TELKOM
NIKA
, Vol. 11, No. 7, July 201
3, pp. 3568 ~ 3575
e-ISSN: 2087
-278X
3568
Re
cei
v
ed
Jan
uary 6, 2013;
Re
vised Ap
ril
3, 2013; Accepted April 1
5
, 2013
Mobile Camera as a Human Vision in Augmented
Reality
Edmund Ng
Gaip Wen
g
*
, Rehman
Ulla
h Khan, Sha
h
ren Ahm
a
d Zaidi Adruc
e
,
Oon Yin Bee
F
a
cult
y
of Cog
n
itive Sci
ences
and Hum
an D
e
vel
opme
n
t (F
SCSHD), Univ
ersiti Mala
ys
ia
Sara
w
a
k
(UNIMAS) 943
00 Kota Sam
a
r
aha
n, Sara
w
a
k
,
Mala
y
s
ia
T
e
l:
+
60(82) 58
149
1, +
60(82)
581
49
2,
+
60(8
2
) 581
49
3, F
a
x: +
60(82) 581
5
67
*Corres
p
onding author, e
-mail: nggiap
w
eng@y
a
hoo.com
A
b
st
r
a
ct
T
he real w
o
rld
objects can b
e
recog
n
i
z
e
d
by
usin
g mark
er base
d
and
mark
er-less a
u
g
mente
d
reality syst
ems. Mostly, the previous
dev
e
lopers
used
m
a
rk
ers based augm
ented reality system
s.
How
e
ver, thos
e systems
actu
ally
hid
e
the r
e
ality a
nd it w
a
s
also
difficult to
keep th
e
mark
ers everyw
her
e.
F
u
rthermore, the previ
ous
mark
er-less a
ppro
a
ches us
e client-serv
e
r
architecture, w
h
ich
is
dr
astical
l
y
affected by
ne
tw
ork latency. Smart
pho
ne
camera
is
mat
u
red
en
oug
h that it can r
e
c
ogn
i
z
e
re
al w
o
rl
d
obj
ects w
i
thout markers. It can gui
de
users
abo
ut their loc
a
tion a
nd the d
i
rectio
n in a co
nven
ient w
a
y. T
h
e
use of Smartp
hon
e is best s
u
ited for o
u
tdo
o
r mo
bil
e
au
g
m
e
n
ted-r
eal
ity app
licati
ons. T
herefor
e, a ma
rker-
less natur
al features bas
ed t
r
acking
system in mobile augm
ented realit
y was form
ulated. In the adapted
framew
ork, the
state-of-the-ar
t al
gor
ith
m
(sp
eed
up r
obust
features)
w
a
s
mo
difi
ed for c
o
mputi
ng i
m
ag
e
features fro
m
li
ve mobi
le ca
mera i
m
a
ge
and
compar
es w
i
th local
l
y store
d
i
m
a
ges fe
ature
s
for recogn
itio
n
.
Moreov
er, the local static
dat
abas
e of locati
on tag
ged i
m
a
ge featur
es usi
ng SQ
Lite w
a
s
imp
l
e
m
ented t
o
bypass t
he s
e
rver. The pr
oposed system
w
a
s tested
in
a m
o
bile
AR-
p
rot
o
type
application us
ing iP
hone
called
UNIMA
S Guide. It was found fr
om the res
u
lts that t
he
adapted
m
a
rker-less system
c
o
uld r
e
cogni
z
e
the rea
l
w
o
rld
obj
ects in s
p
e
edy, easy
an
d
conve
n
i
ent
w
a
y. T
h
is tech
n
o
lo
gy ca
n b
e
app
lie
d i
n
tour
i
s
m
ind
u
stry, surge
r
y and ed
ucati
ona
l fields.
Ke
y
w
ords
:
au
gmente
d
rea
lity, marker-l
ess, outdoor, i
m
a
g
e
features, i
m
a
ge reco
gniti
on,
static databas
e
Copy
right
©
2013 Un
ive
r
sita
s Ah
mad
Dah
l
an
. All rig
h
t
s r
ese
rved
.
1. Introduc
tion
Augmente
d
reality usin
g
emerging te
chnolo
g
ies su
ch a
s
gl
obal
positio
ning
system
(GPS), accel
e
rom
e
ter, gyroscop
e, com
pass an
d mo
bile vision, provides a b
e
st opportunity
to
Smartpho
ne
use
r
s to expl
ore th
eir
su
rroundi
ng
s. Th
e re
al
worl
d
obje
c
ts
ca
n b
e
re
co
gni
zed
by
usin
g marke
r
ba
sed a
n
d
marke
r-le
s
s aug
mente
d
reality sy
stem
s.
Mostl
y
, the previous
develop
ers u
s
ed m
a
rke
r
s
based au
gm
ented reality system
s. Ho
weve
r,
t
h
o
s
e
sy
st
em
s a
c
t
ually
hide the
real
ity and it wa
s al
so difficu
lt to
keep th
e marke
r
s
e
v
erywhe
re. F
u
rthe
rmo
r
e, the
previou
s
marker-le
s
s a
pproache
s u
s
e
client-serve
r a
r
chite
c
tu
re,
which
is drasti
cally affe
cted
by
netwo
rk l
a
te
ncy. The ma
rke
r
s-b
a
sed
augme
n
ted r
eality was a
pplied in diff
erent field
s
l
i
ke
medical visu
alizatio
n, mai
n
tenan
ce
an
d repai
r,
nav
igation
and
entertain
ment
. Ho
weve
r, the
markers
are
not suitable f
o
r o
u
tdoo
r m
obile a
ugm
en
ted re
ality be
cau
s
e
ma
rke
r
s hi
de the
re
ality
and
need
to
kee
p
eve
r
ywhere
[1]. Its range
is also
very limited
a
nd e
nd-users often d
on’t li
ke
them. Marke
r
-less
natu
r
al f
eature
s
ba
se
d app
ro
ac
he
[2], can
re
cog
n
ize
re
al
worl
d obje
c
t
s
, su
ch
as si
ghts, b
u
ilding
s
, and
living beings and ov
e
r
come the
s
e
limitations. Robu
st feature
descri
p
tors
such
as SIFT [
3
], SURF [4],
and G
L
O
H
[5
] are mo
st sui
t
able for
appli
c
ation
s
su
ch
as
image
re
cog
n
ition [6] and
image
re
gistration [7
]. T
hese de
script
ors are
stabl
e und
er
different
viewpoi
nts a
nd lightin
g
condition
s. Th
ese
de
scri
pt
ors a
r
e id
eal
ly suited
for
sea
r
ching
im
age
databa
se
s b
e
c
au
se th
ey are re
pre
s
e
n
tin
g
feature poi
nts a
s
hi
gh-di
mensi
onal ve
ctors. However,
tracking
fro
m
natu
r
al feat
u
r
es i
s
a
com
p
lex pr
oble
m
and usually
p
e
rform
e
d
on
a
remote se
rve
r
[8], [9], [10]. It is th
erefore
a
ch
alle
nging
task to u
s
e
natu
r
al featu
r
e t
r
acking
in
m
obile
Augmente
d
Reality ap
plications.
Ho
we
ver, the m
obi
le pho
ne
s a
r
e very in
exp
ensive,
attractive
targets fo
r outdoor AR. T
he improve
m
ents in
Smartphone capa
bilities
and g
r
eat potential
of
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
e-ISSN:
2087
-278X
Mobile Cam
e
ra as a
Hum
a
n Vision in Au
gm
ented Rea
lity (Edm
und Ng Gai
p
We
n
g
)
3569
comp
uter/mo
b
ile visio
n
m
o
tivated us t
o
implem
ent marker-le
s
s natural
feature
ba
sed mo
bile
augme
n
ted reality.
A re
cent
exa
m
ple of
an A
R
o
b
je
ct tra
c
king
ap
plicati
on i
s
the
Su
doku G
r
ab
[1
1]. This
appli
c
ation
can tra
ck a S
udo
ku pu
zzle
and solv
e it,
adding the
missi
ng nu
m
bers in the e
m
pty
Sudoku sl
ots. Georg Klei
n et al. creat
ed an
appl
i
c
ation which can an
alyze t
he surrou
ndi
ngs,
ma
k
i
ng
it po
ss
ib
le to
re
n
der
d
i
ffe
re
n
t
3D c
h
a
r
ac
te
rs lo
ok li
ke
they a
r
e
sitting o
n
t
he d
e
sk i
n
th
e
physi
cal worl
d [12]. Occi
p
i
tal develope
d Red
La
se
r whi
c
h
can tra
ck it
s environ
ment. It is a g
ood
example to scan a
nd an
al
yze the ca
me
ra view by iPhone [13].
No
kia’
s MA
RA proj
ect
by
Kähäri
an
d M
u
rphy
[14] d
o
e
s
not
perfo
rm any i
m
ag
e
analysi
s
,
instea
d it uses an extern
al GPS for locali
zation
an
d an inertial
sen
s
o
r
to pro
v
ide orientati
on.
Phone Guid
e
[15]
is clien
t
-se
r
ver obje
c
t
re
co
gni
tio
n
sy
stems. T
he sy
stem
e
m
ploys
a ne
ural
netwo
rk train
ed to re
cogni
ze no
rmali
z
e
d
colo
ur
featu
r
es a
nd is u
s
ed as a m
u
se
um guide. Se
ifer
et al. [16] used a m
obile
system
ba
se
d on
a ha
nd-held d
e
vice,
GPS sen
s
o
r
,
and a
came
ra for
road
sid
e
sig
n
detectio
n
and invento
r
y. Their
algo
rithm ha
s go
od quality re
sults in m
o
b
ile
set
t
i
ng
s.
For o
b
ject d
e
tection a
nd
re
cog
n
ition, Frit
z et
al. [17] u
s
ed
a modifie
d
versi
on of the SIFT
algorith
m
. Th
e syst
em u
s
e
s
a
c
lie
nt-server archite
c
tu
re, where
a mobile phon
e
client ca
ptures
an image of an urb
an env
ironm
ent and
send
s it to
the se
rver for
analysi
s
. The
SURF algo
rit
h
m
has be
en u
s
ed su
ccessful
ly in a variety of applic
atio
ns, inclu
d
ing
an intera
ctive
museu
m
gui
de
[18]. Local
de
scripto
r
s
hav
e also be
en
use
d
for
tracking. S
k
rypny
k an
d Lo
we [
19] use the S
I
FT
feature
s
for
reco
gnition, tracki
ng, an
d
virtual obje
c
t
place
m
ent.
Came
ra tracking i
s
do
ne
b
y
extracting SI
FT features from a video f
r
ame, matchi
n
g
them ag
ain
s
t features in
a datab
ase, a
n
d
usin
g the co
rre
sp
ond
en
ce
s to comp
ute the came
ra pose. Takacs et al. [20] applied S
URF
feature
s
usi
n
g video co
der motion vecto
r
s fo
r m
obile
augme
n
ted reality applicat
ions.
Yeh et
al. [21
]
pro
p
o
s
ed
a
system
for
de
terminin
g a
u
s
er’
s
location
from
a m
obil
e
devi
c
e
via image ma
tching. Th
e a
u
thors first b
u
ild a “b
o
o
tst
r
ap data
b
a
s
e
”
of image
s o
f
landmarks a
n
d
train a CBIR algorithm o
n
it. Since the im
age
s in the bootstrap databa
se
are tagged
with
keyword
s
,
when
a q
u
e
r
y image
i
s
m
a
tche
d a
gain
s
t the
boot
st
rap
datab
ase
,
the a
s
soci
a
t
ed
keyword
s
ca
n be
used to
find more tex
t
ually relate
d
image
s throu
gh a
we
b sea
r
ch.
Finally, the
CBIR alg
o
rith
m is ap
plied
to the image
s retu
rn
ed from the web
sea
r
ch to pro
duce only th
ose
image
s that are visually rel
e
vant.
The pu
rpo
s
e
of this re
sea
r
ch is to find
solution for ma
rke
r-l
ess mo
b
ile augme
n
te
d reality
and
solve
th
e cli
ent-se
r
ve
r p
r
obl
em. T
he p
r
op
os
ed
frame
w
o
r
k
can di
scover the
surro
undi
ngs
and p
r
ovide i
n
formatio
n a
bout differe
nt object
s
.
Thi
s
sy
stem ca
n be u
s
ed
e
a
sily in different
fields of life just by cha
ngi
ng image
s fe
ature
s
data
b
a
se. Th
e syst
em, use
s
mo
dified versi
o
n
of
SURF [4] fo
r obje
c
t tra
cki
ng an
d recog
n
ition. In this re
sea
r
ch the
re
sea
r
chers
modified S
U
RF
becau
se origi
nal SURF u
s
es IPLImage
and iP
hone
camera gen
erate
s
UIIma
ge. The mob
ile
memory wa
s
saved by
redu
cing 64-e
l
ement
de
scri
ptor to 3
2
-e
lement
witho
u
t affecting i
t
s
efficien
cy. The numb
e
r of
octave
s wa
s
also
redu
ce
d
to two and n
u
mbe
r
of intervals p
e
r o
c
ta
ve
to three. In the formulate
d
approa
ch, fe
ature p
o
ints
were extra
c
te
d from in
comi
ng video fra
m
es
of iPhone ca
mera at ru
n-ti
me and mat
c
hed ag
ain
s
t a local data
b
a
s
e of feature
points a
nd G
P
S
data sto
r
ed
i
n
mobile. T
h
e GPS data
wa
s u
s
ed
as
index an
d pri
m
ary key in
databa
se
de
sign.
The se
archi
n
g query was optimized b
y
using this
prima
r
y key. After succe
ssful m
a
tchin
g
a
homog
ra
phy
matrix and t
r
ansfo
rmatio
n
matrix
we
re cal
c
ul
ated
from mat
c
hin
g
point
s u
s
ing
comp
uter vi
sion techniq
u
e
s a
nd m
o
b
ile se
nsi
ng t
e
ch
nolo
g
y such
as a
c
celero
meter a
n
d
comp
ass [12]
, [23].
2. Metho
dolog
y
iPhone 3GS
and ope
n su
rf library [24] were
used fo
r feature
s
extractio
n
and
creating
the de
scripto
r
s. Furth
e
rm
ore, Open
CV
2.2 [25] wa
s
use
d
for im
ag
e co
nversion
from UIma
ge
to
IPLImage an
d vice versa.
2.1. Open SURF
Open
-SURF l
i
bra
r
y [24] use
s
the SURF [4]
algorith
m
whi
c
h is o
ne of the best interest
point dete
c
to
rs a
nd d
e
scri
p
tors
cu
rrentl
y
avail
able. It has b
e
st p
e
rform
a
n
c
e a
s
compa
r
e
d
to
other inte
re
st
points
de
scri
ptors like
[3] and G
L
O
H
a
s
sho
w
n
by Mi
kolaj
c
zyk [5]. As mobil
e
ha
s a
limited re
sou
r
ce
s so th
e op
en su
rf library
was m
odifie
d
. The de
scri
ptor si
ze
wa
s redu
ce
d to 3
2
-
Evaluation Warning : The document was created with Spire.PDF for Python.
e-ISSN:
2087
-27
8
X
TELKOM
NIKA
Vol. 11, No
. 7, July 2013
: 3568 – 357
5
3570
element, nu
mber of o
c
ta
ves to two a
nd num
ber o
f
intervals pe
r octave to t
h
ree. Li
ke m
any
feature d
e
scriptor alg
o
rith
ms it also ext
r
act
s
re
gion
s of interest th
at tend to be
repe
atable a
nd
invariant und
er
tran
sform
a
tions su
ch
as brig
htne
ss o
r
p
e
rspe
ctive chang
e
s
. An ima
g
e
is
analyzed at
several
scale
s
, so inte
re
st points
ca
n be
extracte
d fro
m
acro
ss
all
possibl
e scal
es.
Additionally, t
he d
o
mina
nt
orientatio
n
of ea
ch
of
the
interest
poin
t
s is d
e
termi
ned to
sup
p
o
r
t
rotation
-invari
ant mat
c
hing
. An exampl
e imag
e a
n
d
its d
e
tecte
d
intere
st p
o
in
ts a
r
e
sho
w
n
in
Figure 1.
Figure 1. SURF interest p
o
ints dete
c
tio
n
on
iPhone
3GS, squ
a
re
s indicate fou
n
d
feature
s
In the descri
p
tor co
mputa
t
ion step, ea
ch ex
tra
c
ted
interest poi
nt defines a
circula
r
regio
n
from
whi
c
h on
e d
e
scripto
r
is
computed. Ea
ch inte
re
st p
o
int is a
s
soci
ated with a
32-
element
de
scriptor. It
wa
s
found th
at th
e de
script
o
r
size in
case
of iPhone
im
age th
us ran
ges
betwe
en 0KB
and
190KB
per im
age,
wi
th an ave
r
ag
e
of ro
ughly
35KB. In this way a
datab
ase
of more than
one hu
ndred i
m
age
s features
can be lo
a
ded in on
e ap
plicatio
n.
2.2. Feature
Data
base
A locatio
n
ta
gged
feature
s
d
a
taba
se
al
ong
with a
fra
m
ewo
r
k give
n by [2] were
used fo
r
getting GPS
data a
nd
rela
ted info
rmatio
n. The
inform
ation
wa
s
sto
r
ed
in
a SQlit
e data
b
a
s
e.
As
SQLite is a
serve
r-l
ess
static
data
b
a
s
e sy
stem
so
it wa
s
used
in the
re
so
urce fold
er.
The
latitude and l
ongitud
e
of locatio
n
we
re
employed
a
s
a p
r
ima
r
y key and ind
e
x in the datab
ase
desi
gn a
s
sh
own in Fig
u
re
2.
Figure 2. Re
cord for
storin
g image featu
r
es
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
e-ISSN:
2087
-278X
Mobile Cam
e
ra as a
Hum
a
n Vision in Au
gm
ented Rea
lity (Edm
und Ng Gai
p
We
n
g
)
3571
2.3. Sy
stem
O
v
erv
i
e
w
The syste
m
is fully implemented on th
e mobile
devi
c
e, and run
s
at close to re
al-time,
while m
a
intai
n
ing ex
cellen
t
reco
gnition
perfo
rm
an
ce. Whe
n
the
system
starts it gets vid
e
o
frame
s
fro
m
li
ve cam
e
ra. As iPho
ne
use
s
UIImage
cl
ass a
nd
Ope
n
CV lib
ra
ry u
s
e
s
the
IplImage
cla
s
s to h
o
ld
image
data.
Ope
n
CV
wa
s u
s
e
d
to
create a
n
IplI
mage i
m
ag
e
from
a
UIImage
image, and a
UIImage ima
ge from an Ip
lImage imag
e
.
Image features from live came
ra fra
m
es
were extra
c
t
ed an
d at th
e sa
me time
the u
s
er
lo
cation was cal
c
ulate
d
from
GPS data. Next,
these
feature
s
a
r
e
com
p
a
r
ed
with the
feature
s
stored in
datab
a
s
e. T
he q
u
e
r
y optimize
wa
s
optimize
d
by
the usi
ng GP
S position
as a prima
r
y ke
y and index.
In this way the exact record
wa
s a
c
cesse
d
witho
u
t scannin
g
the
whol
e data
b
a
se. T
hen
th
e su
rf d
e
scri
ptor
comp
ari
s
on
function
ality wa
s u
s
e
d
to
recogni
ze
the
obje
c
t. O
n
ce
the
obje
c
t i
s
re
co
gni
zed
then th
e
obje
c
t’s
homg
r
ap
hy matrix and
transfo
rmat
ion matrix
are cal
c
ulate
d
from the matche
d poi
nts
orientatio
ns.
The co
nceptu
a
l diagram of pr
op
osed fra
m
ewo
r
k is sh
own in Fig
u
re
3.
Figure 3. Con
c
eptu
a
l diag
ram of prop
osed frame
w
o
r
k
The re
sea
r
ch
ers inte
nde
d that the whole
pr
ocess to be done directl
y
on a mobile device
for seve
ral re
aso
n
s. Th
e p
r
opo
se
d fram
ewo
r
k
sig
n
if
icantly redu
ce
s the system l
a
tency a
s
it is
indep
ende
nt of serve
r
. Mo
reove
r
, it can also
work at any locatio
n
of the use
r
.
C
a
mer
a
image
and GPS dat
a
Extract features
Comp
are wit
h
feature
s
in
Datab
a
s
e
No
Guide
Use
r
Re
cog
n
ized?
Yes
Cal
c
ulate p
o
se
matrix
Augment re
al
ity
Start
End
Evaluation Warning : The document was created with Spire.PDF for Python.
e-ISSN:
2087
-27
8
X
TELKOM
NIKA
Vol. 11, No
. 7, July 2013
: 3568 – 357
5
3572
3. Resul
t
s
and
Discus
s
ion
The Ope
n
SURF lib
ra
ry wa
s applie
d as a f
eature descri
p
tor in
t
he system for iPhon
e
3GS devi
c
e
s
. Test
s were
con
d
u
c
ted a
n
d
re
sult
s were re
co
rde
d
u
s
ing iP
hone
3GS. The
scale
level pyra
mid
co
nsi
s
ts out
of two
octave
s
with
th
ree
l
a
yers ea
ch.
T
he
system
wa
s te
sted
outsi
de
the cam
p
u
s
. The Steve Job’
s pi
cture
feature
s
we
re compa
r
ed
for re
cogniti
on quality wi
th
Qual
comm S
D
K [26] an
d the propo
se
d
system. It
wa
s foun
d that the pr
opo
se
d frame
w
ork
ov
er
perfo
rms the
Qual
comm S
D
K [26] as sh
own in Fig
u
re
s 4 and 5.
From
Figu
re
s 1
and
5
and
from
the
an
alysis result i
t
is
obviou
s
t
hat Qu
alcom
m
can’t
track well thi
s
image. But
the propo
se
d sytstem
ca
n track this
well even in
different lighti
n
g
con
d
ition
s
and from different viewpo
ints. In
university ca
mpu
s
the prop
o
s
ed sy
stem
can
recogni
ze all
those d
epa
rtments an
d ce
nters
whi
c
h
were incl
ude
d in the feature datab
ase from
different viewpoints a
s
sho
w
n in Figu
re
s 6 and 7.
Figure 4. Analysis re
sult of
Qual
comm S
D
K
Figure 5. Feature
s
dete
c
ted by Qual
co
mm SDK
Figure 6. Ce
nter of excell
enc
e mono
gram re
cog
n
ize
d
by propo
se
d frame
w
ork i
n
bad lightin
g
condition
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
e-ISSN:
2087
-278X
Mobile Cam
e
ra as a
Hum
a
n Vision in Au
gm
ented Rea
lity (Edm
und Ng Gai
p
We
n
g
)
3573
Figure 7. Re
cog
n
ition of Faculty logo by
propo
se
d fra
m
ewo
r
k
A su
rvey of t
he
state of th
e art A
R
m
o
b
ile br
owse
rs
wa
s give
n by
Butcha
rt [27]
in Ma
rch
2011.To
com
pare
the
p
r
op
ose
d
system
with cu
rre
ntly availabl
e b
r
o
w
sers,
sele
ct
ed
colu
mn
s o
f
the survey ta
ble we
re eval
uated.
Table 1. AR
Browse
rs
Co
mpari
s
o
n
Product
Marker based
Mark
er-l
es
s
Offline mode
Platform
La
y
a
r
No
No
Online
onl
y
iPhone,
Android, S
y
mbia
n
Junaio
Y
e
s
Y
e
s
Online onl
y
iPhone, Android,
Nokia (N8)
Wikitde API
No
No
Offline
iPhone, Android
Wikitde Worlds
No
No
cacheable
iPhone, Android,
Sy
m
b
i
a
n
Sekai Camera
No
No
Online onl
y
IPhone, Android,
iPad,
iPodTouch
Libre Ge
osocial
source
plugin Online
onl
y
Android
Two
criteria
such
a
s
ma
rle
r
-less
and
offli
ne mo
de
we
re u
s
ed fo
r
evaluation
of Ta
ble 1. It
wa
s di
scovered that, the
most of
the
b
r
owse
r
can’t
sup
port m
a
rker-l
ess. However, some o
f
the
bro
w
sers can
supp
ort ma
rker-le
s
s whi
c
h depe
nd
s o
n
desktop
po
werful
se
rver using
netwo
rk.
Since the p
r
evious b
r
o
w
sers h
a
ve few limitat
ions such as n
e
t
work laten
c
y, uploading
and
downloadi
ng
of co
ntents.
The p
r
op
ose
d
fram
ew
o
r
k covered
th
e above
m
entio
ned
p
r
o
b
lem
s
by
usin
g local feature data
b
a
s
e.
4.
Conclusion
A
mar
k
e
r-le
s
s v
i
si
on-
ba
se
d A
R
sy
st
em
wa
s
p
r
e
s
e
n
ted that
re
co
gnize an
d track real
worl
d obje
c
t
s
in re
al-tim
e wi
thout m
a
rkers. Signi
ficant im
provement in pe
rforma
nce was
achi
eved by
usin
g stati
c
d
a
taba
se of i
m
age f
eatu
r
e
s
. The
pro
p
o
s
ed f
r
ame
w
o
r
k
explored n
o
vel
resea
r
ch dire
ction
s
. These
were p
r
eviou
s
ly
only possi
ble with de
sktop comp
uters and no
w ca
n
be exe
c
uted
with a m
obil
e
device. Th
e develo
per
and p
r
og
ram
m
ers can a
p
p
ly the pro
p
o
s
ed
frame
w
ork in
tourism indu
stry, games an
d edu
cation
al
fields.
Referen
ces
[1]
Reitmay
r
G, Schmalstieg
D,
editors.
Locati
on base
d
app
li
cations
for mo
bile
a
u
g
m
e
n
te
d
re
ality
. 4th
Australas
i
an U
s
er Interface C
onfere
n
ce. 20
0
3
. Adela
i
d
e
, Australia: Austral
i
an Com
puter S
o
ciet
y, Inc.
Evaluation Warning : The document was created with Spire.PDF for Python.
e-ISSN:
2087
-27
8
X
TELKOM
NIKA
Vol. 11, No
. 7, July 2013
: 3568 – 357
5
3574
[2]
Edmun
d
Ng Gaip W
,
Rehma
n
Ull
ah k, Sha
h
ren
Ahma
d Z
a
idi A, Oon YB. A frame
w
or
k for Outdoo
r
Mobil
e
A
ugm
e
n
ted
Re
alit
y.
IJ
CSI Internati
o
n
a
l J
ourn
a
l
of C
o
mputer
Scie
n
c
e Issues
. 20
1
2
;
9(2): 41
9-
23.
[3]
Lo
w
e
DG. Disti
nctive ima
ge f
eatures from s
c
ale-i
n
var
i
ant k
e
ypo
i
nts.
Internatio
nal
jour
na
l of co
mpute
r
vision
. 2
0
0
4
; 60(2): 91-1
10.
[4]
Ba
y
H, T
u
y
t
e
l
a
a
rs T
,
Van Gool L. Surf: Spe
ede
d up ro
bus
t features.
Co
mp
uter Visi
on
–
E
CCV 20
06
.
200
6: 404-
17.
[5]
Mikola
jcz
y
k K, Schmid C. A p
e
rformanc
e ev
alu
a
tion
of loca
l descri
p
tors.
IEEE transactions on patter
n
ana
lysis an
d machi
ne int
e
lli
ge
nce
. 200
5; 27(
10): 161
5-3
0
.
[6]
Darrell
KGaT
,
editor.
T
h
e
Py
ramid M
a
tch
Kerne
l
: Discri
m
i
nativ
e
Classification with
Sets of Im
age
F
eatures
. ICC
V
. 2005.
[7]
Lo
w
e
MB
aD. A
u
tomatic Pa
nor
amic Image Sti
t
ching Us
in
g In
varia
n
t F
eature
s
.
Internation
a
l
Journa
l of
Co
mp
uter Visi
on
. 200
7; 74: 5
9–7
7.
[8]
M Pielot N
H
, C Nicke
l, C Menke, S Sam
a
di, an
d S Bol
l
editor. Eva
l
uati
on of Cam
e
ra
Phon
e Bas
e
d
Interaction to Access Informa
tion Related to Posters.
Mobil
e
Interaction w
i
t
h the Real W
o
rld
. 200
8.
[9]
Ay
d
ı
n B, Gens
el J, C
a
l
abretto
S, T
e
llez B. A
RCA
MA-3D
–
A
Conte
x
t-A
w
a
r
e
Augm
ente
d
R
ealit
y M
o
b
i
l
e
Platform for Enviro
nmenta
l
Discover
y
.
W
e
b and W
i
re
les
s
Geograp
hic
a
l Informati
on S
ystems
. 20
12:
17-2
6
.
[10]
Reitma
yr
G, Schmalsti
eg
D
,
editors.
Data
man
a
g
e
ment
strategi
es for
mob
ile
a
ugm
ented
re
alit
y.
ST
ARS 2003; T
o
ky
o, Ja
pan.
[11]
Schal
l G, W
a
g
ner D,
Reitma
yr G, T
a
ichma
nn E,
W
i
es
er
M, Schmalstie
g
D, H
o
fman
n-
W
e
llen
hof B
,
editor. Glob
al
Pose Estimation usi
ng Mu
lti-Sens
or F
u
sion for Outdo
o
r Augme
n
ted
Realit
y. 8t
h
IEEE/ACM Internatio
nal Sy
mp
osiu
m o
n
Mi
xe
d and Au
g
m
e
n
ted Re
ality (ISMAR 200
9)
; 20
09.
[12]
Honk
amaa P, Siltan
en S, Jäppi
nen J, W
o
o
d
w
a
r
d
C, Korkalo O, editor.
Interactive out
door
mo
bil
e
aug
mentati
on
usin
g markerl
e
ss tracking a
n
d
GPS
. Virtual
Real
it
y
Inter
n
ation
a
l C
onfer
ence (VRIC).
200
7; Lava
l
, F
r
ance.
[13]
Greeni
ng C. i
P
hon
e Su
dok
u Grab. Jap
a
n
20
08 [cite
d
201
2 Ma
y
16 2
0
1
2
]; Availa
bl
e from:
http://
w
w
w
.
cm
grese
a
rch.com
/
sudoku
g
ra
b/.
[14]
Murray
GKaD,
editor.
P
a
ra
lle
l tracking
a
nd mapp
ing
o
n
a
c
a
mer
a
pho
ne
. I
n
ternatio
nal
S
y
mposi
u
m o
n
Mixed a
nd Au
g
m
ented R
eal
it
y (ISMAR). 2009.
[15]
Occipital L. Re
dlas
er 201
1 [cited 20
12 Ma
y 1
6
201
2]; Availa
ble from: http://redl
aser.com/.
[16]
Greene K. H
y
p
e
rlink
i
n
g
real
it
y via pho
nes.
MIT
T
e
chnolo
g
y Review
, Nove
mber/Dec
emb
e
r. 2006.
[17]
Bruns E, Bimb
er O. Adaptive
traini
ng of vi
de
o sets for imag
e reco
gniti
on o
n
mobi
le
pho
n
e
s.
Person
al
and U
b
iq
uito
us
Computi
n
g
. 20
09; 13(2): 1
65-
78.
[18]
Seifert C, Pa
let
t
a L, Jeitl
e
r A,
Hödl
E, Andr
eu
JP, Lul
e
y
P, et
al. Visu
al
ob
je
ct detectio
n
for
mobi
le r
oad
sign i
n
vent
or
y
.
Mobil
e
Hu
man-
Co
mp
uter Interaction
–
Mob
i
l
e
HCI
. 2004: 5
8
7
-
90.
[19]
Fritz G
,
Seifert C, Paletta L,
editors.
A m
o
bile vision system
for urban
detection with in
formative loc
a
l
descriptors
. ICVS ’06: Fourt
h
IEEE Intern
ational
Conf
er
ence on Computer V
i
sion S
y
stems. 2006:
IEEE.
[20]
Ba
y H, Fas
e
l
B,
Van G
ool
L, ed
itors.
Int
e
ractive
muse
um g
u
id
e: F
a
st and
ro
bust
recog
n
itio
n
of
m
u
se
um
ob
j
e
cts
. 4
t
h
i
n
te
rn
a
t
i
o
na
l
con
f
e
r
e
n
ce
o
n
Ada
p
t
i
v
e
mu
l
t
i
m
ed
i
a
re
tri
e
val
:
u
s
e
r
, co
n
t
ex
t, and
feedb
ack 20
07
; Springer-V
erl
ag Berl
in, Hei
d
elb
e
rg: Citese
e
r
.
[21]
Skry
pny
k I,
Lo
w
e
DG, edit
ors.
Scen
e
mode
lli
ng, r
e
co
gniti
on
an
d tr
ackin
g
w
i
th
in
varia
n
t i
m
a
g
e
features
. ISMA
R ’04:T
h
ird IEEE and ACM
International
S
y
mposium on M
i
xed
an
d Augm
ented Realit
y
(ISMAR’04). 2
004.
[22]
T
a
kacs G, Chandras
ekhar V,
Girod B, Grzeszczuk R, editor
s
.
F
eature tracking for
mo
bil
e
aug
me
nte
d
reality
usi
ng v
i
deo
cod
e
r
moti
on v
e
ctors
. Sixth IEEE and A
C
M Internat
ional S
y
mposium
on Mixed an
d
Augme
n
ted R
e
alit
y
(ISMAR
’0
7). 2007.
[23]
Yeh T
,
T
o
llmar
K, Darre
ll T
,
editors.
S
earc
h
in
g the
w
eb
w
i
th mo
bil
e
images
for l
o
cati
on r
e
cog
n
iti
o
n
.
IEEE. 2004.
[24]
Evans C. Open SURF
Co
m
puter Visi
on
Librar
y. [cited Ma
y
17
, 2012]; Avail
abl
e from:
http://
w
w
w
.
chr
i
sevans
dev.co
m
/computer-vi
s
ion-
ope
nsurf.html.
[25]
Open S
ourc
e
I
.
Open
Com
p
u
t
er
Visi
on
Li
brar
y
.
s
ourc
e
forg
e; [cited
M
a
y 17,
20
12]; Av
aila
bl
e from:
http://sourceforge.net/proje
cts
/
opencv
libr
a
r
y
/f
iles/.
[26]
Devel
o
p
e
rs. Q. Qualcomm SDK.
[cited Ma
y
2
4
,
2012]; Availa
bl
e from:
https://ar.qualc
o
mm.at/qdevnet/sdk/ios.
[27]
Ben Butch
a
rt
E. Augment
ed
Real
it
y for Sm
ar
tpho
nes: A
Guide for
dev
e
l
op
ers an
d co
n
t
ent pub
lis
hers
Rep
o
rt. UK: JISC Observator
y. Ma
y
20
11. R
eport No: 1.1.
Evaluation Warning : The document was created with Spire.PDF for Python.