T
E
L
KO
M
N
I
KA
T
e
lec
om
m
u
n
icat
ion
,
Com
p
u
t
i
n
g,
E
lec
t
r
on
ics
an
d
Cont
r
ol
Vol.
18
,
No.
4
,
Augus
t
2020
,
pp.
1904
~
1916
I
S
S
N:
1693
-
6930,
a
c
c
r
e
dit
e
d
F
ir
s
t
G
r
a
de
by
Ke
me
nr
is
tekdikti
,
De
c
r
e
e
No:
21/E
/KP
T
/2018
DO
I
:
10.
12928/
T
E
L
KO
M
NI
KA
.
v18i4.
14318
1904
Jou
r
n
al
h
omepage
:
ht
tp:
//
jour
nal.
uad
.
ac
.
id/
index
.
php/T
E
L
K
OM
N
I
K
A
S
of
t
w
a
r
e
e
n
gi
n
e
e
r
in
g
m
od
e
l
b
ase
d
sm
a
r
t
i
n
d
oor
lo
c
al
iz
at
io
n
syste
m
u
si
n
g d
e
e
p
-
le
a
r
n
in
g
Z
ain
ab
M
oh
am
m
e
d
Re
s
an
,
M
u
ayad
S
ad
i
k
Cr
o
oc
k
Co
mp
u
t
er
E
n
g
i
n
eer
i
n
g
D
e
p
art
me
n
t
,
U
n
i
v
ers
i
t
y
o
f
T
ec
h
n
o
l
o
g
y
-
Iraq
,
Bag
h
d
a
d
,
Iraq
Ar
t
icle
I
n
f
o
AB
S
T
RA
CT
A
r
ti
c
le
h
is
tor
y
:
R
e
c
e
ived
Oc
t
12
,
2019
R
e
vis
e
d
F
e
b
12,
2020
Ac
c
e
pted
M
a
r
18,
2020
D
u
r
i
n
g
t
h
e
l
a
s
t
fe
w
y
e
ar
s
,
t
h
e
a
l
l
o
c
a
t
i
o
n
o
f
o
b
j
e
c
t
s
o
r
p
ers
o
n
s
i
n
s
i
d
e
a
s
p
e
c
i
f
i
c
building
is
highly
r
e
qui
r
e
d
.
I
t
is
we
ll
known
th
a
t
the
glob
a
l
positioning
syste
m
(
G
PS
)
c
a
n
n
o
t
b
e
a
d
o
p
t
e
d
i
n
i
n
d
o
o
r
e
n
v
i
r
o
n
m
e
n
t
d
u
e
t
o
t
h
e
l
ac
k
o
f
s
i
g
n
a
l
s
.
T
h
e
r
ef
o
re,
i
t
i
s
i
m
p
o
r
t
a
n
t
t
o
d
i
s
c
o
v
er
a
n
e
w
w
a
y
t
h
a
t
w
o
r
k
s
i
n
s
i
d
e.
T
h
e
p
r
o
p
o
s
e
d
s
y
s
t
em
u
s
e
s
t
h
e
d
ee
p
l
e
ar
n
i
n
g
t
e
c
h
n
i
q
u
e
s
t
o
c
l
a
s
s
i
f
y
p
l
ac
e
s
b
a
s
e
d
o
n
c
a
p
t
u
r
i
n
g
i
m
a
g
e
s
.
T
h
e
p
r
o
p
o
s
e
d
s
y
s
t
em
c
o
n
t
a
i
n
s
t
w
o
p
a
r
t
s
:
s
o
f
t
w
a
re
p
a
r
t
a
n
d
h
a
r
d
w
a
re
p
a
r
t
.
T
h
e
s
o
f
t
w
a
re
p
ar
t
i
s
b
u
i
l
t
b
a
s
e
d
o
n
s
o
f
t
w
a
re
e
n
g
i
n
e
er
i
n
g
m
o
d
e
l
t
o
i
n
cr
ea
s
e
t
h
e
r
e
l
i
a
b
i
l
i
t
y
,
f
l
e
x
i
b
i
l
i
t
y
,
a
n
d
s
c
al
a
b
i
l
i
t
y
.
I
n
a
d
d
i
t
i
o
n
,
t
h
i
s
p
a
r
t
,
t
h
e
d
a
t
a
s
e
t
i
s
c
o
l
l
e
c
t
e
d
u
s
i
n
g
t
h
e
R
a
s
p
b
e
rr
y
P
i
I
I
I
ca
me
ra
a
s
t
r
a
i
n
i
n
g
a
n
d
v
a
l
i
d
a
t
i
n
g
d
a
t
a
s
e
t
.
T
h
i
s
d
a
t
a
s
e
t
i
s
u
s
e
d
a
s
a
n
i
n
p
u
t
t
o
t
h
e
p
r
o
p
o
s
e
d
d
ee
p
l
e
a
r
n
i
n
g
m
o
d
e
l
.
I
n
t
h
e
h
ar
d
w
ar
e
p
ar
t
,
R
a
s
p
b
e
rr
y
P
i
I
II
i
s
u
s
e
d
f
o
r
l
o
a
d
i
n
g
t
h
e
p
r
o
p
o
s
e
d
m
o
d
e
l
a
n
d
p
r
o
d
u
c
i
n
g
p
r
e
d
i
c
t
i
o
n
re
s
u
l
t
s
a
n
d
a
ca
me
ra
t
h
a
t
i
s
u
s
e
d
t
o
c
o
l
l
ec
t
t
h
e
i
ma
g
e
s
d
a
t
a
s
e
t
.
T
w
o
w
h
ee
l
s
’
ca
r
i
s
ad
o
p
t
e
d
a
s
a
n
o
b
j
e
c
t
f
o
r
i
n
t
r
o
d
u
c
i
n
g
i
n
d
o
o
r
l
o
c
a
l
i
z
a
t
i
o
n
p
r
o
j
ec
t
.
T
h
e
o
b
t
a
i
n
e
d
t
r
a
i
n
i
n
g
a
cc
u
ra
c
y
i
s
9
9
.
6
%
f
o
r
t
ra
i
n
i
n
g
d
a
t
a
s
e
t
a
n
d
1
0
0
%
f
o
r
v
a
l
i
d
a
t
i
n
g
d
a
t
as
e
t
.
K
e
y
w
o
r
d
s
:
C
NN
De
e
p
lea
r
ning
GPS
I
ndoor
loca
li
z
a
ti
on
R
a
s
pbe
r
r
y
P
i
R
oboti
c
c
a
r
S
of
twa
r
e
e
nginee
r
ing
Th
i
s
i
s
a
n
o
p
en
a
c
ces
s
a
r
t
i
c
l
e
u
n
d
e
r
t
h
e
CC
B
Y
-
SA
l
i
ce
n
s
e
.
C
or
r
e
s
pon
din
g
A
u
th
or
:
Mu
a
ya
d
S
a
dik
C
r
ooc
k,
De
pa
r
tm
e
nt
of
C
omput
e
r
E
nginee
r
ing,
Unive
r
s
it
y
of
T
e
c
hnology
-
I
r
a
q,
Als
inaa
S
tr
e
e
t,
B
a
ghda
d,
I
r
a
q.
E
mail:
M
ua
ya
d.
S
.
C
r
ooc
k@uotec
hnology.
e
du.
iq
1.
I
NT
RODU
C
T
I
ON
L
oc
a
li
z
a
ti
on
s
ys
tems
a
r
e
r
e
c
e
ntl
y
a
ppe
a
r
e
d
to
pr
ovide
the
inf
o
r
mation
of
loca
ti
on
of
di
f
f
e
r
e
nt
objec
ts
,
s
uc
h
a
s
pe
ople,
a
nim
a
ls
a
nd
thi
ngs
.
L
oc
a
ti
on
f
indi
ng
is
nor
mally
a
c
quir
e
d
by
us
in
g
global
pos
it
ioni
ng
s
ys
tems
(
GPS
)
f
or
outdoor
e
nvir
onm
e
nts
.
S
ince
GPS
ha
s
a
lac
k
o
f
the
s
ight
li
ne
to
s
a
telli
te
a
nd
a
lot
of
other
obs
tac
les
,
it
is
not
wo
r
king
f
or
i
ndoor
e
nvir
onments
[
1
]
.
T
he
loc
a
li
z
a
ti
on
is
ve
r
y
i
mpor
tant
f
or
the
pa
r
a
lyze
d
pe
ople,
a
nd
their
f
a
mi
li
e
s
a
s
it
e
a
s
e
s
the
li
ve
f
or
them
a
nd
dis
plac
e
s
their
c
onc
e
r
ns
.
I
n
a
ddit
ion,
it
s
a
ve
s
ti
me
of
s
e
a
r
c
hing
f
or
the
dis
a
bled
pe
r
s
ons
if
they
we
r
e
los
t
f
o
r
a
ny
r
e
a
s
on.
I
n
[
2
]
,
the
a
uthor
s
s
tudi
e
d
f
inger
pr
int
ing
-
ba
s
e
d
indoor
loca
li
z
a
ti
on
in
c
omm
od
it
y
5
-
GHz
Wi
-
F
i
ne
twor
ks
.
T
he
y
p
r
opos
e
d
a
s
ys
tem
c
a
ll
e
d
B
iL
oc
,
whic
h
us
e
s
bi
-
modalit
y
de
e
p
lea
r
ning
f
or
loca
li
z
a
ti
on
in
the
indoor
e
nvir
on
ment
us
ing
of
f
-
the
-
s
he
lf
W
i
-
F
i
de
vice
s
.
T
he
e
xpe
r
im
e
ntal
r
e
s
ult
s
va
li
da
ted
the
s
upe
r
ior
pe
r
f
or
manc
e
of
B
iL
oc
ove
r
s
e
ve
r
a
l
be
nc
hmar
k
s
c
he
mes
.
I
n
[
3]
,
the
a
utho
r
s
de
s
igned
a
n
indoor
loc
a
li
z
a
ti
on
s
ys
tem
ba
s
e
d
on
c
r
owe
d
s
our
c
ing
,
whic
h
c
a
n
c
ons
tr
uc
t
the
map
of
W
iF
i
r
a
dio
us
ing
it
s
da
ta,
the
map
of
indoo
r
loca
ti
on
is
tu
r
ne
d
to
indi
c
a
ti
ve
map,
t
he
a
uthor
s
s
how
r
e
s
ult
s
that
their
p
r
opos
e
d
ne
t
wor
k
c
a
n
mi
nim
ize
the
va
r
it
y
pr
oblems
c
a
us
e
d
by
c
ha
ngi
ng
e
nvir
onment.
I
n
[
4]
,
a
c
onvolut
ional
ne
u
r
a
l
ne
twor
k
(
C
NN
)
model
f
or
indoor
loca
li
z
a
ti
on
in
mu
lt
i
-
f
l
oor
buil
dings
wa
s
pr
oduc
e
d
us
ing
W
iF
i
r
e
c
e
ived
s
ignal
s
tr
e
ngth
va
lues
.
T
he
s
e
va
lues
ha
ve
be
e
n
tak
e
n
f
r
om
the
a
c
c
e
s
s
point
s
in
wir
e
les
s
L
A
N.
T
he
r
e
s
ult
s
f
or
t
his
Evaluation Warning : The document was created with Spire.PDF for Python.
T
E
L
KO
M
NI
KA
T
e
lec
omm
un
C
omput
E
l
C
ontr
o
l
Soft
w
ar
e
e
nginee
r
ing
mode
l
bas
e
d
s
mar
t
indoor
lo
c
ali
z
ati
on
s
y
s
tem
…
(
M
uay
ad
Sadik
C
r
ooc
k
)
1905
model
s
howe
d
a
c
c
ur
a
c
y
f
or
tr
a
ini
ng
model
a
bout
100%
.
I
n
[
5]
,
the
a
uthor
s
us
e
d
pe
r
va
s
iv
e
W
i
-
F
i
a
nd
a
t
tr
a
c
ti
ve
f
i
e
l
d
da
ta
f
or
in
do
o
r
l
im
i
tat
i
on
.
E
x
p
lo
r
a
to
r
y
o
utc
o
mes
d
e
m
ons
t
r
a
ted
th
a
t
pr
o
f
o
un
d
n
e
t
wo
r
k
m
o
de
ls
c
o
ns
o
li
da
ti
ng
a
tt
r
a
c
ti
ve
f
ie
ld
a
nd
W
i
-
F
i
f
i
ng
e
r
p
r
in
t
in
gs
im
p
r
o
ve
d
in
do
o
r
lo
c
a
li
z
a
t
i
on
p
r
e
c
is
i
on
.
T
h
e
t
r
a
in
in
g
p
ha
s
e
o
f
d
e
e
p
pos
it
i
on
in
g
wa
s
c
om
pu
tat
i
on
a
l
ly
in
tens
i
ve
,
t
he
t
e
s
t
in
g
ph
a
s
e
wa
s
f
a
s
t
a
n
d
s
u
it
a
b
le
f
o
r
r
e
a
l
t
i
me
i
nd
oo
r
l
oc
a
l
iza
ti
on
.
I
n
[
6]
,
the
a
u
th
or
s
de
s
i
gne
d
W
iF
i
de
e
p
,
w
hi
c
h
wa
s
a
W
i
-
Fi
-
b
a
s
e
d
in
do
o
r
f
i
ng
e
r
p
r
in
ti
ng
lo
c
a
li
z
a
ti
on
s
ys
t
e
m
t
ha
t
c
a
n
a
c
hi
e
v
e
r
o
bus
t
a
nd
h
ig
h
a
c
c
u
r
a
c
y
t
r
a
c
ki
ng
i
n
t
he
pr
e
s
e
nc
e
o
f
de
vi
c
e
h
e
t
e
r
o
ge
n
e
i
ty
.
T
he
s
ys
te
m
e
mp
lo
ye
d
a
m
ode
l
r
e
gu
la
r
iza
t
io
n
t
o
e
n
a
b
le
t
he
ne
t
wo
r
k
t
o
ge
ne
r
a
li
z
e
a
n
d
a
vo
id
o
ve
r
-
f
i
t
ti
ng
,
lea
di
ng
t
o
a
m
or
e
r
ob
us
t
a
nd
s
tab
le
m
o
de
ls
.
T
he
r
e
s
u
lt
s
s
ho
we
d
tha
t
t
he
W
I
de
e
p
c
om
e
s
w
it
h
a
l
oc
a
l
iza
t
io
n
a
c
c
u
r
a
c
y
b
e
t
te
r
t
ha
n
t
he
s
t
a
t
e
-
of
-
t
he
-
a
r
t
s
ys
te
ms
b
y
a
t
le
a
s
t
5
3%
a
nd
2
9
.
8
%
i
n
the
l
a
r
ge
a
n
d
s
ma
l
l
e
nv
i
r
o
nm
e
n
ts
r
e
s
pe
c
ti
ve
ly
.
I
n
[
7
]
,
the
y
p
r
op
os
e
d
a
m
o
de
l
t
ha
t
l
e
a
r
ns
p
r
o
f
o
un
d
lea
r
ni
ng
s
c
e
n
e
a
c
kn
ow
le
dg
me
nt
t
e
c
hn
iq
ue
.
T
his
d
e
pe
nds
o
n
l
i
mi
ta
ti
on
i
mp
r
ov
e
me
nt
u
t
il
iz
in
g
the
t
e
c
h
ni
qu
e
f
o
r
e
x
c
ha
ng
e
lea
r
n
i
ng
on
in
c
e
pt
io
n
V
3
ne
tw
o
r
k
.
M
o
de
l
f
e
a
t
u
r
e
i
n
f
o
r
mat
i
on
wa
s
a
dd
e
d
to
a
s
s
is
t
in
s
c
e
ne
r
e
c
o
gn
it
io
n
[
8
]
.
I
n
[
9
],
T
he
a
ut
hor
s
uti
li
z
e
d
a
n
a
dva
nc
e
d
mobi
le
phone
-
ba
s
e
d
moveme
nt
a
c
knowle
dgment
f
or
indoo
r
r
e
s
tr
iction
uti
li
z
ing
a
c
onv
olut
ional
ne
ur
a
l
s
ys
tem.
T
he
s
e
e
xe
r
c
is
e
s
c
a
n
be
uti
li
z
e
d
a
s
the
tour
is
t
s
pots
f
or
indoo
r
loca
li
z
a
ti
on.
T
he
s
e
e
xe
r
c
is
e
s
c
ould
be
uti
li
z
e
d
a
s
the
tour
is
t
s
pots
f
or
indoor
c
o
nf
ineme
nt.
Anothe
r
c
onvolut
ional
ne
ur
a
l
s
ys
tem
h
a
s
be
e
n
int
e
nde
d
to
be
c
ome
f
a
mi
l
iar
with
the
be
s
t
pos
s
ibl
e
f
e
a
tur
e
s
c
ons
e
que
ntl
y.
I
n
thi
s
a
r
ti
c
le,
a
n
indoor
objec
t
loca
li
z
a
ti
on
model
is
p
r
opos
e
d
ba
s
e
d
on
de
e
p
lea
r
ning
[
10
]
a
nd
de
e
p
C
NN
[
11
]
.
T
he
p
r
opos
e
d
model
c
on
tains
two
leve
ls
.
F
ir
s
t
leve
l,
or
s
of
twa
r
e
leve
l,
is
us
e
d
f
or
c
oll
e
c
ti
ng
the
im
a
ge
s
da
tas
e
t,
done
us
ing
r
a
s
pbe
r
r
y
pi
I
I
I
c
a
mer
a
.
T
his
c
a
mer
a
is
f
ixed
on
a
s
im
ple
r
obot
of
two
whe
e
ls
’
c
a
r
.
T
he
da
tas
e
t
is
c
oll
e
c
ted
a
nd
p
r
e
pa
r
e
d
by
ga
ther
ing
a
nd
pr
e
-
ha
ndli
ng
im
a
ge
s
inf
o
r
mation.
I
t
us
e
s
a
s
input
f
or
the
p
r
opos
e
d
model.
T
he
mode
l
is
tr
a
ined
unti
l
it
r
e
a
c
he
s
to
a
s
a
ti
s
f
ying
a
c
c
ur
a
c
y
f
or
both
tr
a
ini
ng
a
nd
va
li
da
ti
ng
da
tas
e
t.
T
he
n
the
model
is
s
a
ve
d.
I
n
the
s
e
c
ond
leve
l,
or
c
a
n
be
c
a
ll
e
d
a
s
h
a
r
dwa
r
e
leve
l,
is
us
e
d
f
o
r
ga
ther
ing
a
nd
c
ontr
ol
li
ng
the
e
lec
tr
ic
r
obot
c
a
r
that
c
a
r
r
ies
on
the
r
a
s
pbe
r
r
y
pi
I
I
I
a
nd
it
s
c
a
mer
a
.
T
his
c
a
mer
a
is
no
t
only
us
e
d
f
o
r
pr
e
pa
r
in
g
the
da
tas
e
t,
it
is
a
ls
o
us
e
d
to
take
a
r
e
a
l
ti
me
im
a
ge
whe
n
a
c
e
r
tain
c
ondit
ion
ha
ppe
ne
d
or
whe
n
a
n
a
utho
r
ize
d
us
e
r
is
a
s
king
a
bout
a
n
objec
t's
plac
e
.
I
n
t
his
c
a
s
e
,
the
r
a
s
pbe
r
r
y
pi
I
I
I
c
a
ptur
e
s
the
im
a
ge
a
nd
s
e
nd
it
to
the
p
r
opos
e
d
ins
ide
the
R
a
s
pbe
r
r
y
P
i.
T
he
n,
the
pr
opos
e
d
s
ys
tem
wa
it
s
f
o
r
the
r
e
s
pon
s
e
.
T
he
p
r
e
dict
ion
o
f
the
model
wi
th
c
e
r
tain
a
c
c
ur
a
c
y
is
s
uppos
e
d
to
be
the
pr
e
diction
of
the
plac
e
it
s
e
lf
[
5
,
12]
a
nd
[
13
]
.
2.
P
ROP
OS
E
D
S
OF
T
WAR
E
E
NGI
NE
E
R
I
NG
M
ODE
L
I
t
is
we
ll
known
that
the
pr
ovided
a
lgor
i
thm
s
a
r
e
r
e
quir
e
d
to
be
mor
e
r
e
li
a
ble
a
nd
ha
ve
a
ll
the
a
bil
it
y
o
f
be
ing
e
xpa
nding
a
nd
f
lexible.
I
n
thi
s
wor
k
,
a
s
of
twa
r
e
e
nginee
r
ing
model
is
pr
opos
e
d
a
s
a
ba
s
e
s
tr
ictur
e
f
or
the
uti
li
z
e
d
a
lgor
it
hm
.
T
h
is
is
to
e
ns
ur
e
th
e
r
e
li
a
bil
it
y,
s
c
a
labili
ty
a
nd
f
lexibil
it
y
of
the
p
r
opos
e
d
a
lgor
it
hm.
T
he
mentioned
a
lgo
r
it
hm
is
c
onc
e
pts
of
pe
r
f
or
mi
ng
the
s
ugge
s
ted
method
of
de
tec
ti
ng
the
a
f
f
e
c
ted
a
r
e
a
s
in
the
br
a
in
.
F
igur
e
1
s
hows
the
block
d
iagr
a
m
of
p
r
opos
e
d
model
that
e
xplains
s
tage
s
of
met
hod
a
s
a
wor
k
f
low
.
F
igur
e
1.
B
lock
diagr
a
m
of
the
de
s
igned
s
of
twa
r
e
e
nginee
r
ing
model
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
1693
-
6930
T
E
L
KO
M
NI
KA
T
e
lec
omm
un
C
omput
E
l
C
ontr
o
l
,
Vol.
18
,
No
.
4
,
Augus
t
2020
:
1904
-
1916
1906
I
t
i
s
s
h
o
w
n
f
r
o
m
F
i
g
u
r
e
1
t
h
a
t
t
h
e
s
y
s
t
e
m
r
e
q
u
i
r
e
m
e
n
t
s
a
r
e
c
o
l
l
e
c
t
e
d
b
y
t
h
e
f
i
r
s
t
r
o
u
n
d
f
o
r
p
r
e
p
a
r
i
n
g
t
h
e
s
t
r
u
c
t
u
r
e
o
f
t
h
e
d
e
s
i
g
n
e
d
a
l
g
o
r
i
t
h
m
.
A
t
t
h
e
s
e
c
o
n
d
r
o
u
n
d
,
t
h
e
i
n
i
t
i
a
l
d
e
s
i
g
n
o
f
t
h
e
s
u
g
g
e
s
t
e
d
a
l
g
o
r
i
t
h
m
i
s
p
e
r
f
o
r
m
e
d
b
a
s
e
d
o
n
t
h
e
s
y
s
t
e
m
r
e
q
u
i
r
e
m
e
n
t
s
.
D
i
f
f
e
r
e
n
t
l
e
v
e
l
s
o
f
d
e
v
e
l
o
p
m
e
n
t
s
h
a
v
e
b
e
e
n
d
o
n
e
i
n
t
h
e
t
h
i
r
d
r
o
u
n
d
d
e
p
e
n
d
i
n
g
o
n
t
h
e
i
n
i
t
i
a
l
d
e
s
i
g
n
o
f
t
h
e
p
r
o
p
o
s
e
d
a
l
g
o
r
i
t
h
m
i
n
i
t
e
r
a
t
i
o
n
w
a
y
.
T
h
e
l
a
s
t
v
e
r
s
i
o
n
o
f
t
h
e
p
r
o
p
o
s
e
d
a
l
g
o
r
i
t
h
m
i
s
t
e
s
t
e
d
i
n
t
h
e
l
a
s
t
r
o
u
n
d
f
o
r
e
n
s
u
r
i
n
g
t
h
e
r
e
l
i
a
b
i
l
i
t
y
,
e
x
p
a
n
d
a
b
i
l
i
t
y
a
n
d
f
l
e
x
i
b
i
l
i
t
y
o
f
t
h
e
i
t
.
3.
P
ROP
OS
E
D
S
Y
S
T
E
M
As
mentioned
e
a
r
li
e
r
,
the
pr
opos
e
d
s
ys
tem
c
lea
r
ly
c
ontains
two
dif
f
e
r
e
nt
pha
s
e
s
,
de
s
igned
ba
s
e
d
on
the
pr
opos
e
d
s
of
twa
r
e
e
nginee
r
ing
model
.
T
he
f
ir
s
t
one
,
c
a
n
be
c
a
ll
e
d
a
s
the
o
f
f
li
ne
pha
s
e
,
whic
h
c
o
ntains
a
ll
the
pr
ogr
a
mm
ing
a
lgo
r
it
hms
s
tar
ti
ng
with
c
r
e
a
t
ing
the
da
tas
e
t
a
nd
pr
oc
e
s
s
ing
it
,
e
nding
with
s
a
ving
a
we
ll
-
tr
a
ined
model.
T
he
s
e
c
ond
pha
s
e
,
or
the
onli
ne
pha
s
e
,
in
whic
h
,
a
r
e
a
l
ti
me
im
a
ge
s
a
r
e
take
n,
s
e
c
ur
it
y
whe
r
e
loca
li
z
a
ti
on
c
a
n
s
igni
f
ica
ntl
y
im
p
r
ove
s
e
c
ur
it
y
c
ondit
ions
the
wo
r
ld
ove
r
.
C
li
e
nt
ve
r
s
a
ti
li
ty
e
xa
mpl
e
s
a
nd
c
omm
unica
ti
on
c
a
n
be
uti
li
z
e
d
to
dis
ti
nguis
h
c
onc
e
ivable
da
nge
r
s
that
may
pr
e
s
e
nt
s
e
c
ur
it
y
r
is
ks
.
S
im
il
a
r
ly,
in
wa
r
z
one
,
the
mi
li
tar
y
c
a
n
f
ol
low
it
s
a
dva
ntage
s
thr
ough
a
r
e
s
tr
iction
f
r
a
mew
or
k
that
c
a
n
im
pr
ove
the
ge
ne
r
a
l
tas
k
a
nd
inc
r
e
ment
the
odds
of
f
r
uit
f
u
l
a
c
ti
vit
y
.
T
he
f
inal
c
las
s
if
ica
ti
on
is
then
a
s
s
igned
to
it
s
e
s
ti
mate
d
loca
ti
on.
As
s
hown
in
F
igu
r
e
2
[
2]
.
F
igur
e
2
.
P
r
opos
e
d
s
ys
tem
block
diag
r
a
m
[
2
]
3.
1.
P
r
op
os
e
d
m
e
t
h
od
ology
I
n
th
is
s
e
c
ti
on,
the
da
tas
e
t
is
c
oll
e
c
ted
a
nd
pr
e
pa
r
e
d.
I
t
a
ls
o
int
r
oduc
e
s
the
C
NN
model
a
lgor
it
hm
a
nd
tr
a
ini
ng
pr
oc
e
dur
e
.
M
or
e
ove
r
,
thi
s
s
e
c
ti
on
pr
ovides
the
we
bs
it
e
de
s
ign
that
ia
ba
s
e
d
on
the
pr
opos
e
d
a
lgor
it
hm.
At
the
o
ther
s
ide,
the
objec
t
loca
li
z
a
ti
on
s
ys
tem
is
int
r
oduc
e
d
with
the
r
e
late
d
pr
opos
e
d
a
lg
or
it
hms
a
nd
f
lowc
ha
r
ts
.
T
he
pr
opos
e
d
ove
r
a
ll
s
ys
tem
is
s
tar
ted
with
the
s
of
twa
r
e
pa
r
t
a
nd
f
oll
owe
d
by
the
ha
r
dwa
r
e
pa
r
t
a
s
s
hown
in
F
igu
r
e
3
.
T
he
f
oll
owing
s
teps
de
s
c
r
ibe
the
method
us
e
d
to
buil
d
the
model
br
ief
ly:
−
S
tar
ti
ng
the
c
oll
e
c
ti
on
of
a
da
tas
e
t
f
or
tr
a
ini
ng
the
model.
T
h
is
da
tas
e
t
mus
t
c
ove
r
a
ll
the
de
s
ir
e
d
a
r
e
a
s
that
is
wa
nted
to
be
a
dde
d
to
the
c
las
s
if
ica
ti
on
model
.
−
P
r
e
p
a
r
i
ng
th
is
da
tas
e
t
by
a
d
di
ng
t
he
ne
c
e
s
s
a
r
y
o
pe
r
a
t
io
ns
,
s
u
c
h
a
s
c
r
op
pi
ng
t
he
im
a
ge
s
t
o
b
e
a
l
l
o
f
th
e
s
a
me
s
ize
.
M
o
r
e
o
ve
r
,
a
dd
i
ng
a
f
il
l
ipe
d
c
op
y
o
f
th
e
s
a
m
e
im
a
g
e
s
f
o
r
inc
r
e
a
s
in
g
t
he
ir
nu
mb
e
r
i
n
e
a
c
h
c
las
s
.
−
De
s
igni
ng
the
C
NN
laye
r
s
of
the
model
with
K
e
r
a
s
li
br
a
r
y.
T
he
n
,
the
s
ys
tem
tr
a
ins
the
model
unti
l
r
e
a
c
hing
the
de
s
ir
e
d
a
c
c
ur
a
c
y
a
nd
s
a
ve
them
in
the
R
a
s
pbe
r
r
y
P
i
f
or
late
r
us
e
.
−
T
he
s
of
twa
r
e
pa
r
t
e
nds
with
c
r
e
a
ti
ng
a
s
im
pl
e
we
bs
it
e
that
c
ontains
the
c
ontr
oll
ing
butt
on
s
on
the
ha
r
dwa
r
e
pa
r
ts
.
T
his
we
bs
it
e
is
buil
t
by
us
ing
t
he
HT
M
L
,
P
HP,
C
S
S
a
nd
python
s
c
r
ipt
s
.
−
T
he
f
ir
s
t
s
tep
in
the
ha
r
dwa
r
e
pa
r
t
c
ontains
ga
the
r
ing
a
nd
buil
ding
a
s
im
ple
p
r
otot
ype
of
a
r
oboti
c
two
whe
e
ls
’
c
a
r
with
it
s
ba
tt
e
r
y
a
nd
mo
tor
d
r
iver
.
−
C
ombi
ning
the
R
a
s
pbe
r
r
y
P
i
a
nd
r
e
late
d
a
c
c
e
s
s
or
ies
that
include
c
a
mer
a
,
L
E
Ds
,
br
e
a
dboa
r
d,
volt
a
ge
s
tep
r
e
gulator
,
two
ult
r
a
s
onic
s
e
ns
or
s
.
All
thes
e
c
o
mponents
a
r
e
c
onne
c
ted
togethe
r
a
nd
plac
e
d
in
a
blac
k
box
f
or
a
r
r
a
nge
ment.
Evaluation Warning : The document was created with Spire.PDF for Python.
T
E
L
KO
M
NI
KA
T
e
lec
omm
un
C
omput
E
l
C
ontr
o
l
Soft
w
ar
e
e
nginee
r
ing
mode
l
bas
e
d
s
mar
t
indoor
lo
c
ali
z
ati
on
s
y
s
tem
…
(
M
uay
ad
Sadik
C
r
ooc
k
)
1907
−
T
he
R
a
s
pbe
r
r
y
P
i
is
the
c
or
e
of
c
ontr
oll
ing
the
h
a
r
dwa
r
e
pa
r
t.
I
t
is
c
onne
c
ted
by
W
iF
i
to
the
we
bs
it
e
.
W
he
n
the
s
e
a
r
c
h
butt
on
is
c
li
c
ke
d
in
the
we
b,
a
c
e
r
tain
pr
oc
e
dur
e
is
f
oll
owe
d
by
movi
ng
th
e
c
a
r
a
nd
c
a
ptur
ing
r
e
a
l
ti
me
im
a
ge
s
.
T
his
p
r
oc
e
dur
e
is
i
ntr
oduc
e
d
in
de
tails
in
the
ne
xt
s
e
c
ti
ons
.
−
T
he
las
t
s
tep
is
f
indi
ng
a
n
a
ppr
opr
iate
c
las
s
if
ica
ti
on
a
c
c
ur
a
c
y
a
nd
a
s
s
igni
ng
it
to
it
s
c
las
s
that
is
r
e
late
d
to
it
s
c
or
r
e
s
ponding
c
las
s
if
ied
i
ndoor
loca
ti
on
.
F
igur
e
3
.
T
he
pr
opos
e
d
s
ys
tem
methodology
f
lowc
ha
r
t
3.
1.
1
.
Dat
as
e
t
c
oll
e
c
t
in
g
T
he
da
tas
e
t
is
c
hos
e
n
ve
r
y
c
a
r
e
f
ull
y,
with
not
if
ying
the
a
ngle
of
c
a
ptur
ing
the
im
a
ge
s
a
nd
the
number
of
im
a
ge
s
that
c
ove
r
s
the
include
d
a
r
e
a
s
that
a
r
e
ne
e
de
d
to
be
c
las
s
if
ied.
I
n
thes
e
im
a
ge
s
,
dif
f
e
r
e
nt
s
it
ua
ti
ons
a
nd
dif
f
e
r
e
nt
pos
it
ions
a
r
e
take
n
in
c
on
s
ider
a
ti
on,
a
s
s
hown
in
T
a
ble
1.
I
n
thi
s
pa
r
t,
the
da
ta
s
e
t
c
oll
e
c
ti
on
a
nd
c
hoos
ing
a
r
e
dis
c
us
s
e
d.
M
or
e
ove
r
,
a
s
we
ll
a
s
the
p
r
e
pr
oc
e
s
s
ing
ope
r
a
ti
ons
,
s
uc
h
a
s
c
utt
ing,
z
oomi
ng
,
a
nd
s
he
a
r
ing
a
r
e
a
ls
o
in
tr
oduc
e
d.
T
he
c
a
mer
a
is
loca
ted
a
bout
20
c
m
f
r
om
gr
oun
d
on
the
r
oboti
c
c
a
r
with
ve
r
ti
c
a
l
a
ngle
of
15
de
gr
e
e
s
.
A
f
t
e
r
a
ll
da
tas
e
ts
a
r
e
c
oll
e
c
ted,
e
a
c
h
c
las
s
is
a
s
s
igne
d
to
one
r
oom.
T
his
is
to
r
e
duc
e
the
c
ompl
e
xit
y
a
s
it
is
no
t
s
e
ns
or
y
to
include
a
ll
wa
ll
s
in
the
da
tas
e
t
be
c
a
us
e
s
ome
im
a
ge
s
in
the
c
oll
e
c
ted
da
ta
s
e
t
a
r
e
take
n
in
a
wa
y
t
ha
t
c
ove
r
s
mor
e
than
o
ne
c
or
ne
r
ins
ide
the
r
oom
.
T
a
ble
1
.
Ove
r
view
of
the
c
oll
e
c
ted
da
tas
e
t
L
oc
a
ti
on (
W
a
ll
)
N
o.
C
or
r
e
s
ponding
R
oom Na
me
N
umbe
r
of
T
r
a
in
in
g da
ta
s
e
t
N
umbe
r
of
V
a
li
da
ti
ng da
ta
s
e
t
S
iz
e
O
f
I
ma
ge
s
(
w
id
th
x
he
ig
ht
)
L
oc
a
ti
on 1
M
e
e
ti
ng R
oom
264
60
480 x 270
L
oc
a
ti
on 2
M
e
e
ti
ng R
oom
268
61
480 x 270
L
oc
a
ti
on 3
E
a
ti
ng R
oom
255
59
480 x 270
L
oc
a
ti
on 4
E
a
ti
ng R
oom
260
60
480 x 270
L
oc
a
ti
on 5
A
dmi
n R
oom
208
54
480 x 270
L
oc
a
ti
on 6
A
dmi
n R
oom
202
60
480 x 270
L
oc
a
ti
on 7
A
dmi
n R
oom
298
67
480 x 270
L
oc
a
ti
on 8
E
a
ti
ng R
oom
267
61
480 x 270
3.
1.
2
.
Dat
as
e
t
p
r
e
p
ar
in
g
T
he
ne
xt
s
tep
is
to
pr
e
pa
r
e
thes
e
da
ta
s
e
ts
to
be
r
e
a
dy
f
or
us
e
in
the
tr
a
ini
ng
model.
T
he
da
ta
a
ugmenta
ti
on
is
us
e
d
a
s
a
tool
that
is
a
va
il
a
ble
i
n
a
ny
de
e
p
lea
r
ning
tool
box
.
I
t
is
done
with
Ke
r
a
s
li
br
a
r
y
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
1693
-
6930
T
E
L
KO
M
NI
KA
T
e
lec
omm
un
C
omput
E
l
C
ontr
o
l
,
Vol.
18
,
No
.
4
,
Augus
t
2020
:
1904
-
1916
1908
us
ing
a
c
las
s
c
a
ll
e
d
im
a
ge
da
ta
ge
ne
r
a
tor
[1
4
]
.
T
his
c
las
s
a
ppli
e
s
a
r
a
ndom
invar
ianc
e
,
s
uc
h
a
s
im
a
ge
s
c
r
opping,
s
c
a
li
ng,
s
he
a
r
ing,
a
nd
f
li
pping
a
s
s
hown
in
T
a
ble
2
.
T
a
ble
2
.
Ove
r
view
of
the
us
e
d
da
ta
a
ugmenta
ti
on
t
e
c
hniques
N
a
me
V
a
lu
e
S
c
a
li
ng
(
1. /
255)
Z
oomi
ng
20%
F
li
ppi
ng
H
or
iz
ont
a
l
C
r
ops
224*224
3.
1.
3
.
M
od
e
l
t
r
ain
in
g
T
he
model
is
tr
a
ined
f
r
om
s
c
r
a
tch,
be
c
a
us
e
ther
e
is
no
pr
e
-
tr
a
ined
model
us
e
d
f
or
ini
ti
a
li
z
a
ti
on.
T
he
main
s
our
c
e
of
da
ta
is
the
im
a
ge
s
a
va
il
a
ble
with
R
a
s
pbe
r
r
y
P
i
c
a
mer
a
.
T
he
pr
opos
e
d
C
NN
with
de
e
p
lea
r
ning
is
de
s
igned
to
tr
a
in
a
number
of
im
a
ge
s
a
nd
tes
t
other
s
[
15
]
a
nd
[
16
]
.
E
a
c
h
e
nter
e
d
im
a
ge
to
the
C
NN
model
pa
s
s
e
s
thr
ough
many
leve
ls
of
c
onvolut
ion
laye
r
s
.
S
ome
o
f
them
c
ontains
a
f
il
ter
s
a
nd
pooli
ng,
while
the
other
s
include
f
ul
ly
c
onne
c
ted
laye
r
s
.
F
inally
,
s
uc
h
a
n
im
a
ge
pa
s
s
e
s
thr
ough
a
s
of
t
max
f
il
ter
[
1
7
]
.
T
he
f
oll
owing
s
teps
il
lus
tr
a
te
the
b
uil
ding
of
the
C
NN
model:
−
I
n
the
f
ir
s
t
c
onvolut
ional
laye
r
,
20
f
il
ter
s
o
f
s
iz
e
5x5
pixels
a
r
e
a
ppl
ied
to
the
input
im
a
ge
of
s
ize
224x224,
f
oll
owe
d
by
a
r
e
c
ti
f
ied
li
ne
a
r
uni
t
(
R
e
L
U)
.
A
max
pooli
ng
laye
r
is
us
e
d
f
o
r
ta
king
the
maximal
va
lue
of
2x2
r
e
gions
with
s
tr
ides
of
2x
2
.
−
T
he
output
of
the
pr
e
vious
laye
r
is
then
pr
oc
e
s
s
e
d
by
the
s
e
c
ond
c
onvolut
ional
laye
r
.
I
t
c
ontains
50
f
i
lt
e
r
s
of
a
s
ize
5x5
p
ixels
.
Aga
in,
that
is
f
oll
owe
d
by
a
R
e
L
U
a
nd
a
max
pooli
ng
laye
r
.
−
T
his
c
onvolut
ional
laye
r
's
output
is
pa
s
s
e
d
to
a
f
ul
ly
c
onne
c
ted
laye
r
a
nd
c
ontains
500
ne
ur
ons
,
f
oll
o
we
d
by
a
R
e
L
U.
−
At
las
t,
the
y
ield
of
the
las
t
c
ompl
e
tely
a
s
s
oc
iate
d
laye
r
is
nour
is
he
d
to
a
s
of
t
max
laye
r
that
a
ppo
int
s
a
li
ke
li
hood
f
or
e
a
c
h
c
las
s
of
8
c
las
s
e
s
.
T
he
n
it
a
s
s
igns
the
c
la
s
s
number
to
it
s
c
or
r
e
s
pon
ding
loca
ti
on
in
one
of
the
thr
e
e
r
ooms
mentione
d
e
a
r
li
e
r
,
with
a
pr
oba
bil
i
ty
goe
s
f
r
om
0
to
100
%
.
As
s
hown
in
F
igu
r
e
4
.
T
r
a
ini
ng
the
C
NN
model
is
the
ne
xt
s
tep,
s
e
e
T
a
ble
3,
it
take
s
a
ti
me
r
a
nge
s
f
r
om
88
to
94
s
e
c
ond
f
or
e
a
c
h
e
poc
h.
T
his
is
be
c
a
us
e
the
de
e
p
lea
r
ning
model
tr
a
ini
ng
r
e
quir
e
s
a
high
a
bil
it
y
GPU
.
T
he
p
r
opos
e
d
model
is
t
r
a
ined
onli
ne
in
Google
C
olab,
whic
h
gives
a
f
r
e
e
GPU
on
a
f
r
e
e
c
loud
s
e
r
vice
f
o
r
de
ve
lopi
ng
a
ppli
c
a
ti
ons
of
de
e
p
lea
r
ning
[
18]
a
nd
[
19]
.
T
hi
s
is
done
by
loading
the
da
tas
e
t
to
the
Google
dr
ive
f
or
ope
ning
the
Google
C
olab
a
nd
making
the
model
.
T
r
a
ini
ng
tr
ies
a
r
e
done
with
di
f
f
e
r
e
nt
number
of
c
las
s
e
s
f
r
om
4
to
8,
in
bo
th
laptop
a
nd
onli
ne
on
Google
C
olab,
but
the
r
e
is
no
ne
e
d
to
mention
the
r
e
s
ult
s
of
thes
e
tr
ies
.
T
he
im
po
r
tant
thi
ng
to
know
is
that,
it
h
a
s
be
e
n
r
e
a
c
he
d
to
a
point
in
whic
h
only
s
mall
number
of
e
poc
hs
is
e
nough
to
tr
a
in
thi
s
model
wi
th
s
a
t
is
f
ying
r
e
s
ult
s
.
T
he
r
e
f
or
e
,
only
5
e
poc
hs
a
r
e
tak
e
n
e
a
c
h
of
whic
h
p
r
oduc
e
s
dif
f
e
r
e
nt
a
c
c
ur
a
c
y
a
nd
los
s
metr
ics
f
or
both
tr
a
ini
ng
a
nd
va
li
da
ti
ng
da
tas
e
t.
F
igur
e
4
.
C
NN
a
r
c
hit
e
c
tur
e
o
f
the
pr
opos
e
d
model
T
a
ble
3
.
M
ode
l
tr
a
ini
ng
r
e
s
ult
s
c
or
r
e
s
ponding
to
e
a
c
h
e
poc
h
E
poc
hs
T
im
e
i
n
s
e
c
onds
T
r
a
in
in
g
V
a
li
da
ti
on
A
c
c
ur
a
c
y
L
os
s
A
c
c
ur
a
c
y
L
os
s
1
94s
0.7923
0.7252
0.9146
0.2176
2
89s
0.9836
0.0785
1.0000
0.0301
3
90s
0.9960
0.0276
1.0000
0.0180
4
88s
0.9965
0.0170
1.0000
0.0096
5
90s
0.9965
0.0152
1.0000
0.0145
Evaluation Warning : The document was created with Spire.PDF for Python.
T
E
L
KO
M
NI
KA
T
e
lec
omm
un
C
omput
E
l
C
ontr
o
l
Soft
w
ar
e
e
nginee
r
ing
mode
l
bas
e
d
s
mar
t
indoor
lo
c
ali
z
ati
on
s
y
s
tem
…
(
M
uay
ad
Sadik
C
r
ooc
k
)
1909
3.
1.
4
.
We
b
s
it
e
d
e
s
ign
Af
ter
the
model
ha
d
be
e
n
t
r
a
ined
a
nd
r
e
a
c
he
d
to
a
s
a
ti
s
f
ying
a
c
c
ur
a
c
y
r
e
s
ult
s
,
whic
h
wa
s
99
.
6%
f
or
tr
a
ini
ng
da
tas
e
t
a
nd
100%
f
or
va
li
da
ti
on
da
tas
e
t.
T
his
model
is
s
a
ve
d
to
the
R
P
i
I
I
I
memor
y
to
be
us
e
d
in
the
ha
r
dwa
r
e
leve
l.
T
he
s
a
ve
d
model
c
a
nnot
be
us
e
d
dir
e
c
tl
y
f
or
pr
e
dictions
,
be
c
a
us
e
it
is
s
a
ve
d
a
s
a
python
s
c
r
ipt
.
S
o,
a
n
int
e
r
f
a
c
ing
mec
ha
nis
m
mus
t
be
a
dde
d
f
or
the
us
e
r
to
be
a
ble
to
c
ontac
t
with
the
model
in
s
ome
wa
y.
F
igur
e
5
s
hows
a
f
lowc
ha
r
t
f
or
the
de
s
igned
we
bs
it
e
,
whic
h
c
ontains
the
s
teps
that
mus
t
be
f
oll
owe
d
to
ope
r
a
te
thi
s
s
ys
tem.
F
igur
e
5
.
W
e
bs
it
e
f
lowc
ha
r
ts
T
he
f
oll
owing
s
teps
a
r
e
de
s
c
r
ibed
in
the
ne
xt
li
ne
s
:
−
A
s
im
ple
we
b
pa
ge
is
c
r
e
a
ted
ins
ide
the
R
P
i.
F
or
s
e
c
ur
it
y
r
e
a
s
ons
,
thi
s
we
b
pa
ge
c
ontains
a
logi
n
butt
on
a
t
the
s
tar
ti
ng
home
pa
ge
.
−
W
it
hout
logi
n
,
the
nor
mal
us
e
r
c
a
nnot
s
e
e
mor
e
th
a
n
s
im
ple
de
s
c
r
ipt
ion
on
the
pr
ojec
t
a
nd
the
c
opy
r
ight
s
a
nd
the
ye
a
r
o
f
modi
f
ica
ti
on.
−
W
he
n
the
us
e
r
f
il
ls
up
the
logi
n
f
o
r
m,
a
nd
he
/s
he
is
a
uthor
ize
d
to
e
nter
to
thi
s
we
b.
T
he
we
b
then
m
ove
s
the
us
e
r
to
the
indoor
loca
li
z
e
r
pa
ge
.
−
T
he
indoor
loca
li
z
a
ti
on
pa
ge
c
ontains
number
o
f
pe
r
s
ons
li
s
ted,
a
nd
f
o
ll
owe
d
by
two
butt
ons
;
the
s
e
a
r
c
h
a
nd
s
howing
butt
ons
.
−
T
he
us
e
r
s
hould
s
e
lec
t
one
objec
t
a
t
a
t
im
e
to
s
e
a
r
c
h
f
or
it
s
plac
e
.
I
n
thi
s
pa
pe
r
,
onl
y
one
obj
e
c
t
is
a
va
il
a
ble
but
f
or
the
f
utur
e
wo
r
k,
thi
s
pr
ojec
t
c
a
n
c
ove
r
lar
ge
number
of
them.
−
Af
ter
c
hoos
ing
the
pe
r
s
on's
na
me,
the
us
e
r
s
ho
uld
c
li
c
k
on
s
e
a
r
c
h
loca
ti
on
butt
on
,
whic
h
c
on
tains
the
ha
r
dwa
r
e
c
ontr
oll
ing
a
nd
model
pr
e
diction
r
e
s
ult
s
.
−
T
he
o
ther
butt
on,
loca
ti
on
butt
on,
dis
plays
the
c
a
ptur
e
d
im
a
ge
s
that
ha
d
be
e
n
int
e
r
r
e
d
to
the
model
f
or
tes
ti
ng
r
e
a
s
ons
.
3.
2.
H
ar
d
war
e
i
n
t
e
gr
at
io
n
l
e
ve
l
T
his
s
e
c
ti
on
c
lar
if
ies
how
the
e
quipm
e
nt
pa
r
ts
a
r
e
ga
ther
e
d,
a
s
s
oc
iate
d
a
nd
uti
li
z
e
d
togethe
r
a
s
a
tot
a
l
wor
king
f
r
a
mew
or
k
f
o
r
indoor
it
e
m
r
e
s
tr
iction
r
obot
.
I
t
c
ontains
two
leve
ls
,
the
ini
ti
a
l
s
e
gment
is
c
onne
c
ted
f
o
r
s
tr
uc
tur
e
the
ve
hicle
s
e
gments
,
whi
le
the
s
ubs
e
que
nt
leve
l
is
in
c
ha
r
ge
of
the
c
o
ntr
oll
ing
s
ys
tem
f
or
the
ve
hicle
a
nd
the
r
e
s
t
s
e
gments
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
1693
-
6930
T
E
L
KO
M
NI
KA
T
e
lec
omm
un
C
omput
E
l
C
ontr
o
l
,
Vol.
18
,
No
.
4
,
Augus
t
2020
:
1904
-
1916
1910
3.
2.
1
.
Car
p
ar
t
s
gat
h
e
r
in
g
an
d
ar
r
an
ge
m
e
n
t
T
he
pr
inciple
s
e
gments
uti
li
z
e
d
f
o
r
s
e
tt
ing
up
th
e
mec
ha
nica
l
ve
hicle
a
r
e
a
ppe
a
r
e
d
in
F
igu
r
e
6
.
T
he
s
e
pa
r
ts
a
r
e
R
a
s
pbe
r
r
y
P
i
I
I
I
model
B
,
t
he
L
2
98N
mot
or
d
r
iver
,
t
he
ult
r
a
s
onic
s
e
ns
or
,
t
he
vol
tage
s
teps
down
r
e
gulator
,
t
he
L
i
-
P
o
ba
tt
e
r
y
,
a
nd
t
he
R
P
i
c
a
m
e
r
a
[
20
-
25]
.
F
igur
e
6
.
T
he
ha
r
dwa
r
e
r
e
quir
e
ment
f
or
c
a
r
modul
e
3.
2.
2
Ro
b
ot
ic
c
ar
m
ove
m
e
n
t
an
d
ob
s
t
ac
les
d
e
t
e
c
t
ion
T
wo
ult
r
a
s
onic
s
e
ns
or
s
,
f
ixed
a
t
the
f
r
ont
a
nd
ba
c
k
of
the
c
a
r
,
a
r
e
us
e
d
to
de
tec
t
the
obs
tac
les
in
the
c
a
r
's
pa
s
s
a
ge
wa
y.
T
he
de
tec
ti
on
is
a
c
c
ompl
is
h
e
d
by
mea
s
ur
ing
the
d
is
tanc
e
be
twe
e
n
the
c
a
r
a
nd
t
he
f
a
c
e
d
obs
tac
les
.
I
t
is
im
po
r
tant
to
note
that
thel
s
of
twa
r
e
a
lgor
it
hms
a
r
e
de
s
igned
ba
s
e
d
on
the
p
r
opos
e
d
s
of
twa
r
e
e
nginee
r
ing
model.
T
he
pr
opos
e
d
a
lgor
it
hm
of
c
a
r
moveme
nt
a
nd
obs
tac
les
de
tec
ti
on
is
s
hown
a
s
a
f
l
owc
ha
r
t
in
F
igur
e
7
a
nd
the
ne
xt
li
ne
s
de
s
c
r
ibe
it
in
de
tails
:
−
F
or
f
or
wa
r
d
moveme
nt,
the
R
P
i
f
u
ll
y
s
tops
the
c
a
r
,
if
the
mea
s
ur
e
d
dis
tanc
e
is
s
maller
than
or
e
qua
l
to
(
100
c
m)
.
T
he
c
a
r
is
ke
pt
wa
it
ing
f
or
5
s
e
c
onds
(
s
e
c
s
)
.
−
Af
ter
5
s
e
c
s
,
a
ga
in
i
t
c
he
c
ks
if
the
mea
s
ur
e
d
di
s
tanc
e
is
incr
e
a
s
e
d.
T
his
mea
ns
that
the
obs
tac
les
a
r
e
moved,
then
the
R
a
s
pbe
r
r
y
P
i
or
de
r
s
the
c
a
r
to
dr
i
ve
a
ga
in,
un
ti
l
the
mea
s
ur
e
d
dis
tanc
e
is
s
maller
th
a
n
or
e
qua
l
to
(
100
c
m)
.
−
I
n
c
a
s
e
of
the
5
s
e
c
onds
a
r
e
pa
s
s
e
d
a
nd
the
dis
tan
c
e
s
ti
ll
the
s
a
me,
thi
s
mea
ns
that
thi
s
obs
tac
le
ma
y
be
a
wa
ll
or
a
ny
s
table
ob
jec
t.
−
Now
,
the
c
a
r
is
s
topped,
a
nd
it
s
tar
ts
the
pr
oc
e
dur
e
of
c
a
ptur
ing
a
nd
c
las
s
if
ying
the
im
a
ge
s
.
−
I
f
the
mea
s
ur
e
d
dis
tanc
e
is
s
maller
than
(
80
c
m)
,
thi
s
mea
ns
that
the
r
oboti
c
c
a
r
is
too
c
los
e
to
the
wa
ll
,
is
to
take
a
n
im
a
ge
in
a
c
lea
r
wa
y
the
c
a
r
mu
s
t
dr
ive
ba
c
k
a
t
lea
s
t
f
or
20
c
m
,
whic
h
c
a
n
be
e
quippe
d
by
movi
ng
the
c
a
r
ba
c
kwa
r
d
f
or
(
1
s
e
c
)
.
−
F
or
ba
c
kwa
r
d
moveme
nt,
T
he
R
a
s
pbe
r
r
y
P
i
f
u
ll
y
s
tops
the
c
a
r
,
if
the
de
s
ir
e
d
ti
me
f
or
ba
c
k
wa
r
d
moveme
nt
is
pa
s
s
e
d,
or
i
f
the
mea
s
ur
e
d
dis
tanc
e
is
be
low
than
o
r
e
qua
l
to
(
30
c
m)
.
3.
2.
3
.
Clas
s
if
yin
g
i
m
age
s
p
r
oc
e
d
u
r
e
Af
ter
the
r
oboti
c
c
a
r
ha
s
be
e
n
r
e
a
c
he
d
to
the
wa
ll
s
uc
c
e
s
s
f
ull
y,
the
R
a
s
pbe
r
r
y
P
i
s
tar
ts
in
c
a
ptur
ing
im
a
ge
s
.
T
his
pr
oc
e
dur
e
is
s
hown
in
F
igu
r
e
8
a
s
a
f
lowc
ha
r
t
that
e
xplains
the
p
r
opos
e
d
a
lgor
it
hm
.
T
h
e
im
a
ge
c
las
s
if
ica
ti
on
a
lgor
it
hm
is
pr
opos
e
d
ba
s
e
d
on
t
he
s
of
twa
r
e
e
nginee
r
ing
model.
T
he
f
oll
owing
li
ne
s
a
r
e
pr
e
s
e
nted
to
de
s
c
r
ibe
thi
s
a
lgor
it
hm
in
de
tails
:
−
T
he
f
ir
s
t
s
tep
in
thi
s
pr
oc
e
dur
e
is
to
load
the
pr
e
v
ious
ly
s
a
ve
d
model
in
the
R
P
i
a
s
a
python
s
c
r
ipt
,
by
a
s
pe
c
ial
P
HP
f
or
mul
a
in
the
we
bs
it
e
.
Af
ter
the
us
e
r
c
li
c
ke
d
on
the
s
e
a
r
c
h
loca
ti
on
bu
tt
on,
a
nd
the
r
o
boti
c
c
a
r
f
ound
a
wa
ll
(
o
r
a
s
it
thi
nks
,
it
is
a
wa
ll
)
.
−
T
he
R
P
i
o
r
de
r
s
the
c
a
mer
a
to
take
the
f
ir
s
t
im
a
ge
a
nd
s
a
ve
it
in
ga
l
ler
y
f
il
e
ins
ide
it
.
−
T
he
n
it
tur
ns
the
c
a
r
to
the
lef
t
with
a
15
de
gr
e
e
f
o
r
c
ha
nging
the
im
a
ge
a
ngle,
a
nd
a
ga
in
the
c
a
mer
a
t
a
ke
s
a
nother
im
a
ge
a
nd
s
a
ve
it
with
the
f
ir
s
t
one
.
−
Aga
in,
the
R
P
i
tur
ns
the
c
a
r
ba
c
k
to
i
ts
or
igi
n
a
l
loca
ti
on
by
tu
r
ning
it
to
the
r
ight
with
the
s
a
me
de
gr
e
e
(
15)
.
−
T
he
n
it
moves
the
c
a
r
to
the
r
ight
with
the
s
a
me
mentioned
de
gr
e
e
,
f
o
r
taking
the
las
t
im
a
ge
in
a
n
other
dif
f
e
r
e
nt
c
o
r
ne
r
.
−
At
thi
s
point
the
s
ys
tem
ha
s
th
r
e
e
dif
f
e
r
e
nt
i
mage
s
f
or
d
is
ti
nc
t
c
or
ne
r
s
.
T
he
s
e
im
a
ge
s
e
nter
to
the
lo
a
de
d
model
a
s
r
e
a
l
ti
me
tes
t
im
a
ge
s
.
Evaluation Warning : The document was created with Spire.PDF for Python.
T
E
L
KO
M
NI
KA
T
e
lec
omm
un
C
omput
E
l
C
ontr
o
l
Soft
w
ar
e
e
nginee
r
ing
mode
l
bas
e
d
s
mar
t
indoor
lo
c
ali
z
ati
on
s
y
s
tem
…
(
M
uay
ad
Sadik
C
r
ooc
k
)
1911
−
E
a
c
h
im
a
ge
is
pa
s
s
e
d
thr
ough
the
loade
d
model
a
nd
r
e
a
c
h
to
a
c
e
r
tain
c
las
s
if
ica
ti
on
a
c
c
ur
a
c
y.
T
he
obtaine
d
th
r
e
e
a
c
c
ur
a
c
ies
f
or
m
th
r
e
e
im
a
ge
s
a
r
e
c
ompar
e
d
with
e
a
c
h
othe
r
.
T
he
maximu
m
a
c
c
ur
a
c
y
of
thes
e
a
c
c
ur
a
c
ies
is
c
hos
e
n
a
s
the
a
c
tual
r
e
s
ult
of
the
c
las
s
if
ica
ti
on.
−
I
n
the
p
r
opos
e
d
a
lgor
it
hm
,
a
de
s
ir
e
d
a
c
c
ur
a
c
y
is
c
ons
ider
e
d
to
be
70%
.
−
I
f
the
r
e
s
ult
ing
im
a
ge
c
las
s
if
ica
ti
on
a
c
c
ur
a
c
y
is
e
qu
a
l
or
higher
than
thi
s
de
s
ir
e
d
a
c
c
ur
a
c
y,
then
the
c
las
s
if
ica
ti
on
is
a
c
c
e
pted
a
nd
the
im
a
ge
is
a
n
a
c
tual
wa
ll
.
T
he
n
thi
s
im
a
ge
s
hould
be
a
s
s
i
gne
d
to
the
c
or
r
e
s
ponding
r
oom
a
nd
take
s
it
a
s
the
f
inal
r
e
s
ult
.
Othe
r
wis
e
,
the
R
P
i
o
r
de
r
e
d
the
c
a
r
−
to
tur
n
to
a
no
ther
wa
ll
.
T
his
is
done
by
movi
ng
it
t
o
the
lef
t
with
90
de
gr
e
e
s
.
T
he
n
the
p
r
oc
e
dur
e
is
s
tar
ted
f
r
om
the
be
ginni
ng,
s
tar
ti
ng
f
r
om
r
oboti
c
c
a
r
move
ment
a
nd
obs
tac
les
de
tec
ti
on.
F
igur
e
7
.
R
oboti
c
c
a
r
move
ment
a
lgor
i
thm
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
1693
-
6930
T
E
L
KO
M
NI
KA
T
e
lec
omm
un
C
omput
E
l
C
ontr
o
l
,
Vol.
18
,
No
.
4
,
Augus
t
2020
:
1904
-
1916
1912
F
igur
e
8.
C
a
ptur
ing
a
nd
c
las
s
if
ying
im
a
ge
s
a
lgor
it
hm
4.
T
HE
E
XP
E
RM
E
N
T
AL
RE
S
UL
T
S
T
o
de
ter
m
ine
the
e
f
f
icie
nc
y
of
the
pr
opos
e
d
s
ys
tem,
many
e
xpe
r
im
e
nts
a
r
e
int
r
oduc
e
d
on
dif
f
e
r
e
nt
c
a
s
e
s
tudi
e
s
.
As
pr
e
vious
ly
mentioned,
the
s
ys
te
m
c
ontains
a
we
bs
it
e
,
in
whic
h,
the
us
e
r
int
e
r
f
a
c
e
s
with
the
pr
opos
e
d
a
lgor
it
hms
.
I
n
thi
s
s
e
c
ti
on,
di
f
f
e
r
e
nt
c
a
s
e
s
tudi
e
s
of
dif
f
e
r
e
nt
s
it
ua
ti
ons
a
nd
dif
f
e
r
e
nt
l
oc
a
ti
ons
a
r
e
dis
c
us
s
e
d
in
de
tails
.
T
he
f
ir
s
t
s
tep
in
a
ll
thes
e
c
a
s
e
s
tudi
e
s
i
s
to
logi
n
to
the
s
ys
tem
thr
ough
the
lo
gin
pa
ge
,
f
or
a
uthor
it
y
c
onf
i
r
mi
ng
,
s
o
thi
s
s
tep
is
int
r
oduc
e
d
f
ir
s
t
f
ol
lowe
d
by
the
c
a
s
e
s
tudi
e
s
.
4.
1
.
A
u
t
h
or
izat
ion
c
h
e
c
k
in
g
T
he
us
e
r
s
hould
be
a
uthor
ize
d
to
be
a
ble
to
e
nter
to
the
pa
ge
of
loca
li
z
a
ti
on
in
the
we
bs
it
e
.
Othe
r
wis
e
,
the
us
e
r
c
a
nnot
do
a
nythi
ng
with
we
bs
it
e
e
xc
e
pt
s
e
e
ing
the
inf
or
mation
that
a
r
e
s
hown
in
the
home
pa
ge
.
T
he
r
e
f
or
e
,
a
nd
a
s
a
be
ginni
ng
in
t
he
loca
li
z
a
ti
on
pr
oc
e
dur
e
,
a
ny
e
xpe
r
im
e
nt
or
c
a
s
e
s
tudy
in
the
pr
opos
e
d
s
ys
tem
pa
s
s
e
s
thr
ough
the
f
oll
owing
l
ines
:
−
I
n
t
he
s
ta
r
t
in
g
o
f
t
he
w
e
bs
it
e
,
a
lo
gi
n
b
ut
t
on
is
a
pp
e
a
r
e
d
.
T
h
e
us
e
r
s
h
ou
ld
c
l
ic
k
o
n
i
t
to
vi
e
w
th
e
l
og
in
f
o
r
m
.
I
t
ha
s
be
e
n
n
ot
ice
d
t
ha
t
t
he
r
e
is
n
o
s
i
gn
in
b
ut
t
on
.
F
o
r
s
e
c
u
r
it
y
a
n
d
p
r
iva
c
y
pu
r
p
os
e
s
,
no
one
c
a
n
s
ig
n
i
n
t
o
t
he
s
ys
te
m
w
it
ho
ut
a
u
th
o
r
iz
a
t
io
n
.
Evaluation Warning : The document was created with Spire.PDF for Python.
T
E
L
KO
M
NI
KA
T
e
lec
omm
un
C
omput
E
l
C
ontr
o
l
Soft
w
ar
e
e
nginee
r
ing
mode
l
bas
e
d
s
mar
t
indoor
lo
c
ali
z
ati
on
s
y
s
tem
…
(
M
uay
ad
Sadik
C
r
ooc
k
)
1913
−
A
f
te
r
f
i
ll
i
ng
the
us
e
r
na
me
a
nd
pa
s
s
wo
r
d
wi
th
a
n
a
u
t
ho
r
ize
d
a
c
c
ou
n
t
,
a
n
d
t
he
n
c
l
ic
ki
ng
o
n
t
he
lo
gi
n
bu
t
to
n
.
I
f
t
he
a
c
c
o
un
t
i
s
a
u
th
o
r
i
z
e
d
,
th
e
us
e
r
s
ho
ul
d
s
e
e
t
he
lo
c
a
li
z
a
t
io
n
p
a
g
e
,
e
ls
e
t
he
w
e
bs
it
e
w
il
l
r
e
t
ur
n
to
t
he
ho
me
pa
ge
w
i
th
lo
gi
n
e
r
r
o
r
mes
s
a
ge
.
−
I
n
t
his
s
t
e
p
,
t
he
us
e
r
mus
t
s
e
le
c
t
o
ne
pe
r
s
o
n
f
r
o
m
the
i
nt
r
od
uc
e
d
li
s
t
,
in
t
he
p
r
o
pos
e
d
s
ys
t
e
m
o
nl
y
one
pe
r
s
on
i
s
a
c
t
i
ve
.
B
u
t
in
t
he
f
u
tu
r
e
,
o
th
e
r
pe
r
s
o
ns
wi
l
l
be
a
c
ti
ve
a
ls
o
.
I
t
s
h
ou
l
d
be
no
t
ice
d
th
a
t
o
nl
y
on
e
pe
r
s
o
n
s
h
o
uld
b
e
s
e
lec
te
d
a
t
a
ny
ti
me
.
−
A
f
te
r
c
ho
os
i
ng
th
e
na
me
o
f
t
he
p
e
r
s
on
to
s
e
a
r
c
h
f
o
r
,
th
e
us
e
r
s
ho
ul
d
c
li
c
k
o
n
t
he
s
e
a
r
c
h
lo
c
a
ti
on
but
t
on
.
T
h
e
n
t
he
p
r
o
c
e
d
u
r
e
o
f
c
o
nt
r
ol
li
ng
t
he
r
ob
o
ti
c
c
a
r
by
t
he
R
a
s
pb
e
r
r
y
P
i
,
c
he
c
ki
ng
f
o
r
obs
ta
c
l
e
s
,
a
nd
c
a
ptu
r
in
g
t
he
i
ma
ge
s
th
a
t
p
a
s
s
e
s
th
r
ou
gh
d
if
f
e
r
e
n
t
a
l
go
r
it
hm
s
to
o
.
T
h
e
e
va
lua
t
io
n
o
f
t
he
m
w
i
ll
be
i
nt
r
od
uc
e
d
in
t
he
f
ol
lo
wi
ng
s
e
c
ti
ons
wi
t
h
d
if
f
e
r
e
n
t
c
a
s
e
s
tu
dies
.
−
T
h
e
f
i
na
l
r
e
s
u
l
ts
a
r
e
s
ho
wn
a
s
a
t
e
x
t
i
n
t
he
we
bs
i
te
.
−
I
f
t
he
us
e
r
c
li
c
ke
d
o
n
the
s
ho
w
l
oc
a
t
io
n
b
ut
to
n
,
the
las
t
th
r
e
e
ta
ke
n
i
ma
ge
s
a
r
e
s
h
ow
n
i
n
t
he
ga
l
le
r
y
we
bp
a
g
e
.
4.
2.
Cas
e
s
t
u
d
y
on
e
T
he
f
ir
s
t
e
xpe
r
im
e
nt,
whic
h
c
a
n
be
na
med
a
s
the
n
or
mal
c
a
s
e
s
tudy
with
d
is
tanc
e
les
s
than
or
e
qua
l
to
100
c
m
is
int
r
oduc
e
d.
He
r
e
,
a
f
ter
the
us
e
r
s
e
lec
ts
the
r
e
quir
e
d
pe
r
s
on
to
s
e
a
r
c
h
f
o
r
,
the
f
oll
owi
ng
s
teps
a
r
e
f
oll
owe
d:
a.
T
he
R
a
s
pbe
r
r
y
P
i
or
de
r
s
the
r
oboti
c
c
a
r
to
move
f
o
r
wa
r
d.
T
he
two
blue
L
E
Ds
a
r
e
on
,
indi
c
a
te
the
f
o
r
wa
r
d
movi
ng,
a
s
s
hown
in
F
igur
e
9
(a
)
.
b.
T
he
c
a
r
moves
f
or
wa
r
d
a
nd
c
he
c
ks
the
dis
tanc
e
,
whe
n
it
be
c
a
me
les
s
than
or
e
qua
l
to
100
c
m
a
nd
mor
e
than
80
c
m
,
a
s
the
a
lgor
it
hm
s
a
ys
,
the
R
a
s
pbe
r
r
y
P
i
or
de
r
s
the
c
a
r
to
s
top
a
nd
wa
it
f
o
r
5
s
e
c
.
T
he
n,
a
ga
in
the
dis
tanc
e
is
mea
s
ur
e
d,
if
it
c
ha
nge
s
,
the
c
a
r
c
onti
nue
s
to
move
f
or
wa
r
d
a
nd
wa
it
the
dis
tanc
e
to
be
c
ome
les
s
than
or
e
qua
l
to
100c
m
a
s
s
hown
in
F
igur
e
9
(b
)
.
Othe
r
w
is
e
a
nd
a
s
s
hown
in
F
igur
e
10,
the
R
a
s
pbe
r
r
y
P
i
s
tar
ts
c
a
ptur
ing
th
r
e
e
tes
t
im
a
ge
s
in
thr
e
e
dif
f
e
r
e
nt
a
ngles
,
with
15
de
gr
e
e
s
be
twe
e
n
e
a
c
h
c
or
ne
r
.
T
he
n
the
c
a
r
r
e
tu
r
ns
to
it
s
o
r
igi
na
l
loca
ti
on.
c.
T
he
s
e
im
a
ge
s
a
r
e
s
e
nt
to
the
we
bs
it
e
to
be
us
e
d
ins
ide
the
pr
ogr
a
m
of
c
las
s
if
ica
ti
on.
T
he
c
las
s
if
ica
ti
on
pr
oc
e
dur
e
take
s
a
bout
3
-
5
s
e
c
.
T
he
r
e
s
ult
s
a
r
e
s
ho
wn
in
the
we
b
pa
ge
a
s
a
text
a
c
c
or
ding
to
F
igur
e
1
1.
(
a
)
(
b)
F
igur
e
9
.
(
a
)
T
he
c
a
r
is
mov
ing
f
o
r
wa
r
d
,
(
b
)
T
he
n
e
a
r
e
s
t
wa
ll
is
on
90
c
m
dis
tanc
e
F
igur
e
10
.
T
he
pr
oc
e
dur
e
o
f
c
a
ptur
ing
im
a
ge
s
Evaluation Warning : The document was created with Spire.PDF for Python.