
e
sug
g
e
s
ted in
this pap
er
a
r
e
early wa
rnin
g of icing, forest fire, lighti
ng,
insul
a
tor flashover, co
ndu
ctor gall
oping
and invasi
o
n
. Only som
e
of those function
alities
are
sele
cted a
n
d
used a
c
cording to different area
s in pra
c
tice.
Among natu
r
al disa
sters
that
transmissio
n lines suffe
r,
icing
i
s
one o
f
the
mo
st
se
riou
s threats.
The
extra
weight cau
s
ed
by
icing
will
cau
s
e
seve
re
sa
g [2], and the
n
the di
scha
rge to the
obj
ects
belo
w
wi
ll cau
s
e
the
short
circuit. Me
an
while, the
con
ducto
r g
a
llopi
ng
would
po
ssibly o
c
cur when th
e shrou
ded i
c
e fall
s
o
f
f.
Even more, severe i
c
ing
wi
ll cau
s
e th
e rupture
of
the transmissio
n lines and
th
e tower
collap
s
e.
Traditio
nal m
e
thod
s for ici
ng mo
nitorin
g
an
d corre
s
pondi
ng d
e
vice
s
can
be
divided into
two
categ
o
rie
s
:
One type
u
s
es
se
nsors [
3
, 4] like
g
r
avity sen
s
or,
ca
pa
citan
c
e
se
nsor, vibration
sen
s
o
r
and
climate se
nso
r
, which h
a
s
disa
dvanta
g
e
s
of less info
rmation a
nd
accuracy. Th
e
other u
s
e
s
m
a
chi
ne like de
icing robot [5] and airc
raft [6], which i
s
e
x
pensive a
n
d
not all-wh
eth
e
r.
The pa
pe
r is outlined
as
follows. Secti
on 2
introdu
ces the
co
nst
r
uction
of the
online
early wa
rnin
g
system for o
v
erhea
d tran
smissio
n
line
s
and
schem
es of six pra
c
tical online e
a
rly
warning. Se
ction 3
cho
o
ses the
worst
disa
ster
intro
duced in
sect
ion 2
and
propo
se
s the i
c
ing
image
re
cog
n
i
tion ba
sed
o
n
PDF a
s
wel
l
as
comp
re
h
ensive i
c
in
g
warning
meth
od. The l
a
st t
w
o
se
ction
s
a
r
e
the expe
ri
ment re
sult
s of
the ici
n
g imag
e re
cognition
ba
sed on P
D
F
and
con
c
lu
sio
n
s.
2. Online Early
Warning Sy
stem for O
v
erhead T
r
an
smission Lines
2.1. Cons
tru
c
tion of the
Online Early
Warni
ng Sy
stem for Ov
erhead Tra
n
s
m
ission Line
The propo
se
d online e
a
rly
warning
syst
em for
the transmi
ssion li
nes
con
s
i
s
ts
of tower
terminal
s in
stalled at e
a
ch towe
r al
on
g the tran
smissi
on li
ne
s, fibre o
p
tic com
m
uni
cat
i
on
system
s an
d
the maste
r
station,
which is
sho
w
n i
n
Figu
re 1. T
he lon
g
-di
s
ta
nce tran
smi
s
sion
lines
are
sup
ported
by the
insul
a
tors, which
are
in
stalled at the
transmi
ssion
towe
rs alon
g
the
route. Th
e to
wer
sp
an i
s
g
enerally seve
ral hu
nd
red
meters, and t
he tower te
rm
inals
are i
n
st
alled
at each to
we
r. The local m
onitorin
g
data
sampl
ed by the towe
r terminals a
r
e transmitted to t
he
maste
r
statio
n throug
h the fibre-o
p
tic comm
uni
cat
i
on system
s based o
n
the optical fi
bre
comp
osite ov
erhe
ad g
r
ou
n
d
wire
s (OPG
Ws).
Vis
u
a
l
Im
ag
e
I
n
f
r
ar
ed
Im
a
g
e
Mi
c
r
o
c
l
i
ma
t
e
Da
t
a
I
n
f
r
ar
ed
De
te
c
t
io
n
I
m
ag
e
P
r
oc
es
s
i
ng
Im
a
g
e
C
onp
r
e
ss
I
m
ag
e
A
nal
y
s
is
Li
n
e
Abn
o
r
m
i
t
y
Wa
r
n
i
n
g
S
i
gna
l
Sa
m
p
l
i
n
g
To
w
e
r
Info
r
m
ati
o
n
OP
G
W
C
o
m
m
uni
c
at
i
o
n
M
a
s
t
er
S
t
at
i
o
n
F
aul
t
C
r
i
t
er
ion
Ag
ori
t
h
m
E
a
rl
y
W
a
rn
i
n
g
f
o
r
Ic
i
n
g
E
a
rl
y
W
a
rn
i
n
g
f
o
r
L
i
gh
t
i
ng
E
a
rl
y
W
a
rn
i
n
g
f
o
r
Fo
r
e
s
t
Fi
r
e
E
a
rl
y
W
a
rn
i
n
g
f
o
r
C
o
n
duc
t
o
r
G
a
ll
opin
g
E
a
rl
y
W
a
rn
i
n
g
f
o
r
I
n
s
u
lat
o
r
F
l
as
hove
r
Wa
r
n
in
g
S
i
gna
l
To
w
e
r
Te
r
m
i
n
a
l
E
a
rl
y
W
a
rn
i
n
g
f
o
r
I
n
va
si
o
n
Figure 1. Onli
ne early warn
ing sy
stem fo
r the tran
smi
s
sion lin
es
The to
we
r terminal
con
s
i
s
ts of the
mea
s
ur
ing
unit
s
, proce
s
sing
unit
,
energy supp
lement
unit an
d o
p
tical net
work un
it (O
NU). Th
e
mea
s
u
r
ing
u
n
its a
r
e
u
s
ed
to sampl
e
th
e de
sired
dat
a,
and the processing u
n
it is used to
colle
ct the monito
ring data, p
r
e
-
process the
data, uploa
d the
data and rece
ive the comm
and sent
from
the master
station.
Fibre
-
opti
c
co
mmuni
cation
is re
alize
d
ba
sed
o
n
the fu
sion
spli
cing t
e
ch
niqu
es of
OPGW
and Ethernet
passive optical net
works (EPONs).
The system
can gua
rante
e
the real-ti
m
e
comm
uni
cati
ons bet
wee
n
each to
wer te
rminal
an
d th
e ma
ster stati
on. Th
e fun
c
t
i
onalitie
s of t
h
e
maste
r
statio
n mainly incl
ude data sto
r
age, dat
a a
nalysi
s
, data
querie
s, re
mote monitoring,
remote
control, system m
anag
em
ent a
nd syste
m
co
nfiguratio
ns.
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
PDF Based Icing Im
age Reco
gnition Ap
plied to Onlin
e Early
Wa
rni
ng Syst
em
fo
r… (Xin Yin)
443
2.2. Online Early
Warning
Scheme for
Ov
erhead Transmission
Line
The
co
ntents and
targets of the
ea
rly wa
rn
i
ng fo
r the tran
smi
ssi
on li
ne
s
should
be
identified ba
sed on the
cha
r
acte
ri
stics of transmi
ssion
lines a
nd the
environm
ent. It should me
et
the early wa
rning re
quirem
ents of all tra
n
smi
ssi
on lin
es catastroph
es, incl
uding t
he cata
stro
ph
es
cau
s
e
d
by typhoo
n, lightni
ng, mou
n
tain
fire, tree
gro
w
th an
d en
gi
neeri
ng
con
s
t
r
uctio
n
. In o
r
der
to reali
z
e the
early wa
rnin
g, the sup
porting sy
stem i
s
nee
ded to
obtain the m
onitorin
g
dat
a in
real tim
e
. Th
rough
analy
z
i
ng the
monito
ring
data,
the
potential
th
re
ats can
be di
stingui
sh
ed, and
then early wa
rning a
nd faul
t alarm ca
n b
e
sent out.
After analyzin
g the influen
ces cau
s
ed by
the di
sa
sters, the content
s
of the early warning
for tra
n
smi
s
si
on line
s
i
s
sugge
sted
as
follows: ea
rly wa
rnin
g of t
he ove
r
he
ad
line i
c
ing, t
h
e
mountain
fire
in the
tran
smissi
on
co
rri
dor, the
li
ghti
ng, the i
n
sul
a
tor fla
s
hove
r
, the
con
d
u
c
tor
gallopin
g
a
n
d
inva
sion.
These
conte
n
ts a
r
e
different a
c
co
rdin
g to the
fiel
ds. Th
e d
e
tailed
principle
will be discussed
as follows.
1) Online earl
y warning for
icing
The i
c
ing
of
overhe
ad li
ne
mainly o
c
curs in
the
sout
h of
China,
a
nd the
meteo
r
ologi
cal
con
d
ition
s
[7] whi
c
h l
ead
to icing
ge
ne
rally in
clud
es: 1. the tem
peratu
r
e
and
the eq
uipm
ent
surfa
c
e tem
p
eratu
r
e is b
e
l
o
w 0d
egree centigra
de; 2.
air rel
a
tive hu
midity is above 85%; 3. win
d
spe
ed i
s
g
r
e
a
ter tha
n
1m
/s. Ho
weve
r,
the exa
c
t rel
a
tionship b
e
twee
n existin
g
po
we
r fa
cili
ties
icing
con
d
itions an
d meteo
r
ologi
cal fa
ctors i
s
still a m
a
jor p
r
oble
m
, and the a
c
curacy of deci
d
in
g
wheth
e
r ove
r
head lin
e is
icing i
s
very
low only by
the meteo
r
o
l
ogical co
ndit
i
on. Whil
e th
e
existing meth
ods,
su
ch a
s
weighi
ng me
thod and
met
hod of lea
d
tilt remain to b
e
improved i
n
terms of a
c
cura
cy. Image
information
of ice
wa
rnin
g is favoure
d
by
sch
olars and o
perat
ion
depa
rtment
s f
o
r it
s intuition
and
reliabilit
y. Howe
ver,
existing m
e
th
ods are
usin
g
photo
s
ta
ke
n
at
the sce
ne
an
d de
cidin
g
th
e ice-cove
rin
g
condition
b
y
edgin
g
det
ection
an
d co
ntradi
stinctio
n of
the photo
s
[8
] before an
d
after the icin
g
.
These
m
e
th
ods h
a
ve disadvantag
es o
f
being affect
ed
by the imag
e
re
solutio
n
se
nsitively and
poor ad
apt
ab
ility under
different i
c
e
co
n
d
itions. In
ord
e
r
to improve th
e accu
ra
cy o
f
early warni
ng of t
he tra
n
smi
ssi
on lin
es i
c
ing,
com
p
reh
e
n
s
ive e
a
rly
warning m
e
t
hod b
a
se
d o
n
the analy
s
es of t
he
re
al-time ima
g
e
s
can
be u
s
ed. Be
cau
s
e the
transmissio
n lines ice cov
e
r is likely to
happen o
n
ly in a particul
a
r climate a
n
d
geog
rap
h
ical
environ
ment,
meteorologi
cal co
ndition
s
combi
ned
wi
t
h
the me
cha
n
ic vari
ation
can
be u
s
e
d
as
the early wa
rning criterio
n, in which met
eorol
ogi
ca
l d
a
ta is provide
d
by the micro-mete
orol
ogi
cal
system. Th
e
n
the PDF
-
b
a
se
d imag
e
identificatio
n
prop
osed in
this pa
per i
s
use
d
for
ea
rly
warning. At the sam
e
time, becau
se t
he temper
atu
r
e of the icin
g position i
s
lowe
r than th
e
norm
a
l temp
eratu
r
e, the i
n
frared
data
according
to t
he mo
nitorin
g
po
sition m
a
rked
in the
visible
image
s ca
n b
e
use
d
to improve the accu
racy of the ea
rly warning.
2) Onlin
e ea
rl
y
warnin
g for
forest fire
Overhe
ad
lin
e corrido
r
s a
r
e often
acro
ss the
mo
untai
ns
and
the
wi
lds.Fo
re
st fire
wo
uld
pose a
se
rio
u
s th
reat to
the saf
e
o
peratio
n of
li
nes. T
he tra
d
itional real
-time fore
st fi
re
monitori
ng b
a
se
d on
3S tech
nolo
g
y [9] can
'
t monito
r at all
-
we
ath
e
r a
nd full ti
me condition.
Its
locatio
n
a
c
cu
racy
is not
hi
gh a
nd th
us t
he
small
ra
n
ge of fi
re
ea
rly wa
rning
ca
nnot b
e
in
time.
The novel idea is as
follows
: firs
tly the infrared dete
ction i
s
used
as the initial
criteria, then
th
e
monitori
ng visible ima
g
e
s
are u
s
ed to p
r
elimin
ary de
cide the m
o
u
n
tain fire ba
sed on the im
age
recognitio
n
tech
niqu
e. Finally, combi
n
ed with
infrared imag
e, pe
riphe
ra
l se
cu
rity
environm
ent
and mete
orol
ogical enviro
n
ment data,
the early
wa
rning
ca
n be
made. It ca
n improve the
accuracy of the tran
smi
ssi
on line corrid
ors fo
re
st fire warning.
3) Onlin
e ea
rl
y warnin
g for
lighting
The lig
hting
occu
rs fre
q
uently in m
a
ny ar
e
a
s. T
he
h
uge ele
c
trom
agn
etic
effects,
mechani
cal effects and heating effe
ct
s
of the lightni
ng current
s
will result in
seri
ous devices
damag
e even
if there are
many lightnin
g
prote
c
tion
device
s
. No
w the method
s of lighting e
a
rly
warning a
r
e
mostly based
on the overvoltage moni
to
ring cau
s
ed b
y
lightning or the overcurre
n
ts
monitori
ng. Other method
s are use
d
to locate
the lig
hting are
a
ba
sed on the li
ghting se
nsors
and GPS. T
he sugge
ste
d
method
ab
out the lighti
ng ea
rly wa
rning contain
s
two aspe
cts: a)
based on the
lighting mon
i
toring data a
nd the loca
l meteorologi
cal
data,
the frequ
en
cy of the
lighting
can
be obtai
ned.
Whe
n
the f
r
e
quen
cy is
off-limits, the
o
p
timal control
mea
s
ureme
n
ts
sho
u
ld be taken, su
ch as redu
cing the transmi
ss
ion lo
ad; b) the cu
rrent in the overhe
ad groun
d
wire
s is m
onit
o
red to lo
cate
the lighting locatio
n
.
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 13, No. 3, March 2
015 : 441 – 4
4
8
444
4) Onlin
e ea
rl
y warnin
g for
insul
a
tor flashover
In the sno
w
melting or ot
her severe weat
her
con
d
itions, the solid
, liquid and the ga
s-
borne parti
cl
es will cause
the decr
ease of the insul
a
tor
electric
strength. Then
a flashover will
occur on the insulator [10] and
the
power outage
will appear.
At present, t
he early
warning
method
s
of the in
sul
a
tor flash
o
ver are
mostly b
a
sed
on
the l
e
a
k
a
ge
cu
rre
nt m
onitorin
g
or t
h
e
equivalent
sa
lt depo
sit de
n
s
ity monito
rin
g
, the reli
abili
ty of whi
c
h i
s
not e
noug
h.
Method
s b
a
sed
on the local
meteorologi
cal data and the method
s based on the
sound
wave
are sug
g
e
s
ted to
improve the a
c
cura
cy of the early wa
rni
ng of the insu
lator flash
o
ve
r.
5) Onlin
e ea
rl
y warnin
g for
con
d
u
c
tor gal
loping
The condu
ct
or gallo
ping
[11] would o
c
cur
u
nde
r the co
ndition
s of the stro
ng win
d
(typhoon
), tra
n
smi
ssi
on li
n
e
s i
c
in
g etc.
The
con
d
u
c
to
r gall
opin
g
ca
n cau
s
e inte
rpha
se fla
s
ho
ver,
fitting damage, etc. Then
line tr
ip and blackout
will occur.
The accordi
ng sensors should be
use
d
for ea
rly
wa
rnin
g of th
e cond
ucto
r g
a
llopi
ng
with
t
he transmissi
on lin
e mo
del
. The
se
nsors
inclu
de the vibration
sen
s
o
r
s, di
spla
cem
ent
sen
s
o
r
s,
ac
cele
rat
i
on sen
s
o
r
s,
et
c.
6) Onlin
e ea
rl
y warnin
g for
invasio
n
Tree
g
r
o
w
th
and
the
p
a
ssing
ma
ch
inerie
s may
ca
use th
e
sh
ort
ci
rcui
t of the
transmissio
n lines. Th
en the po
wer
out
age o
r
the
transmi
ssion li
ne dama
ge
will occu
r. But the
early warni
n
g
of the invasi
on ha
s not b
een g
r
eat
ly a
ddre
s
sed. Th
e sug
g
e
s
ted
method
s for t
h
e
early
wa
rnin
g
of the
inva
si
on
contai
n two a
s
pe
cts:
firstly the
co
nd
uctor
sag
wo
uld b
e
id
entified
based on the
operating pa
rameters of
the tran
smissio
n
lines a
nd th
e local mo
nitoring d
a
ta, su
ch
as the tem
perature, then th
e video surve
illance is
u
s
e
d
for the e
a
rl
y warni
ng of i
n
vasio
n
. For t
h
e
gro
w
th of the
tree b
e
lo
w th
e co
ndu
cto
r
, the meth
od
s b
a
se
d on
the g
r
owth
spee
d
can
be
used
as
the initial criteria.
3. Icing Image Reco
gnitio
n
for Ov
erhead Po
w
e
r Tr
ansmission
Lines
In this sectio
n, icing imag
e recognitio
n
for overhe
ad
power tran
sm
issi
on line
s
b
a
se
d on
PDF is explai
ned an
d a co
mpre
hen
sive
icing
warning
method for o
v
erhea
d tran
smissio
n
line
is
prop
osed. Th
e discu
ssi
on
are ma
de in two sub-ch
apt
ers.
3.1. Icing Image Recogni
tion Based
on Probabilit
y
Density
Function
In orde
r to si
mplify the proce
s
sing, we
usua
lly u
s
e
gray imag
e. And we e
s
ta
blish the
coo
r
din
a
te
system, in
which defin
e d
o
wn di
re
cti
on
a
s
X
po
sitive
axis, the
rig
h
t
dire
ction
a
s
Y
positive axis.
Then a ray image of M
line, N line ca
n be
expresse
d a
s
(0
,
0
)
(
0
,
1
)
(0
,
-
1
)
(1
,
0
)
(
1
,
1)
(
1
,
N
-
1
)
,
(
M
-1
,
0
)
(
M
-
1
,
1
)
(
M
-1
,
N
-1
)
ff
f
N
ff
f
ff
f
fx
y
()
(1)
Whe
r
e
(
,
)
[0
,
2
1
]
k
fx
y
,
[0
,
M
1
]
x
,
(,
)
x
yS
S
,
y[
0
,
N
1
]
.The
gray level is
2
k
L
，
the
dynamic ra
ng
e of ima
ge i
s
[0
,
L
1
]
. The hi
ghe
r
resol
u
tion the
spatial
and
g
r
ay have the
i
m
age i
s
mo
r
e
de
lica
t
e.
Then th
e PDF [12] for a di
screte va
ria
b
le gray scal
e
i
is
defined
a
s
Equation
(2
)
and th
e
CDF
(cumula
t
ive distributi
on func
tion
) i
s
defined a
s
Equation (3
),
where the t
h
re
shol
d of th
e
icing target is
(,
)
sf
x
y
t
.
(x,
y
)
1
1
(i
)
f
pi
MN
(2)
0
()
(
i
)
f
i
H
fp
(3)
The ste
p
s of the PDF ba
se
d icing ima
g
e
reco
gnition a
r
e as follo
ws.
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
PDF Based Icing Im
age Reco
gnition Ap
plied to Onlin
e Early
Wa
rni
n
g Syst
em
fo
r… (Xin Yin)
445
1) Im
age pre
p
ro
ce
ssi
ng:
the image
preproc
e
s
s
i
ng inc
l
udes
RGB to gray process
i
ng and
median
filter.
Conve
r
ting
RGB to gray re
sults
f
r
om th
a
t
the gray im
age n
e
ed
s le
ss
sto
r
ag
e an
d
can b
e
pro
c
e
s
sed qui
ckly with low pe
rfo
r
man
c
e h
a
rd
ware. What’
s
more, the det
ail of gray image
is sufficient f
o
r ici
ng d
e
ci
sion. The m
e
d
i
an filteri
ng i
s
a method
of
ran
k
ing th
e
pixel gray val
u
e
within a sli
d
i
ng win
d
o
w
a
nd usi
ng the
median in
st
ead of the central g
r
ay p
i
xel value. The
method i
s
a
nonlin
ear
sm
oothing m
e
th
od, and it ca
n dedu
ce im
pulse interfe
r
ence of salt
and
pepp
er n
o
ise
effectively. It can
also pro
t
ect t
he ed
ge
from dimin
g
effectively. Con
s
id
erin
g the
pra
c
tical p
e
rf
orma
nce of the median filte
r
, 7×7
ma
skin
g median filte
r
wa
s u
s
ed in
this pape
r.
(a) T
he preproce
s
sed ima
g
e
(b) T
he pa
rtition blo
c
ks
Figure 2. Dia
g
ram of spa
c
e-ba
se
d imag
e segm
entati
o
n
2) Partitionin
g
:
The image
is partitione
d into P×Q b
l
ocks in
spa
c
e sho
w
n in F
i
gure 2,
block is d
enot
ed as
symbol
A(i, j), where
[1
,
]
,
[
1
,
]
iP
j
Q
.
3) Cal
c
ulatio
n:
The
imag
e
re
co
gnition
need
s to
det
e
ct the
comp
arability of
gray scale
distrib
u
tion
of two a
d
ja
cent
blocks. Supp
ose th
at
the
segmentatio
n
has
bee
n d
o
n
e
in a
rul
e
an
d
the image is s
e
gmented
into m×n differ
ent
blocks (
k×k
)
whi
c
h
are
not ov
erlap
ped. T
h
en
cumul
a
tive grayscale di
stri
bution
hi
stog
ram, whi
c
h
is t
he
Cumul
a
tive Di
strib
u
tion
Fun
c
tion i
n
the
sense
of probability, can
be calcul
ated according to (2)
and
(3). Actually, it
is the gray level
distrib
u
tion fu
nction (CDF).
4) Icin
g criterion:
To de
cid
e
the area of
icing, the
si
milarity must
be an
alysed
amon
g
neigh
bou
r’s
pixels
blo
c
ks of a
pictu
r
e, or th
e
sa
me blo
c
ks i
n
pi
cture
s
. T
he a
c
cumul
a
tive
histog
ram
fun
c
tion
cu
rve of
adja
c
ent
blo
c
ks
1
()
Hf
an
d
2
()
Hf
are
carrie
d out
as sh
own in
Fig
u
re
3. If the gray
distrib
u
tion
of the two
curv
es i
s
sim
ila
r, it is illu
strated
that the two
curves bel
ong
to
the sam
e
are
a
. For the growin
g and
co
mbining of a
r
eas, it is ne
cessary to giv
e
a stand
a
rd
to
measure the
simila
rity. The most
com
m
only used
standard is Ko
lmogo
rov-Sm
imov crite
r
ia
and
Smoothed
-Dif
feren
c
e criteri
a
, as sh
own in Equation (4
) and Equ
a
tio
n
(5).
Figure 3. The
grayscale di
stribution
featu
r
es of the n
e
ighbo
ring a
r
e
a
s
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 13, No. 3, March 2
015 : 441 – 4
4
8
446
12
1
1
()
()
ma
x
f
Hf
H
f
regi
on
mergi
n
g
stay
t
Th
e
a
m
e
T
s
(
4
)
12
2
2
()
()
f
Hf
H
f
re
gi
on
m
e
r
g
i
n
g
s
ta
y
t
h
e
sa
m
e
T
T
(
5
)
If we ch
oo
se
as the i
n
itial sea
r
ch poi
nt, we can
sea
r
ch th
e eig
h
t adja
c
ent blo
c
ks
a
s
sho
w
n in Fig
u
re 2(b). If the block satisfies the
K-S
criteria, it wi
ll be combin
ed with the start
block. T
hen
we
ch
oo
se th
e ne
w
blo
c
k
as th
e
cent
ral
blo
c
k an
d
se
arch it
s ei
ght
adja
c
ent
blo
c
ks
until no blocks me
et the crite
r
ia. This
pape
r uses
Kolmogo
rov-Simimov crit
eria to judge
the
simila
rity of blocks.
3.2. Compre
hensiv
e Icin
g Warning M
e
thod
for Ov
erhead T
r
an
smission Line
In orde
r to
improve th
e accu
ra
cy of t
he tran
smissi
on lin
e
s
ice ea
rly warning,
comp
re
hen
si
ve
ea
rly
warning method
based on
t
h
e
analy
s
e
s
of
the
real-tim
e image
s can
be
use
d
. Becau
s
e the
tran
smissi
on lin
es ice
cove
r is
likely to ha
pp
en only in
a
particula
r cli
m
ate
and geo
grap
hical envi
r
on
ment, meteorologi
cal co
ndi
tions co
mbin
ed with the mech
ani
c varia
t
ion
can
be
u
s
ed
as the
ea
rly wa
rnin
g
crit
erion,
in
whi
c
h
meteo
r
olo
g
ical
data
is
provide
d
by
the
micro-m
e
teorologi
cal sy
ste
m
. Then the
PDF-b
a
s
ed i
m
age id
entification p
r
o
p
o
s
ed in this
pa
per
is used for e
a
rly wa
rning.
At the same time, bec
au
se the tempe
r
ature of the icing p
o
sition
is
lowe
r tha
n
t
he n
o
rm
al te
mperature, t
he infr
ared
data a
c
cordi
ng to th
e m
onitorin
g
p
o
sition
marked in th
e visible ima
ges
can
be u
s
ed to imp
r
o
v
e the accuracy of the ea
rly warning.
The
s
p
ec
ific
pr
oc
es
se
s
ar
e
s
h
ow
n
in
F
i
g
u
r
e
4
.
Lo
ad
i
m
a
g
e
P
r
e
p
r
o
ce
ss
i
n
g
I
m
ag
e
re
c
ogn
i
t
i
o
n
te
ch
n
o
l
o
g
y
M
o
ni
t
o
r
i
n
g
da
t
a
a
c
q
u
i
s
it
io
n
J
udg
e w
h
e
t
h
e
r
ic
i
n
g
e
x
is
t
s
St
a
r
t
W
het
h
e
r t
h
e
m
e
t
e
or
a
l
og
i
c
a
l
d
a
t
a
m
e
et
t
h
e
i
c
e
c
ond
i
t
i
ons
I
n
f
r
a
r
e
d
i
m
ag
e
t
e
m
p
er
a
t
ure
m
e
s
u
rem
ent
d
a
t
a
No
Ye
s
C
o
m
p
r
e
h
ens
i
v
e pr
e
-
wa
r
n
in
g
ic
i
n
g
ju
dgm
e
n
t
Ye
s
No
Figure 4. The
comp
reh
e
n
s
i
v
e early wa
rn
ing method fo
r tran
smi
ssio
n
line icin
g
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
PDF Based Icing Im
age Reco
gnition Ap
plied to Onlin
e Early
Wa
rni
ng Syst
em
fo
r… (Xin Yin)
447
4. Experiment Re
sults
To verify the
icing ima
ge reco
gnition
ba
sed on PDF, experim
ents
have been m
ade as
follows. In Fi
gure
5, the i
c
ing
tran
smi
s
sion
li
ne
i
s
e
x
tracted by
calcul
ating CDF
(Cumul
ative
Distri
bution F
unctio
n
) of a
d
jacent
blo
c
ks having be
e
n
divided. We divide Figu
re 5(a) into 3
×
3
blocks
a
n
d
e
a
ch
blo
c
k ha
s
a
resolution
of 200×26
6. Thre
e adja
c
e
n
t blocks hav
e been labell
ed
as
2,
1
A
,
2,
2
A
, and
3,
1
A
in F
i
gure
5
(
a) by
sp
atial
segm
entati
on.
Usi
ng the
alg
o
rit
h
m of p
a
rtitio
ned
matrix, the CDF of three bl
ocks a
r
e
sho
w
n in Figu
re
5(b
)
.
(a) di
agram o
f
the adjace
n
t blocks
(b)
CDF
of block A(2,1
)
, A(2,2) and A(3,1)
Figure 5. The
partitioning a
nd CDF cal
c
u
l
ation re
sults
The previou
s
experime
n
t based on 3
×
3 blocks
of segmentatio
n
can ju
dge the
area, but
roug
hly. To
g
e
t the a
c
cu
ra
te area
whe
r
e the i
c
in
g i
s
, there
are m
o
re
se
gme
n
tation bl
ocks
are
made in Fig
u
re 6.
Figure 6. Partitioning into
more bl
ocks and u
s
ing i
c
in
g crite
r
ion
In Figu
re 6, t
he ima
ge i
s
divided into
8
×
11
blo
c
ks
a
nd the
se
arching u
n
it is
a
4×4 a
r
ea i
n
each block. T
he starte
r of the se
arch alg
o
rithm can be
one arbitr
ary
point identified in block A
2,
1
and A
2,2
as sh
own in Fig
u
re
5(a).
5. Conclusio
n
The ba
si
c task and
prin
cipl
e of the early
warning i
s
propo
sed in the
pape
r. Base
on this,
the sup
portin
g
system a
n
d
the
hierarchical
con
s
tru
c
tion of the
early wa
rni
n
g techn
o
logy
is
discu
s
sed. T
he real-tim
e
image
of t
he tra
n
smission line
an
d
other relate
d information
is
colle
cted by the remote m
easurin
g and
controlli
ng
terminal
s install
ed at each to
wer. Th
e ma
ss
informatio
n is tran
sferre
d
timely by p
a
ssiv
e optica
l
fibre co
mm
unication techniqu
e ba
se
d on
50
100
150
20
0
25
0
0
0.2
0.4
0.6
0.8
1
Gr
ey
L
e
v
e
l
P
r
o
b
a
b
ilit
y
A
(
2,
2)
A
(
2,
1)
A
(
3,
1)
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 13, No. 3, March 2
015 : 441 – 4
4
8
448
OPGW. An
effective earl
y
warni
ng
mech
ani
sm
i
s
e
s
tabli
s
he
d by synthe
tically analyzing
monitori
ng d
a
t
a. Based
on
the su
ppo
rtin
g syste
m
, th
e
mission a
nd
prin
ciple
of si
x early wa
rni
n
g
are di
scusse
d.
Based
on
th
e characte
ri
stic of i
c
ing
e
a
rl
y warning,
the P
D
F-ba
sed
imag
e
re
cog
n
ition
techn
o
logy i
s
appli
ed fo
r icing
early
warning.
T
he syntheti
c
icing criteri
on is e
s
tabli
s
he
d
combi
ned
with the micro
c
l
i
mate, mech
anics mea
s
u
r
ement and in
frare
d
therm
a
l image, whi
c
h
can
enh
an
ce
the accu
ra
cy and real-ti
m
e pe
rform
a
nce
of the e
a
rly wa
rni
ng.
The
simulati
on
results
ba
sed
on the
actu
al
icing
imag
e from th
e
sce
n
e
prove the v
a
lidity of the
prop
osed P
D
F
-
based imag
e recognitio
n
m
e
thod.
Ackn
o
w
l
e
dg
ements
This
resea
r
ch
was
su
ppo
rted by the Nat
i
onal Natural Scien
c
e Fo
u
ndation of
Ch
ina (No.
5127
7084
).
Referen
ces
[1]
Qu Jun-su
o. A securit
y
co
mmunicati
on s
c
heme for R
e
alT
i
me EPON.
T
E
LKOMNIKA Indon
esia
n
Journ
a
l of Elec
trical Eng
i
ne
eri
n
g
. 201
2; 10(5)
: 1073-1
0
7
6
.
[2]
F
a
rzane
h M, Savad
jiev K. Sta
t
istical an
al
ys
is
of
field d
a
ta fo
r precip
itatio
n i
c
ing accr
etio
n
on ov
erhe
ad
po
w
e
r lin
es.
IEEE Trans on Power Deliv
ery.
200
5; 20(2): 10
80-1
087.
[3]
Ma GM, Li
C
R
, Jia
ng J,
et
al. A
n
o
vel
optica
l
l
o
a
d
c
e
ll
use
d
in
ici
ng m
onit
o
rin
g
on
ov
erhe
ad
transmissio
n
li
nes.
Col
d
Reg
i
ons Scie
nce a
nd T
e
chn
o
l
ogy
. 2012; 71: 67-
72. 201
2.
[4]
Gan Yan, Liu
Cha
o
, Ruan Ji
ang
jun, et al. Anal
ys
is of Weak Positi
on in
Overhead li
ne
under He
a
v
y
Icing Co
nditi
o
n
b
y
F
i
n
i
te Eleme
n
t Method.
T
E
LKOMNIKA Indones
ian Jo
urna
l of Electrica
l
Engi
neer
in
g
. 2014; 12(
7): 501
1-50
21.
[5]
Yang YM, W
ang YN, Yu
a
n
XF
, et al. Neura
l
net
w
o
rk-bas
ed self
-learn
i
n
g
cont
rol for po
w
e
r
transmissio
n
li
ne de
icin
g rob
o
t.
Neural C
o
mputin
g & Appl
ic
ations
. 20
13; 2
2
(5): 969-
98
6.
[6]
Liu
X, Yan
g
JJ, Z
hu M, et al.
System D
e
sig
n
of T
r
ansmissi
on Li
ne Icin
g Detectio
n Bas
ed on R
e
mot
e
Control Model
Aircraft
. Proceedi
ngs of 2
0
1
2
2nd Inter
nati
o
nal
C
onfer
enc
e on C
o
mp
uter
Scienc
e And
Net
w
ork T
e
chn
o
lo
g
y
. 2
0
1
2
; 424-4
28.
[7]
Savad
jiev K, F
a
rzan
eh M. M
ode
lin
g of ici
n
g
an
d ice s
h
e
ddi
ng o
n
ov
er
hea
d po
w
e
r l
i
nes b
a
sed
on
statistical an
al
ysis of meteorol
ogic
a
l dat
a.
IEEE Trans on Power Deliv
ery.
200
4; 19(2): 71
5-72
1.
[8]
Maciej B. Di
git
a
l ima
ge pr
oce
ssing i
n
meas
ureme
n
t of ice
thickness o
n
po
w
e
r transmi
ssion l
i
nes:
a
roug
h set appr
oach. W
i
nn
ipe
g
: Univers
i
t
y
of
Manitob
a
; 200
2.
[9]
W
ang Ji
a, F
eng Z
h
o
ngke,
Study on F
o
r
e
st F
i
re Spre
ad Mo
del B
a
s
ed o
n
MODIS Imag
e
.2
010
Internatio
na
l C
onfere
n
ce o
n
Multimed
ia
T
e
chno
log
y
(ICMT
)
. 2010; 1(5): 2
9
-31.
[10]
F
ontana E, Ma
rtins JF
, Oliveira SC, et al. Se
nsor
Net
w
o
r
k for Monitor
i
n
g
the
State of Po
ll
ution
of Hig
h-
Voltag
e Insul
a
tors Via Satel
lit
e.
IEEE Transactions on Power Delivery
. 2
0
1
2
; 27(2): 95
3-9
62.
[11]
Ma GM, Li
CR, Jian
g J, et
al.
A Passiv
e
Op
ti
cal F
i
b
e
r An
e
m
ometer for W
i
nd S
p
e
ed M
e
asurem
ent o
n
High-V
o
lta
g
e
Overhea
d
T
r
ansmission
Lin
e
s.
IEEE Transactions on Instr
u
mentation
and
Measur
e
m
ent
. 201
2; 61(2): 53
9-54
4.
[12]
H W
ang. Boun
ded D
y
n
a
mic S
t
ochastic Distri
butio
ns Mod
e
ll
i
ng an
d Co
ntrol
,
Springer-V
erl
ag. 199
9.
Evaluation Warning : The document was created with Spire.PDF for Python.