Indonesi
an
Journa
l
of
El
ect
ri
cal Engineer
ing
an
d
Comp
ut
er
Scie
nce
Vo
l.
12
,
No.
3
,
Decem
ber
201
8
, p
p.
1349
~
1357
IS
S
N: 25
02
-
4752, DO
I: 10
.11
591/ijeecs
.v1
2
.i
3
.pp
1349
-
1357
1349
Journ
al h
om
e
page
:
http:
//
ia
es
core.c
om/j
ourn
als/i
ndex.
ph
p/ij
eecs
Improve
ment
o
f Au
to
-
T
racking
Mobi
le Robot
b
ased on
HSI
C
olor
M
odel
Suresh
Sun
daraj
oo
1
,
Ah
m
ad Sh
ah
ri
z
an
Ab
d
ul Gh
an
i
2
1
DN
C
Autom
at
ion
(M) Sdn.
Bhd
.
,
Ta
m
an
Industr
i
Mera
n
ti Ja
y
a
,
4
7100
Puchong,
Sela
ngor
,
Ma
lay
si
a
2
Facul
t
y
of
Man
ufa
ct
ur
ing
En
gin
ee
ring
,
Univ
ersiti
Mal
a
y
s
ia Pahang
,
26600
Pekan
,
Pahang
,
M
al
a
ysia
Art
ic
le
In
f
o
ABSTR
A
CT
Art
ic
le
history:
Re
cei
ved
A
ug
9,
2018
Re
vised Oct
2, 2018
Accepte
d Oct
30, 201
8
Auto
tra
ck
ing
m
obil
e
robo
t
is
a
devi
c
e
th
at
abl
e
to
detec
t
and
track
a
ta
rge
t
.
For
an
aut
o
tr
acking
device
,
the
m
ost
cru
ci
al
p
ar
t
of
the
s
y
s
te
m
i
s
the
object
ide
nti
f
icati
on
an
d
tracki
ng
of
th
e
m
oving
t
arg
ets
.
In
orde
r
to
i
m
prove
th
e
ac
cur
acy
of
id
en
ti
ficat
ion
of
object
in
diff
ere
n
t
ill
um
ina
ti
on
and
b
ac
kground
condi
ti
ons
,
th
e
i
m
ple
m
ent
at
ion
o
f
HS
I
col
or
m
ode
l
is
used
in
image
proc
essing
al
gorit
hm
.
In
th
i
s
proje
ct
HS
I
-
ba
sed
col
or
enha
n
c
ement
a
lgori
thm
were
used
for
object
ide
n
tification.
Th
is
is
because
HS
I
p
ara
m
et
er
ar
e
m
o
re
stab
le
in
diffe
ren
t
l
ight
a
nd
bac
kground
condi
ti
ons
,
so
i
t
is
sel
ec
t
ed
as
the
m
ai
n
par
amete
rs
of
th
is
s
y
stem.
Pix
y
CMU
ca
m
5
is
us
ed
as
th
e
vision
sensor
whil
e
Arduino
Uno
as
the
m
ai
n
m
ic
roc
ontrol
ler
th
at
co
ntrol
s
a
ll
the
inp
ut
and
outpu
t
o
f
the
dev
ic
e
.
M
ore
over
,
two
se
r
vo
m
otors
were
used
to
cont
rol
t
he
pan
-
ti
l
t
m
ovement
of
th
e
v
ision
sensor
.
Expe
riment
al
re
sults
demons
tra
t
e
tha
t
whe
n
HS
I
col
or
-
base
d
fil
t
eri
ng
al
gor
ithm
is
appl
i
ed
to
visual
tracki
ng
it
improve
s
the
ac
cur
acy
a
nd
stabi
l
ity
of
tracki
ng
und
e
r
the
condition
of
var
y
i
ng
bright
ness,
or
e
ven
in
the
low
-
li
ght
-
l
evel
envi
r
onm
ent
.
B
esides
tha
t
,
thi
s
al
gorit
hm
al
so
pre
vent
s
tracki
n
g
loss
due
to
o
bje
c
t
col
or
app
ea
rs
in
the
bac
kground.
Ke
yw
or
d
s
:
C
olo
r
trac
king
HS
I
co
l
our
-
bas
ed fil
te
ring
PixyC
MUcam
5
M
ob
il
e r
obot
Copyright
©
201
8
Instit
ut
e
o
f Ad
vanc
ed
Engi
n
ee
r
ing
and
S
cienc
e
.
Al
l
rights re
serv
ed
.
Corres
pond
in
g
Aut
h
or
:
Ah
m
ad
S
hahri
zan
Abd
ul Gha
ni,
Faculty
of Ma
nufactu
rin
g
E
nginee
rin
g,
U
niv
ersit
i M
al
ay
sia
Pah
a
ng,
26600 Pe
kan,
Paha
ng,
Ma
la
ysi
a
.
Em
a
il
:
sh
ahr
iz
an@um
p.
ed
u.m
y
1.
INTROD
U
CTION
The
c
ontrib
ution
of
r
obots
is
rap
i
dly
increa
s
ing
day
by
day
as
r
obots
are
sta
rting
to
substi
tute
hum
ans
in
e
ver
yday
ta
sk
s.
At
t
he sa
m
e
tim
e
robo
ts
c
an
do
ta
sks
tha
t
would
ei
the
r
be i
m
po
ssible
f
or
hum
an bei
ng
to
do
or
it
wo
uld
tak
e a longer
ti
m
e
f
or h
im
o
r
he
r t
o
com
plete
it
.
Th
us
, ro
bots m
ake hum
an’
s
w
ork
easi
er a
nd
m
or
e
pro
du
ct
ive
.
T
he
functi
on
of
r
obot
ca
n
be
f
urt
her
increa
sed
by
gi
ving
visi
on
t
o
the
r
obot
.
Colo
r
vision
base
d
m
ob
il
e
ro
bot
pa
th
tracki
ng
is
pro
po
se
d
by
Luo
et
al
.
[1]
.
In
this
pro
po
se
d
syst
e
m
,
path
f
ol
lowing
al
gorit
hm
is
pro
po
se
d
base
d
on
t
he
data
e
xtracted
f
ro
m
HS
I
(
Hue,
Sat
ur
at
io
n,
i
ntensi
ty
)
colo
r
m
od
e
l
us
in
g
fu
zzy
c
on
t
ro
l.
As
the
im
age
i
s
captu
re
d,
it
will
be
co
nvert
ed
int
o
H
SI
c
ol
or
m
od
el
an
d
us
in
g
se
gm
entat
ion
m
et
ho
d,
the
path
is
extrac
te
d
fro
m
the
i
m
age
us
ing
optim
al
th
reshold
i
n
H
SI
m
od
ule.
The
pa
th
sk
el
et
on
is
then
e
xtracted
us
i
ng
sk
el
et
on e
xtrac
ti
on
m
et
ho
d. I
n ad
diti
on
,
fuzz
y con
t
ro
l i
s
us
e
d for
rob
ot p
at
h
trac
king c
ontrol.
So
a
ns
et
al
.
[2]
us
ed
a
da
ptiv
e
colo
r
th
res
hold
m
et
ho
d
w
hich
is
e
qu
i
pp
ed
wit
hin
a
m
ob
il
e
r
obot
t
o
detect
a
nd
fo
ll
ow
a
par
ti
cular
col
or
of
a
n
ob
je
ct
.
As t
he
r
obot
e
quip
ped
with
m
echan
ic
al
arm
, t
he m
ob
il
e
r
obot
is
able
to
pic
k
the
ta
rg
et
obj
e
ct
.
Ba
sed
on
t
he
pro
po
se
d
m
eth
od,
t
he
ca
ptu
r
ed
im
ages
are
trans
ferre
d
to
a
c
olor
thres
ho
l
ding
al
gorithm
to
detect
the
ta
rg
et
.
T
he
noise
s
ar
e
fi
lt
ered
out
an
d
f
inall
y
the
colored
ob
j
ect
is
det
ect
ed
and
picke
d
up.
I
n
ad
diti
on,
as
m
entioned
by
So
a
ns
et
al
.
[
2],
on
e
of
the
m
a
in
pro
blem
in
colo
r
trac
ki
ng
is
th
e
insig
nificant
of
current
c
olor
thres
holdin
g
te
chn
i
qu
e
w
hich
cause
d
by
e.
g.
ref
le
ct
ion
of
s
m
oo
th
gro
und
plane.
This
re
flect
ion
res
ults
i
n
fals
e
ob
j
ect
detect
ion
a
nd
c
on
s
e
qu
e
nces
le
ad
s
to
false
c
olor
tracki
ng.
O
n
t
he
ot
her
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2502
-
4752
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci,
Vo
l.
12
, N
o.
3
,
Dece
m
ber
2
01
8
:
1349
–
1357
1350
hand,
the
var
ia
ti
on
of
s
urr
oundin
g
li
ghti
ng
c
onditi
ons
al
s
o
le
ads
t
o
false
c
olor
detect
ion
and
trac
king.
These
pro
blem
s ar
e the
fo
c
us
obj
ect
ives in
this
paper.
People
detect
ion
a
nd
trac
king
syst
em
base
d
on
real
-
ti
m
e
RGB
-
D
f
or
m
ob
il
e
robo
t
is
pr
opos
e
d
by
Fang
et
.
al
.
[
3].
In
t
he
pr
opose
d
syst
em
,
an
op
e
n
s
ource
r
ob
ot
op
e
rati
ng
s
yst
e
m
(ROS
)
is
i
m
ple
m
ented
to
a
m
ob
il
e
ro
bot
t
o
trac
k
a
ta
r
get
.
I
n
the syst
e
m
,
the f
eat
ure o
f
t
he
ta
r
get
is
ext
racted b
e
fore
t
he
de
pth
i
nform
at
ion
is
colle
ct
ed
a
nd
us
e
d
t
o
tr
ack
the
ta
rg
et
base
d
on
t
he
near
e
st
poi
nt
posit
ion
in
form
ation
.
Th
en
,
t
he
i
m
ple
m
entat
io
n
of
C
AM
-
S
hi
ft
al
gorithm
w
hich
is
base
d
on
R
GB
i
nfor
m
at
ion
is
ap
plie
d
t
o
im
pr
ove
t
he
a
nti
-
interfe
ren
ce
abi
li
t
y.
For
t
he
pur
po
se
of
im
age
c
ontrast
im
pr
ove
m
ent
of
in
hom
og
e
ne
ous
il
lu
m
inati
on
,
A
bdul
Gh
a
ni
a
nd
Ma
t
Isa
[
4]
ha
ve
ap
plyi
ng
rec
ur
si
ve
a
dap
ti
ve
histo
gr
am
m
od
ific
at
ion
w
hic
h
f
oc
us
es
on
a
pp
ly
in
g
cl
ip
-
li
m
it
and
gr
ay
-
le
vel
m
app
in
g
of
the
ca
pt
ur
e
d
im
ages.
Fr
om
the
res
ults,
the
ou
t
pu
t
i
m
ages
show
t
he
sign
ific
a
nt
outp
ut
by
pro
du
ci
ng
a
ho
m
og
eneous
il
lum
inati
on
of
t
he
im
ag
es.
T
his
m
et
ho
d
c
ou
l
d
r
edu
ce
t
he
noise
le
vel
in
the
ca
ptured
i
m
age
of
the
a
ut
o
-
trac
king
m
ob
il
e
r
obot.
I
n
a
dd
it
io
n
t
o
t
his
a
lgorit
hm
,
un
s
uper
vised
c
on
t
ra
st
co
rr
ect
io
n
t
hro
ugh
integrate
d
-
i
ntensity
stret
che
d
-
Ra
yl
ei
gh
Hist
ogram
[4
]
,
co
ul
d
be
c
om
e
anot
her
opti
on
t
o
address
lo
w
co
ntrast
and no
n
-
hom
og
ene
ous il
lum
i
nation i
m
ages.
This
a
uto
trac
ki
ng
de
vice
ca
n
be
us
e
d
t
o
e
nh
ance
t
he
s
urvei
ll
ance
syst
em
.
F
or
in
sta
nce,
it
can
be
us
e
d
to
trac
k
valua
bl
e
things.
F
or
e
xam
ple,
this
syst
e
m
can
act
as
a
m
ov
ing
CC
TV
as
it
not
on
ly
able
to
ob
s
er
ve
bu
t
it
al
so
able
to
track
t
hings.
B
esi
des
that,
ne
w
te
ch
nolo
gy
inv
e
nts
a
n
el
ect
ric
w
heelchair
wh
e
re
th
e
use
r
can
con
t
ro
l
the
m
ov
em
ent
of
wheel
chair.
N
ow
,
with
a
uto
tra
ckin
g
te
c
hnol
ogy
,
t
he
el
ect
ric
w
heelc
hair
c
an
be
conve
rt
into
a
n
aut
om
atic
wh
eel
chair
w
herea
s
the
m
ov
em
ent
of
the
w
heelchair
t
o
a
certai
n
point
i
s
done
autom
at
ic
ally
base
d
on
li
ne
or
m
ov
in
g
ob
je
ct
’s
colo
r.
In
add
it
io
n,
the
t
echnolo
gy
al
s
o
can
be
a
pp
li
ed
to
a
luggage
or s
ho
pp
i
ng tr
olley
where
the syste
m
w
ill autom
atical
ly
t
rack
and
fo
ll
ow it
s
owner.
In
this
pro
j
ect
,
the
cam
era
will
be
at
ta
ched
at
t
he
m
ob
il
e
r
obot
as
a
c
om
po
ne
nt
to
m
ov
e
f
r
om
a
locat
ion
to
an
oth
e
r
l
oca
ti
on
base
d
on
t
he
m
ov
in
g
ta
r
ge
t.
M
ic
ro
c
on
t
r
oller
with
im
a
ge
processi
ng
i
m
ple
m
entat
ion
play
s
an
im
po
rtant
r
ole.
A
c
olo
r
-
ba
sed
filt
ering
al
gorithm
is
us
ed
f
or
obj
ect
tra
ckin
g.
C
olor
-
ba
sed
filt
ering
m
et
ho
ds
are
popula
r be
cause t
hey are
fast, e
ff
ic
ie
nt,
and relat
ively
robust.
Nex
t,
a
m
ic
ro
con
t
ro
ll
er
is
al
s
o
a
n
esse
ntial
par
t
i
n
a
ut
o
tra
ckin
g
de
vice.
Mi
cro
co
ntr
oller
ac
t
a
s
t
he
br
ai
n
of
the
de
vice.
On
ce
t
he
i
m
age
has
bee
n
capt
ur
e
d
by
the
cam
era,
the
i
m
age
will
be
deco
m
po
se
d
in
to
it
s
ind
ivi
du
al
col
or
c
hannel.
Ba
s
ed
on
the
se
c
olo
r
cha
nnel
s,
t
he
total
inte
ns
it
y
an
d
the
m
axim
u
m
intensit
y
values
of
eac
h
col
or
channel
are
c
al
culat
ed.
T
he
m
axi
m
u
m
intensit
y
val
ues
betwee
n
these
c
olo
r
c
ha
nn
el
s
will
determ
ine
the
do
m
inant
col
or
of
the
obj
ect
f
or
t
he
pur
po
se
of
t
he
dev
ic
e
or
m
ob
il
e
rob
ot
to
ide
ntify
it
. B
esi
des
that,
the
m
ic
ro
con
t
ro
ll
er
al
so
sen
ds
el
ect
rica
l
sign
al
t
o
t
he
m
ob
il
e
ro
bot’s
act
uato
rs
t
o
re
act
to
the
m
oti
on
of
tracke
d object.
The
directi
on
of
this
pro
j
ect
is
to
fa
br
ic
at
e
a
sim
ple
an
auto
t
r
ackin
g
m
ob
il
e
rob
ot
that
is
a
bl
e
to
detect
an
obj
ect
an
d
tr
acks
it
.
T
he
m
ai
n
f
oc
us
of
t
his
researc
h
pro
j
ec
t
is
to
im
pr
ov
e
the
detect
i
on
a
nd
trac
king
abi
li
ti
es
of
t
he
ta
r
get
co
lor,
a
s
m
entione
d
by
So
a
ns
et
al
.
[2
]
.
The
al
gorithm
of
im
age
processi
ng
to
detect
the
obj
e
ct
is
base
d
on
the
c
olor
of
the
obje
ct
.
A
(
Hu
e
,
S
at
ur
at
io
n
a
nd
I
ntensity
)
HSI
colo
r
-
base
d
filt
erin
g
al
go
rith
m
wer
e
us
e
d
in
t
his
pr
oj
ect
t
o
detect
the
ta
r
get.
First
ly
,
the
cam
era
will
captu
re
th
e
i
m
age
an
d
th
e
m
ic
ro
co
ntr
oller
will
process
t
he
i
nfor
m
at
ion
an
d
m
on
it
or
the
be
hav
i
or
of
the
obj
ect
.
Ne
xt,
wh
e
n
t
he
obj
e
ct
sta
rts
to
m
ov
e,
t
his
m
ob
il
e
ro
bot
al
so
will
f
ollow
a
nd
t
rack
s
t
he
obj
ect
.
At
the
sa
m
e
tim
e,
the
m
ob
il
e
r
obot
al
w
ay
s
m
ai
ntain
distance
with
t
he
ob
j
ect
to
pr
e
ven
t
f
r
om
colli
sion
with
t
he
ob
j
ect
.
Be
sides
t
hat,
the
visi
on
of
the
r
obot
is
up
t
o
180°
i
n
x
-
a
xis
a
nd
y
-
a
xis
as
t
he
cam
era
is
at
ta
che
d
to
a
pan
-
ti
lt
w
hich
is
a
ble
to
ro
ta
te
.
T
he
r
otati
on
of
t
he
ca
m
era
is
con
t
ro
l
by
two
servo
m
oto
rs
wh
e
reas
on
e
s
ervo
m
oto
r
is
us
e
d
to
co
ntr
ol
x
-
a
xis
m
ov
em
ent
an
d
an
oth
e
r
ser
vo
m
oto
r
is use
d f
or y
-
axis
m
ov
em
ent.
2.
RESEA
R
CH
MA
TE
RIA
L
S
AND
METH
OD
OL
OG
Y
Fr
om
the
li
te
ra
ture
re
view
,
sui
ta
ble
com
pone
nts
f
or
th
e
proj
ect
wer
e
sel
e
ct
ed.
T
he
ci
rcui
t
desig
n
of
the
pro
j
ect
is
de
velo
ped.
T
his
fo
ll
owe
d
by
s
oft
war
e
de
velo
pm
ent
of
the
pro
j
ect
.
A
fter
this,
the
syst
e
m
is
t
est
ed
un
ti
l
the
syst
e
m
wo
r
ks
acc
ordin
g
to
the
ob
je
ct
ive
of
t
his
pro
j
ect
.
Fi
gure
1
sh
o
ws
t
he
m
eth
od
ology
flo
w
char
t
of
the pr
oj
ect
of a
uto
-
tracki
ng m
ob
il
e
rob
ot.
Evaluation Warning : The document was created with Spire.PDF for Python.
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci
IS
S
N:
25
02
-
4752
Impr
ovem
e
nt
of
Au
to
-
Tr
ackin
g
M
obil
e Robot
b
ase
d o
n H
SI Color
M
od
el
(
Su
r
esh
Su
ndarajo
o
)
1351
Figure
1. Me
th
odology
flo
wc
har
t
of aut
o
-
tr
a
ckin
g
m
ob
il
e r
obot
a.
Pixy C
M
Ucam5
The
visi
on
se
nsor
sel
ect
e
d
f
or this p
r
oject
is Pixy CM
Ucam
5.
T
he
Pixy C
MUcam
5
vision
senso
r
is
a
fast
im
age
sens
or
t
hat
trac
ks
obj
ect
a
nd
it
can
directl
y
co
nne
ct
to
A
rduin
o
Uno
t
hroug
h
I
CSP
po
rt
on Arduin
o
Uno
boar
d.
Be
sides
that,
Pix
y
has
it
s
ow
n
powe
rful
proc
essor
to
proces
s
the
im
age.
Since
Pi
xy
has
it
s
ow
n
process
or,
it
wi
ll
process
the
c
aptu
re
d
im
ages
f
ro
m
the
sen
sor
a
nd
e
xtract
th
e
use
f
ul
in
f
or
m
at
ion
.
Be
sid
es
t
hat,
Pixy
c
om
e
with
a
c
olor
al
gorithm
to
detect
ob
j
ect
’s
col
or.
Norm
al
l
y,
RGB
(r
e
d,
gr
ee
n,
an
d
bl
ue)
use
d
to
represe
nt
c
olor
s.
B
ut,
Pixy
ca
lc
ulate
s
the
hue
c
olo
r
a
nd
sat
ur
at
io
n
of
each
R
GB
pix
el
f
rom
the
i
m
age
s
ens
or
and
us
es
these
as
the
pri
m
ary
filt
ering
pa
ram
et
ers.
Th
us
,
c
onve
rsion
al
gori
thm
to
co
nv
e
rt
RGB
to
HSI
c
olor
base
d
is
not
re
qu
i
red
in
the
pro
gr
am
m
ing
pa
rt
as
t
he
al
gorithm
is
al
ready
integrate
d
i
n
the
Pixy
CM
Ucam
5
i
m
age
sens
or
m
od
ule.
Ne
vert
heless,
Pi
xy
proces
ses
a
n
ent
ire
640x
400
i
m
age
fr
am
e
ev
ery
1/50th
of
a
seco
nd.
This m
eans th
e
cam
era d
et
ect
ed object
s'
posit
ion
s e
ver
y
20
m
illi
secon
ds.
b.
Ardu
in
o Un
o
In
this
pro
j
ect
of
de
velo
pi
ng
an
a
uto
-
trac
king
de
vice,
Ard
ui
no
U
no
was
s
el
e
ct
ed
as
t
he
con
t
ro
ll
er
of
the
syst
em
.
Th
e
A
rduin
o
U
no
boar
d
is
a
m
ic
ro
c
ontrolle
r
ba
sed
on
ATm
ega3
28.
It
has
14
di
gital
input/
outp
ut
pin
s
in
w
hich
6
can
be
us
ed
as
P
W
M
outp
uts,
a
16
M
Hz
cera
m
ic
reso
nato
r,
an
ICS
P
hea
de
r,
a
U
SB
co
nn
e
ct
ion,
6
a
nalo
g
in
puts
, a
powe
r jack
and a
reset b
utt
on.
T
he
cam
era can
be direct
l
y connect
ed
to
ICSP hea
der.
c.
L293D
Moto
r
Driver
L2
93D
m
oto
r dr
i
ver
is a
n
int
egr
at
e
d
ci
rc
uit chip w
hich
is
usual
ly
u
se
d
to
con
t
ro
l. M
otor
dr
i
ver
act a
s
an
i
nterfac
e
bet
ween
Ard
uino
and
the
m
oto
rs
.
T
h
ese
ICs
are
desig
ne
d
t
o
c
ontr
ol
tw
o
DC
m
otors
sim
ultane
ou
sly
.
L2
93D
c
onsist
of tw
o
H
-
br
i
dge. H
-
br
id
ge
is
the sim
plest ci
rcu
it
for
c
ontr
olli
ng
a l
ow curr
ent r
at
e
d
m
oto
r
.
d.
Circui
t Desig
n
The
m
ai
n
hard
war
e
c
om
po
ne
nts
sel
ect
e
d
t
o
bu
il
d
t
he
protot
ype
for
aut
o
-
tr
ackin
g
m
ob
il
e
rob
ot
ar
e
t
he
Ardu
i
no
U
no,
Pixy
CM
Ucam
5
cam
era,
A
rduin
o
M
oto
r
S
h
ie
ld,
ser
vo
m
oto
r
an
d
DC
m
oto
r
.
T
he
dev
el
opm
ent
of
ci
rc
uit
desig
n
f
or
this
proj
e
ct
was
sta
rte
d
with
a
blo
c
k
di
agr
am
as
s
how
n
i
n
Fig
ur
e
2.
A
blo
c
k
diagr
a
m
is
a
diag
ram
o
f
a
s
yst
e
m
in
w
hich
the
pri
ncipal
parts are
r
e
prese
nted
b
y
blo
c
ks
and c
onnected
by li
nes
to sh
ow t
he
inputs a
nd out
pu
ts
of a
syst
em
.
L
it
e
ra
ture
re
view
De
fine
prob
lem
state
ment?
De
ter
mi
ne
objec
ti
ve
De
c
lar
e
proje
c
t scope
C
omponent se
lec
ti
on
C
irc
uit
De
ve
lopm
e
nt
S
y
stem
wor
k?
F
a
bric
a
te of
the pr
ojec
t
No
No
Ye
s
Ye
s
S
TA
R
T
S
oftw
a
re
De
ve
lopm
e
nt
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2502
-
4752
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci,
Vo
l.
12
, N
o.
3
,
Dece
m
ber
2
01
8
:
1349
–
1357
1352
The
com
plete
schem
at
ic
c
ircuit
diagram
of
auto
-
tracki
ng
m
ob
il
e
ro
bo
t
de
vice
is
sh
ow
n
in
Figure
3.
The
m
ai
n
com
pone
nt
of
the
c
ircuit
is
Ardu
i
no
U
no
as
it
rec
ei
ves
data
a
bout
the
po
sit
io
n
of
tracke
d
obj
ec
t
fr
om
the
Pixy
CM
U
ca
m
5.
The
n,
A
rduin
o
will
cont
ro
l
the
m
ov
em
ent
of
DC
m
otor
acco
r
ding
to
t
he
m
otion
of
tr
acke
d
obj
ect
. Besi
des
that, the
serv
o m
oto
rs
are
dire
ct
ly
co
ntr
olled
by P
ixy C
M
Uc
a
m
5
it
sel
f.
Figure
2. Bl
oc
k diag
ram
o
f
a
uto
-
tracki
ng m
ob
il
e
rob
ot
Figure
3. Sc
he
m
at
ic
circuit diagr
am
o
f
a
uto
-
trackin
g
m
ob
il
e r
obot
e.
Co
m
pleted
h
ar
dware de
sign
The
f
ollo
wing
Figure 4
s
how
s
the co
m
plete
m
ob
il
e ro
bo
t e
qu
i
pp
e
d wit
h
P
ixy ca
m
er
a fo
r t
he
pur
pose
of
c
olo
r
detect
ion
an
d
trac
ki
ng.
I
n
a
ddit
ion
to
the
cam
era,
the
m
ob
il
e
r
obot
is
al
s
o
e
quipee
d
with
t
wo
DC
m
oto
rs,
m
oto
r dr
i
ver
,
and
pa
n
-
ti
lt
m
echan
ism
w
hich
co
ns
i
sts of tw
o
i
nter
-
co
nnect
ed
se
r
vo m
oto
rs.
Figure
4. Final
h
a
rdwar
e
d
e
sign
Ar
d
u
in
o
Un
o
P
ixy
CM
U ca
m
5
L
2
9
3
D
M
o
to
r
D
riv
er
D
C
M
o
to
r
D
C
M
o
to
r
Serv
o
M
o
to
r
Serv
o
M
o
to
r
P
o
we
r Su
p
p
ly
9V
Evaluation Warning : The document was created with Spire.PDF for Python.
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci
IS
S
N:
25
02
-
4752
Impr
ovem
e
nt
of
Au
to
-
Tr
ackin
g
M
obil
e Robot
b
ase
d o
n H
SI Color
M
od
el
(
Su
r
esh
Su
ndarajo
o
)
1353
f.
Pro
gra
m
mi
ng flowch
art
Figure
5
sho
ws
the
pro
gr
am
m
ing
flo
wc
har
t
of
a
uto
-
tracki
ng
m
ob
il
e
rob
ot.
Firstl
y,
w
he
n t
he
de
vice
is
tur
n
on,
the
sy
stem
will
be
init
ia
li
zed.
Af
te
r
that,
the
Pixy
CM
Ucam
5
vision
se
nsor
will
find
t
he
sig
nat
ur
e
of
the
tracki
ng
ob
j
ect
.
O
nce
t
he
obj
ect
is
detect
ed
the
visio
n
s
ens
or
will
cal
culat
e
the
area
a
nd
x
-
co
ordinat
e
of
t
he
trackin
g
obj
ect
for
eve
ry
20
m
illi
secon
ds.
To
m
ake
su
re
the
tracke
d
obj
ect
al
ways
pa
rall
el
with
the
visio
n
sens
or
an
op
ti
m
u
m
area
an
d
x
-
c
oor
din
at
e
a
r
e
set
as
re
fer
e
nc
e
val
ue
w
her
e
the
re
fer
e
nce
values
are
def
i
ned
as
,
,
.
If
>
m
eans
t
he
ob
j
ect
is
t
oo
near
to
the
de
vi
ce.
Th
us,
t
he
dev
ic
e
will
m
ov
e
ba
ck
wards
to
kee
p
the
de
vice
a
nd
ob
j
e
ct
at
op
tim
u
m
range.
Sam
e
goes
if
<
,
this
m
eans
the
obj
ect
this
fa
r
f
ro
m
the
dev
i
ce.
Th
us
,
t
he
m
ob
il
e
ro
bot
will
m
ov
e
f
orward
un
ti
l
it
is
at
an
opti
m
u
m
range.
Be
side
s
that,
x
-
c
oor
di
nates
are
us
e
d
to
determ
ine
the
m
ov
em
ent
of
t
he
obj
ect
in
x
-
directi
on.
I
f
th
e
ob
j
ect
<
m
ea
ns
t
he
obj
ect
is
at
le
ft.
Th
us,
t
he
pro
gr
am
will
set
the
rig
ht
m
oto
r
on
a
nd
le
ft
m
oto
r
off s
o
that
th
e
m
ob
il
e
r
obot
can
t
urn
le
ft.
S
a
m
e
goes
f
or
>
c
onditi
on
w
her
e
f
or
t
his con
diti
on
t
he object
is t
o
t
he rig
ht.
Figure
5. Pro
gra
m
m
ing
flo
wc
har
t
of aut
o
-
tr
a
ckin
g
m
ob
il
e r
obot
3.
RESU
LT
S
A
ND
DI
SCUS
S
ION
The
a
uto
-
track
ing
m
ob
il
e
r
obot
able
to
t
racki
ng
obj
ect
s
e
ffec
ti
ve
ly
as
pro
po
s
ed
at
norm
al
conditi
on
.
Seve
ral
e
xp
e
rim
ents
we
re
ca
r
ried
out
on
the
pro
po
se
d
syst
em
to
ensure
the
ob
j
ect
ive
of
th
e
pro
j
ect
is
ac
hi
eved.
First
ex
per
im
e
nt
was
to
deter
m
ine
the
m
axim
u
m
ta
rg
et
locking
distan
ce.
Anothe
r
ob
j
ec
ti
ve
of
this
pro
j
ect
to
ev
al
uate
the
ac
cur
acy
of
c
ol
our
ide
ntific
at
io
n
i
n
dif
fer
e
nt
i
ll
u
m
inati
on
a
nd
bac
kgr
ound
conditi
ons
,
an
d
t
hus
i
m
pr
oves
it
e
f
fici
ency
in
tra
ckin
g
within
var
i
ou
s
e
nv
i
ronm
ent
includi
ng
lo
w
co
ntra
st
an
d
al
m
os
t
si
m
il
ar
backg
rou
nd
-
ta
r
get
col
or
en
vir
on
m
ent
.
T
hu
s
,
th
e
sec
ond
ex
per
im
ent
is
car
ried
out
to
det
erm
ine
the
effe
ct
of
STA
RT
F
i
nd O
b
j
ec
t
Sig
nat
ur
e
O
b
j
ec
t
s
i
g
nat
u
r
e
de
t
e
ct
ed
?
>
<
<
>
C
a
lcula
te
obj
ec
t
’
s
a
r
e
a a
nd
x
-
coor
di
nat
e
I
nit
ializ
ing
the s
y
stem
Rig
h
t
M
o
t
o
r &
L
eft
M
o
to
r
in
fo
r
ward
d
irectio
n
Onl
y
L
eft
M
o
t
o
r
in
fo
rward
d
ir
ect
i
o
n
Rig
h
t
M
o
t
o
r &
L
eft
M
o
to
r
in
b
ack
wards
d
irectio
n
Onl
y
R
ig
h
t
M
o
t
o
r
in
fo
rward
d
ir
ect
i
o
n
No
v
No
No
No
No
Yes
Yes
Yes
Yes
Yes
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2502
-
4752
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci,
Vo
l.
12
, N
o.
3
,
Dece
m
ber
2
01
8
:
1349
–
1357
1354
backg
rou
nd
co
lour
durin
g
ob
j
ect
trac
king.
Thir
d
e
xperim
ent
wa
s
to
exe
rm
ine
the
eff
e
ct
of
il
lum
inati
on
on
detect
ion an
d
t
arg
et
tra
cki
ng.
a.
Maximum
tar
get lockin
g di
sta
n
ce
The
ob
j
ect
ive
of this
exp
e
rim
ent is
to
d
et
e
r
m
ine the
m
axi
m
u
m
d
ist
ance
of tar
get
from
the
de
vice t
hat
giv
es
acc
ur
a
cy
and
e
ff
ic
ie
nt
tracki
ng.
Th
e
re
su
lt
of
t
he
ex
pe
rim
ent
is
in
T
able
1
.
Ba
se
d
on
Table
1
it
is
fou
nd
that
wh
e
n
t
he
de
te
ct
ion
ra
ng
e
is
increase
d
as
the
sta
bili
ty
of
the
t
arg
et
loc
ki
ng
dec
reases.
At
sig
natur
e
ra
ng
e
of
8.5,
t
he
visi
on
sens
or
a
ble
to
detect
ta
rg
et
th
at
240cm
away
from
the
visio
n
se
ns
or.
H
owever,
at
this
ra
ng
e
of
detect
ion
the
tr
ackin
g
l
os
s
occ
ur
s
f
re
qu
e
ntly
du
e
to
un
sta
ble
ta
r
get
loc
king
as
s
how
n
i
n
Figure
8.
At
si
gnat
ur
e
range
of
1.5
th
e
ta
rg
et
loc
king
is
ve
ry
sta
ble
as
sh
ow
n
in
Fi
gure
6.
But,
t
he
range
of
de
te
ct
ion
is
ve
ry
shor
t.
The
m
os
t
m
axi
m
um
ta
rg
et
locki
ng
distance
an
d
the
m
os
t
sta
ble
ta
rg
et
l
ock
is
giv
e
n
at
sig
natu
re
ra
nge
of
5.5
a
s
sh
ow
n
in
Fi
gur
e 7
a
nd
this
is the
op
ti
m
u
m
sett
ing
of si
gn
at
ure ra
nge
for be
st detec
ti
on and e
ff
ect
ive
trac
king.
Table
1.
Res
ult o
f
E
xp
e
rim
ent
A
Sig
n
atu
re
rang
e
Ran
g
e of
Detec
tio
n
(c
m
)
Tar
g
et
Lock
Stab
ility
1
.5
90
Hig
h
2
.5
125
Hig
h
3
.5
150
Hig
h
4
.5
170
Hig
h
5
.5
190
Hig
h
6
.5
210
Mod
erate
7
.5
225
Mod
erate
8
.5
240
Low
Figure
6.
Pix
y
o
utput
at
sing
at
ur
e
ran
ge
of
1.
5
Figure
7.
Pix
y
o
utput
at
sing
at
ur
e
ran
ge
of
5.
5
Figure
8.
Pix
y
o
utput
at
sing
at
ur
e
ran
ge
of
8.
5
b.
Effect
of
ba
c
k
ground
du
ri
n
g objec
t
tr
acki
ng
The
ob
j
ect
ive
of
t
his
e
xperim
e
nt
was
to
deter
m
ine
the
ca
pabi
li
t
y
of
visio
n
s
ens
or
to
diff
e
re
ntiat
e
ta
r
get
and
bac
kgr
ound
if
both
hav
e
alm
os
t
si
m
i
la
r
colo
r.
Be
side
s
that,
the
acc
ur
a
cy
of
detect
ion
an
d
trac
king
i
n
this
conditi
on
is
stu
died.
In
this
e
xperim
ent
a
gr
e
en
ball
is
use
d
as
a
tracki
ng
obj
ect
an
d
the
ba
ckgr
ound
c
ol
our
al
so
set
as
green
c
olo
ur.
T
his
e
xper
i
m
ent
is
co
nduc
te
d
by
m
anipula
ti
ng
t
he
si
gn
a
ture
ra
ng
e
par
a
m
et
er.
T
he
res
ult
of
the expe
rim
ent
is
ta
bula
te
d
in
Table
2.
Table
2.
Res
ult o
f
E
xp
e
rim
ent
B
Sig
n
atu
re
rang
e
No
ise Level
Tar
g
et
Lock
Stab
ility
1
.5
No
n
e
Hig
h
2
.5
No
n
e
Hig
h
3
.5
No
n
e
Hig
h
4
.5
Low
Hig
h
5
.5
Mod
erate
Mod
erate
6
.5
Hig
h
Low
7
.5
Hig
h
Low
8
.5
Hig
h
Low
Evaluation Warning : The document was created with Spire.PDF for Python.
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci
IS
S
N:
25
02
-
4752
Impr
ovem
e
nt
of
Au
to
-
Tr
ackin
g
M
obil
e Robot
b
ase
d o
n H
SI Color
M
od
el
(
Su
r
esh
Su
ndarajo
o
)
1355
Figure
9. Pixy
ou
t
pu
t at
sig
na
ture
range
of 3.5
Figure
10. Pi
xy
o
ut
pu
t at
sign
at
ur
e
ra
ng
e
of
5.5
Figure
11. Pi
xy
o
ut
pu
t at
sign
at
ur
e
ra
ng
e
of
8.5
Ba
sed
on
Tabl
e
2
the
best
set
ti
ng
f
or
sig
natu
r
e
ra
ng
e
par
am
et
er
is
at
3.5.
T
hi
s
is
beca
us
e
at
this
value
the
PixyC
MUc
a
m
5
giv
es
the
best
detect
io
n
and
ta
r
get
loc
ki
ng
without
a
ny
noise
.
At
the
sam
e
tim
e,
the
gree
n
backg
rou
nd
al
so
d
id not
inter
r
up
t
t
he
trac
king
as sho
wn
i
n
Figure 1
0.
Be
s
ides
that,
it
al
so
g
ives
t
he
m
axim
u
m
range
of
detect
ion
with
out
no
ise
w
hich
is
a
bout
150
cm
.
This
m
eans
t
he
aut
o
-
t
rack
i
ng
m
ob
il
e
r
obot
able
t
o
track
an
obj
ect
from
a
distance
of
150
cm
.A
t
sign
at
ure
ra
nge
of
5.5
ther
e
is
ve
ry
sli
gh
t
bac
kgr
ound
inter
ruption
as
show
n
Fig
ure
10.
Be
si
des
that
,
at
sig
na
ture
range
of
8.5
t
he
Pixy
C
MUcam
5
un
a
bl
e
to
dif
fer
e
ntiat
e
the
backg
rou
nd
an
d
ta
r
get
when
bo
t
h
a
re
i
n
sa
m
e
color
as
show
n
i
n
Fig
ur
e
11
,
this
will
cause
t
rack
i
ng
loss
due
m
isi
nterp
et
ion
betwee
n
ta
r
get
and
obj
ect
.
Th
us
,
si
gn
at
ur
e
ra
ng
e
of
3.5
giv
e
s
the
best
trac
ki
ng
wh
e
n
the
obj
ect
colo
r
a
pp
ea
r
i
n bac
kgr
ound.
c.
Effect
of ill
u
m
inat
i
on
durin
g
object
t
r
acki
ng
The
s
urr
ounding
bri
ghtness
al
so
play
s
a
n
i
m
po
rtant
ro
le
in
pe
rfo
rm
ance
an
d
se
ns
it
ivit
y
of
Pixy
CM
Ucam
5 visi
on
se
nsor
. Fi
rst
ly
, t
he a
uto
-
tra
ckin
g m
ob
il
e
was t
est
ed i
n
va
rio
us
li
gh
ti
ng
conditi
on. It
is
fou
nd
that
the
dev
ic
e
loss
trac
king
wh
e
n
the
re
is
dr
am
at
ic
change
i
n
li
gh
ti
ng
conditi
on.
H
oweve
r,
the
trac
king
was
sti
ll
sta
ble
wh
e
n
the
re
is
on
ly
sli
gh
t
cha
nge
i
n
il
lum
inati
on
.
Th
us
,
a
n
e
xp
e
ri
m
ent
was
c
onduct
ed
at
five
di
ff
e
ren
t
su
r
rou
nd
i
ng
li
gh
ti
ng
c
onditi
on
s
.
At
the
sa
m
e
tim
e
PixyC
MUcam
5
cam
era’s
bri
ghtn
ess
ke
pt
co
ns
t
ant.
T
he
resu
lt
of
the
e
xperim
ent
ta
bu
la
te
d
in
Ta
ble
3
.
At
this
point,
we
awa
re
with
the
i
m
pr
ovem
e
nt
pro
posed
by
Abd
ul
Gh
a
ni
(
2018
)
a
nd
A
bdul
G
ha
ni
an
d
Ma
t
Isa
(20
15)
for
t
he
en
han
cem
ent
of
im
age
co
ntr
ast
.
This
m
et
ho
d
will
be
im
ple
m
ente
d
in
ou
r next e
nh
a
ncem
ent syst
e
m
f
or a
bette
r
c
olo
r
d
et
ect
i
on w
it
h va
rio
us i
ll
umi
nations.
Table
3.
Res
ult o
f
E
xp
e
rim
ent
C (bef
or
e
cali
br
at
in
g)
Su
rr
o
u
n
d
in
g
Brightnes
s
Pix
y
Brigh
tn
ess
No
ise Level
Tar
g
et
Lock
Stabil
ity
Ver
y
brig
h
t
80
Hig
h
No
n
e
Brig
h
t
80
Mod
erate
Low
No
r
m
al
80
No
n
e
Hig
h
Less Brig
h
t
80
Less
Hig
h
Dark
80
Mod
erate
Low
Ba
sed
on
Ta
bl
e
2
the
facto
rs
that
dif
fer
e
ntiat
e
al
l
this
cond
it
ion
are
th
e
sta
bili
ty
of
the
ta
rg
et
loc
k
by
Pixy
visio
n
se
nsor
a
nd
the
no
i
se
ge
ner
at
ed
from
the
back
gr
ound
of
the
tra
ckin
g
ob
j
ect
.
H
ow
e
ve
r,
this
prob
le
m
can
be
overc
om
e
by
cal
ibrati
ng
the
bri
ghtne
ss
of
Pixy
CM
Ucam
5
visio
n
sens
or
unti
l
th
e
ta
rg
et
loc
k
is
sta
ble.
Hen
ce
,
the
Pix
y
CM
Uca
m
5
vi
sion
se
nsor
m
us
t
be
cal
ib
rate
d
eve
ry
tim
e
the
m
ob
il
e
rob
ot
exp
e
rience
sig
nificant
su
r
rou
nd
i
ng
bri
gh
tne
ss
c
hang
e.
T
hu
s
,
the
ex
per
im
ent
is
repea
te
d
by
ad
justi
ng
t
he
br
i
gh
t
ne
ss
of
Pi
xy
ca
m
era
un
ti
l t
he
ta
rg
et
lock
is
stable a
nd the
noise
l
e
vel is re
duced
. T
he res
ult i
s tabu
la
te
d i
n Tabl
e 4
.
Table
4.
Res
ult o
f
E
xp
e
rim
ent
C (a
fter cali
br
at
ing
)
Su
rr
o
u
n
d
in
g
Brightnes
s
Pix
y
Brigh
tn
ess
No
ise Level
Tar
g
et
Lock
Stabil
ity
Ver
y
brig
h
t
20
Less
Mod
erate
Brig
h
t
50
No
n
e
Hig
h
No
r
m
al
80
No
n
e
Hig
h
Less Brig
h
t
100
No
n
e
Hig
h
Di
m
120
No
n
e
Mod
erate
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2502
-
4752
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci,
Vo
l.
12
, N
o.
3
,
Dece
m
ber
2
01
8
:
1349
–
1357
1356
Table
3
sho
ws
the
data
c
ollec
te
d
afte
r
cal
ibra
ti
ng
the
Pixy
br
igh
tne
ss
wit
h
the
s
urrou
nd
i
ng
br
i
gh
t
ness
.
Re
su
lt
s
pro
ves
that
the
PixyC
MUcam
5
able
to
detect
an
d
tr
ack
ob
j
ect
at
va
rio
us
il
lum
ina
ti
on
.
T
his
will
m
ake
the
a
uto
-
trac
kin
g
m
ob
il
e
r
ob
ot
ine
ff
e
ct
ive
as
it
will
lo
ss
tracki
ng
wh
e
n
the
bri
ghtness
changes
dr
am
at
ic
al
ly
un
ti
l i
t i
s cali
brat
ed
agai
n. The
r
es
ults o
f
ta
r
ge
t l
ock
in
g an
d no
ise
level a
re
sh
ow
n
in
Fi
gur
e
s
12
-
15.
Figure
12. Pi
xy
ou
t
pu
t at
ve
r
y br
i
gh
t c
ondit
ion
.
(b
e
fore a
nd aft
er cali
brat
ing)
Figure
13 Pi
xy outp
ut at bri
gh
t condit
ion.
(b
e
fore
and after
cali
brat
ing
)
Figure
14. Pi
xy
o
ut
pu
t at
nor
m
al
an
d
le
ss
br
igh
t
conditi
on. (be
f
or
e
and a
fter c
al
ibrati
ng)
Figure
15.
Pi
xy outp
ut at da
r
k condit
ion.
(b
e
f
or
e
and
after cali
brat
in
g)
4.
CONCL
US
I
O
N
The
im
ple
m
ent
at
ion
of
en
ha
nc
e
m
ent
integra
te
d
with
filt
ering
m
et
ho
ds
in
auto
-
tracki
ng
m
ob
il
e
ro
bot
giv
es
a
sat
isfac
tory
s
olu
ti
on
to
the
pr
ob
le
m
s
that
discusse
d
i
n t
he
pro
blem
sta
tem
ent.
T
o
i
m
pr
ov
e
the
acc
ur
acy
of
ide
ntific
at
io
n
of
obj
ect
c
olo
r
in
diff
e
re
nt
il
lu
m
inati
on
a
nd
bac
kgrou
nd
co
ndit
ion
s
,
th
e
H
SI
c
olor
m
od
el
is
us
e
d
in
im
age
processi
ng
al
gorithm
.
The
enh
a
ncem
ent
and
filt
erin
g
processes
in
HS
I
col
or
m
od
el
ha
s
su
ccess
fu
ll
y
in
te
gr
at
ed
t
o
so
l
ve
the
pro
blem
s
of
l
os
s
of
c
olo
r
trac
king.
Ex
per
im
ental
resu
lt
s
dem
on
stra
te
that
wh
e
n
H
SI
col
or
-
base
d
filt
erin
g
al
gorithm
is
app
li
ed
t
o
visua
l
tracki
ng
it
im
pr
ov
es
the
ac
cur
acy
an
d
sta
bili
ty
of
colo
r
trac
king
unde
r
the
co
nd
it
ion
of
var
yi
ng
br
ig
ht
ness,
or
ev
en
i
n
the
lo
w
-
c
ontrast
e
nviro
nm
ent
,
as
di
scusse
d
in the res
ults
. B
esi
des
that,
th
is al
gorithm
al
so
pr
e
ve
nts tra
ckin
g
lo
ss
du
e
to
sim
i
la
r
bac
kgr
ound
c
olor
.
Althou
gh
t
he
t
rack
i
ng
l
os
s
is
su
e
du
e
s
ud
de
n
li
gh
ti
ng
c
ha
nge
ca
n
be
s
ol
ve
d
by
cal
ib
rati
ng
t
he
vision
sens
or
.
One
of
the futur
e
work that ca
n
be d
on
e t
his projec
t i
s to
m
ake
hig
h
acc
uracy
col
or
detect
ion
of
visi
on
sens
or
es
pecia
ll
y
wh
e
n
the
bri
ghtness
c
hanges.
Be
side
s
t
hat,
this
pro
bl
e
m
al
so
can
be
s
olv
e
d
by
re
placi
ng
PixyC
MUcam
5
wit
h
a
m
or
e
reli
able
cam
e
ra
that
a
ble
to
ad
j
us
t
cam
era
br
i
gh
t
ness
w
hen
the
s
urr
ound
i
ng
il
lu
m
inati
on
va
ries.
F
or
f
uture
i
m
pr
ovem
ent,
the
in
vestigat
i
on
on
t
he
reli
able
an
d
r
obus
t
al
gorithm
fo
r
bette
r
colo
r
trac
king
will
be
inclu
de
d.
T
he
rob
us
t
c
olor
trac
king
w
il
l
be
base
d
on
the
c
urren
t
fi
nding
w
hich
a
re
r
el
at
ed
to
the
prob
le
m
s
of
in
hom
og
ene
ous
il
lum
i
nation,
va
ri
ou
s
bac
kgr
ound
colo
r,
an
d
sim
il
arit
y
of
ta
rget
an
d
backg
rou
nd col
or
.
ACKN
OWLE
DGE
MENTS
Au
t
hors
w
ou
l
d
li
ke
to
t
hank
a
ll
rev
ie
we
rs
for
the
c
on
tri
bu
ti
on
to
ward
im
pr
ovin
g
the
pa
per.
This
w
ork
is
sup
ported
by
internal
gra
nt
U
niv
er
sit
i
Ma
la
ysi
a
Paha
ng
(
UMP),
Au
t
om
ot
ive
E
ng
i
ne
erin
g
Ce
ntre
(
AEC)
,
RDU
180313
1
entit
le
d
“
De
velo
pm
ent
of
Mult
i
-
Visio
n
G
uid
e
d
Ob
st
acl
e
Avoida
nc
e
Syst
em
fo
r
Gro
un
d
Veh
ic
le
”.
Evaluation Warning : The document was created with Spire.PDF for Python.
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci
IS
S
N:
25
02
-
4752
Impr
ovem
e
nt
of
Au
to
-
Tr
ackin
g
M
obil
e Robot
b
ase
d o
n H
SI Color
M
od
el
(
Su
r
esh
Su
ndarajo
o
)
1357
REFERE
NCE
S
[1]
Luo
B.
,
Ding
Y
.
,
Hao
K.,
Li
u
J.
(2015)
R
ese
ar
c
h
on
Mobile
rob
ot
pat
h
tracki
ng
base
d
on
col
or
vision.
Chin
ese
Autom
at
ion
Con
gre
ss
.
W
uhan, C
hina
.
27
-
29
Nov
ember
2015.
[2]
Fang
F.,
Qian
K.,
Zhou
B.
,
M
a
X.
(2017)
.
R
e
al
-
T
ime
RGB
-
D
base
d
Peopl
e
Dete
c
ti
on
and
T
rac
king
S
y
s
te
m
for
Mobile
Robo
ts.
Proce
edi
ng
of
20
17
IEEE,
Int
ern
a
ti
onal
Conf
ere
nc
e
on
Mec
h
at
roni
cs
and
Autom
at
i
on
,
Augus
t
6
–
9
,
Ta
kamatsu, Japa
n.
[3]
Abdul
Ghani
A.
S.,
R
aj
a
Aris
R
.
S.N.A.
,
Muhd
Z
ai
n
M.
L.,
Uns
up
erv
ised
Con
tra
st
Corre
ct
ion
for
Underwat
er
Im
age
Quali
t
y
Enha
n
cem
ent
through
In
t
egr
ated
-
Int
ensity
Stretche
d
-
Ra
y
l
e
igh
Histograms
.
Journal
of
Te
l
ecom
m
uni
ca
ti
on,
El
e
ct
roni
c and
C
om
pute
r
Engi
n
e
eri
ng.
ISS
N:
218
0
–
1843
e
-
ISS
N:
2289
-
8131
Vol
.
8
No.
3
.
[4]
Cla
rk,
M.
,
Feldp
ausc
h,
D.
,
&
T
e
wolde,
G.
S.
(20
14).
Microsoft
ki
nec
t
sensor
for
r
e
al
-
ti
m
e
col
or
tr
a
cki
ng
robot
.
IEEE
Inte
rna
ti
ona
l
Co
nfe
ren
c
e
on
E
le
c
tro
Inform
at
ion
Te
chno
log
y
,
416
–
421.
ht
tps:/
/do
i.org/10.
1109/
EIT.2014.
6871800
[5]
Gam
al
,
B
.
E.,
Ouda,
A. N
.
,
E
lha
l
wag
y
,
Y.
Z.,
&
El
nasha
r
,
G
.
A.
(
2016).
Embedde
d
ta
rg
et
de
tecti
o
n
s
y
st
em
base
d
o
n
raspbe
rr
y
pi
sy
stem.
2016
1
2th
In
te
rn
at
ion
al
Com
pute
r
Engi
ne
eri
ng
C
onfe
ren
c
e
(ICE
NCO
),
154
–
15
7.
htt
ps://
do
i.
org/1
0.
1109/ICE
NCO
.
2016.
7856461
[6]
Sendar
i,
S
.
,
Le
st
ari
,
D
.
,
Kus
um
ohadi
,
C
.
U.
,
W
ib
owo,
F.
S.,
&
A
nam,
K.
(2017). Int
egr
a
ti
ng
embedde
d
col
or
visio
n
to Bio
loi
d
robot
for pl
a
y
ing
socc
er.
Proce
edi
ngs
-
Int
ern
ational
Co
nfe
ren
c
e
on
S
ign
al
s
and S
y
stems
,
ICSigS
y
s
2017
,
297
–
302.
ht
tps:/
/
doi.
org/10
.
1109/
ICSIG
SYS.2017.
7967060
[7]
W
an,
M.,
Gu,
G
.
,
Qian
,
W
.
,
R
en
,
K.,
&
Chen,
Q
.
(2017).
Hue
pr
ese
rva
t
ion
base
d
col
or
spac
e
tr
a
nsform
at
ion
for
bright
ness
-
robust
tracki
ng.
Op
ti
k
-
Inte
rn
ati
onal
Journal
f
or
L
ight
and
E
lectr
on
Opti
cs,
144,
26
–
3
3.
htt
ps://
do
i.
org/1
0.
1016/j.ijleo
.
20
17.
06.
073
[8]
Abdul
Ghani
A.
S.,
Mat
Is
a
N.A
.
(2017).
Autom
at
i
c
s
y
st
em
for
improving
unde
rwate
r
image
co
ntra
st
and
col
or
th
rough
rec
u
rsive
ada
p
ti
v
e
histog
ram
m
odifi
ca
t
io
n.
Com
pute
rs
an
d
El
e
ct
roni
cs
in
Agric
ult
ur
e
141
(2017)
181
–
195.
htt
p://dx.doi.org/10.1016/j.c
om
pa
g.
2017.
07
.
021
[9]
W
ang,
Y.
(2009
).
2009
As
ia
-
Pa
c
ifi
c
Confer
ence
on
Inform
at
ion
Proce
ss
ing
A
no
vel
m
et
hod
for
o
bje
c
t
r
ec
ogn
it
io
n
under
var
y
i
ng
il
l
um
ina
ti
on,
210
–
213.
ht
tps:/
/do
i.
o
rg/10.
1109/APC
IP.2009.
188
[10]
Soans
R.
V.,
He
gde
A.
,
Kum
ar
A.,
Ranjith
,
Si
ngh
C.
(2017).
Object
Tra
ck
in
g
Robot
Us
ing
Adapti
v
e
Colo
r
Thre
sholding
.
Proce
ed
ings
of
the
2nd
Inte
rna
ti
ona
l
Confer
en
ce
on
Com
m
unic
at
ion
and
Elec
tron
ic
s
S
y
stems
(ICCES
2017)
IEEE
Xpl
ore
Com
pli
an
t
-
Part
Num
ber
:
C
FP
17A
W
O
-
AR
T,
ISBN
:978
-
1
-
5090
-
5013
-
0
[11]
Zha
o,
S.
,
Li
u
,
B.
,
R
en,
Y.
,
&
Ha
n,
J
.
(2009).
Col
or
tr
ac
k
ing
v
isio
n
s
y
s
te
m
for
the
aut
onom
ous
rob
ot.
ICEMI
2009
-
Proce
edi
ngs
of
9th
Internat
io
nal
Confer
ence
on
El
e
ct
ron
ic
Mea
surem
ent
and
Instrum
e
nts,
3182
–
3185
.
htt
ps://
do
i.
org/1
0.
1109/ICE
MI.2
009.
5274360
[12]
Abdul
Ghani,
A.
S
.
2018.
Im
age
c
ontra
st
enh
ance
m
ent
using
an
in
te
gra
ti
on
of
r
ec
u
rsive
-
over
l
appe
d
cont
rast
li
m
it
e
d
ada
pt
ive
histogr
am
spec
ifica
t
ion
and
dua
l
-
image
wave
let
fusion
for
the
h
igh
visi
bil
ity
of
de
ep
u
nder
wate
r
image
.
El
sevi
er:
Oc
ea
n
Eng
inee
ring,
Volum
e
162
(2
018)
pp.
224
-
238.
ISI
-
Q1.
DO
I:
htt
ps://
do
i.
org/1
0.
1016/j.oc
ea
ne
ng.
2018.
05
.
027
[13]
Abdul
Ghani
A.
S.,
N.
A.
Mat
Isa
.
2015
.
Hom
om
orphic
Fi
lt
er
ing
w
it
h
Im
ag
e
Fus
ion
for Enhan
ce
m
e
nt
of
Det
ai
ls
and
Hom
ogene
o
us
Contra
st
of
Under
wate
r
Im
ag
e
–
Special
Iss
ue
.
Ind
ia
n
Journa
l
of
Geo
-
Marine
Sci
en
ce
s
(IJM
S).
Vol
44(12),
pp
.
1904
-
1919(ISI
-
Q4).
h
tt
p://nopr.n
isca
ir
.
res.
in
/ha
nd
le
/12
3456789/34939
.
BIOGR
AP
HI
ES OF
A
UTH
ORS
Suresh
Sundara
j
oo
is
a
Proj
ec
t
E
ngine
er
at
DN
C
Autom
at
ion
(M)
Sdn.
Bhd.
whi
c
h
is
lo
ca
t
ed
at
Puchong,
Ma
la
u
y
sia
.
He
has
b
een
with
DN
C
Aut
om
at
ion
sin
ce
2
017.
His
m
aj
or
work
scope
is
in
the
f
ie
ld
of
pr
ogra
m
m
abl
e
lo
gic
cont
rol
le
r
(
PLC
).
Previous
l
y
,
h
e
obt
ai
n
ed
Bac
h
el
or
of
Mec
hat
ron
ic
(Hons
) at U
niv
ersiti
Mal
a
y
s
ia
Paha
ng
in 2017
.
His
f
ina
l
y
ea
r
proj
ect
is
relate
d
to
th
e
are
as
o
f
au
tono
m
ous robot
,
ima
ge
proc
essing
an
d
computer
visio
n.
Ahm
ad
Shahriz
a
n
Abdul
Ghani
i
s
a
senior
lectur
er
a
t
Univer
si
ti
Malay
s
ia
Pah
an
g,
Mal
a
y
s
ia
.
He
rec
e
ive
d
M.E
ng
.
degr
ee
in
m
ec
h
atronics
from
Univ
ersity
of
Appli
ed
Scie
n
ce
s
Augs
burg,
Germ
a
n
y
in
2009.
He
r
ecei
ved
the
Ph.D.
degr
ee
in
image
proc
essing
and
computer
visio
n
s
y
stem
from
Univer
si
ti
Sains
Malay
s
ia
(US
M)
,
Mal
a
y
sia
in
20
15.
His
cur
ren
t
i
nte
rests
inc
lud
e
m
ec
hat
roni
cs,
col
or
image
p
roc
essing,
computer
vision
s
y
stem,
r
oboti
cs,
sensor
a
nd
instrumentati
on
s
y
st
em,
and
aut
om
at
ion
.
Evaluation Warning : The document was created with Spire.PDF for Python.