TELKOM
NIKA
, Vol.13, No
.2, June 20
15
, pp. 624 ~ 6
3
1
ISSN: 1693-6
930,
accredited
A
by DIKTI, De
cree No: 58/DIK
T
I/Kep/2013
DOI
:
10.12928/TELKOMNIKA.v13i2.1189
624
Re
cei
v
ed
No
vem
ber 2
8
, 2014; Re
vi
sed
March 3, 201
5; Acce
pted
March 27, 20
15
Object Detector on Coastal Surveillance Radar Using
Two-Dimensional Order-Statistic Constant-False Alarm
Rate Algoritm
Da
y
a
t Kurni
a
w
a
n*, Pur
w
oko Adhi, Ar
ief Sur
y
adi
Sat
y
a
w
a
n
, Iqb
a
l Sy
amsu, T
e
guh Praludi
Rese
arch Ce
nter F
o
r Electron
ics and T
e
leco
mm
unicati
on, Indo
nesi
an Insti
t
ute of Science
Kampus LIPI, Jl. Sangkur
ia
ng,
GD. 20, Bandung 4
0
1
35, Ind
ones
ia
*Corres
p
o
ndi
n
g
author, e-ma
i
l
: da
ysdk6
3
@g
mail.com
A
b
st
r
a
ct
T
h
is pa
per d
e
s
c
ribes th
e dev
elo
p
m
ent of ra
dar o
b
ject d
e
te
ction usi
ng tw
o
di
mens
io
nal c
onstant
false a
l
ar
m rat
e
(2D-CF
AR).
Objective
of thi
s
deve
l
op
m
ent
is to mini
mi
z
e
nois
e
detecti
on
if compar
ed w
i
th
the previ
ous al
gorith
m
that
us
es one di
mens
ion
a
l consta
nt false alar
m rat
e
(1D-CF
AR) a
l
gorit
hm suc
h
as
order-statistic (
O
S) CF
AR, cell-aver
ag
ing (
C
A) CF
AR
, AND log
i
c (AND)
CF
AR an
d var
i
abi
lity i
ndex (
V
I)
CF
AR w
here h
a
s bee
n i
m
p
l
e
m
e
n
ted o
n
coa
s
tal survei
lla
nc
e radar. T
he o
p
timu
m detecti
on resu
lt in coa
s
ta
l
surveil
l
a
n
ce ra
dar testin
g w
h
e
n
Pfa set to 1e-
2, K
th
set to 3/4*Nw
ind
o
w
and
Guard Ce
ll set
to 0. Princip
l
e
of
2D-CF
A
R al
go
rithm is co
mbin
ing of tw
o CF
AR alg
o
rith
ms fo
r each array d
a
t
a of a
z
i
m
uth a
nd ran
ge. Orde
r
statistic (OS) C
F
AR algorit
m i
s
imp
l
e
m
e
n
ted
on this 2D-CF
A
R by fusion r
u
le of AND lo
gi
c.T
he algor
ith
m
of
2D-CF
A
R
is d
e
vel
ope
d
usin
g Micros
oft Vi
sual
C+
+
20
08
an
d the
o
u
tp
ut of 2
D
-CF
A
R is
plotte
d o
n
PPI
scope r
a
d
a
r us
ing GDI+
l
i
brar
y. T
he resu
lt o
f
2D-C
F
A
R
de
velo
p
m
ent s
h
o
w
s that 2D-CF
A
R can
min
i
mi
ze
nois
e
d
e
tected
if co
mp
ared
w
i
th 1D-CF
A
R
w
i
th the
sa
me
para
m
eter of
CF
AR. Best p
e
rformanc
e of
2D-
CF
AR in
o
b
ject
detecti
on
w
h
e
n
Nw
in
dow
set
to 1
28.
T
h
e ti
me
of s
o
ftw
are proc
essin
g
of
2D-CF
A
R
is a
b
out
tw
o times lon
g
e
r than the 1
D
-
C
F
A
R.
Ke
y
w
ords
: Object Detectio
n, Noise E
n
vironm
ent, OS-CFA
R, 2D-CFAR
1. Introduc
tion
Obje
ct o
r
ta
rget dete
c
tion
in
noi
se
en
vir
onme
n
t is very im
port
ant p
r
obl
em
in rada
r
system. On
e
techni
que to
detect of obj
ect in noi
se
environ
ment i
s
u
s
ing
con
s
t
ant false al
arm
rate (CFAR).
This
detec
tion refers
to
a c
o
mm
on form of ada
ptive algorithm
used in rad
a
r
system
s to d
e
tect target return
s a
gain
s
t a ba
ckgr
o
u
n
d
of noi
se,
cl
utter an
d inte
rfere
n
ce [1]. Cell
averag
e (CA
)
and
o
r
de
r statistic (OS) CFAR
ha
s b
een i
m
plem
e
n
ted o
n
coa
s
tal su
rveillan
c
e
rada
r wh
ere has be
en test
ed in Tanjun
g
Pasir B
each locate
d in Banten Province
, Indonesia a
nd
the result as
sho
w
n o
n
Fig
u
re 1.
(a)
C
A-
CFA
R
(b)OS-CFAR
Figure 1. Ech
o
sign
al of ob
ject and n
o
ise plotted on
PPI scope
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
Object Detector on Coastal
Surveill
ance
Rada
r Using Two-Dim
e
nsi
onal .... (Dayat Kurniawan)
625
The optimum
re
sult of
co
astal ra
dar
d
e
te
ction
in range
1 NM
(Na
u
tical
Mil
e
) wh
en
Nwi
ndo
w
set
to 64, Pfa
set
to 1e
-2
but n
o
ise
si
gnal
f
r
om surrou
ndi
ng a
r
ea
is stil
l dete
c
ted.
CA-
CFAR ha
s g
ood
perfo
rma
n
ce
on
hom
o
geno
us env
ironment and
the
oth
e
r han
d
OS-CFA
R ha
s
good p
e
rfo
r
m
ance on non
homog
eno
us
environ
ment
and multiple t
a
rget
s [1]-[6]. Figure 2
sho
w
s
the performa
c
e dete
c
tion
of the CA and OS
CFAR algoritm whe
r
e have been
impleme
n
ted on
coastal radar. The detection proba
bility Pd of CA CFAR is de
scribe in [7],[8] and Pd of OS
CFAR is
des
c
ribe in [8],[9].
Figure 2. Performa
nce dete
c
tion of CA a
nd OS CFA
R
Two
-
dime
nsi
onal
con
s
tant
-false
ala
r
m
rate (2D-CFA
R
)
will b
e
de
veloped to
m
i
nimize
noise dete
c
te
d rath
er th
an
usin
g on
e di
mensi
onal
CFAR. Prin
cipl
e of two
dime
nsio
nal
CFA
R
is
combi
n
ing of
two cfar alg
o
rithm [10] to compa
r
e
cel
l
under te
st with array da
ta of azimuth
bin
cell and array data of ra
nge bin
cell as shown in Fi
gure 3. OS
-CF
A
R will be im
plemented and
tested on thi
s
developm
ent
of 2D-CFAR.
It is
chosen
becau
se it ha
s
goo
d pe
rformance on n
o
n
homog
ene
ou
s e
n
viron
m
e
n
t and
for m
u
ltiple tar
get
s. Th
e bl
ock diag
ram
of
OS-CFAR i
s
as
s
h
ow
n
on
F
i
gu
r
e
4
.
Figure 3. Prin
ciple of 2
D
-CFAR
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 13, No. 2, June 20
15 : 624 – 63
1
626
Figure 4. The
Block dia
g
ra
m of the OS-CFAR al
gorit
m.
The
perfo
rma
n
ce
dete
c
tion
of OS
-CFA
R dep
endi
ng
o
n
sele
ct K
th
whi
c
h
K
th
= 3
/
4*N
i
s
the optimum
value [11]
and probability
of false alarm
(Pfa) val
ue.
The K
th
value is
es
timated
to
be the means of c
l
utter [12]. Sc
aling fac
t
or (Tos
) c
a
lc
ulated by following formula
[8],[9],
[13] :
1
!
!
!
(1)
Whe
r
e:
Pfa
= Prob
ality False Ala
r
m
K
= S
e
lect
e
d
ce
ll unt
er t
e
st
T
= Scalin
g Fa
ctor
N
= Sliding Wi
n
dow
This pa
pe
r focu
s on the de
velopment of 2D-CFA
R ba
sed o
n
com
b
i
ne of two OS-CFAR
algorith
m
. 2D OS-CFA
R d
e
tector
algo
rithm wa
s devel
oped u
s
in
g Micro
s
oft Visu
a
l
Studio C++
2008.
2. Rese
arch
Metho
d
The
2D-CFAR impl
eme
n
ted by
co
mbi
nes of two
OS-CFA
R al
goritm as
sh
own
on
Figure 5.
First step
is calculate OS
-CA
F
R fo
r
azim
u
t
h and
the
n
calcul
ate OS
-CFAR for ran
ge
with o
u
tput
o
f
each
step
i
s
conve
r
t int
o
bin
a
ry
num
ber which lo
gic
1
rep
r
e
s
e
n
t as o
b
je
ct
and
logic 0 rep
r
e
s
ent as noi
se
.
Each oup
ut
from
OS-CF
A
R-Azi
m
uth and OS-CFA
R-Ran
ge will
be
comp
ared using AND logi
c rule to get
output
of 2D-CFAR a
s
sho
w
n on T
able 1. Base
on
optimum setting of previous experiment
of coas
tal surveillance radar,
so
setting parameter
of
each OS-CF
A
R is Pfa=1
e
-2, Nwind
o
w
=64, K
th
=3/
4
*Nwind
ow,
array of bin cell of ra
nge
=10
24
and array of bin cell of a
z
i
m
uth=3
60.
Figure 5. Block di
agram 2
D
OS-CFAR
with AND fu
si
on
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
Object Detector on Coastal
Surveill
ance
Rada
r Using Two-Dim
e
nsi
onal .... (Dayat Kurniawan)
627
The
key fa
ctor of
CFA
R
algorith
m
s lie
s in
setting the threshold
adaptively by
estimatin
g
t
he
backg
rou
nd n
o
ise p
o
wer in
clud
ed in a te
st cell [14]. T
he scalin
g factor (T) d
e
cre
a
se
s when Pf
a
or K
th
incre
a
se [15],[16] for fixed Nwi
nd
ow valu
e. Array data of ra
nge a
nd a
z
i
m
uth is
coll
e
c
ted
from rada
r re
ceiver a
s
be
at sign
al thro
ugh An
alog t
o
Digital
Co
n
v
erter
(ADC) usin
g firm
ware
wa
s install
e
d
in PC/Lapto
p
as sho
w
n
on Fi
gure 6. Raw d
a
ta
from ADC
need to be
pre
-
pro
c
e
ssi
ng a
nd FFT p
r
o
c
e
ssi
ng first before throug
h
to CFAR
pro
c
essor. Imple
m
entation
co
de of
OS-CFAR u
s
i
ng Mi
cro
s
oft
Visual
C++
2
008 a
nd t
he para
m
eter se
tting
of
OS-CFAR is
as sh
own
on Figu
re 7.
Table 1. The
rule of AND logic 2
D
-CFA
R
OS-C
FAR-Azimu
t
h OS-C
FAR-
Rang
e
Ouput
2
D
-C
FAR
Description
0 0
0
Noise
0 1
0
Noise
1 0
0
Noise
1
1
1
Object or ta
rget
Figure 6. Coll
ect data of array bin cell
ra
nge an
d azim
uth
Figure 7. Parameter Settin
g
OS-CFAR
Raw Data
CFA
R
Th
r
e
shold
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 13, No. 2, June 20
15 : 624 – 63
1
628
3. Results a
nd Analy
s
is
Simulation fo
r 2D
OS-CF
A
R from
ra
w data
coa
s
tal surveill
an
ce rada
r is
done
by
config
urin
g
CFAR p
a
ramet
e
r
on
win
d
o
w
software a
s
sho
w
n
on
Fig
u
re
7. Fi
rst
si
mulation
is o
ne
dimen
s
ion
a
l CFAR u
s
in
g
OS algorith
m
for ran
ge bin
cell an
d the
n
two dime
nsional CFAR
with
parameter
setting is radial
distan
ce set
to 0.5NM, sliding window
set to 64, probability of fal
s
e
alarm
set to 1e-2. Fig
u
re
8(a
)
sh
ow th
e result of one dimen
s
ion
CFAR
whe
r
e
PPI scope sho
w
many clutte
r or noi
se in th
ere. Fig
u
re
8(b) sh
o
w
the result of two
d
i
mensi
onal
CFAR where PPI
scope
sho
w
redu
ction
of
clutter
or noi
se i
n
t
he
sa
me region
of
Figu
re
8(a
)
.
Figu
re
9 sh
ow
detectio
n
performan
ce of
1D OS-CFA
R and
2
D
O
S
-CFA
R wh
e
r
e 2D
OS-CFAR ha
s bet
ter
detectio
n
rath
er than 1D O
S
-CFA
R for same S
NR val
ue. Figure 1
0
show differe
nt perform
an
ce
of 2D-CFA
R
with differe
nce Nwi
ndo
w
setting whe
r
e
radial
dista
n
ce set to 1 n
a
u
tical mile a
n
d
probability of false alarm set to 1e-2.
(a)one dim
e
n
s
ion
a
l CFA
R
(
b
)
t
w
o
d
i
mens
io
na
l C
F
AR
Figure 8. Minimize noi
se d
e
tected
CFA
R
Figure 9. Performa
nce dete
c
ti
on of 1D-O
S and 2D-OS
CFAR
Better perfo
rmance of 2D-CFA
R achived whe
n
Nwi
ndo
w set to 128 a
s
sho
w
n on Figure 11.
Obje
ct
or
ta
rget
dete
c
ted can be see
n
clea
rly
fr
om t
he othe
r its
becau
se
scal
ing facto
r
(T) is
N
oi
se
re
dect
i
o
n
N
oi
se
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
Object Detector on Coastal
Surveill
ance
Rada
r Using Two-Dim
e
nsi
onal .... (Dayat Kurniawan)
629
decrea
s
e
s
when Nwind
o
w incre
a
ses a
s
sho
w
n on Ta
ble 2. Clutter
or noi
se is im
pact of intern
al
or exte
rnal i
n
terferen
ce li
ke
som
e
u
n
w
ante
d
p
r
od
uct of
digital
synthe
si
s, p
o
we
r
conve
r
t
e
r o
r
reflectio
n
co
ming from ne
arby buildi
n
g
s
and
sea [17
]
.
Table 2. Scali
ng facto
r
(T) f
o
r OS-CFA
R for Pfa=1e
-2
and K
th
= 3/4*
Nwi
ndo
w
N
w
indo
w
Scaling Factor (T
)
16 4.29
32 3.89
64 3.69
128 3.59
.
(a)Nwind
ow=16
(b)Nwind
ow=32
(
c
)
N
w
i
nd
ow
=6
4
(d)Nwind
ow=128
Figure 10. 2D-CFA
R pe
rformanc
e with di
fference Nwin
dow
Table 3
shows the diffe
ren
c
e of p
r
o
g
ra
m pro
c
e
s
sing
time of one
d
i
mensi
onal
CFAR an
d
two dimen
s
io
nal CFA
R
wit
h
pro
c
e
s
sor I
n
tel co
re 2 du
o @ 2.1 G
H
z, 32 bit opera
t
ion system a
n
d
3GB RAM in
stalled o
n
lap
t
op. The time of progr
am p
r
ocessin
g
is
getting high
er when
Nwi
nd
ow
set to high
er
and p
r
og
ram
pro
c
e
ssi
ng ti
me of 2D-CF
A
R is ab
out two time
s lon
ger tha
n
the 1D-
CFAR. The efficien
cy
of sele
cted
CF
AR
pe
rfor
ms dep
end
s o
n
value
s
of th
e len
g
th of
cell
(Nwind
ow)
[1
8]
co
rrelated
with rotation
spe
ed
of ant
enna
for on
e
degree. If the
time p
r
o
c
e
s
sing
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 13, No. 2, June 20
15 : 624 – 63
1
630
of 2D-CFAR is m
o
re lon
gger than
ro
tation
spee
d
of ante
nna
so
Nwindo
w CFA
R
mu
st
be
decrease to reduction time proc
essing but
it’s will
decrease
perf
o
rm of
radar object
detecti
on
too.
Figure 11. Dif
f
eren
ce pe
rfo
r
man
c
e 2
D
-O
S-CFA
R
ba
se on numbe
r of sliding wi
n
dow
Table 3. The
time of program pro
c
e
s
sin
g
of 1D-CFA
R vs 2D-CFA
R
N
w
i
ndo
w
1D-C
FAR (ms
)
2D-C
FAR (ms
)
16 1
3
32 2
5
64 5
10
128 10
21
4. Conclusio
n
Software
dev
elopme
n
t of
2D-CFA
R
wo
rks a
s
expe
cted. Oup
u
t of
2D-CFAR
a
nd 1
D
-
CFAR
ha
s d
i
fferent pe
rfo
r
man
c
e
on o
b
ject o
r
targ
et detectio
n
in noi
se inve
ronm
ent ra
d
a
r
system. P
e
rf
orma
nce of
o
b
ject
dete
c
tio
n
of
2D
-CFAR i
s
better when
Nwind
o
w set
to 1
28.
The
time of software
pro
c
e
s
sin
g
of 2
D
-CFA
R i
s
ab
out two times lon
g
e
r tha
n
the
1
D
-CFAR. In
the
future optimi
z
e of algorithm
to redu
ce sof
t
ware p
r
o
c
e
s
sing time is n
eede
d.
Ackn
o
w
l
e
dg
ement
Than
ks to
Re
sea
r
ch
Center for Electro
n
ics an
d Tel
e
comm
unication
s-In
done
sia
n
Institute of Science, Bandu
ng for
all the sup
port an
d coorp
e
ration.
Referen
ces
[1]
Yuhu
a Q, Huil
i
G,
T
i
ng L. A n
e
w
CF
AR D
e
te
ctor base
d
o
n
Automatic C
e
n
s
orin
g Ce
ll Av
erag
ing
an
d
Cell Av
erag
ing.
T
E
LKOMNIKA Indones
ia
n Jo
urna
l of Electri
c
al Eng
i
ne
eri
n
g
. 2012; 1
1
(6): 329
8-33
03.
[2]
Sung W
H
, Do
n
g
SH.
Perfor
mance A
nalys
is of an Envi
r
o
n
m
ental A
daptiv
e
CF
AR Detector
. 2014: 1-7.
http://dx
.
doi.
o
rg/10.
1155/2014/615704.
[3]
H Mansour
i, M Hamado
uch
e
, F
Youcef E.
Infl
ue
nce of Ordere
d
and W
e
i
ghted An
al
ysis
W
i
ndo
w
s
o
n
Detectio
n in a
CFAR Rad
a
r
. Internati
o
n
a
l Jo
urna
l of Circuit
,
System an
d S
i
gn
al Process
i
ng
. 20
14; 8:
109-
115.
[4]
Moham
ed B,
F
aouzi S.
P
e
rformanc
e Co
mparis
on of S
o
me
CF
AR D
e
tectors in
Ho
moge
no
us an
d
Non-H
o
mo
gen
eous C
l
utter.
IEEE International Confer
ence
on
Signal and Image Precessing
Appl
icatio
ns(IC
S
IPA). 2013: 1
01-1
05.
[5]
Moham
ed BE
M. Performanc
e Anal
ys
is of CF
AR De
tectio
n of F
l
uctuatin
g Ra
d
a
r T
a
rgets in Noni
de
a
l
Operatin
g Envi
ronme
n
t.
Internation
a
l Jo
urna
l of Aerospac
e
Scienc
e
. 201
2; 1(3): 21-35.
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
Object Detector on Coastal
Surveill
ance
Rada
r Using Two-Dim
e
nsi
onal .... (Dayat Kurniawan)
631
[6]
Lon
g C, Xi
aoc
hua
n M, Qi X,
Bin L, Shi
w
e
i
R.
Performanc
e Anal
ysis o
n
Some Ne
w
CF
AR Detector
s
und
er Clutter
. Journ
a
l of Co
mputers.
201
1; 6
(
6): 1278-
12
85
.
[7]
In-K
yu K, et
al. T
a
rget Det
e
ction Pr
ob
abi
lit
y
Simu
latio
n
in th
e H
o
mo
gen
eo
us Grou
nd C
l
utter
Enviro
nment.
KSAS International Journal.
2
005; 6(1): 8-1
6
.
[8]
B Magaz, A Belo
uchra
n
i,
M Hamad
ouc
h
e
. A Ne
w
Ad
aptive L
i
n
i
er
Combi
n
e
d
CF
AR Detector i
n
Presenc
e of Interferin
g T
a
rgets.
Progress in
Electro
m
a
gneti
cs Researc
h
B
. 2011; 3
4
: 367
–38
7
.
[9]
Nadav L. Detection Loss
D
u
e
to Interferin
g
T
a
rgets in Ord
e
red Ststistic
CF
AR.
IEEE Transaction
on
Aerosp
ace a
n
d
Electronics Sy
stems.
19
88; 2
4
(6): 678-
68
1.
[10]
Matthias K, Rohli
ng H. F
a
st
T
w
o D
i
mens
i
ona
l CF
AR Proced
ure.
IEEE Transactions
on Aeros
pac
e
and El
ectron
ic Systems
. 20
13
; 49(3): 181
7–1
823
.
[11]
B Magaz, A
Belo
uchra
n
i,
M Hamad
ouc
he. Au
tomatic
T
h
reshold S
e
lecti
on i
n
OS-CF
A
R Rad
a
r
Detectio
n usi
n
g Informatio
n
T
heoritic Criter
ia.
Progr
ess In
Electro
m
a
gne
tics Rese
arch
B.
2011;
30:
157
–1
75
.
[12]
Xi
an
g S, Ran T
,
Xia B.
A
F
a
st Order Methode on OS-CF
A
R Detector in SAR
I
m
ag
e
. 2
009:
72
5-72
8
.
http://dx
.
doi.
o
rg/10.11
09/APS
AR.200
9.5374218.
[13]
Rohl
in
g H. Rad
a
r CF
AR
T
h
reshol
din
g
in Cl
utter and Multi
p
l
e
T
a
rget Situatio
ns.
IEEE Trans
actions on
Aerosp
ace a
n
d
Electronic Sys
t
ems
. 19
83; 19
: 608–6
21
.
[14]
Aja
y
KY,
La
xmi K. Movi
ng
T
a
rget Dete
ct
ion Us
in
g V
I
-CF
A
R Algor
i
t
hm on M
a
tla
b
Platform.
Internatio
na
l Journ
a
l of Adv
a
nced
Rese
arch
in Co
mputer
Scienc
e an
d S
o
ftw
are Engin
e
e
rin
g
. 201
3;
3(12): 91
5-9
1
8
.
[15]
H Mansour
i, M Hamado
uch
e
,
F
Youcef E.
Influe
nce of Ordere
d
and W
e
i
ghted An
al
ysis
W
i
ndo
w
s
o
n
Detectio
n in a
CFAR Rad
a
r
. Internati
o
n
a
l Jo
urna
l of Circu
its Systems a
n
d
Signa
l Proces
sing
. 20
14
;
8: 109-1
15.
[16]
Waleed KAA,
Najim
AU. Evaluation
of AND-CF
AR and O
R
-CFAR Proc
essor under
Different Clutter
Mode
ls.
Engin
eeri
ng & T
e
chn
o
lo
gy Journ
a
l
.
201
3; 31(5): 96
4-97
5.
[17]
P Papr
ocki. Impact of Inter
n
a
l
an
d E
x
ter
nal
Interferenc
es
on th
e Perfor
mance
of a F
M
CW
Rad
a
r
.
Internatio
na
l Journ
a
l on M
a
ri
ne Nav
i
gat
i
on
and Safety of
Sea T
r
ansp
o
rtation
. 20
11; 5(
3): 325-3
28.
[18]
R
y
sz
ard W
.
Experime
n
tal
Eva
l
uati
on
of T
he
Consta
nt F
a
ls
e
Alarm
Rat
e
(
C
FAR) Al
gorith
m
s use
d
i
n
Marine F
M-CW
Radars
. Scie
n
t
ific Journa
ls Mariti
me Un
ivers
i
ty of S
z
c
z
e
c
i
n
.
2013; 3
6
(10
8
): 177-1
81.
Evaluation Warning : The document was created with Spire.PDF for Python.