TELKOM
NIKA Indonesia
n
Journal of
Electrical En
gineering
Vol.12, No.6, Jun
e
201
4, pp. 4671 ~ 4
6
7
8
DOI: 10.115
9
1
/telkomni
ka.
v
12i6.545
0
4671
Re
cei
v
ed
De
cem
ber 2
9
, 2013; Re
vi
sed
March 6, 201
4; Acce
pted
March 20, 20
14
Water Level Int
e
lligent System of Data Acquisition and
Early Warning
Lei Li*, Fenglian Cao, Zh
eng Li
Schoo
l of Ph
ys
ics and El
ectro
n
ic Electric
al E
ngi
neer
in
g/Hua
i
yin Norm
al U
n
i
v
ersit
y
111 C
h
a
ngj
ian
g
Roa
d
Hu
aia
n
Jiangs
u ROC/+
86-05
17- 8
3
5
251
80
*Corres
p
o
ndi
n
g
author, e-ma
i
l
: home0
51
6@
126.com
A
b
st
r
a
ct
In order to imp
r
ove w
o
rk efficiency of hydro
m
et
ric statio
n, guar
ante
e
the accuracy a
nd ti
me
lin
ess
of w
a
ter level data acqu
isitio
n, process w
a
ter level d
a
ta
in a t
i
mely
man
ner
and
mak
e
earl
y
w
a
rning of risk,
mo
bil
e
dev
ices
are pro
pose
d
as coll
ectio
n
, send
ing
and
ear
ly w
a
rning ter
m
inal
of w
a
ter le
vel data, a
nd t
h
e
upp
er co
mput
er as th
e rec
e
iving
an
d pr
oc
essin
g
te
r
m
in
a
l
of w
a
ter lev
e
l data. T
h
e w
a
ter lev
e
l
acc
e
ss
alg
o
rith
m is
st
udi
ed, and
the
strippi
ng meth
od is used
to o
b
tain
the
hi
gh-
accuracy
w
a
ter
lev
e
l infor
m
ati
on.
T
he i
m
age
thi
n
nin
g
a
l
g
o
rith
m
is a
ppl
ie
d to
o
p
timi
z
e
, s
o
as
to red
u
ce
the
compl
e
xity of
the
alg
o
rith
m. T
h
e
progr
a
m
min
g
i
s
abl
e to c
ontr
o
l re
al-ti
m
e
dat
a sen
d
i
ng, w
h
i
c
h rea
l
i
z
e
s
the
data
exch
ang
e betw
e
e
n
ser
v
e
r
termi
nal
an
d c
o
mmunic
a
tio
n
mo
du
le. Micro
s
oft Access da
tabase
is a
ppl
i
ed to th
e op
er
ation
of w
a
ter l
e
vel
infor
m
ation and the design
of m
a
n-
m
a
chine interfac
e. Experim
ental
study shows that this system
c
an
accurate
ly acq
u
ire i
n
for
m
ati
o
n in re
al ti
me,
und
ertake
stati
s
tical an
alys
is of data
w
i
th many functi
ons s
u
c
h
as ear
ly w
a
rni
ng, effective
l
y
improv
e the
w
o
rk
efficienc
y of w
a
ter lev
e
l
mo
nitori
ng
and
i
m
pr
ove t
h
e
ma
na
ge
me
nt level of hydr
ol
o
g
ical i
n
d
u
stry.
Ke
y
w
ords
:
water level
m
o
nitoring system
, di
ssection
m
e
thod, im
age thinning,
prewitt, m
i
c
r
osoft access
Copy
right
©
2014 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
Wate
r level
monitori
ng i
s
the ba
sis
of the water
con
s
erva
ncy
wo
rk. Witho
u
t water level
monitori
ng, there would be
no wate
r co
n
s
erva
ncy co
n
s
tru
c
tion o
r
other infr
as
tructure. Currently,
throug
h re
al-t
ime monitori
n
g
on the wat
e
r level of rivers a
nd reservoirs, wate
r l
e
vel monitori
ng
room
s
can
gi
ve out the
ea
rly wa
rnin
g o
n
the p
o
ss
ibl
e
flood
s a
nd
other
disaste
r
s. Howeve
r,
most
of the dom
estic hydrologi
cal stat
ions
still use the artificial m
onitori
ng method at
present, whi
c
h
not only
po
se
s a
threat to
the
safety of t
e
st p
e
rso
nnel
but
also resu
lts in
rel
a
tively larg
e e
r
ror i
n
monitori
ng d
a
ta that
can
not meet th
e
req
u
ir
ement
s of
accu
ra
cy and timeli
n
e
ss [1]. With
the
developm
ent
of com
pute
r
technol
ogy and n
e
tw
o
r
k commu
nication technol
ogy, wate
r l
e
vel
detectio
n
ha
s also devel
op
ed fro
m
si
ngl
e field d
e
tect
i
on to m
u
lti-sit
e
re
mote
real
-time d
e
tectio
n,
with the
cro
s
s-regi
onal,
all
-
we
athe
r, rea
l
-time
ch
a
r
a
c
t
e
risti
cs.
With
the d
e
velop
m
ent of m
obi
le
comm
uni
cati
on technol
og
y and the
po
pulari
z
atio
n o
f
person
a
l m
obile te
rminal
, real-tim
e water
level informat
ion ca
n be
se
nt to the serv
er an
d
the st
aff who can keep ab
rea
s
t
of its informat
ion
whe
r
eve
r
the
y
are. Furthe
rmore, the water le
vel ch
ange
s withi
n
a ce
rtain pe
ri
od ca
n al
so
be
analyzed thro
ugh lon
g
-time
reco
rd
s of the softwa
r
e,
in
orde
r to make the timely prepa
re
dne
ss.
The ba
si
c d
e
sig
n
structu
r
e of
water l
e
vel monito
ri
ng sy
stem i
s
basi
c
ally th
e sa
me.
Ho
wever, va
rious produ
cts have th
eir o
w
n
ch
ara
c
te
ri
stics a
c
co
rdin
g to the
ma
rket dema
nd
a
nd
their e
m
ph
ases
on
differen
t
asp
e
ct
s, whi
c
h
mainly
ref
e
rs to th
at th
ey have
different a
c
cesse
s
to
water l
e
vel in
formation. Th
e existing p
r
odu
cts
on th
e market mai
n
ly have the followin
g
de
si
gn
thought
s in te
rms
of the a
c
ce
ss to
wate
r level info
rm
a
t
ion: float type, ultr
a
s
oni
c t
y
pe, lase
r typ
e
,
came
ra
colle
ction type, et
c., and
their
respe
c
ti
ve a
d
vantage
s a
nd di
sadva
n
tage
s a
s
well
as
appli
c
ation
s
are a
s
follows: float type
is widely u
s
e
d
in rese
rvoi
rs an
d rivers becau
se of
its
advantages such as
dat
a collection
stability, easy insta
llation and low
cost, but it has the
disa
dvantag
e
su
ch
a
s
complex m
a
chining,
sh
ort
life an
d
a
sig
n
ifica
n
t
decrea
s
e
in
the
measurement
accu
ra
cy
when th
er
e
i
s
se
dimentati
on; ultras
oni
c type
ma
kes
use of
the
aero
a
cou
s
tic
measurement
and control
prin
ciple, a
n
d
obtains th
e
water l
e
vel b
y
calcul
ating
the
time differen
c
e between transmi
ssion
a
nd re
ce
pt
ion
of ultrasoni
c
wave
s. It has an out
standi
ng
advantag
e and ca
n obtai
n a very
goo
d measurem
ent result. Curr
ently, a few ultra
s
oni
c
typ
e
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 12, No. 6, June 20
14: 4671 – 4
678
4672
prod
uct
s
hav
e been u
s
ed
in the water level moni
toring. Ho
weve
r, it also has a disadvant
age
that it is more obviously i
n
fluen
ced by the
weathe
r. Whe
n
stro
ng
wind
s and bi
g waves a
p
p
ear,
an error m
e
a
s
ureme
n
t will
be cau
s
e
d
and the
data
can not be
discrimi
nated
and se
nt ba
ck,
whi
c
h
will in
crease the
wo
rkloa
d
of the
p
r
oces
so
r; the
comp
uter te
chnolo
g
y deve
l
opment m
a
kes
it possible th
at the intellig
ent image
se
nso
r
can
b
e
i
n
trodu
ce
d in
the wate
r lev
e
l dete
c
tion.
The
intelligent im
age
se
nsor n
o
t only h
a
s a
high
er de
gre
e
of
re
cog
n
ition, but
also
can p
r
o
c
e
s
s d
a
ta
intelligently a
nd se
nd b
a
ck the on-the
-
spot situation
i
n
the form of
image o
r
vide
o, facilitating the
analysi
s
of st
aff.
2. Rese
arch
on Wa
ter Le
v
e
l Information Acquisi
tion Algorith
m
Wate
r level i
n
formatio
n a
c
qui
sition al
g
o
rithm refers
to an algo
rith
m of rega
rdi
n
g water
gaug
e imag
e
placed in th
e wate
r a
s
the re
se
arch
obje
c
t, and
mainly usi
n
g
the scale v
a
lue
whe
r
e
wate
r
line di
sa
ppea
rs in th
e
wat
e
r
gaug
e fo
r
referen
c
e. T
h
e auxilia
ry lin
e is ad
opted
to
calib
rate
and
adju
s
t the
relative po
sition bet
wee
n
came
ra
and
water ga
uge.
Du
ring th
e i
n
itial
installatio
n
, the po
sition
s
of wate
r ga
u
ge an
d the
mobile p
hon
e
may not be
parall
e
l, or
water
gaug
e is not in the middle
of the image, whi
c
h nee
ds t
he third auxili
ary line to adj
ust the cam
e
ra
angle an
d the
relative posit
ion between
came
ra a
nd water g
aug
e in orde
r to en
sure the imag
ing
quality. The value
s
and
Let
ter "E" scal
e
in the wate
r g
auge pl
ate are importa
nt referen
c
e
s
wh
en
the water lev
e
l is read.
Ho
wever,
the va
lues o
r
sc
ale
s
are bl
urred
subje
c
t to
a va
riety of exte
rn
al
factors such as stai
ned fo
ot plate. The
n
, this al
go
rithm will di
scard the digital informatio
n in
the
water g
aug
e plate and fin
d
the relative
position wh
ere the wate
rline di
sap
p
e
a
rs in the water
gaug
e panel
to calcul
ate
the water l
e
vel . The algorithm flo
w
is sho
w
n in
Figure 1: af
ter
colle
cting
wat
e
r ga
uge i
m
a
ge, use imag
e pro
c
e
s
si
ng
method to
ge
t the scale va
lue where wa
ter
line disappe
a
r
s
so a
s
to ob
tain accurate water level inf
o
rmatio
n.
Figure 1. Obtain Wate
r Le
vel Informatio
n Algorithm F
l
ow
Whe
n
u
s
in
g
this al
go
rithm to p
r
o
c
e
s
s the
ima
ge,
the im
age
shall
first be
grayed,
becau
se the
r
e is a lot of
colo
r informa
t
ion
on color
image, which
has hi
ghe
r requireme
nts
for
system m
e
m
o
ry and th
e
pro
c
e
ssi
ng
speed
of pr
o
c
essor,
while
the graye
d
color im
age
can
better a
dapt t
o
the
statu
s
o
f
smalle
r
sma
r
t
pho
ne me
mory.
Unde
r
t
he weig
hted averag
e
m
e
th
od,
R, G and B are en
do
wed
with differen
t
weights a
c
cordin
g to the different req
u
irem
ents of the
importa
nce. Becau
s
e the
human eye
s
have a diffe
rent sen
s
itivity to red, green and blu
e
, the
graying fo
rmu
l
a is obtain
e
d
as follows after testing:
0.29
9
0
.587
0.11
Gr
a
y
R
G
B
(1)
The main p
u
rpose of imag
e prep
ro
ce
ssi
ng is
to elimi
nate the noi
se interferen
ce, which
is achieved
by adopting
the spatial fil
t
ering
meth
o
d
. This meth
od ca
n be d
i
vided into two
asp
e
ct
s: smo
o
thing a
nd sharp
enin
g
a
c
cording to th
e different fu
nction
s. Smo
o
thing
can b
e
achi
eved by a
lo
w-p
a
ss
filt
er,
aimi
ng at removin
g
blu
r
red
detail
s
o
r
co
nne
cting
small inte
rvals in
the target pi
cture togeth
e
r before
extra
c
ting la
rg
e
r
target
s in o
r
d
e
r to a
c
hieve
the purpose
of
eliminating t
he noi
se i
n
terferen
ce; sharp
enin
g
re
fers to
ma
king u
s
e of h
i
gh-p
a
ss filte
r
to
enha
nce the
informatio
n o
f
blurred
deta
ils. The i
n
te
rf
eren
ce
of tin
y
line type inf
o
rmatio
n in t
h
e
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Wate
r Le
vel Intelligent System
of Data Acqui
sition a
n
d
Early Warnin
g (Lei Li
)
4673
water g
aug
e plate ca
n be
eliminated a
n
d
the Lette
r “E” scale an
d nume
r
ical informatio
n ca
n
be
retaine
d
by condu
cting the
smoothin
g
proce
s
si
ng of i
m
age at first to remove
sm
all details.
The m
o
st
co
mmonly u
s
ed
smo
o
thing
filter
is divide
d
into two types: lin
ear sm
oothin
g
filter and
n
onl
inear smoothi
ng filter,
of which
t
he li
nea
r
smoothi
ng fi
lter a
dopt
s th
e field
averag
e
method to set
templates. F
o
r exampl
e, take the
co
efficient of 3*3 t
e
mplate a
s
1.
But in this way,
some details
in the image
will be bl
urred out
w
hen the noi
se i
s
removed. A m
e
dian filter
can be
adopte
d
to
condu
ct the
smoothing
p
r
o
c
e
ssi
ng
of
i
m
age i
n
o
r
d
e
r to
achiev
e the
purpo
se of
saving the i
m
age detail
s
simultane
ou
sly with elim
i
nating the n
o
ise. Thi
s
p
u
rpo
s
e i
s
m
a
inly
achi
eved by
arrangi
ng the
gray value
s
of the
co
rrespondi
ng pixel
s
un
der t
he t
e
mplate
s fro
m
small to big,
taking the middle value
and assigni
ng it to the
pixel of the
center of the
corre
s
p
ondin
g
template.
1
X
、
2
X
、、、
n
X
rep
r
e
s
ent the gray values i
n
the
template. Th
ey are a
r
ran
ged by
size
as
follows
:
1
i
X
、
2
i
X
、、、
in
X
. The median
Y
is cal
c
u
l
ated by the Formul
a 2:
2
12
12
e
v
e
n
num
be
r
1
2
odd num
be
r
,
....
..
in
in
in
XX
n
nX
n
YM
i
d
X
X
X
(2)
Isolated n
o
ise points
ca
n be elimin
ated
by
using the
above filters. Becau
s
e th
e scal
e
line
o
n
wate
r
g
auge
p
a
n
e
l
in
th
e water gau
ge picture ha
s obviou
s
stro
ke
i
n
form
atio
n,
informatio
n o
n
the wate
r line po
sition can be obtai
ne
d throu
gh the
edge dete
c
ti
on wh
en ma
king
the image
an
alysis. As the
scale lin
es
a
r
e no
rmally
d
i
splaye
d in a
hori
z
ontal
wa
y, the scal
e
i
s
rich
in the
ed
ge of the
hori
z
ontal
dire
cti
on. The
edg
e
informatio
n i
n
the h
o
ri
zont
al direction
of
the
cha
r
a
c
ter
can
be extracted
by using the
operator in
th
e hori
z
ontal d
i
rectio
n of Prewitt operator or
Sobel op
erator [2-5]. The
pre
s
ent al
go
ri
thm use
s
th
e
Prewitt op
era
t
or. The tem
p
late is
sho
w
n
in
Figure 2.
(a)
(b)
Figure 2. Pre
w
itt Operator
Algorithm
Those in Fig
u
r
e 2(a) a
nd Fi
gure
2(b
)
respective
ly refe
r to the ope
ra
tors of the h
o
rizo
ntal
and ve
rtical t
e
mplate
s.
When
usi
ng th
e ho
ri
zontal
operator to e
x
tract the
ed
ge info
rmatio
n on
the scale line,
the edge in the vertical di
rect
ion
can b
e
ignore
d
, to redu
ce interfe
r
ence.
In this
algo
rithm, the hi
gh
pre
c
isi
on
of scale
is
re
quired. Th
e a
c
cu
racy
of po
siti
oning
the
intermediate
point will
hav
e a large i
m
pact on t
he performance of
the algorithm
.
Furthermore
,
the scale
on
water ga
uge
board h
a
s it
s inhe
rent
ch
a
r
acte
ri
stics, for exam
ple:
scale lin
es are
evenly di
strib
u
ted, an
d th
e same
scal
e line
is
ho
ri
zontal.
T
he refinement proce
s
s can
first
con
s
id
er a
ddi
ng p
r
ior
kn
owledge
of wate
r ga
uge
plate
to give the refined a
nd p
r
e
c
ise p
o
sitio
n
in
g
to the wi
der
edge
s. In the
algo
rithm di
ssectio
n
met
hod i
s
u
s
e
d
mainly to di
sse
c
t edg
e pi
xels
from outsi
de to insid
e
.
The main
ste
p
s of disse
c
ti
on are a
s
foll
ows:
1) Fo
reg
r
o
u
n
d
image val
u
e of bina
rize
d scale li
ne i
s
set as
255,
and the
ba
ckgroun
d
value i
s
0. T
a
ke
a poi
nt of t
he verti
c
al
ce
nter lin
e, with
prio
r o
r
l
a
ter
pixel 0-255
a
s
the
bo
unda
ry
and the initial
point;
2) Se
arch
fro
m
the i
n
itial p
o
int to th
e di
rect
ion
of
255
pixel, del
ete
the initial
poi
nt wh
en
N
foreg
r
o
und points su
ccessively
emerg
e
, and elimin
ate the noise interferen
ce;
3)
sea
r
ch d
o
w
n
w
ard fro
m
the initial p
o
int in the v
e
rtical
directi
on to d
e
term
ine the
positio
n of the uppe
r and l
o
we
r edg
e po
ints;
-1
1
-1
1
-1
1
1
1
1
-1
-1
-1
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 12, No. 6, June 20
14: 4671 – 4
678
4674
4) Set the
midpoint of t
he upp
er a
n
d
lowe
r ed
g
e
points
as
the skeleton
baseline
positio
n;
5) Kee
p
the
hori
z
ontal
ce
nter poi
nts
wi
th the
erro
r of
less than
3; con
n
e
c
t ba
se
point to
determi
ne its
positio
n.
The study
o
n
scale
line po
sition sh
ows that
skeleto
n
positio
n a
nd t
he u
ppe
r a
n
d
lower
boun
dari
e
s
of the scale
lin
e are sub
s
ta
ntially parall
e
l
to ea
ch oth
e
r
. The
width
of the scal
e li
ne
remai
n
s u
n
ch
ange
d, so the
image refinin
g
algo
rithm can acco
rdi
ngl
y be optimize
d
, redu
cin
g
the
compl
e
xity of
the algorith
m
[6]. The spe
c
i
f
ic pro
c
e
s
ses
are a
s
follows:
1) Sca
n
line b
y
line, and mark the
cu
rre
nt
foregroun
d
as the to-b
e-determi
ned p
o
int;
2) M
a
rk the l
a
st p
o
int on
the
corre
s
po
n
d
ing
colu
mn
of the to-be-d
e
termin
ed p
o
i
n
t the a
s
backg
rou
nd
point, label t
he next point
as the fo
reg
r
oun
d poi
nt, and then
det
ermin
e
the u
pper
bord
e
r of th
e scal
e o
n
t
h
is
point, a
n
d
put th
e p
o
s
ition i
n
form
ation into
th
e corre
s
po
nd
ing
memory;
3)
Clu
s
ter th
e
po
sition info
rmation o
n
th
e up
per
bo
rd
er to
derive
the p
r
e
c
ise lo
cation
of
the up
per bo
rder, a
nd th
en
su
btra
ct 1/2
of the
scale li
ne
width in
the o
r
igin
al im
age to
get th
e
exact po
sition
of the water
gaug
e scale.
As illustrated in Figure
3(b), the preci
s
e
position of water l
e
vel
scale m
a
rk
can
be
obtaine
d a
s
l
ong a
s
th
e nu
mber
of iterati
ons
of
op
erati
ons i
s
red
u
ce
d and
the o
p
e
r
ation
sp
eed i
s
improve
d
after optimizi
ng the algo
rithm.
Although the
positio
ns
of scale
mark a
n
d
simila
r
scal
e mark
noise
have be
en
o
b
tained
throug
h the i
n
itial pro
c
e
s
si
ng of image,
the interfe
r
en
ce of the figu
res
with
strai
ght lines
ca
n
not
be eliminate
d
.
Therefo
r
e, short straight l
i
nes
a
nd figu
res m
u
st be
removed to re
duce the noi
se
interferen
ce.
The scale
ma
rk
on
the wat
e
r gaug
e scal
e
pl
ate
ge
ne
rally ha
s a
fixed le
ngth
whi
c
h
ca
n
be u
s
ed a
s
a
referen
c
e to
define a too
sho
r
t or too l
ong
straig
ht line a
s
noi
se a
nd then a
naly
z
e
length inform
ation of the straight line in t
he image,
thu
s
exclu
d
ing th
e straig
ht line
interfere
n
ce.
By adopting the conn
ecte
d comp
one
nt anal
ysis m
e
thod, the indepen
den
ce o
f
each
straig
ht line
can
be dete
r
mined, the n
u
mbe
r
of
co
nne
cted regi
ons
ca
n be
cou
n
ted, an
d
the
statistics of l
ength info
rm
ation ca
n be
con
d
u
c
ted in
orde
r to elimi
nate the interferen
ce of scale
marks.
The
so-called
conn
ected
compo
nent a
nalys
i
s
method
refe
rs to extra
c
ti
ng the
si
ze
a
nd
positio
n info
rmation of th
e conn
ecte
d
regi
on
s
by
che
c
king
the
co
nne
ctivity between
va
riou
s
pixels an
d their neig
hbo
rin
g
pixels.
In pra
c
tical a
pplication
s
, the co
nne
cted
compo
nent
analysi
s
met
hod is divid
e
d
into two
types of m
e
thod
s: pixel l
abele
d
an
alysis meth
o
d
a
nd run
co
nn
ectivity analysis metho
d
.
The
algorith
m
ad
o
p
ts the
pixel l
abele
d
an
alysis meth
o
d
to
mark th
e current
scann
ed
pixels th
ro
ug
h
scanni
ng a b
i
nary image f
r
om left to right and from
top to botto
m. It is requi
red to che
c
k the
con
n
e
c
tivity betwe
en th
e
curre
n
t sca
n
ned
pixels
an
d
the previou
s
scann
ed ne
ighbo
ring
pix
e
ls.
This alg
o
rith
m will take th
e water g
aug
e scale ma
rk as the analy
s
is o
b
je
ct. If
the width of the
water g
aug
e
scale
mark is 1, the
nu
mb
er
of
iteration
s
can
b
e
red
u
ce
d, and the
operatin
g sp
eed
can b
e
faste
r
. As illustrate
d
in Figure 3
(
c), can b
e
too long or to
o sh
ort strai
ght lin
e noise ca
n b
e
remove
d after the co
nne
cted com
pon
en
t analysis.
The
scale in t
he water
gau
ge plate
ca
n
be divide
d int
o
two
part
s
, o
f
which the
ri
ght pa
rt
only ha
s
unif
o
rm
scale
ma
rks
whil
e the
left part
have
both
unifo
rm
scale
line
s
a
nd
scale val
u
es.
The followi
ng
part is to co
ndu
ct statistical identif
icatio
n mainly aiming at
scale m
a
rks. The
sca
l
e
values may
be treate
d
a
s
scale lin
es in the su
b
s
e
quent p
r
ocessing if t
hey are not re
mov
ed,
whi
c
h will affect
the
measurem
ent
result. Therefore,
it
is necessary to
desi
gn an algorithm
to
eliminate th
e
interfe
r
en
ce
of value
s
in
the
su
b
s
e
que
nt pro
c
e
s
sing
. The
desi
g
n
thoug
ht of t
h
is
algorith
m
is to dire
ctly proj
ect the right p
a
rt onto
the le
ft part in orde
r to remove the origi
nal scale
lines in the lef
t
area. The p
r
oce
s
s is a
s
follows:
1) Ma
ke u
s
e
of the symme
tric rel
a
tion to
get the positi
on of middle
marking;
2) Co
ndu
ct a cluste
r analy
s
is of the interval
betwe
en lines in the ri
ght part acco
rding to
a set thre
shol
d value;
3) P
r
oje
c
t th
e po
sition
of
scale
ma
rk
in the
rig
h
t
part
onto th
e
left pa
rt in
orde
r to
eliminate the
interferen
ce o
f
the or
iginal
scale in the p
r
oje
c
tion a
r
ea
;
4) T
h
e
effect after
remo
ving the i
n
te
rfere
n
ce of
short
strai
ght
lines an
d fig
u
re
s i
s
illustrated in Figure 3(d).
An effective
wate
r g
aug
e scal
e ma
rk
can
be
o
b
tained
ba
si
cally after th
e ab
ove
pro
c
e
ssi
ng. b
u
t it is ne
ce
ssary
to
contin
ue p
r
o
c
e
ssi
n
g
the
scale
m
a
rk d
ue to
a
highe
r a
c
cu
ra
cy
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Wate
r Le
vel Intelligent System
of Data Acqui
sition a
n
d
Early Warnin
g (Lei Li
)
4675
requi
rem
ent, and then q
u
a
n
tify the analysis result
to dra
w
a ne
w staff gauge.
As illustrated
in
Figure 3
(
e), t
h
is p
ape
r o
b
t
ains th
e inte
rval len
g
th b
y
clu
s
terin
g
t
he water ga
uge
scale
m
a
rk
intervals, thu
s
obtaini
ng a
n
effective wa
ter gaug
e sca
l
e mark.
Dra
w
a
ne
w staff gauge, an
d use it to find t
he minim
u
m scale m
a
rk of water
gau
ge, thus
obtainin
g
the
water level. The po
sition
whe
r
e water l
e
vel disa
ppe
ars in Fi
gure 3(f) is the
wa
ter
level informa
t
ion which i
s
ne
ce
ssary t
o
obtai
n. Thi
s
al
go
rithm t
a
ke
s l
o
o
k
ing
for th
e p
o
sit
i
on
whe
r
e
water
level disa
ppe
ars
as
a trai
n of thought.
The key lies in establi
s
hi
ng a ne
w
wa
ter
gaug
e mo
del
throug
h the
i
m
age
pro
c
e
s
sing
algo
rith
m for
pre
c
i
s
e
po
sitioning
o
f
the scale
m
a
rks
in order to redu
ce e
r
rors,
thus m
eetin
g the a
c
cura
cy re
quireme
nt req
u
ire
d
b
y
the pra
c
tical
appli
c
ation.
(a)
(b)
(c
)
(d)
(e)
(f)
Figure 3. Algorithm Pro
c
e
ssi
ng
Aim at algo
rithm a
c
cura
cy and
stab
ility to perform the
wat
e
r level recognition
experim
ent, with the re
sul
t
s sho
w
n i
n
F
i
gure
4.
Take
124 water ga
uge pi
ctures
of different water
levels und
er t
he labo
ratory’
s
man
-
mad
e
scene a
s
test
library, and the test re
sult
s are: un
der t
h
e
con
d
ition of
t
he
all
o
wable
error
of ±
1,
124 pictu
r
e
s
are
corre
c
tly identified, a
n
d
the
re
cog
n
ition
rate is
100%.
Figure 4. Experime
n
tal Re
sults
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 12, No. 6, June 20
14: 4671 – 4
678
4676
3. Sy
stem design
Wate
r level monitori
ng sy
stem is
com
posed
of the
control se
rv
er an
d the di
spe
r
sed
water level
monitori
ng p
o
ints. The se
rver is mai
n
l
y
respo
n
si
ble
for pro
c
e
ssi
ng and an
alyzing
water l
e
vel informatio
n, and giving ea
rly warning
on ha
za
rd in
formation; ea
ch water lev
e
l
monitori
ng p
o
int mainly use
s
the imag
e analysi
s
sy
stem develo
p
ed on the mobile termin
al
to
analyze wat
e
r ga
uge pi
cture
s
a
nd
get the
wat
e
r level info
rmation, an
d
adopts
GS
M
comm
uni
cati
ons n
e
two
r
k and serve
r
to transmit
the informa
t
ion [7]. Server analy
z
e
s
and
pro
c
e
s
ses d
a
t
a, with the system structu
r
e sho
w
n in Fi
gure 5.
Figure 5. Wat
e
r Level Mo
ni
to
ring System
Structure ch
art
Information collectio
n mod
u
le take
s sm
art the
mobile
terminal as the platform, requiri
ng
WP op
eratin
g syste
m
, du
al co
re
1.5G
HZ p
r
o
c
e
s
so
r an
d ab
ove, 1GB RAM
a
nd 32
GB RO
M,
came
ra
(e
qui
pped
with th
e flashli
ght)
resol
u
tion of
1280
* 76
8 p
i
xels, whi
c
h
sho
u
ld h
a
ve
the
stron
g
ca
pabi
lities of data pro
c
e
ssi
ng a
nd imagin
g
.
3.1. Design
of Ac
quisitio
n
Module
Acco
rdi
ng to
the appli
c
at
ion re
quirem
ents,
the m
odule
shall
have the fu
nction
of
receiving
and
se
ndin
g
info
rmation. Its
pri
n
cipl
e
i
s
that host co
mpute
r
via RS-232
asyn
chrono
u
s
seri
al
comm
u
n
icatio
n u
s
e
s
AT comma
n
d
control
SIM300
GSM mo
dule to
comp
lete the
wh
ole
function. SIM
300 i
s
eq
uipp
ed with
a
sta
ndard RS
-2
3
2
se
rial i
n
terf
ace,
whi
c
h vi
a the data
ca
ble
can b
e
co
nne
cted directly to a comp
uter
seri
al po
rt.
A compl
e
te water l
e
vel reco
rd
sho
u
ld
contai
n the
informatio
n a
r
ea
cod
e
, wa
ter level
time, real
-tim
e water level
heig
h
t, an
d
the text fo
rmat is u
s
ed
to improve t
he effici
en
cy of
encodin
g
an
d tran
smissio
n
. The abov
e informatio
n
is repl
aced
by letters an
d numbe
rs. Its
comm
and format is sh
own
in Table 1.
Table 1. Orde
r Form
at
time
monitoring station
Water level information
tw
elve characte
r
s
four c
haracte
rs Seven
character
s
An entire co
mmand
cont
ains 23
cha
r
acters. The
written o
r
de
r of water le
vel time
informatio
n inclu
d
e
s
year, month, day, hour,
minu
te, seco
nd.
Monitori
ng p
o
int informati
on
identifier ID
shall add th
e
area
co
de. Water level in
fo
rmation i
s
co
nstituted by t
he identifie
r
WL
plus water lev
e
l. After recei
v
ing commands, the
receiver will recover
data
and archive
accordi
ng
to rules fo
r writing instru
ctio
ns. The follo
wi
ng i
s
the co
mmand form
at used in the
system.
Information
from tra
n
smitting te
rminal
i
s
1
304
101
20
906ID05
W
L0
0097,
co
nsi
s
t
i
ng of
a
total of 23 ch
ara
c
ters. Tim
e
format: 130
4101
2090
6 is obtained by the tran
smittin
g
terminal
wh
en
obtainin
g
the
system
time, whi
c
h rep
r
e
s
ents 12:09:
0
6
on
10 Ap
ril,
2013; l
o
catio
n
form
at: ID0
5
is
the location
identifier; 0
5
re
pre
s
e
n
ts the po
sitio
n
marke
d
b
y
the tran
smitting termi
nal;
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Wate
r Le
vel Intelligent System
of Data Acqui
sition a
n
d
Early Warnin
g (Lei Li
)
4677
informatio
n fo
rmat: WL
000
97 is th
e wat
e
r level
id
enti
f
ier, and 0
0
0
97 re
present
s the
water l
e
ve
l
of 9.7 meters.
The tran
smitting terminal
first extract
s
t
he fr
o
n
t 12
chara
c
te
rs to
obtain th
e inf
o
rmatio
n
about
water
level time, then obtai
ns t
he monito
ri
n
g
point p
o
siti
on informatio
n from the
rear
cha
r
a
c
ter
of ID, and fin
a
lly rea
d
s
out th
e wate
r level
data after
WL
, thus
compl
e
tely transmitti
ng
and receivin
g the water level inform
ation [8]. In pra
c
tical u
s
e, the ch
ang
e in water l
e
vel
informatio
n u
s
ually i
s
not
p
a
rticul
ar
ly sig
n
ificant
i
n
a short
time.
In
order to reduce the cos
t,
s
h
ort
messag
e ca
n
be sent on
ce
every 10 min
u
tes.
3.2. Ser
v
er Soft
w
a
re
Desi
gn
Server
program is mainly
to
re
ceive, store data, con
t
rol
the com
m
unication
m
o
dule
a
n
d
make the d
a
ta analysi
s
. System softwa
r
e ca
n be divi
ded into thre
e
levels, as sh
own in Fig
u
re
6.
Figure 6. Upp
e
r Co
mpute
r
Software Flo
w
Huma
n-com
p
uter inte
ra
ctio
n platform i
s
to
serve
cu
st
omers, an
d p
r
ovide a
wid
e
range
of menu
inte
rface
a
c
cordi
ng to
user
n
eed
s,
an
d
convert the
u
s
er ne
ed
s in
to the m
a
chine
langu
age
to
sen
d
to th
e
data p
r
o
c
e
s
sing laye
r; d
a
ta p
r
o
c
e
ssin
g
layer is mai
n
ly to de
al
with
instru
ction
s
from huma
n
-compute
r
interaction pl
atform and send
to the hard
w
are d
r
iver lay
e
r,
and p
r
o
c
e
ss
data fed ba
ck from th
e up
per a
nd lo
we
r layers; ha
rd
ware d
r
iver l
a
yer controls the
comm
uni
cati
on modul
e through the
seri
al port.
With the
chara
c
te
risti
c
s of low h
a
rd
wa
re
req
u
irem
ents,
easy o
p
e
r
ati
on, low
developm
ent co
st, desktop
database
ru
n
s
on a perso
n
a
l comp
uter a
nd is pop
ular
with individua
l
use
r
s.
Co
mm
on de
sktop
d
a
taba
se p
r
o
d
u
cts in
cl
ude
Paradx, Fox
B
ase, Vi
sual
FoxPro, Acce
ss
and
so on. M
i
cro
s
oft Acce
ss i
s
a
relatio
nal datab
as
e. Acce
ss data
base is
save
d in the form
of
file, and the file extension i
s
MDB. Serv
er datab
ase use
s
Acce
ss,
mainly base
d
on the following
advantages:
a singl
e storage mo
de, good object-ori
ented
compat
ibility. Access database used
in this syste
m
can fully meet the requi
re
ments.
Server soft
ware
is de
sign
ed to
process m
a
in
ly a
ccordin
g to
the
data.
Data
source
s of
the serve
r
po
rt mainly inclu
de su
ch three
parts
a
s
seri
al port data, d
a
taba
se data,
and user inp
u
t
data. Accordi
ng to the data
source, the
server fun
c
tion is
mainly c
l
as
s
i
fied into three parts
:
1) Data exch
ange b
e
twe
e
n
se
rver a
nd
comm
uni
cati
on mod
u
le
s. This fun
c
tion
is mainly
to pro
g
ra
m serial
port, u
s
e the
seri
al
port to a
c
hi
e
v
e comm
uni
cation bet
wee
n
ha
rd
ware
and
human
-comp
u
ter inte
ractio
n platform, re
ceive
comma
nds
and d
a
ta
transmitted from the hum
a
n
-
comp
uter i
n
te
ractio
n platfo
rm for ap
pro
p
r
iate tre
a
tme
n
t, and tran
smit feedba
ck
data ba
ck. T
h
is
function i
s
m
a
inly to comp
lete thre
e fun
c
tion
s of u
p
lo
ading
data, d
o
wnl
oadin
g
d
a
ta and
setting
para
m
eter.
2) Mani
pulat
e data in the databa
se. The role
of da
tabase is to store d
a
ta. The se
rve
r
prog
ram
ne
e
d
s to
man
age
the d
a
taba
se
. This fu
nctio
n
in
clude
s
m
odificatio
n
s to
the d
a
ta, qu
ery,
input an
d out
put. First, it is ne
ce
ssa
r
y to co
nne
ct m
anag
ement
p
r
ocedu
re
s a
n
d
datab
ases.
In
VC6.0 d
e
velo
pment e
n
viro
nment, this p
aper u
s
e
s
A
D
O m
e
thod
to a
c
hieve
co
nne
ction b
e
twee
n
manag
eme
n
t prog
ram a
n
d
databa
se. Di
fferent from the tradition
al
layers of dat
a obje
c
ts, ADO
can
be i
nde
pend
ently created, an
d
can
create
a "Co
nne
ctio
n" obje
c
t, b
u
t multiple
and
indep
ende
nt "Re
c
ordset" object
s
ca
n use it [9-10].
3)
De
sign
of huma
n
-com
puter i
n
tera
ct
ion inte
rface. The m
a
in fu
nction
of wat
e
r level
monitori
ng
sy
stem i
s
to
m
onitor th
e
wa
ter level, a
n
d
ca
n b
e
divid
ed into
two
functio
n
s such a
s
the real-tim
e
alarm and
water level i
n
formatio
n analysi
s
. The operator can
pre-d
e
si
gn
water
level wa
rning
value, and g
i
ve out the alarm when
da
ta value at the monitori
ng
point is la
rge
r
than the wa
rning value; water level info
rmation a
nal
y
s
is i
s
mainly to inquire the histori
c
al
wat
e
r
level informat
ion stored in
the database and use t
he curve to illustrate the g
eneral dire
cti
on.
This
pap
er p
r
ovide
s
two
ways of
curv
e re
present
at
ion. On
e i
s
t
he
curve
tre
n
d
of
water le
ve
l
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 12, No. 6, June 20
14: 4671 – 4
678
4678
informatio
n within the cu
rrent one ho
ur; the other
is to inquire th
e cu
rve tren
d of each h
o
u
r
within on
e da
y.
4. Sy
stem Realizatio
n
Data
storage
types i
n
th
e data
b
a
s
e
are
sh
own
belo
w
. The
entire i
n
form
ation i
s
comp
osed
of
the lo
catio
n
informatio
n, time info
rmati
on, an
d
wate
r level
inform
ation. When
t
h
e
serve
r
re
ceiv
es
a m
e
ssa
g
e
, it will fi
rst
judge
the
wa
ter level,
and
will
give o
u
t the al
arm
if
it
exceeds t
he
warning water level. All the data
w
ill be
stored in the
dat
abase. T
h
e curve
data
are
read from the
databa
se.
(a)
Wate
r lev
e
l cha
r
t of a day
(b)
Wate
r level cha
r
t before one ho
ur
Figure 7. Wat
e
r Level Te
nd
ency Chart
Server program
can
ma
nually
sele
ct
the
q
u
e
r
y
way, an
d p
r
ovide the
water l
e
vel
informatio
n of each ho
ur wi
thin one day, as sho
w
n in Figure 7 (a) a
nd one ho
ur real-time wate
r
level informa
t
ion as sho
w
n Figu
re 7
(
b) b
e
lo
w, and within th
e displ
a
y area ca
n dra
w
the
corre
s
p
ondin
g
cu
rve. The
system p
r
ov
ides the
real
-time alarm fu
nction. You
can pre-set th
e
alarm level.
Whe
n
the water level mo
nitoring val
u
e excee
d
s th
e alarm val
u
e, early wa
rn
ing
informatio
n column
will sh
ow the i
n
formation ab
out
water l
e
vel
positio
n, time and
water l
e
vel
,
and give out the alarm.
The
system
u
s
e
s
the
GSM
comm
uni
cati
on m
odule
to
transmit info
rmation o
b
tain
ed from
data acq
u
isiti
on modul
es
of each mo
ni
toring stati
o
n
to water level monitorin
g
center d
a
tab
a
se,
and a
dopt
s t
he mo
nitorin
g
software fo
r re
al-time
m
onitorin
g
on
data of mo
nitoring
statio
ns. It
has the
fun
c
tions
of
warning,
settlem
ent & a
c
cu
m
u
lation, statistical
a
nalysi
s
and
so on to
achi
eve the scene p
a
ram
e
ter colle
ctio
n, real-tim
e data co
mmu
nicatio
n
co
ntrol of the co
ntrol
room
an
d th
e hydrologi
ca
l station,
whi
c
h
ca
n
effe
ctively improve
the a
u
tomati
on a
nd
co
ntrol
level of water level monito
ring, an
d hel
p to im
prove
the mana
ge
ment level of the hydrolo
g
i
cal
indu
stry.
Referen
ces
[1]
Kun W
a
ng,
Xi
nzhi
Ch
en.
Re
search
on
D
y
n
a
mic
Leve
l
of Ground
w
a
ter Monitori
ng
S
y
s
t
em
Base
d on
GPRS.
Compu
t
er Measure
m
e
n
t & Control
. 2
011; 19(
2): 263
-265.
[2]
Bai
x
i
an Z
o
u, Ran
Z
h
a
ng,
Jun M
i
ao. I
m
prove
d
Pre
w
itt M
e
tho
d
for Image
Ed
ge D
e
tectio
n
.
Microel
ectron
ic
s & Comp
uter.
201
3; 30(5): 23
-26.
[3]
Lame
i
Xu, Ch
unj
u Yi. Ap
plic
ation
of Gr
e
y
S
y
stem T
heor
y i
n
Imag
e Ed
ge D
e
tectio
n.
Geomatics a
n
d
Information Sci
ence of W
u
h
a
n
University
. 20
12; 56(8): 9
29-
931.
[4]
X
u
e-lan Zhang, Jun Li, Pei-hong
Xin.
Paper
Defect
Detection Based on
Pe
w
i
tt Operator and
Mathematic
al Morph
o
lo
g
y
.
P
aper a
nd Pa
pe
r Making.
20
12
; 31(8): 25-28.
[5]
Xi
e Z
hao
li, Bai Ying
jie. F
P
GA
implem
entati
o
n
of pre
w
itt ed
g
e
detectio
n
an
d edg
e thinn
i
n
g
.
Applic
ati
o
n
of Electronic T
e
chn
i
qu
e
. 201
0; 35(6): 39-4
1
.
[6]
Shi Sh
ao
qia
n
g
.
An Improved
T
h
inning A
l
g
o
ri
thm of Chi
nes
e Ideo
gra
ph I
m
age.
C
o
mput
er T
e
chn
o
lo
g
y
and D
e
vel
o
p
m
ent.
200
7; 17(9
)
: 88-91.
[7]
Yuan-s
o
n
g
Pe
ng, D
uan
Pe
n
g
. W
a
ter
Leve
l
Moni
tori
ng an
d Measur
emen
t Sy
stem
of Riv
er
Bas
ed
o
n
Z
i
gBee.
Instru
me
nt T
e
chni
qu
e and Se
nsor.
201
2; 42(7): 68
-70.
[8]
Lei L
i
, Hua
b
a
o
Che
n
, An
yi
W
ang. Desi
gn
an
d
Implem
entatio
n of Mo
bile T
e
rmina
l Market T
r
acing
Service Bas
ed
on SMS.
Mode
rn Electron
ics T
e
chni
que.
2
0
08; 15(1
3
): 73-
75.
[9]
Z
H
AO Huimin
g, YU Ming
hui
. Design
a
nd
r
ealiz
atio
n of da
tabase of
w
a
t
e
r and sed
i
ment
characters
o
f
Che
ngl
ing
ji co
nflu
x reac
h. En
gin
eeri
ng Jo
ur
nal
of W
uha
n
Univers
i
t
y
. 2
0
0
7
; 51(1): 58-6
0
.
[10]
Jing C
ao, Jia
n
f
eng Z
h
i. Impl
ementati
on of
Data Sh
arin
g Bet
w
e
e
n
La
b
W
indo
w
s
/C
VI and Micr
osoft
Access.
Industrial Co
ntrol C
o
mp
uter.
200
9; 22(1
0
): 58-6
3
.
Evaluation Warning : The document was created with Spire.PDF for Python.