Estimation of the Relation between Weibull Distributed Sea Clutter and the CA-CFAR Scale Factor

Abstract

When radars operate in coastal or offshore environments, an undesired signal known as clutter appears as
background in the measurements. The CA-CFAR detector is the classic solution for the detection of targets inside clutter, typically using a fixed value for its adjustment factor the entire period of operation. Using MATLAB, the author simulates the CA-CFAR response under amplitude samples from the classical sea clutter Weibull distribution. As a result, a relation between operating conditions (Weibull shape parameter) and the most efficient adjustment factor is obtained for various false alarm probabilities. Thus, the implementation of a new detector that achieves the adaptation to changing environments through the use of a variable adjustment factor is suggested. Note that sea environment is present in most radar scenarios both in military and meteorological applications.

Full text

Ingeniería 25 (2): 19-28, ISSN: 1409-2441; 2015. San José, Costa Rica
ESTIMATION OF THE RELATION BETWEEN WEIBULL
DISTRIBUTED SEA CLUTTER AND THE CA-CFAR
SCALE FACTOR
1. INTRODUCTION
The task of primary radars is to detect
objects within the observation area and
estimate their position (Barton and Leonov,
1998). Target detection would be an easy task
if objects that produce echoes were located
on a non-reflecting background. In that case,
the echo signal could simply be compared
with a fixed threshold, and targets would be
detected when the received signal exceeded the
threshold (Kouemou, 2009).
However, in real life radar applications,
targets almost always appear embedded in a
background lled with clutter, which is a random
signal. Frequently, the clutter signal’s behavior is
subject to time and position variations. Therefore,
the application of adaptive processing techniques
becomes necessary to calculate constantly
changing detection thresholds that correspond
Esta obra está bajo una Licencia de Creative Commons. Atribución - No Comercial - Compartir Igual
Abstract
When radars operate in coastal or offshore environments, an undesired signal known as clutter appears as
background in the measurements. The CA-CFAR detector is the classic solution for the detection of targets inside
clutter, typically using a fixed value for its adjustment factor the entire period of operation. Using MATLAB, the author
simulates the CA-CFAR response under amplitude samples from the classical sea clutter Weibull distribution. As a
result, a relation between operating conditions (Weibull shape parameter) and the most efficient adjustment factor is
obtained for various false alarm probabilities. Thus, the implementation of a new detector that achieves the adaptation
to changing environments through the use of a variable adjustment factor is suggested. Note that sea environment is
present in most radar scenarios both in military and meteorological applications.
Keywords: sea clutter, Weibull distribution, CA-CFAR, false alarm probability, radar targets detection.
Resumen
Cuando un radar opera en ambientes costeros o en alta mar, aparece una señal indeseable denominada clutter
marino como fondo en las mediciones. El esquema CA-CFAR es la solución clásica a la detección de blancos
sobre fondos de clutter, contando con un factor de ajuste que se mantiene fijo durante todo el período de operación.
Empleando la herramienta MATLAB, el autor simula mediciones de amplitud del clásico clutter Weibull bajo la
operación CA-CFAR. Como resultado, se obtiene una relación entre las condiciones de operación (parámetro de
forma Weibull) y la configuración más eficiente del factor de ajuste para varias probabilidades de falsa alarma. Así,
es sugerida la implementación de un detector con factor variable, adaptado a los cambios del ambiente. Nótese que el
entorno marino está presente en la mayor parte de los escenarios tanto en aplicaciones militares como meteorológicas.
Palabras clave: clutter marino, distribución Weibull, CA-CFAR, Probabilidad de Falsa Alarma, Detección de
Blancos de Radar.
Recibido: 20 de Febrero 2015 Aprobado: 30 de Julio 2015
José Raúl Machado Fernández

Ingeniería 25 (2): 19-28, ISSN: 1409-2441; 2015. San José, Costa Rica
20
with the clutter’s local situation (Skolnik, 2008).
The techniques are even more necessary on widely
variable backgrounds such as sea clutter, which is
the signal obtained from the radar’s echo reected
at the sea surface (Machado and Bueno, 2012).
In order to obtain the necessary local
information, schemes with sliding windows
around the analyzed cell are commonly used
(Rohling, 1983; Nagle, 1991). According on the
application, the number of cells to be used in
the window may vary, being the larger amounts
responsible for a better estimate of the clutter
average and the smaller ones more effective
at eliminating critical situations, such as: the
presence of multiple nearby targets and the
occurrence of abrupt changes in the background’s
level. When such situations occur, the clutter
is said to be heterogeneous. Otherwise, it’s
categorized as homogeneous (Bacallao, 2003).
When detectors are designed for situations
where targets appear inside sea clutter, the well-
known Neyman-Pearson theorem is applied. This
means that the designer rst seeks to ensure a
given false alarm probability (Pf) and then tries
to maximize the probability of detection (Pd).
Thus, the most popular clutter level estimation
mechanisms are known as CFAR (Constant
False Alarm Rate) because they ensure that the
detection will occur under the guarantee of a
constant false alarm (Skolnik, 2008).
Conceived at rst under the assumption of
Gaussian distributed clutter, several types of
CFAR algorithms can be found, all based on the
sliding window mechanism. The most popular
are the CA-CFAR (Cell Averaging), the GO-
CFAR (Greatest-Of), the SO-CFAR (Smallest-
Of) and OS-CFAR (Ordered Statistics). These
detectors have been treated in the literature by
several authors (Rohling, 1983; Farina, Studer,
1986; Weingberg, 2004). and are often used as a
reference on recent researches (Takahashi, 2010;
Caso, De Nardis, 2013; de Figueiredo, 2013;
Qin, Gong, 2013). . In addition, each year new
alternatives and contributions appear in various
international journals. Some proposals try to
introduce new processing methods (Van Cao,
2012; Qin, Gong, 2013), while others focus on
improving the existing ones (Kumar, Kant, 2013;
Magaz, Belouchrani, 2011). However, all CFAR
implementations have in common that they allow
the adjustment of the false alarm probability by
means of the modication of a scale or adjustment
factor (K), which has an inverse relationship
with the probability of detection (Rohling, 1983;
Farina, Studer, 1986).
The preliminary statement that claimed the
clutter was Gaussian distributed was quickly
proven as false by several (Haykin, Bakker,
Currie, 2002; Antipov, 1998). Specically in the
case of sea clutter, numerous studies have shown
that the family of heavy-tailed distributions is
the best suited for representing measurements
made on the sea surface. While many others have
been proposed, the following distributions are
generally the most accepted by the community:
Rayleigh, Log-Normal, K, Weibull and Log-
Weibull (Antipov, 1998; Oyedokun, 2012; Totir,
Rador, Anton, 2008; Jian-bo Hu, Wen-wen, 2009).
1.1 Motivation and objetives
Recent studies have reinforced the theory
that sustains that the Weibull distribution is one
of the best sea clutter models (Ishii, Sayama,
Mizutani, 2011; Sayama, Ishii, 2013). Likewise,
it has been noted that the average wave height
inuences the selection of the shape parameter
of the distribution. Additionally, the shape
parameter varies when using S-band radars
instead of X-band. For S-band radars, the β shape
Figure 1. Block Diagram of a CA-CFAR Detector.

MACHADO: Estimation of the Relation between Weibull ... 21
parameter from clutter samples, the systems
requires having knowledge about its relation with
the CA-CFAR K scale factor, in order to control
the variation of the detector’s behavior. Therefore,
the conceived scheme is useless if a table, with the
possible occurrences of β and its corresponding
K values, is not available for several false alarm
probabilities. The creation of such a table is the
objective of the author of this paper.
The CA-CFAR scheme is preferred for the
analysis over the SO-CFAR, GO-CFAR and OS-
CFAR alternatives because it’s the internationally
recognized reference model for comparing new
implementations. A great amount of recent
articles support this statement (Caso, De Nardis,
2013; Qin, Gong, 2013; Kumar, 2013; Ranjan,
Krishna Moorthy, 2013; Mashade, 2013).
The current project is intended to solve
problems founded in previous researches (Machado,
Bueno, 2012; Machado, García, 2014; Machado,
García Delgado, B., 2014). where sea clutter and
CFAR detectors were studied independently but
both approaches were not associated. Neither was
conducted a thorough analysis on the behavior of
echoes received from sea surface.
Clearly establishing the purpose of the
investigation, it should be noted that critical
points can’t be determined for all false alarm
probabilities. This would consume a huge amount
of time and efforts. Instead, the author searches to
nd the combination of values of the CA-CFAR
K and the β parameter for which the Pfs are equal
parameter settles around 4,5 and for X-band
around 2,5. Besides, if the inuence of other
not fully specied climatic factors is taken into
account, a variation interval which goes from
1,75 to 6,25 may be assumed for the β parameter.
The formula for the Weibull Probability Density
Function (PDF) is given below (O’Connor, 2011).
and draws for several combinations of Weibull
parameters are shown in Figure 2.
f (x
|
α,β) = (βxβ-1 / αβ) exp [-(x/a)β] (1)
The previous statements raised questions about
whether the selection of a scale invariant factor
truly allowed maintaining a constant Pf for the
entire operation period of a CFAR detector. Taking
as a priori information that the β parameter varies
in a known range (Ishii, Sayama, Mizutani, 2011),
the ISPJAE radar research group has proved, by
performing a number of experiments in MATLAB,
that a detector that uses a xed scale factor must
operate inefciently in order to ensure a constant
Pf. On the contrary, if the scale factor would vary
according to the value of the β shape parameter
from the Weibull sea distribution, the inefciency
will disappear (Machado y Bacallao, 2014).
Thus, the need for a system capable of
identifying the β Weibull parameter becomes
evident. A project for its creation has already been
conceived (Machado, 2014). However, there is an
essential add-on that has not been considered as a
part of the research. Once identied the Weibull
Figure 2. PDF draws from the Weibull Distribution.

Ingeniería 25 (2): 19-28, ISSN: 1409-2441; 2015. San José, Costa Rica
22
to and. The selection of such Figures is a classic
choice in radar radars issues (Kouemou, 2009;
Skolnik, 2008).
2. MATERIALS AND METHODS
To achieve the aimed objective, the author
worked with the MATE-CFAR (MAtlab Test
Environment for CFAR detectors) testing
environment created in MATLAB/Simulink 2011
by himself. The software allows the simulation of
clutter, targets and CFAR detectors by adjusting
the simulation variables in a quickly and
intuitively way (Machado, Bacallao, 2014). The
multiple blocks that compose MATE-CFAR are
shown in Figure 3.
The test environment has a total of 12
congurable parameters that were arranged, in
the current research, in a way that Weibull clutter
was generated, with no targets present. Besides,
samples were processed by a 64 cells CA-CFAR
architecture with no guard cells. The previous
conguration was maintained all simulation long.
In contrast, the value of the K adjustment
factor and the α and β Weibull clutter parameters
Figure 3. Structure of the MATE-CFAR Test Environment
Table 1. α and β Values Selected for the Simulation.
Beta 1,75 2 2,25 2,5 2,75 3
Alfa 1,1228 1,1284 1,129 1,127 1,1237 1,1198
Beta 3,25 3,5 3,75 4 4,25 4,5
Alfa 1,1156 1,1114 1,1072 1,1033 1,0994 1,0958
4,75 5 5,25 5,5 5,75 6 6,25
1,0924 1,0891 1,086 1,0831 1,0804 1,0779 1,0755

MACHADO: Estimation of the Relation between Weibull ... 23
Figure 4. Two possible thresholds obtained by selecting different K values.
were changed until the False Alarm Probabilities
of 10-2, 10-3 and 10-4 were found. The procedure
was performed as follows. Firstly, 19 values of
β were chosen from the range between 1,75 and
6,25 including the edges, so there was a difference
of 0,25 units between consecutives values. The α
Weibull parameter was selected for making the
average of the samples equal to one, by using the
known Weibull mean formula (O’Connor, 2011).
given bellow in Equation 2. In addition, Table 1
shows the and values selected for the simulation.
mean = αΓ (1+1/β) (2)
Then, using Table 1 as a reference, the rst (α,
β) pair was placed in MATE-CFAR, and the K was
set to a small value such as K = 1,5. Afterwards,
one million Weibull clutter samples were processed
under the established conditions emulating a real
detector. After nishing, the author calculated the
Pf according to the following equation:
Pf = false positives / total amount of samples (3)
Where false positives are those clutter
samples mistakenly identied by the CA-CFAR
detector as targets. Generally, if the simulation
starts by assuming a small K value, the obtained
Pf is high. Then, as K is increased, in later
executions, the Pf will decrease.
The described behavior can be understood if
Figure 4 is examined. The reader will notice there
are two thresholds resulting from the selection of
two different K gures. The lower threshold is
associated with a high Pf because it often forces
to identify clutter samples as targets. Conversely,
the higher threshold, corresponding to a higher K,
provides a smaller Pf since it’s rare to nd a clutter
value exceeding the established level. However,
when a too high threshold is chosen, the detector
starts to miss targets that actually exist; so the
excessive elevation of K is not recommended
because it decreases the Probability of Detection.
Consequently, the experiment’s goal was to nd
the exact value for which the scale factor (K)
guaranteed a Pf of 10-2, 10-3 and 10-4 for each
pair of Weibull parameters with the best possible
value of probability of detection.
As the reader may deduce, after obtaining
the Pf for the rst K value, the procedure was
repeated changing the K value until the desired Pf
was found. About 20 executions were necessary
as an average to ensure a good accuracy on the
results extracted from the experiment (less than
0,5 % of error).
After nishing with one (α, β) pair, the next
one was be included in the trials, performing
simulations until all the required Ks were
founded. The procedure was completed when all
pairs were processed. A summary of the algorithm
is described below.

Ingeniería 25 (2): 19-28, ISSN: 1409-2441; 2015. San José, Costa Rica
24
Repeat 19 times
In each Repetition: Select a (α,
β) pair
Repeat until the Pf is found
(approximately 20 repetitions)
In each repetition: Modify
K looking for the desired Pf
End of Repetitions
End of Repetitions
3. RESULTS
After performing the procedure described in
the preceding section, the data shown in Figure
5 was obtained. The X axis provides the CA-
CFAR K value and the Y axis the calculated Pf.
Each draw corresponds to one β Weibull shape
parameter. Thus, 19 draws are visible covering the
range from 1,75 to 6,25. The draw to the extreme
right shows the result for the smaller β (β = 1,75).
Draws appearing to the left plot lines related with
the gradual increase of the shape parameter.
As it may be observed in Figure 5, the
slope was always negative in what constitutes
an expected result if the effect of raising the
adjustment factor is taken into account. Note that
by increasing K, the calculated threshold gets
higher, so fewer targets will be classied as clutter
obtaining as a consequence a smaller false alarm
probability.
Figure 6 shows one of the draws from Figure
5 (the one highlighted) a little more in detail. In
this case, besides the solid line, some dots in a
cross shape were plotted. Each dot represents a
measurement of Pf obtained by performing an
experiment with one million samples using the
procedure described in the previous section.
As it can be seen, most of the crosses are
distributed in three concentration areas. Indeed,
the groups of measurements are around the Pf
values of 10-2, 10-3 and 10-4. The reader may
understand then that the search for K was not
made in a uniform manner but it was performed
efciently by getting closer and closer to the
points of interest. However, outliers appeared as
a result of the lack of information of the initial
executions.
The result of the measurements was the
extraction of the desired critical point where the
required Pf was obtained. Table 2 relates each
occurrence of the β Weibull parameter with the
better CA-CFAR K. The second column displays
the values of the shape parameter used in the
simulation, while the rest to the right show the
values of the best possible K for each tabulated
Pf. Note that each of the amounts shown was
extracted from Figure 5.
In order to maximize the practical application
of the values shown in Table 2, the author
recommends focusing future research efforts in
the development of a mathematical expression
that will allow the generalization of the results.
With this expression, it will be possible to
estimate the K value necessary to maintain the
desired Pf for any β in the range from 1,75 to
6,25, and not only for the values tabulated in
Table 2.
Figure 5. Relation between the K Adjustment Parameter and the False Alarm Probability for several.

MACHADO: Estimation of the Relation between Weibull ... 25
Figure 6. Successive Executions for the precise measuring of the adjustment parameter.
Table 2. Best k Scale Factor for Several Clutter Congurations.
N, β Weibull
Shape Par K for Pf = 10-2 K for Pf = 10-3 K for Pf = 4
1 1,75 2,74 3,50 4,16
2 2,00 2,47 3,05 3,55
3 2,25 2,26 2,73 3,13
4 2,50 2,11 2,50 2,82
5 2,75 1,99 2,32 2,82
6 3,00 1,89 2,18 2,41
7 3,25 1,81 2,06 2,27
8 3,50 1,74 1,96 2,15
9 3,75 1,68 1,88 1,99
10 4,00 1,63 1,82 1,99
4. DISCUSSION
The founded relation between the K and the
Weibull shape parameter achieved an important
step toward the creation of an improved CA-CFAR
detector, which will vary its adjustment factor
according to the conditions of the environment. The
new detector would really guarantee that a constant
Pf is maintained when facing variable operating
conditions. However, the conception of the scheme
contains other complexities not addressed in the
current research such as the number of samples
to be taken from real clutter measurement for
guarantying stability in the β Weibull parameter.
Making a preliminary analysis of the observed
data, it was convenient to plot the difference

Ingeniería 25 (2): 19-28, ISSN: 1409-2441; 2015. San José, Costa Rica
26
between each of the values of table II, that is, the
subtraction of each value with the subsequent.
So, Figure 7 shows the derivative of the K value
for each of the addressed Pfs. Note that this time
the X axis shows the 18 values resulting from the
subtraction of the 19 K measurements.
As it can be seen, plots were decreasing and
followed a non-linear structure. This indicates that
it’s highly unlikely that small order polynomial
approximations will t the data correctly.
However, the derivative seemed to stabilize after
the 10th measurement which corresponds to β =
4 (see Table 2). Consequently, an estimator of the
measurements could be divided into two parts if
good results are not reached by using a single one.
The area with greater linearity is also visible
in Figure 8 where data is shown before performing
the derivation. Note that this Figure illustrates the
inuence of β on K.
From Figure 7 and 8, it was visible that as
β increases, its effect on the modication of K
becomes reduced. The fact is remarkable when
trying to establish expressions to generalize the
obtained data. Selected expressions must have a
smooth behavior beyond the β = 6,25 limit, as it
was suggested by the shapes of the draws. Besides,
the expressions must have a tendency to converge
for the higher βs and to divergence for smaller βs.
On another note, observe that the shapes of the
draws in Figure 8 were very similar. This suggests that
Figure 7. Derivative of the K founded values.
Figure 8. β inuence on K for several False Alarm Probabilities.

MACHADO: Estimation of the Relation between Weibull ... 27
a single expression could reproduce the measurements
of the three Pfs. An approach which could prove to
be satisfactory is to select a base function which will
be added to an auxiliary function that will change
according to the selected Pf. The author did not tried
to nd expression to generalize the results because
he thought that higher amounts of samples should
be used to calculate better false alarm probabilities
before offering mathematical expressions.
5. CONCLUSIONS
The best values for the K scale factor where
found for a CA-CFAR detector and false alarm
probabilities of 10-2, 10-3 and 10-4, considering
several states of Weibull distributed clutter. The
founded gures guarantee that a CA-CFAR
scheme will operate maintaining its theoretic
false alarm probability while the probability
of detection is maximized, even when facing
variations of the β clutter parameter from 1,75 to
6,25. In addition, the study showed that for the
upper values of β, the clutter inuence decreases
over the correction of the K factor. The current
investigation concluded an important piece of a
system designed to operate as an improved CA-
CFAR detector that will adapt to variations in the
environment.
6. RECOMMENDATIONS
The author recommends nding a
mathematical expression through which the
founded results may be reproduced. Likewise,
it will be necessary to increase the amount of
samples involved in the experiment in order to
calculate higher order false alarm probabilities.
The reproduction of the study for the classic
OS-CFAR detector is also recommended. This
would begin creating a general methodology for the
design of detectors adapted to clutter conditions.
It will be also necessary, for the further
development of the proposed scheme, to estimate the
convergence of the Weibull clutter to its statistical
mean as more samples are taken into consideration.
This will allow establishing the necessary quantity
of samples to gather for performing the proposed K
correction. Note that the amount of samples should
be inferior to one million.
REFERENCES
Antipov, I. (1998). Simulation of Sea Clutter
Returns. Australia: DSTO Electronic and
Surveillance Research Laboratory.
Bacallao Vidal, J. C. (2003). Un modelo Teórico
de la Técnica DRACEC. Metodología del
Proceso de Adaptación al Fondo: Instituto
Técnico Militar “José Martí”.
Barton, D.K., Leonov, S. A.(1998). Radar
Technology Encyclopedia (Electronic
Edition) London: Artech House.
Caso, G. , De Nardis, L.(2013). Cooperative
Spectrum Sensing based on Majority decision
under CFAR and CDR constraints. Workshop
on Cognitive Radio Medium Access Control
and Network solutions. IEEE 24th International
Symposium on Personal, Indoor and Mobile
Radio communications. London, UK.
de Figueiredo, F. (2013). AP. LTE Random Access
Detection Based on a CA-CFAR Strategy.
Sao Pablo, Brazil: Convergent Networks
Department, Research and Development
Center campinas.
Farina, A., Studer, F. A. (1986) A Review of CFAR
Detection Techniques in Radar Systems.
Horizon House-Microwave Inc.
Haykin, S., Bakker, R., Currie, B. W. (2002)
Uncovering Nonlinear Dynamics—The Case
Study of Sea Clutter. Proceedings of the
IEEE, 90(5).
Ishii, S., Sayama, S., Mizutani, K. (2011).
Effect of Changes in Sea-Surface State on
Statistical characteristics of Sea Clutter with
X-band Radar. Wireless Engineering and
Technology, 2(3).
Jian-bo Hu, J. T., Wen-wen G. (2009). Modeling
sea clutter as a nonstationary and nonextensive
random process. IEEE Conference on Radar
Issues.
Jian-bo Hu, J. T., Wen-wen, G. (2009) A New
Way to Model Nonstationary Sea Clutter.
IEEE Signal Processing Letters, 16(2).
Kouemou, G.(2009). Radar Technology: I-Tech.
London: Artech House.

Ingeniería 25 (2): 19-28, ISSN: 1409-2441; 2015. San José, Costa Rica
28
Kumar Yadav, A., Kant, L. (2013). Moving
Target Detection using VI-CFAR Algorithm
on MATLAB Platform. International Journal
of Advanced Research in Computer Science
and Software Engineering.
Machado Fernández, J. R., Bueno González,
A.(2012) Clasicación del Clutter Marino
utilizando Redes Neuronales Articiales.
Habana: Instituto Superior Politécnico José
Antonio Echeverría (ISPJAE).
Machado Fernández, J. R. (2014). Empleo de
las Redes Neuronales Articiales en la
estimación de los Parámetros del Clutter
Marino distribuido Weibull. Congreso
Internacional de Telecomunicaciones y
Telemática. Convención Cientíca de
Ingeniería y Arquitectura; Palacio de
Convenciones de la Habana, Cuba.
Machado Fernández, J. R., Bacallao Vidal, J. C.
(2014). MATE-CFAR: ambiente de Pruebas
para detectores CFAR en Matlab. Telem@
tica, Revista Digital de las Tecnologías de
la Información y las Comunicaciones, 13
(3):86-98.
Machado Gil, A., García Delgado, B. (2014).
Reconocimiento de Parámetros Asociados a
distribuciones del Clutter Marino con Redes
Neuronales Articiales. La Habana, Cuba:
Instituto Superior Politécnico José Antonio
Echeverría (ISPJAE).
Magaz, B., Belouchrani, A. (2011). Automatic
Threshold Selection in OS-CFAR Radar
detection using Information Theoretic
Criteria. Progress In Electromagnetics
Research,p. 157-75.
Mashade, M. B. (2013). Performance Analysis
of the Modied Versions of CFAR Detectors
in Multiple-Target and Nonuniform Clutter.
Radioelectronics and Communications
Systems, 56(8).
Nagle, D. T. (1991). Analysis of Robust Order
Statistic CFAR Detectors. Illiniois: Illinois
Institute of Technology.
O’Connor, A. N. (2011). Probability Distributions
Used in Reliability Engineering. Maryland:
University of Maryland.
Oyedokun, O. (2012). Sea Clutter Simulation.
Cape Town: University of Cape Town.
Qin, Y., Gong, H. (2013). A New CFAR Detector
based in Automatic Censoring Cell Averaging
and Cell Averaging. TELKOMNIK,11(6):
3298 - 303.
Ranjan, D. S., Krishna Moorthy, H. (2013).
Development of Adaptive Algorithm for
CFAR in non-homogenous environment.
International Journal of Engineering and
Innovative Technology (IJEIT), 3(1).
Rohling, H.(1983). Radar CFAR Thresholding in
Clutter and Multiple Target Situations. IEEE
Transactions on Aerospace and Electronic
Systems,19(4).
Sayama, S., Ishii, S. (2013). Suppression of
Log- Normal Distributed Weather Clutter
Observed by an S- and Radar. Wireless
Engineering and Technology, 4(3).
Skolnik, M. I.(2008). Radar Handbook. New
York: McGraw-Hill.
Takahashi, S. (2010) A CFAR Circuit of Detecting
Spatially Correlated Target for Automotive
UWB Radars. Hiroshima City University:
Faculty of Information Sciences.
Totir, F., Rador, E., Anton, L. (2008) Advanced
Sea Clutter Models and their Usefulness for
Target Detection. MTA Review.
Van Cao, T-T (2012). Non-homogeneity Detection
in CFAR Reference Windows using the
Meanto- Mean Ration Test. Australia: DSTO
Defence Science and Technology Organisation.
Wawruch, R. (2013). Experimental evaluation
of the Constant False Alarm Rate (CFAR)
algorithms used in maritime FM-CW radars.
Scientic Journals, Maritime University of
Szczecin, 36.
Weingberg, G. V. (2004) Estimation of False
Alarm Probabilities in Cell Averaging
Constant False Alarm Rate Detectors via
Monte Carlo Methods. Australia: DSTO
Systems Sciences Laboratory.
ABOUT THE AUTHOR
José Raúl Machado Fernández received his
engineering degree in Telecommunications and
Electronics from the Instituto Superior Politécnico
José Antonio Echeverría (ISPJAE). Since 2013, he
has been receiving a Master in Telecommunications
course at the same institution. He’s interested in
researches related to pattern recognition, articial
intelligence, sea clutter and radar detection schemes.