TELKOM
NIKA Indonesia
n
Journal of
Electrical En
gineering
Vol.12, No.7, July 201
4, pp
. 5522 ~ 55
2
8
DOI: 10.115
9
1
/telkomni
ka.
v
12i7.517
0
5522
Re
cei
v
ed
No
vem
ber 1
9
, 2013; Re
vi
sed
Jan
uar
y 6, 20
14; Accepted
February 2, 2
014
Mathematical Programming Model on Joint and
Recovery of Paper Scrap
Jiang Zhixia
*
1
, Yang Jianjun
2
, Luo Haifeng
3
, We
n Zhengqian
g
4
1
Departme
n
t of Appli
ed Math
e
m
atics, School
of
Science, Ch
angc
hu
n Univ
e
r
sit
y
of Scie
nce
and
T
e
chnolog
y, C
han
gch
un, Jili
n
,
13002
2, Chi
n
a
2-
4
School of Co
mputer Scie
nc
e and T
e
chno
l
o
g
y
, Cha
ngc
hu
n Univ
ersit
y
of Scienc
e an
d T
e
chn
o
lo
g
y
,
Cha
ngch
un, Jil
i
n, 130
02
2, Chi
n
a
*Corres
p
o
ndi
n
g
author, e-ma
i
l
: zhixia
_ji
ang
@12
6
.com
A
b
st
r
a
ct
T
he jo
int
and
r
e
covery
of p
a
p
e
r scrap
w
a
s a
n
o
p
timal
matc
hin
g
iss
ue. T
h
e a
u
to
matic r
e
storation
techno
lo
gy for
the
broke
n
d
o
cu
me
nt w
a
s
desi
gne
d
acco
rdin
g to
differ
ent cuttin
g
w
a
ys of the
pa
p
e
r
shred
der. As the frag
me
nt d
a
ta w
a
s an o
ne-si
de pr
i
n
t file, a no
nli
near
progra
mming
mo
del w
i
th n
o
constrai
nt con
d
itio
ns w
a
s establis
he
d for th
e straig
ht
-cutting br
oken
pa
p
e
r scrap fro
m
a giv
en pr
int fil
e
of
the sa
me
pag
e
,
and
the
no
nl
i
near
pro
g
ra
mmi
n
g
mod
e
l
w
i
t
h
restra
int pr
o
g
ra
mmin
g
mod
e
l w
a
s
estab
lis
hed
for the broke
n
scrap w
i
th straight cut and cr
o
ss cut. F
o
r the
pap
er scrap d
a
t
a from a o
ne-
pag
e pri
n
t file i
n
Engl
ish pri
n
ted
on both si
des, the joi
n
t w
a
s
accomplis
he
d b
y
the ant colo
n
y
algor
ith
m
.
Ke
y
w
ords
:
jo
i
n
t, the least sq
uare
meth
od, a
n
t colony a
l
g
o
ri
thm
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 . Introducti
on
Curre
n
tly, a lot of re
sea
r
ches
on the
a
u
toma
tic
re
storation te
ch
n
o
logy for the
bro
k
en
document ha
ve been don
e in most of the develop
ed cou
n
trie
s internation
a
l
l
y. However,
in
Chin
a, the rese
arch on t
h
is a
r
ea
started t
oo late, and few
re
search a
c
com
p
lishm
ents
were
repo
rted.
The
joint of
the
b
r
oken
do
cum
ent wa
s wi
del
y applie
d in
the ju
dici
al p
h
y
sical
evide
n
c
e
recovery, hi
st
orical
docum
ent re
storation and military inform
ati
on
acqui
rem
ent. T
r
aditionally, the
joint and
re
co
very wo
rk
ne
eded to b
e
d
one ma
nually
, which wo
uld
lead to hig
h
accuracy b
u
t low
efficien
cy. Especi
a
lly a
s
th
e fra
g
ment
a
m
ount
wa
s
h
uge, the
ma
n
ual joi
n
t woul
d cost
a l
o
t
o
f
human a
nd m
a
terial resource
s.
The auto
m
atic joint issu
e
belon
ged to t
he co
mpute
r
vision an
d pa
ttern re
co
gnition area,
whi
c
h was a
c
compli
she
d
by the comp
uter proc
essi
ng to obtain
the informati
on of the pa
per
scrap, li
ke
shape
and
color. T
hen t
he pa
pe
r
scrap wa
s re
stored autom
atically
o
r
semi-
automatically. Curre
n
tly, most of the
pape
r
scrap
joint wo
rks
were a
c
com
p
lish
ed ma
n
ually.
Although so
me re
sea
r
ch
ers h
a
ve bee
n done ab
ro
a
d
, little resea
r
ch findi
ng
s could be foun
d for
the ap
plicatio
n ba
ckg
r
oun
d
sp
ecifi
c
ity of the p
ape
r
scrap
autom
atic re
st
oration t
e
ch
nolo
g
y. T
h
e
descri
p
tion o
n
the sh
ape
matchin
g
al
g
o
rithm of
t
he
simila
r i
s
sue
wa
s fou
nd i
n
some
a
r
ticle
s
[1-
7]. For exam
ple, in literatu
r
e [7], a meth
od to a
c
comp
lish the
pap
er scra
p joint b
y
judging if t
w
o
profile
s mat
c
hed by b
oun
dary a
nd a
r
e
a
pri
n
cipl
es
from extra
c
ting
the contou
r l
i
ne of the
pa
per
scrap
wa
s
given to achie
v
e t
he pap
er scra
p auto
m
atic resto
r
a
t
ion ba
sed
o
n
the comp
u
t
er
as
sist
a
n
c
e
.
2. Broke
n
Paper Scrap o
f
One-side Print Do
cumen
t
w
i
th
Only
Straight Cut
For a given b
r
oken pa
pe
r scra
p from the
same p
r
int d
o
cum
ent with
only straight
cut, the
fragme
n
t dat
a of the d
o
cu
ment with
19
items o
n
two
page
s in
Chi
nese an
d En
glish
re
spe
c
ti
vely
wa
s given by the comp
uter
pro
c
e
ssi
ng, a
nd then the jo
int and re
cov
e
ry we
re do
n
e
.
The lea
s
t sq
uare met
hod
wa
s a mathe
m
atic optimi
z
ation techn
o
l
ogy, which m
i
nimize
d
the e
rro
r
sq
u
a
re
an
d fou
n
d
the
optimal
functio
n
m
a
tch. T
he
un
kn
own
data
cou
l
d be
cal
c
ulat
ed
conve
n
iently
and th
e e
r
ror sq
uare b
e
tween th
e
cal
c
ulated and a
c
tual data was
th
e
mini
m
u
m.
Nonli
nea
r le
ast squa
re
method
co
ul
d be
see
n
as a
spe
c
ial
situation
fo
r the u
n
restraint
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Mathem
atical
Program
m
i
ng Model on
Joint and Re
cove
ry of Pap
e
r Scrap (JIA
NG Zhi
x
ia
)
5523
minimization,
which wa
s
widely ap
plie
d in the area
s like d
a
ta fitting, para
m
et
er estim
a
tion
and
func
tional approximation
[8]. The optimal mat
c
hin
g
solution
for th
e p
a
pe
r
scrap
in
the
attachme
nt by the least sq
uare m
e
thod
and its value
wa
s determin
e
d.
From th
e b
a
sic
kno
w
led
g
e
of the im
age
pro
c
e
s
sing, i
t
wa
s
kno
w
n
that the pixel
value
rang
e of
ea
ch pixel p
o
int
on the
pap
er scra
ps to
be
jointed wa
s 0-25
5,acco
rdi
n
g
to whi
c
h, the
figure
coul
d b
e
analyzed. T
he figure wa
s firstly bro
k
en
up into the p
o
int colle
ction
,
and the pixe
l
matrix of the
point
colle
ct
ion of the
fig
u
re
wa
s
esta
blish
ed fo
r a
ran
dom fig
u
r
e, which
wa
s
assume
d a
s
j
th
figure. It was
kn
own
tha
t
the matrix o
f
the j
th
figure
wa
s 1
980*
7
2
, and th
e pi
xel
matrix of the i
th
column of the j
th
figure was:
hij
ij
ij
ij
p
p
p
P
2
1
,
0
i
72
,0
j
19
Firstly, for all
the pa
per
scrap to
be
joint
ed,
the fi
rst
o
ne o
n
the
left
wa
s
dete
rmi
ned if
all
of the left bounda
ry point
s we
re
white,
namely t
he pixel wa
s 25
5, from whi
c
h the first pa
per
scrap to be jo
inted wa
s fou
nd.
Obviou
sly, for a ra
ndom f
i
gure, the
co
rre
sp
ondi
ng
pixel point color differen
c
e of the
other figu
re
a
nd it wa
s n
o
t the sam
e
. Th
e scre
enin
g
functio
n
q
p
Q
,
was
defined, a
n
d t
o
cal
c
ul
ate
the squ
a
re
sum of the pixel point colo
r value
differe
nce b
e
twe
e
n
the right bo
unda
ry of the p
th
figure an
d co
rrespon
ding le
ft boundary of
the q
th
figure, the expressi
on wa
s a
s
follows:
2
,
1
,
1980
1
,
72
,
,
)
(
q
h
h
p
h
q
p
p
p
Q
Whe
r
e,
hij
p
mean
t the h
th
row a
nd i
th
column
of the pixel point colo
r value matrix of the j
th
figure.
For Fi
gure p
,
as oth
e
r fi
gure
s
and it
match
ed, e
a
ch
q
p
Q
,
was
ob
tained, an
d
what
need
ed was the optimal
match, na
m
e
ly the nonl
i
near
un
rest
raint mathem
atic prog
ram
m
ing
model.
Min
q
p
Q
,
Then Fig
u
re
p and q could
be jointed.
The program
was
written
with the Ma
tlab so
ftwa
r
e
in the Acer compute
r
of
5750G
c
o
n
f
igu
r
ed
w
i
th
i5
pr
oc
es
so
r
,
th
e
r
u
nn
ing
time
wa
s 0.
1s, the
joint
wa
s
su
ccessf
ul and
the fig
u
re
of the recovery res
u
lt was
as
follows
:
Figure 1. In Chine
s
e
Figure 2. In English
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 12, No. 7, July 201
4: 5522 – 55
28
5524
3. Broke
n
Paper Scrap o
f
One-side Print Do
cumen
t
w
i
th Straigh
t
and Cr
oss
Cut
For a give
n b
r
oken p
ape
r
scrap f
r
om th
e sam
e
pri
n
t document
with strai
ght an
d cross
cut, the
frag
ment d
a
ta of
the
docume
n
t with
209
it
ems on
two
page
s i
n
Chi
nese a
nd E
n
glish
respe
c
tively were given b
y
the comput
er proc
essin
g
,
and then the
joint and re
covery wa
s do
ne.
3.1. Establis
hment of
No
nlinear Prog
ramming Model
In the situati
on of the
cro
ss
cut, the m
e
t
hod a
bove
wa
s imp
r
ove
d
. Firstly, ea
ch p
ape
r
scrap i
n
the
attachme
nt was an
alyze
d
, and the
mini
mum vertical
distan
ce b
e
twee
n the whi
t
e
pixel
point on
the
top
of e
a
ch pap
er scrap and
th
e f
i
rst n
on-white
pixel poi
nt was utili
zed
an
d
analyzed in the com
pute
r
to roughly j
udge the
sh
ape match o
f
the font. Accordi
ng to the
literatures, th
e geomet
rica
l characte
ri
stic of
the ro
w wh
ere the
paper
scra
p lay could
be
obtaine
d by e
x
isted te
chn
o
l
ogy, like
the
row hei
ght
of
the characte
r and
the
sp
ace bet
wee
n
th
e
c
h
arac
ter lines
[9].
Ran
dom pap
er scrap
s
we
re
pi
cked as the
i
th
and j
th
page, a
nd if t
he minim
u
m
vertical
distan
ce
s b
e
twee
n the
whi
t
e pi
xel point
on the top
of the i
th
and j
th
pape
r scrap
and the first n
on-
white pixel po
int were
mi
d
and
mj
d
,
mj
mi
d
d
d
-
ij
.
Acco
rdi
ng to
the figure
s
in this pa
p
e
r, the bo
un
dary value
o
f
d
was
0 an
d 5
respe
c
tively. If the value of
d
wa
s between
0 and 5, it wa
s
con
s
ide
r
e
d
that the i
th
and j
th
pape
r
scrap
could
b
e
jointed
in th
e same lin
e,
or el
se
the two could
not b
e
jointed
in th
e same lin
e f
o
r
the big differe
nce of the
ch
ara
c
ter
stru
ct
ure, wh
ich m
eant that the
two pap
er
scrap could n
o
t be
jointed. Th
en
the dista
n
ces of the p
e
a
ks
of i
th
pape
r
scrap
on the
lef
t
and
right a
n
d
the first no
n
-
white pixel
p
o
int ho
rizo
ntally were
cal
c
ulate
d
an
d
analyzed to f
i
nd the mi
ni
mum di
stan
ce
si
x
and
ti
x
, the bou
ndary valu
e o
f
which was chosen a
s
1
2
, 16 an
d 6 to
28. By the sa
me way, the
maximum character space
i
y
max,
bet
ween
the top a
n
d
bottom p
o
i
nt in anoth
e
r
re
straint
con
d
ition was determin
ed.
Figure 3. The
Characte
r of
the ith Paper
Scra
ps
In the
pro
c
e
s
s a
bove, th
e
binary
z
ation
method
was
adopte
d
. By
setting th
e th
reshold
value, the
co
nversi
on f
r
o
m
the g
r
ay l
e
vel imag
e t
o
the
bina
ryzation ima
ge
wa
s a
c
compli
she
d
[10]. The
gra
y
level value
of the
white
pixel poi
nt
was
marke
d
a
s
255,
whil
e th
e gray value
s
of
the othe
r pixe
l points were
marked
as 0.
In this
pro
c
e
ss, th
e poi
nt, who
s
e
pixel v
a
lue
wa
s
clo
s
e
to that of the
white pixel,
was
con
s
id
ere
d
as
the
whit
e point. Th
e
Form
co
uld b
e
expresse
d
as
follows
:
255
,
0,
e
l
s
e
u
t
he po
i
n
t
i
s
w
hi
t
e
Then, ba
sed
on the theory
above, for all the
paper
scrap to be joint
ed, the prog
ram wa
s
firstly compil
ed to ran
k
the pape
r scraps in
the d
e
scen
ding way accordi
n
g
to the minimum
prin
ciple fo
r the sum of
mi
d
and
si
x
, the pap
er
scra
p with th
e
minimum val
ue was
cho
s
e
n
as th
e
first eleme
n
t on the left top. Then ea
ch of t
he pa
per scrap
s was joine
d
by the recu
rren
ce
d
mi
y
m
a
x,i
x
si
x
ti
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Mathem
atical
Program
m
i
ng Model on
Joint and Re
cove
ry of Pap
e
r Scrap (JIA
NG Zhi
x
ia
)
5525
algorith
m
ba
sed on th
e pri
n
cipl
e of from
left to right a
nd from to
p to bottom. In the joint p
r
o
c
e
s
s
the least squ
a
re meth
od i
n
the first mo
del wa
s u
s
ed
to screen th
e pape
r scra
ps to be joint
e
d
with the
obj
e
c
tive pa
per scra
p, the
Min
q
p
Q
,
was then
cal
c
ul
ated, which me
ant the
difference
of paper
scra
p p and q wa
s the minimu
m, namely pa
per scrap
s p
and q were th
e best mat
c
h.
The optimi
z
e
d
model was
a nonlin
ear
re
straint optim
a
l
proble
m
:
Min
q
p
Q
,
ma
x
,
05
12
16
..
62
8
38
42
mi
mj
si
ti
i
dd
x
st
x
y
As the pape
r
scrap p a
nd q
were joi
n
ted
cro
s
swi
s
e, th
e obje
c
tive function
wa
s:
min
2
,
1
,
1980
1
,
72
,
,
)
(
q
h
h
p
h
q
p
p
p
Q
As the pape
r
scrap p a
nd q
were joi
n
ted
l
ength
w
ays, the obje
c
tive functio
n
wa
s:
min
72
2
18
0
,
,
1
,
,
1
()
ip
i
q
i
pp
Acco
rdi
ng to the model e
s
t
ablishment a
nd analy
sis, the pro
g
ram was compile
d with
Matlab software an
d the in
itial joint was
obtaine
d.
3.2. Artificial
Interfe
r
enc
e
Solution
In many situa
t
ions, as t
w
o
pape
r scrap
s
were both
co
nsid
ere
d
a
s
the suita
b
le o
ne to be
jointed with
the
obj
ective pape
r scrap by
the
co
mpu
t
er, the comp
uter
would
au
tomatically
ch
ose
one of the
m
whi
c
h
wa
s co
nsid
ere
d
the
better to join
t
without the l
o
gical an
alysi
s
. At this time
, it
wa
s ea
sy to produ
ce a la
rg
e-scal
e and e
x
tensive join
t error. He
nce the artificial int
e
rferen
ce wa
s
of signifi
can
c
e as the
solu
tion erro
r of the
comp
uter
happ
ened. A
s
sho
w
n in
F
i
gure
4 a
nd
5,
obviou
s
ly the
two fig
u
re
s could
be totall
y match
ed
wi
th the o
b
je
ctive figure,
and
in thi
s
situati
on,
if the wron
g f
i
gure
was pi
cked
by the
compute
r
(as
sho
w
n
in Fi
g
u
re
2), th
e ot
her
pap
e
r scrap
woul
d
b
e
joi
n
ted with
oth
e
r pape
r scra
ps wro
ngly
eith
e
r
, which
woul
d result in th
e
domi
n
o
effect.
Mean
while, if the artificial in
terfere
n
ce wa
s used
with t
he logi
cal thin
king a
nd an
alysis of huma
n
,
it could
be
re
alize
d
that th
e word
“cate
n
s” in th
e fig
u
re to
be
join
ted in Fi
gure
5 didn’t
actu
a
lly
exist, hen
ce t
he joint
erro
r
woul
d be
fou
nd a
nd
co
rre
cted,
whi
c
h
would le
ad to
a high
a
c
cura
cy
and rig
h
t re
storation
of the
literatures.
Figure 4. Join
ted Corre
c
t
F
igure 5. Join
ted Erro
r
Hen
c
e, the artificial interference nee
ded
to
be added
after the initia
l joint was o
b
t
ained,
and th
e inte
rf
eren
ce
way
wa
s: the
joint
co
uple
of
th
e pa
pe
r
scra
ps i
n
the
run
n
ing
re
sult
of the
prog
ram
was firstly foun
d, and th
en th
e
se
rial
numb
e
r a
nd th
e joi
n
t ord
e
r
we
re
then
re
corde
d
.
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 12, No. 7, July 201
4: 5522 – 55
28
5526
Another
se
cti
on of the
progra
m
was then a
dde
d
i
n
the initial j
o
int program
to achieve t
he
desce
nding o
r
de
r of the sum of
mi
d
and
si
x
for each pa
per scrap, the second
se
quen
cin
g
value obtai
ne
d wa
s
cho
s
e
n
as th
e first
element o
n
the left top to find the p
a
per
scra
p to
be
jointed ag
ain,
the seri
al nu
mber a
nd join
t order
we
re rec
o
rded, then the third, the forth etc
.
were
found until n
o
single
pap
er scrap
wa
s le
ft or after the
prog
ram
wa
s run for th
re
e
times o
r
more,
no n
e
w pa
p
e
r
scra
ps we
re fo
und
to
be joi
n
ted
wi
th othe
r p
a
p
e
r
scra
ps.
T
hen th
e a
r
tificial
interferen
ce
wa
s finished,
and th
e final
right
re
sult
wa
s a
c
hieve
d
by the a
r
tificial joi
n
t with
the
logical analysis on the character a
nd lan
guag
e.
In additio
n
, for En
glish d
o
c
ume
n
ts, th
e
variabl
e val
ue rang
e of t
he rest
raint
condition
coul
d be
ch
ange
d app
ro
priately be
ca
use of the
differen
c
e b
e
t
ween En
glish and
Chin
e
s
e
cha
r
a
c
ter, the
following
wa
s used:
42
23
28
6
16
12
16
0
ma
x
y
x
x
d
d
t
s
mj
mi
The rig
h
t joint result obtai
n
ed wa
s a
s
Figure 6. an
d F
i
gure 7:
Figure 6. In Chine
s
e
Figure 7. In English
4. Broke
n
Paper Scrap o
f
both Sides P
r
int Doc
u
me
nt
For
a give
n b
o
th si
de
s p
r
in
t document
of
English ve
rsi
on fro
m
the
same p
a
ge
wit
h
both
straig
ht an
d
cro
s
s
cut by
the shredde
r, the jo
i
n
t a
nd resto
r
atio
n of the
pap
er
scrap
were
accompli
sh
ed
with th
e hel
p of the
both
sid
e
s
frag
m
ent data
of 2
09 d
o
cument
s give
n by th
e
comp
uter.
The
pape
r
scrap
coll
ectio
n
of the fi
rst li
ne &
colu
mn
and
the l
a
st
line &
col
u
mn after
being j
o
inted
wa
s firstly judged. Fo
r
wh
ich, a n
e
w v
a
riabl
e
ni
d
was
introduc
e
d to indicate the
minimum dist
ance
b
e
twee
n
the white p
i
xel
point
of
all the p
e
a
k
s at the
botto
m of the i
th
p
a
per
and the
first non-white
pix
e
l point ve
rtically. From th
e compute
r
p
r
og
ram, the
ni
d
,
mi
d
,
si
x
and
ti
x
could be o
b
tained an
d the four matrixes were e
s
tabli
s
he
d for the pape
r scrap give
n.
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Mathem
atical
Program
m
i
ng Model on
Joint and Re
cove
ry of Pap
e
r Scrap (JIA
NG Zhi
x
ia
)
5527
Acco
rdi
ng to the prin
ciple
that
mi
d
of
the first line elem
ents was big
g
er than oth
e
r eleme
n
ts,
and the
difference bet
wee
n
mi
d
and t
he el
e
m
ents i
n
ea
ch line
sho
u
ld
not be too
big
,
and by the
artificial i
n
terf
eren
ce
to ob
serve
the val
ue of e
a
ch el
ement in th
e
matrix, the 1
9
pap
er
scra
ps in
the first line,
the la
st li
ne
, the first
col
u
mn
and
the
last
column
co
uld
be
fo
und
so
on. T
hen
according to the prin
cipl
e that only one
comm
on ele
m
ent existed
in a rand
om li
ne or colum
n
, the
first eleme
n
t in the first lin
e coul
d be d
e
termin
ed so
on, and the
optimal match for ea
ch p
aper
scrap
coul
d b
e
found a
c
cording to the pri
n
cipl
e of fr
om
left to right and from top to bottom. In the
same
way, the la
st elem
en
t of the first li
ne, the
first el
ement of th
e l
a
st lin
e an
d t
he la
st ele
m
e
n
t
of the l
a
st li
n
e
were d
e
termined,
and
e
a
ch
mat
c
h re
sult coul
d be
obtain
e
d
by taking
the
three
element
s as t
he initial obje
c
tive pape
r scra
p to be ma
tched.
For the rest
ri
ction of the b
o
th side
s p
r
in
t, t
he algorith
m
accuracy
wa
s lo
we
red. Hence, in
the matchin
g
process of the obje
c
tive pape
r sc
rap and othe
r pa
per scrap
s, the ant colo
n
y
algorith
m
wa
s ado
pted.
The ant colo
ny algorithm
wa
s a prob
a
b
ilisti
c an
d o
v
erall situatio
n sea
r
ch method, and
the uncertai
n
ty property in
this algo
rith
m woul
d le
a
d
to more o
p
p
ortunitie
s
to
find the optimal
solutio
n
in
th
e overall
situ
ation [11
-
13].
The
ant
col
ony alg
o
rith
m co
uld i
n
crease the
ma
tch
accuracy for
the pape
r scrap, it
s optimi
z
ation p
r
o
c
e
s
s did not re
ly on the rigo
rous m
a
them
atic
prop
erty of itself, and it ha
d the potenti
a
l parall
e
lism
,
which in
cre
a
se
d the efficiency an
d instan
t
rea
c
tion a
b
ility of the algo
rithm. In the
matchin
g
p
r
oce
s
s, the p
aper
scrap n
u
mbe
r
be
ca
me
large
r
, an
d th
e pixel of th
e
figure
s
wa
s
smaller, h
e
n
c
e
the ant
col
o
n
y
algorithm
was
use
d
in
thi
s
pape
r to pro
c
ess optimally and in
cre
a
se the match a
c
curacy.
The four m
a
tchin
g
re
sults from the proc
e
s
s above
were
ob
served, and the
artificial
interferen
ce
wa
s a
dde
d at
this time, wh
ich
wa
s to
re
cord th
e
right
joint
re
sults,
adju
s
t the
wrong
jointed
pap
er scrap
s
whi
c
h was a
r
tifici
ally re
cog
n
ized, an
d
re-m
atch th
e
wro
ng joi
n
ted
pa
per
scrap
s
whi
c
h
could
not be
artificially re
cog
n
ized fro
m
the pape
r
edge. Th
e proce
s
s sh
ould
be
repe
ated until
the right matchin
g
re
sults
were obtain
e
d
:
Figure 8. Join
t Result
s
5. Conclusi
on
The technolo
g
y
of
joint and re
covery of
paper
scra
p
are con
s
ide
r
ed in sh
ape
before.
This pa
pe
r give a method com
b
ine
d
ant colony
algorith
m
an
d the least
squ
a
re m
e
th
od
according
to
the word
s m
e
ssag
e in
th
e pa
per
sc
ra
p. Ho
weve
r,
for the
pa
per scrap
with
big
informatio
n, a great amo
u
n
t
of artificial interfere
n
ces
was ne
ede
d in this model.
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 12, No. 7, July 201
4: 5522 – 55
28
5528
Referen
ces
[1]
Da Gama Le
i
t
ao, HC Stolfi
J. A Multiscale
Meth
od f
o
r the Reass
e
mbl
y
of T
w
o
-
dimens
io
na
l
F
r
agmente
d
Ob
jects.
IEEE. Pa
ttern Analysis
and Mac
h
in
e Intelli
ge
nce
. 20
02; 24(9): 1
239
–12
51.
[2
]
Yi
g
i
tsoy
M, Na
va
b N
.
Structu
r
e
Prop
ag
a
t
ion for
Imag
e R
egistrati
on
. Medic
a
l
Imagi
ng.
IEEE
T
r
ansactio
n
s o
n
. 2013; 3
2
(9): 165
7 – 16
70.
[3]
Kong WX
, BB
Kimia.
On
S
o
lv
ing 2D
an
d
3
D
Pu
zz
l
e
s Usi
ng Curve
M
a
tchi
n
g
.
Proce
edi
ngs
of Com
pute
r
Visio
n
and P
a
ttern Rec
ogn
itio
n, Ha
w
a
i
i
. USA. 2001; (2): 583-5
90.
[4]
Z
hu LJ, Z
hou ZT
, Hu DW
.
Globa
ll
y Co
nsi
s
t
ent Reconstr
uction of Ri
pp
ed-U
p
Docum
ents.
Pattern
Analys
is and M
a
chi
ne Intel
lig
e
n
ce, IEEE Transactions o
n
, Issue: 200
8; 3.1 – 13.
[5]
Liu
HR, Ca
o
SJ, Yan SC
. Automated
Assembl
y
of Shred
d
e
d
Pie
c
es F
r
om Mu
ltiple
Phot
os,
Multimed
ia.
IEEE Transactions on
. 20
11; 1
3
( 5): 1154
– 1
162.
[6]
Richter F
,
Ri
es C
X
, Ce
br
on N. L
i
en
ha
rt R.
Learn
i
n
g
to Re
asse
mble S
h
red
d
e
d
Doc
u
ments,
Multimed
ia
, IEEE Transactions on
. 20
13; 1
5
(3): 582 –
593
.
[7]
Jia HY, Z
hu LJ, Z
hou ZT
, Hu DW
.
A Shape Matchin
g
Method for Auto
matic Reass
e
mbl
y
of Pape
r
F
r
agments.
Co
mp
uter Si
mu
lat
i
on
. 20
06; 23(
1
1
): 180-1
83.
[8]
Yuan Y
X
, Sun
W
Y
,
Optimi
z
a
t
i
on theory a
nd
meth
od
, Bei
jin
g: Science pr
e
ss, 1997.
[9]
Luo Z
Z
.
Sem
i
-auto stitch
in
g of scra
ppe
d
pa
per
bas
e
d
on
char
acter char
acterist
ic.
Co
mp
ute
r
Engi
neer
in
g an
g Appl
icatio
ns
. 201
2; 48(5): 20
7-21
0.
[10]
Yao Y
H
, Ch
en
g Y. A Bi
nar
iza
t
ion M
e
t
hod
to
Process Sc
ann
ed Ima
ges
of ID Car
d
. Mac
h
i
ne B
u
il
din
g
&
Autom
a
tion
. 20
02; (5): 65-68.
[11]
Yang JF
. Ant colo
n
y
alg
o
rithm
and its ap
plic
a
t
ion rese
arch. Z
heji
ang U
n
iv
e
r
sit
y
, doctor
a
l thesis. 20
07.
[12]
Qi CM. Vehicl
e Routi
ng Opti
mizatio
n
in L
o
g
is
tics Distrib
u
t
ion Usi
ng H
y
brid Ant Co
lo
n
y
Al
gorithm
.
T
E
LKOMNIKA Indon
esi
an Jou
r
nal of Electric
al Eng
i
ne
eri
ng
. 2013; 11(
9): 5308-
531
5 .
[13]
Len
in Ka
nag
as
aba
i, B Ravin
d
r
anath R
e
d
d
y
,
M Sur
y
a K
a
lav
a
thi.
Optimal P
o
w
e
r F
l
ow
using Ant Colo
ny
[14]
Search
Alg
o
rit
h
m to Eva
l
u
a
te L
oad
Curta
i
lment
Inc
o
rpor
ating
Volta
g
e
Stabil
i
t
y
M
a
rg
i
n
Criter
ion.
Internatio
na
l Journ
a
l of Electr
ical
a
nd Co
mp
uter Engi
ne
erin
g (IJECE)
. 201
3; 3(5): 603-6
1
1
.
Evaluation Warning : The document was created with Spire.PDF for Python.