Internati
o
nal
Journal of Ele
c
trical
and Computer
Engineering
(IJE
CE)
V
o
l.
5, N
o
. 1
,
Febr
u
a
r
y
201
5,
pp
. 46
~54
I
S
SN
: 208
8-8
7
0
8
46
Jo
urn
a
l
h
o
me
pa
ge
: h
ttp
://iaesjo
u
r
na
l.com/
o
n
lin
e/ind
e
x.ph
p
/
IJECE
A Robotic Assistance Machine Vision Technique for an
Effective Inspection and Analysis
San
t
osh
Kum
a
r S
a
h
o
o*, B.
B. Ch
ou
dhur
y**
* Depart
em
ent o
f
El
ectron
i
cs
&
Tel
ecom
m
unicat
ion Eng
i
n
eer
ing, IGIT, Sarang
, U
t
kal
University
, Odisha,
India
**Departm
ent
of
M
echan
ic
al
Eng
i
neer
ing,
IGIT,
S
a
rang
Article Info
A
B
STRAC
T
Article histo
r
y:
Received Oct 19, 2014
Rev
i
sed
D
ec 17
, 20
14
Accepte
d Ja
n
4, 2015
An Inspection
is a stud
y
of
methods
and
tech
niques th
at
can
be suitab
ly
emplo
y
ed in pr
actical
appl
icatio
ns. In this pap
e
r
,
a n
e
w activity
is proposed
and an
al
ys
is
fr
am
ework to fa
cili
tat
e
the
ins
p
ect
ion of
an o
b
jec
t
us
in
g
machine vision
techniqu
e in which maxi
m
u
m
effici
enc
y
c
a
n be
a
c
hiev
ed.
B
y
using LABVIEW software and
vision build
er s
o
ftware
the qu
ality
of ou
tput
im
ages such as im
age com
p
re
ssion, im
age re
storation and
m
u
ltim
edia
streaming are achiev
ed successfully
.
So the proposed design makes use of
various image processing functions like
spe
c
ia
l filte
r
s a
nd cla
ssifie
r
s to
com
pute th
e op
t
i
m
u
m
results. U
s
ing sm
art c
a
m
e
ra in
th
e inspe
c
t
i
on s
y
st
em
the static as well
as
the d
y
n
a
mic object is capt
ured in fr
action
of seconds
without an
y
b
l
urs; as a result th
e optimum i
m
age quality
w
ithout an
y
distortion is ob
tain
ed for better anal
y
s
is. The proposed s
y
stem is v
e
r
y
precis
e
, a
ccur
a
te
and fl
exibl
e
wit
h
reas
onabl
e dev
e
lopm
ent cos
t
c
o
m
p
ared t
o
other model. W
ith the
aid of
an Indus
trial rob
o
tic s
y
s
t
em with simulation
software the ob
ject
is re
placed
immediately
when the sa
me is rejected b
y
th
e
machine vision
model. Apart from
this, the proposed model can be
implemented fo
r
an
y
ty
pe of Automation work
Keyword:
I
ndu
str
i
al Robo
r
t
LABV
IE
W s
o
f
t
ware
Sm
art Cam
e
ra
Vision
b
u
ild
er
Copyright ©
201
5 Institut
e
o
f
Ad
vanced
Engin
eer
ing and S
c
i
e
nce.
All rights re
se
rve
d
.
Co
rresp
ond
i
ng
Autho
r
:
San
t
osh Ku
m
a
r
Saho
o
Depa
rt
em
ent
of El
ect
r
oni
cs
&
Tel
ecom
m
uni
cat
i
on E
n
gi
nee
r
i
n
g
Ig
it, Saran
g
, Utk
a
l
Un
iv
ersity,
Odisha, India
Em
a
il: san
t
o
s
h.kr.sah
oo@g
m
ail.co
m
1.
INTRODUCTION
It is an
ob
ligato
r
y step to
ev
alu
a
te th
e
q
u
a
lity of ou
t
p
u
t
im
ag
es in m
a
n
y
i
m
ag
e p
r
o
cessi
n
g
ap
p
lication
s
such
as im
ag
e acq
u
i
sition
,
co
mp
ressi
on
, re
storatio
n, transm
i
ssio
n
, etc. Since h
u
m
an
b
e
ing
s
are
th
e u
lti
m
a
te o
b
serv
ers o
f
the p
r
o
c
essed
i
m
a
g
es and
th
us th
e ju
dg
es
o
f
imag
e q
u
a
lity, it is
h
i
g
h
l
y d
e
sired
to
d
e
v
e
l
o
p au
t
o
matic ap
p
r
o
a
ch
es th
at can
p
r
edict p
e
rcep
t
u
al imag
e qu
ality c
o
n
s
isten
tly with
hu
m
a
n
sub
j
ectiv
e
evaluation.
Machine visi
on
is
a branch of
en
gi
ne
eri
n
g t
h
at
us
es com
put
er
vision in the
context
of
manufact
uri
n
g, whe
r
e the im
a
g
es analysis is done to
ex
tract
d
a
ta for co
n
t
ro
llin
g
a
p
r
o
cesso
r activ
ity. Mach
ine
vision process
e
s are targete
d
at r
ecogn
izing th
e actu
a
l
ob
jects in
an im
a
g
e an
d assign
i
n
g prop
erties to
tho
s
e
ob
ject
s-
u
nde
rst
a
ndi
ng
w
h
at
t
h
ey
m
ean. En
gi
nee
r
s are a
d
di
n
g
M
achi
n
e
vi
si
o
n
sy
st
em
s t
o
a num
ber
of
industrial applications to
re
duce co
st
s, i
n
cre
a
se t
h
r
o
u
g
h
p
u
t
,
an
d i
m
prove
cu
sto
m
er
satisfactio
n
.
All
m
a
ch
ine
vision system
s
include a c
o
m
b
ination of
hardware a
n
d
soft
ware t
o
ac
qui
re an
d
pr
o
cess i
m
ages, usu
a
l
l
y
resulting in a
response
from
a seconda
ry sy
stem
connecte
d
to t
h
e ins
p
ec
tion system
. There a
r
e two
possible
designs for thi
s
refere
nce arc
h
itecture.
One
include
s
a l
o
w
-
cost
,
ru
g
g
ed
, em
bedde
d s
o
l
u
t
i
on w
h
i
l
e
t
h
e ot
he
r
m
a
kes use of t
h
e p
o
we
r of a com
put
er t
o
acqui
re and
pr
oc
ess im
ages at
a hi
ghe
r rat
e
and re
sol
u
t
i
o
n
.
Vi
si
o
n
syste
m
s can
b
e
u
s
ed
t
o
precisely
measu
r
e
n
u
m
b
e
r of v
a
riab
les su
ch
as len
g
t
h, ang
l
e,
p
o
s
ition
,
o
r
ientatio
n
,
color and so
on. T
h
e m
a
in a
dva
ntage
of a
vision m
easur
in
g
system
in
t
h
ese app
lication
s
is its n
o
n
-
co
n
t
act
featu
r
e,
wh
ich
is i
m
p
o
r
tan
t
in
cases wh
en
it is
d
i
fficu
lt to im
p
l
e
m
en
t co
n
t
act
m
easu
r
em
e
n
t.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
J
ECE
I
S
SN
:
208
8-8
7
0
8
A Robo
tic Assi
sta
n
c
e Ma
ch
ine Visio
n
Tech
niq
u
e
f
o
r An
Effective In
sp
ectio
n and
Ana
l
ysis
(S
an
to
sh
K.S
.
)
47
Thi
s
pa
per p
r
o
pos
ed t
h
e de
vi
ces and m
e
t
hod t
o
devel
op a
n
effect
i
v
e i
n
s
p
ect
i
on sy
st
em
. The sy
st
em
co
nsists of
Sm
ar
t Cam
e
r
a
1
7
2
2
, a m
o
tio
n
platf
o
r
m
, Lab
view
W
i
n
dow
an
d NI
v
i
sion
dev
e
lop
m
en
t
mo
du
le.
In t
h
i
s
pr
o
pose
d
m
odel
t
h
e i
n
vest
i
g
at
i
o
n
of t
h
e ci
rc
ul
ar l
i
ne
ari
t
y
by
usi
n
g
t
h
e
NI
Vi
si
o
n
s
y
st
em
whi
c
h c
onsi
s
t
o
f
th
e
fo
llowing
Parts su
ch
as
[A]
Ill
u
m
i
natio
n
[B]
Ob
ject
[C]
Sm
art Cam
e
ra
[D]
Im
age
pr
o
cesso
r
[E]
Co
ntr
o
ller
(Ro
botic)
Fi
gu
re
1.
Sc
he
m
a
t
i
c
B
l
ock di
agram
of
O
b
je
ct
Ins
p
ect
i
o
n S
y
st
em
A. Illu
m
i
n
a
tio
n
:
Lig
h
ting
in
clud
es u
s
e
o
f
b
o
t
h
artificial lig
h
t
so
u
r
ces
su
ch
as la
m
p
s an
d
n
a
tural illu
m
i
n
a
tio
n
of in
teriors
fro
m
d
a
ylig
h
t
. Artificial lig
h
tin
g is m
o
st co
mm
o
n
l
y prov
id
ed
t
o
d
a
y
b
y
electric l
i
g
h
t
s.
Wh
en
t
h
e
inform
ation from
the sa
m
p
les is accu
m
u
lated and a
n
alyzed, with res
p
ect to the specific sa
m
p
le
and
inspection
requirem
ents, we
can ac
hi
eve t
h
e prim
ary goal of
NI
vision
l
i
ght
i
ng a
n
al
y
s
i
s
- t
o
p
r
o
v
i
d
e
sam
p
le appropriate lightin
g t
h
at m
eets three accepta
nce cri
t
eria consistent
ly:
Maximize the cont
rast on t
h
ose features
of i
n
terest
Minim
i
ze the cont
rast elsewhere
Provi
d
e for a
measure
of robustness
.
B. Obj
ect:
The
part
whi
c
h i
s
goi
ng t
o
be i
n
s
p
ect
e
d
i
n
t
e
rm
s of ci
r
c
ul
ar l
i
n
ea
ri
t
y
. In
t
h
i
s
part
i
c
ul
ar
pape
r t
h
e
cylindrical ba
r
is placed for inve
stigating its
circular li
nearity.
C. Sm
ar
t
Ca
m
e
r
a
1
722
The selection
of cam
era is heavily de
pe
nde
n
t on our appl
ication. If sele
ct an appropri
ate ca
m
e
ra, lens
and
l
i
g
ht
ni
n
g
set
u
p
,
t
h
e
ef
f
o
rt
s ca
n t
h
e
n
b
e
f
o
cus
e
d
o
n
devel
opi
ng
ou
r s
o
l
u
t
i
o
n,
rat
h
er t
h
a
n
w
r
est
l
i
n
g
wi
t
h
po
o
r
i
m
age
dat
a
a
n
d
sa
ves
pr
ocessi
ng
t
i
m
e
at
execut
i
on. He
re
NI s
m
art ca
m
e
ra1722 is selected
and
h
a
v
i
n
g
r
e
so
lu
ti
o
n
of
12
80x
102
4 w
ith 7.4x7
.4
µm
p
i
x
e
l size.
D. Im
age
process
o
r
It
i
s
used
f
o
r
pr
ocessi
ng
t
h
e
acq
ui
re
d i
m
age
fr
om
th
e smart ca
m
e
ra fo
r resu
lts
with actu
a
l
p
r
ed
efi
n
ed
param
e
ters.
E.
Co
n
t
ro
ller (Rob
o
tic):
The R
o
bot
i
c
cont
rol
l
e
r gi
ves
t
h
e com
m
a
nd si
gnal
t
o
t
h
e Robot after successful analysis by
m
eans of
circularity
m
a
tching
with pre
d
efi
n
ed param
e
ters;
as
a result the defective
objects can be replaced
form
t
h
e t
e
st
be
nc
h.
The B
l
oc
k
di
a
g
ram
of
O
b
ject
I
n
spect
i
o
n
Sy
s
t
em
present
e
d i
n
Fi
gu
re
1.
2.
ALGO
RITH
M
The al
go
ri
t
h
m
com
p
ri
ses t
h
e
vari
ous
anal
y
z
i
n
g
an
d
p
r
oces
s
i
ng
f
unct
i
o
ns i
l
l
u
st
rat
e
d
bel
o
w
.
Illu
m
i
n
a
tio
n
Sm
art ca
m
e
ra
[N
I-
17
2
2
]
Object to be
inspected
Im
age process
o
r
Ro
bo
t
(AR
I
ST
O-
2.
1)
Con
t
ro
ller
Refere
nce Im
age
Si
gnal
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
SN
:
2
088
-87
08
IJEC
E V
o
l
.
5, No
. 1, Feb
r
uar
y
20
1
5
:
4
6
– 54
48
Fi
gu
re
2.
Al
go
ri
t
h
m
chart
f
o
r
Ins
p
ect
i
o
n
of
a
n
ob
ject
The im
age is
c
a
ptured
by the NI
SMA
R
T CA
MERA
(N
I-17
22)
.
V
i
sion
Assistant offe
rs
three types
of
acq
ui
si
t
i
ons
:
sna
p
,
g
r
ab
, a
n
d se
q
u
ence
.
A
sna
p
ac
qui
re
s a
n
d dis
p
lays a
s
i
ngle im
age. T
h
e
gra
b
ac
quires and
d
i
sp
lays im
ag
es in
a co
n
tinuou
s m
o
d
e
at m
a
x
i
m
u
m
r
a
te, w
h
ich
is
u
s
ef
u
l
w
h
en
you
n
e
ed to
fo
cus th
e ca
m
e
r
a
.
The seque
n
ce
acqui
res im
ages according t
o
settings t
h
at speci
fy in t
h
e Seque
n
ce ta
b
of the
Ac
quisition
Interface
window and s
e
nds t
h
e im
ages to the Im
age Brows
e
r.
Figure
3. Act
u
al im
age
captured by
Sm
art Ca
m
e
ra
The origin im
age’s size is 1360 × 10
24, it is so big that it will waste
long tim
e
to proc
ess. So it is
necessa
ry to
resize the
ori
g
in im
age
b
y
scale o
f
0.2. B
u
t th
e resu
lt
m
u
st
m
u
lt
ip
ly
th
e sam
e
scale of
5
co
rresp
ond
ing
l
y to
recov
e
r its real
p
o
s
ition
i
n
o
r
i
g
in
im
ag
e.
Fi
gu
re
4.
Sel
e
c
t
i
ng t
h
e
de
si
re
d m
a
rk
poi
nt
s
Actual im
age to inspect
S
e
le
c
tin
g
s
e
ar
ch
ar
e
a
s
Fi
nd
m
easurem
ent
poi
nt
s
Convert pixel
value
t
o
Real world
coo
r
di
nat
e
val
u
es
Object Classifi
cation
result
Evaluation Warning : The document was created with Spire.PDF for Python.
I
J
ECE
I
S
SN
:
208
8-8
7
0
8
A Robo
tic Assi
sta
n
c
e Ma
ch
ine Visio
n
Tech
niq
u
e
f
o
r An
Effective In
sp
ectio
n and
Ana
l
ysis
(S
an
to
sh
K.S
.
)
49
Classificatio
n
id
en
tifies an
u
n
k
nown
sam
p
le
b
y
co
m
p
ar
ing
a set of its significant features
to a set of
feat
ure
s
t
h
at
c
once
p
t
u
al
l
y
re
prese
n
t
cl
asses
o
f
kn
ow
n
sa
m
p
les. The object classifier
uses
feature
vectors t
o
identify sam
p
les based
on t
h
eir shape
.
Col
o
r classifier
u
s
es co
lo
r
features to
id
en
tify sa
m
p
les b
a
sed
o
n
th
ei
r
color. Classific
a
tion i
n
volves
two phas
es: t
r
a
i
ning a
n
d class
i
fying. Trai
ning
is a pha
se during whic
h we teach
th
e
m
ach
in
e v
i
sio
n
so
ftware th
e typ
e
s o
f
sam
p
les
wan
t
to
classify duri
n
g the classi
fying phase. It can train
any num
b
er of sam
p
les to create a
set o
f
classes, wh
ich, later co
m
p
are to
un
kno
wn sam
p
les d
u
r
i
n
g
t
h
e
classifying
phase and store
d
in the cl
asses
in a classifier file. Trai
ning
might be a
one-tim
e proces
s, or it
might be
an increm
ental proc
ess re
peat t
o
a
d
d ne
w sam
p
les to e
x
isting
classes or t
o
c
r
eate se
veral c
l
asses
,
thus broa
de
ning
the scope of sam
p
le
s to classify. The
classifying
phase
cla
ssifies a sam
p
le accordi
ng t
o
how
si
m
ilar th
e samp
le feat
u
r
es are to
th
e
sam
e
features
of the
trained
sam
p
les.
The
nee
d
t
o
cl
assify is common in
m
a
ny
m
achi
n
e
vi
si
o
n
a
ppl
i
cat
i
ons
. Ty
pi
cal
ap
pl
i
cat
i
ons i
n
v
o
l
vi
ng
cl
assi
fi
ca
t
i
on i
n
cl
ud
e t
h
e f
o
l
l
o
wi
ng:
• So
r
t
i
n
g
• I
n
sp
ection
Here t
h
e g
e
om
et
ri
c
m
a
t
c
hi
ng al
go
ri
t
h
m
i
s
used t
o
det
ect
t
h
e
m
a
rk of ci
rcl
e
. Geom
et
ri
c
m
a
t
c
hi
ng i
s
an i
m
port
a
nt
t
ool
f
o
r
m
achine
vi
si
o
n
a
p
pl
i
cat
i
ons;
i
t
m
u
st
w
o
r
k
rel
i
a
bl
y
u
nde
r
vari
o
u
s,
som
e
t
i
m
e
s ha
rs
h,
co
nd
itio
ns.
In
au
to
m
a
ted
m
a
ch
in
e
v
i
sion
ap
p
lication
s
–
esp
ecially th
o
s
e in
corporated in
to
m
a
n
u
f
actu
ring
pr
ocess
– t
h
e
vi
sual
a
ppea
r
a
n
ce
of m
a
t
e
ri
al
s or c
o
m
pone
nt
s u
n
d
er i
n
s
p
ect
i
on ca
n c
h
a
nge
beca
use
of
fact
o
r
s
su
ch
as
v
a
rying
p
a
rt
o
r
ien
t
atio
n, scale, and
l
i
g
h
ting
.
Th
e geo
m
etric
m
a
tc
h
i
ng
too
l
m
u
st main
tain
its abilit
y to
lo
cate th
e tem
p
late p
a
ttern
s d
e
sp
ite th
ese ch
an
g
e
s.
The
geom
et
ri
c m
a
t
c
hi
ng p
r
oc
ess co
nsi
s
t
s
o
f
t
w
o st
a
g
es:
l
earni
ng a
n
d m
a
t
c
hi
n
g
.
Du
ri
n
g
t
h
e l
ear
ni
n
g
stage, the
ge
ometric features
from
th
e te
m
p
late image. The algorithm
orga
nizes a
n
d stores t
h
ese feat
ures a
nd
th
e sp
atial relatio
n
s
h
i
p
s
b
e
tween
th
ese
features in
a m
a
n
n
e
r th
at facilitates faster search
i
n
g
i
n
th
e i
n
sp
ectio
n
im
age. Duri
ng the m
a
tching stage, t
h
e
ge
om
etric
m
a
t
c
hi
ng al
go
ri
t
h
m
ext
r
act
s
geom
etric feature from
the
in
sp
ection
im
a
g
e th
at co
rrespo
nd
to
th
e feat
u
r
es in
th
e temp
late i
m
ag
e. Th
en, th
e alg
o
rith
m
fin
d
s
m
a
tc
h
e
s b
y
lo
catin
g
reg
i
on
s in
t
h
e in
spectio
n
im
ag
e wh
ere featu
r
es alig
n
th
em
selv
es in
sp
atial p
a
ttern
s simila
r to
th
e
sp
atial p
a
ttern
s of th
e feat
ures in
th
e tem
p
lat
e
s.
In La
b vi
ew
W
i
n
d
o
w
s/
C
V
I,
t
h
e funct
i
o
n i
m
aq Det
ect
C
i
rcl
e
s base
d ge
om
et
ri
c
m
a
t
c
hing i
s
use
d
t
o
p
e
rform
th
e tas
k
. In
th
e fu
n
c
ti
o
n
, th
e th
reshold
wh
ich
sp
eci
fies th
e
m
i
n
i
mu
m
co
n
t
rast a seed
po
in
t m
u
st
h
a
v
e
i
n
or
der t
o
beg
i
n a curve m
u
st
be set
careful
l
y
.The i
n
spect
i
on re
sul
t
i
s
l
i
s
ted i
n
Tabl
e I.
The co
or
di
nat
e
ori
g
i
n
is in
to
p
left corn
er of th
e i
n
spected
i
m
ag
e b
y
d
e
fau
lt. Th
ere
ex
its a little d
i
fferen
ce correspo
n
d
i
ng
to
d
i
fferen
t
resi
ze scal
e. T
h
e reas
o
n
l
i
e
s i
n
t
h
e i
n
t
e
rp
ol
at
i
on m
e
t
hod w
h
e
n
resi
ze
t
h
e ori
g
i
n
i
m
age usi
ng R
e
s
a
m
p
l
e
f
u
n
c
tion
in Lab
V
i
ew
W
i
ndow
s. B
u
t it
d
o
e
sn
’
t
af
f
ect t
h
e i
n
spectio
n r
e
su
l
t
if
w
e
conv
er
t
th
e coo
r
d
i
n
a
te to
the
benc
hm
ark.
Tab
l
e 1
.
Mark in
sp
ection
result
Scale 0.
6
0.
5
0.
4
0.
35
0.
3
0.
2
X (
p
ixel)
883.
33
880
882.
5
887.
14
833.
33
815
Y (
p
ixel)
685
688
652.
5
652.
86
656.
67
660
T
i
m
e
(
m
s)
22.
2
16.
9
16.
8
12.
2
11.
8
9.
2
2.
1
C
o
o
r
d
i
nate Co
nv
ers
ion
Before in
sp
ect
io
n
,
th
e system
m
u
st b
e
ab
l
e
to
relate reg
i
o
n
s
o
f
p
a
rticu
l
ar i
m
ag
es to
be in
sp
ected.
Howev
e
r, in
an
in
lin
e in
sp
ectio
n
system
, wh
en an
obj
ect m
o
v
i
n
g
in
t
o
the in
sp
ecti
o
n
positio
n
,
there are so
m
e
min
o
r
p
o
sitio
nal
o
r
o
r
ien
t
atio
n
errors b
ecau
s
e o
f
t
h
e
m
e
ch
an
ical
facto
r
and
th
e i
n
terp
o
l
ation
facto
r
. To
eliminate the deviation, it is necessary
to
alig
n
t
h
e obj
ect prio
r t
o
in
sp
ectio
n. Th
is p
a
p
e
r
p
r
esen
ts an
algo
rith
m
to rectify t
h
e error
from
the above
fact
or.
As s
h
o
w
n i
n
F
i
gu
re 5
,
P a
n
d
Q are m
a
rks i
n
dat
a
base
wi
t
h
coo
r
di
nat
e
(
X
p,
Y
p)
an
d (
X
q,
Y
q)
,
P
’
and
Q
’ a
r
e m
a
rks inspected from
a sam
p
l
e
objec
t
wi
t
h
co
or
di
na
t
e
(
X
p’
,
Y
p’
) a
nd
(
X
q’
,
Y
q’
).T
h
e m
i
dpoi
nt
o
f
P
and
Q
is
O
(
X
o,
Y
o)
, an
d t
h
e m
i
dp
oi
nt
of
P
’ a
nd
Q
’
is O’ (
X
o’
,
Y
o’). T
h
eir coordinate ca
n be calc
u
lated as the
fo
llowing
equ
a
tio
n
s
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
SN
:
2
088
-87
08
IJEC
E V
o
l
.
5, No
. 1, Feb
r
uar
y
20
1
5
:
4
6
– 54
50
2
X
X
X
q
p
O
[1]
2
Y
Y
Y
q
p
O
[2]
and
2
X
X
X
/
/
p
O
/
q
[3]
2
Y
Y
Y
/
q
/
p
O
/
[4]
O
/
O
X
X
X
P
=
/
/
p
X
X
q
q
p
X
X
[5]
O
/
O
Y
Y
Y
P
=
/
/
p
Y
Y
q
q
p
Y
Y
[6]
According to
above e
q
uation,
we ca
n
conv
ert th
e coo
r
din
a
te syste
m
o
f
th
e
real ob
j
e
ct in
to
th
e
dat
a
base
co
or
di
nat
e
sy
st
em
. A
pi
xel
p
o
i
n
t
i
n
a
n
ob
ject
wi
t
h
c
o
o
r
di
nat
e
(
P
’,
Q
’
)
ca
n
co
nve
rt
t
o
da
t
a
base
coo
r
di
nat
e
poi
nt
(
P
,
Q
)
as
fol
l
owi
n
g e
q
uat
i
o
n
P=
X
P
/
P
,
Q =
Y
P
/
Q
[7]
2.
2 P
a
t
t
ern
M
a
tc
hi
ng
a
nd I
n
spect
the
Wr
on
g P
a
rt
s
Pat
t
e
rn m
a
t
c
hing
qui
c
k
l
y
l
o
cat
es regi
o
n
s o
f
a gray
scale image that m
a
tch a
kn
ow
n re
fere
nce pat
t
e
r
n
,
al
so refe
rre
d t
o
as a
m
odel
or t
e
m
p
l
a
t
e
. Wh
en usi
n
g pa
ttern
m
a
tch
i
n
g
,
first create a te
mp
late th
at rep
r
esen
ts
the objects for which you ar
e searching. Pattern m
a
tching applicati
o
n
th
e
n
s
e
ar
c
h
es
fo
r
in
s
t
an
c
e
of th
e
te
m
p
late in ea
ch acquire
d image, calculating a score
for each m
a
tch. This score
re
lates how clos
ely the
te
m
p
late resem
b
les th
e lo
cated
m
a
tch
e
s. Pattern
m
a
tch
i
n
g
fi
n
d
s
tem
p
late
m
a
tch
e
s reg
a
rd
less
o
f
lig
h
t
i
ng
vari
at
i
o
n, bl
ur
,
noi
se, a
n
d ge
om
et
ri
c t
r
ansform
a
t
i
on suc
h
as shi
f
t
i
n
g, r
o
t
a
t
i
on,
or scal
i
n
g o
f
t
h
e t
e
m
p
lat
e
. N
I
pat
t
e
rn m
a
t
c
hi
ng t
e
c
hni
que
s
i
n
cl
ude
n
o
rm
al
i
zed cros
s-c
o
r
r
el
at
i
on,
py
ra
m
i
dal
m
a
t
c
hi
ng, scal
e-a
n
d r
o
t
a
t
i
on-
i
nva
ri
ant
m
a
t
c
hi
n
g
, a
nd i
m
age un
de
rst
a
n
d
i
n
g.
No
rm
al
i
zed
cross
-
c
o
r
r
el
at
i
on i
s
t
h
e m
o
st
com
m
on
m
e
t
hod
fo
r
b
u
ild
i
n
g a temp
late in
an
imag
e.
Th
e fo
llowing
is th
e b
a
sic con
cep
t
o
f
co
rrelatio
n
:
co
n
s
i
d
er a su
b
i
m
ag
e
Evaluation Warning : The document was created with Spire.PDF for Python.
I
J
ECE
I
S
SN
:
208
8-8
7
0
8
A Robo
tic Assi
sta
n
c
e Ma
ch
ine Visio
n
Tech
niq
u
e
f
o
r An
Effective In
sp
ectio
n and
Ana
l
ysis
(S
an
to
sh
K.S
.
)
51
R(
p
,
q)
of
size
k
×
l
with
i
n
an
im
ag
e
f
(p,q) of size
m
×
m
, where
k
≤
m
and
l
≤
n
. T
h
e
correlation be
twee
n
R(
p
,
q)
an
d
f
(
p
,
q
) at po
in
t
(
i
,
j
)
i
s
gi
ven
by
1
0
1
0
)
,
(
)
,
(
)
,
(
l
p
k
q
j
q
i
p
f
q
p
R
j
i
C
[8]
Whe
r
e
i
= 0, 1…
m
-1
,
j
= 0,
1, …
n
-1
, and
th
e
su
mm
at
i
o
n is tak
e
n
over th
e
reg
i
on
in
th
e im
age
whe
r
e
R
a
nd
f
o
v
e
rlap
. Basic co
rrelation
is v
e
ry sensitiv
e to
am
p
litu
d
e
ch
ang
e
s in th
e i
m
ag
e, su
ch
as
in
ten
s
ity, and
i
n
th
e tem
p
late.
For exam
ple, if the in
ten
s
ity o
f
t
h
e im
ag
e
f
is doubled. So are t
h
e
values of
C
.
W
e
ca
n
ov
er
c
o
me
sen
s
itiv
ity b
y
co
m
p
u
tin
g
t
h
e
no
rm
alized
co
rrelatio
n
co
efficien
t,
wh
ich
is
defin
e
d
as
)]
,
(
[
)
(
)
,
(
)
,
(
j
i
f
D
R
D
j
i
j
i
S
[
9]
)
,
(
)
,
(
)(
)
,
(
(
)
,
(
1
0
1
0
j
i
f
j
q
i
p
f
R
q
p
R
j
i
l
p
k
q
[1
0]
)
)
,
(
(
)
(
1
0
1
0
l
p
k
q
R
q
p
R
R
D
[1
1]
1
0
1
0
2
))
,
(
)
,
(
(
)]
,
(
[
l
p
k
q
j
i
f
j
q
i
p
f
j
i
f
D
[1
2]
whe
r
e
R
(calcu
lated
on
ly o
n
c
e) is th
e av
erage in
ten
s
ity v
a
lu
e of th
e
p
i
x
e
l
s
in
th
e tem
p
la
te
R
. The
vari
a
b
l
e
f
is
the avera
g
e value of f in the re
gion
co
in
ci
d
e
nt with
th
e cu
rren
t
lo
catio
n
of
R
. The val
u
e of
S
(
i
,
j
)
lies in
th
e rang
e -1
t
o
1 and is in
d
e
p
e
n
d
e
nt o
f
scale ch
ang
e
s in th
e in
ten
s
ity v
a
lu
es
of
f
and
R
. Normalized
cross
-
c
o
r
r
el
at
i
on i
s
a go
od t
e
chni
que f
o
r fi
n
d
i
n
g pat
t
e
r
n
s in an im
age when the pa
tterns
in the image are not
scaled or rotated. Typically,
cross
-
c
o
rrelation can detect
patterns
of t
h
e
sam
e
size
u
p
to
a ro
tation
of 5
°
t
o
10
°.
Scal
e i
n
v
a
ri
ant
m
a
t
c
hi
ng a
d
d
s
a
si
g
n
i
f
i
cant
am
ount
o
f
c
o
m
put
at
i
on
t
o
m
a
t
c
hi
ng
pr
ocess.
T
o
resol
v
e t
h
e
pr
o
b
l
e
m
,
NI i
m
proves t
h
e c
o
m
put
at
i
on t
i
m
e of pat
t
e
r
n
m
a
t
c
hi
ng
by
r
e
duci
ng t
h
e si
ze of t
h
e i
m
ag
e and t
h
e
te
m
p
late b
y
pyramid
al
m
a
tc
h
i
ng
.
In
p
y
ramid
a
l
m
a
tch
i
n
g
,
bo
th
t
h
e imag
e an
d th
e tem
p
la
te are sam
p
led
to
sm
a
ller spatial res
o
lutions. B
ecause t
h
e image is
sm
aller, m
a
tching is faster.
When matching is c
o
mplete,
onl
y
a
r
eas
wi
t
h
hi
g
h
m
a
t
c
h sc
ores
nee
d
t
o
be
co
nsi
d
e
r
ed
as
m
a
t
c
hi
ng a
r
eas
i
n
t
h
e
o
r
i
g
i
n
al
im
age.
3.
R
E
SU
LTS AN
D ANA
LY
SIS
We pe
rf
o
r
m
e
d t
h
e p
r
o
p
o
sed
al
go
ri
t
h
m
and m
e
t
hod wi
t
h
La
b Vi
e
w
W
i
n
d
o
w
s/
C
V
I
t
o
det
ect
t
h
e
q
u
a
lity o
f
an
circu
l
ar
o
b
j
ect. Fig
u
res 6-8
are th
e soft
ware u
s
er in
terface wh
ich
are desig
n
e
d
to
d
e
tect th
e
d
i
fferen
t
edg
e
p
o
s
ition
.
All th
e aforesai
d tech
n
i
qu
es
are fo
llowed
b
y
u
s
ing
th
e
NI v
i
sio
n
i
n
which
th
e
in
v
e
stig
ation
of th
e circu
l
arity o
f
an obj
ect
as fo
llo
ws
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
SN
:
2
088
-87
08
IJEC
E V
o
l
.
5, No
. 1, Feb
r
uar
y
20
1
5
:
4
6
– 54
52
Fi
gu
re
6.
Si
m
u
l
a
t
e
d o
b
j
ect
by
t
h
e s
o
ft
wa
re
Fig
u
re
7
.
Circular edg
e
p
a
rameters settin
g poin
t
s
Evaluation Warning : The document was created with Spire.PDF for Python.
I
J
ECE
I
S
SN
:
208
8-8
7
0
8
A Robo
tic Assi
sta
n
c
e Ma
ch
ine Visio
n
Tech
niq
u
e
f
o
r An
Effective In
sp
ectio
n and
Ana
l
ysis
(S
an
to
sh
K.S
.
)
53
Fig
u
re
8
.
Circular edg
e
p
a
rameters settin
g poin
t
s
4.
CO
NCL
USI
O
N
The c
o
m
put
at
i
onal
s
p
ee
d
of i
m
age pr
ocessi
ng
has
bee
n
i
m
prove
d u
s
i
n
g
M
achi
n
e
Vi
si
on
b
u
i
l
d
er
.
A
new
sy
st
em
based
on
m
achi
n
e vi
si
o
n
hel
p
s
t
o
cal
cul
a
t
e
the res
u
lts with
precise a
n
d ac
curate
values
.
Object
Classification
is an e
ffecti
v
e
techni
que to
classify or ins
p
ect the
Circ
ular linearity autom
a
tically and
give
resu
lts t
o
m
a
tc
h
with
t
h
e
p
r
ev
iou
s
ly sav
e
d
te
m
p
late. F
ilte
rs
h
e
lp to im
p
r
o
v
e
th
e qu
ality of an
im
ag
e. Imag
e
q
u
a
lity is b
a
sically d
e
p
e
nd
s up
on
t
h
e illu
m
i
n
a
tio
n
co
nd
itions wh
ich produ
ces no
ise in th
e
i
m
ag
e.
REFERE
NC
ES
[1]
Z. W
a
ng,
et
al
.
,
“
M
odern Im
age Quali
t
y
As
s
e
s
s
m
ent”,
Syn
t
h
e
sis Lectures o
n
Image, Video
,
and Multimed
ia
Proc
e
ssing
, vo
l.
2, no
.1, pp. 1-15
6, 2006
.
[2]
Richard P.
et
al., “A machine-vision
s
y
stem for ir
is recognition D
a
vid Samoff
Research C
e
nter, In
c”, Prin
ceton, N
J
08543-5300, USA ,1996, Page No: 1-8
[3]
F
r
anci Lha
j
nar
,
et al
., “
M
achi
n
e Vis
i
on s
y
s
t
e
m
for ins
p
ecting elec
tri
c
plat
es
”,
Els
e
vi
er
,
co
mputer in industry
,
47(2002), Pag
e
No: 113-122.
[4]
J
o
ze Dergan
c,
e
t
al
.,
“
A
m
achine
vis
i
on s
y
s
t
em
f
o
r m
eas
uring th
e e
c
c
e
ntri
cit
y
of
bear
ings
”,
Els
e
vier
, pag
e
no.10
3-
111.
[5]
Tadhg, Brosnan
,
et al. “Improving quality
inspection
of food products b
y
comp
uter vision––a r
e
view”,
Journal of
Food Eng
i
neerin
g
, 61
, (2004)
, Page No: 3–16
.
[6]
T. S. White ,et
al.,
: “A Mobile
Climbing Robot
for Hi
gh Precision Manufactu
r
e
and Inspection o
f
Aero-structur
e
s”,
The Internationa
l Journal
of
Rob
o
tics
Research
,
2005; DOI: 10
.1
177/0278364905055701, Page N
o
:- 589-598,
[7]
Se-gon Roh et al.
, “
Differential-
Drive In-Pipe Robot for
Moving Inside Urban Gas Pipelines”,
IEEE
transactions on
robotics, vol. 21
, no. 1, Fe
bruar
y
2005, Page No:1
-17.
[8]
Andr´
e
Treptow,
et a
l
.
,
“
R
ea
l-tim
e peopl
e tr
ack
in
g for mobile rob
o
ts using thermal vision”,
Elsevier: Robotics and
Autonomous Sys
t
ems
, 54
(2006),
Page No:
729–7
39.
[9]
Marko Heikkila et al
., “A Texture-Based Meth
od for Modeling
th
e Backgroun
d and Detecting
Moving Objects”,
IEEE transactio
ns on pattern an
alysis and mach
ine in
telligen
ce
,
vol. 28
, no
. 4
,
A
p
ril 2006
, Pag
e
No: 657-662.
[10]
Yakov Fra
y
m
a
n
,
et a
l
., “
M
achi
n
e Vision S
y
ste
m
for Auto
m
a
tic Inspection of
Surface Defe
cts
in Alum
inum
Die
Casting”,
Journ
a
l
of Advan
ced Computational
I
n
telligen
ce and
I
n
telligent Inform
atics
, Vol.10
No.3, 2006
, Pag
e
N
o
-
281-290.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
SN
:
2
088
-87
08
IJEC
E V
o
l
.
5, No
. 1, Feb
r
uar
y
20
1
5
:
4
6
– 54
54
BIOGRAP
HI
ES OF
AUTH
ORS
S.K. Sahoo, obtained his M.
Tech
in Electron
i
cs an
d Instrumenta
tio
n Engineering.
He is currently
a Ph.D Scholar, Dept. of Electronics & Comm
unication Eng
i
neer
ing at Utkal University
,
Bhubaneswar, Odisha, Ind
i
a Mr.
Sahoo
is a member of
ISTE.I
E
and IEEE
B.B. Choudhur
y
,
obt
ained
his
Ph. D.
in T
a
sk
Allocation
strateg
i
es in
Multi –
Robot
Environment. H
e
is currently
a Assistant Profe
ssor at I.G.I.T, Sarang
in the
Department of
M
echani
cal
E
ngineer
ing. Hi
s
current
res
earch
int
e
res
t
includ
es
Ro
botics
,
F
M
S
,
CAD/CAM/CI
M, and Soft
computing. Dr
. Choudhur
y
is a member of ISTE, I
E
(I)
,
IACSIT,
I
AENG.
Evaluation Warning : The document was created with Spire.PDF for Python.