1.

Introduction

Radar Detection in Weibull Clutter

D.C. SCHLEHER, Member, IEEE AIL-Cutler-Hammer Deer Park, NY 11729

Abstract
Radar detection in Weibull clutter is examined from a statistical detection viewpoint. Weibull clutter parameters are determined and related to measured values of land and sea clutter. Optimum performance in Weibull clutter is determined, and practical receivers that approach this performance are identified. Receiver performance in Rayleigh, log-normal, and Weibull clutter is evaluated and compared.

In this paper the detection of a steady target in Weibull clutter is examined from a statistical detection viewpoint. Optimum performance is determined and receivers that approach the optimum are identified. The approach followed is similar to that used in establishing corresponding results for detection performance in a log-normal clutter background [1]. The motivation for considering a Weibull clutter model involves a consideration of how well the model matches the anticipated clutter statistics. Available data do not permit a definitive answer but general guidelines can be deduced. Most previous experimental data have resulted in clutter being modeled using either a Rayleigh or a log-normal distribution. The log-normal model is most frequently used when the radar sees land clutter [2] or sea clutter [3] at low grazing angles (q < 5 degrees), while the Rayleigh model is used for clutter whose amplitude probability distribution encompasses a limited dynamic range. In general, the real value of the Rayleigh and log-normal clutter models is that they apparently provide limiting distributions for most experimentally and theoretically determined clutter distributions. The Rayleigh model generally tends to underestimate the dynamic range of the real clutter distribution while the log-normal model tends to overestimate the dynamic range. The Weibull clutter model [4] offers the potential to accurately represent the real clutter distribution over a much wider range of conditions than either the log-normal or Rayleigh model. By appropriately adjusting its parameters, the Weibull distribution can be made to approach either the Rayleigh (a member of the Weibull family) or log-normal distribution. To illustrate the capability of the Weibull clutter model, sea clutter data [5] taken with a relatively high-resolution airborne radar were plotted on log-Weibull probability paper [4] for grazing angles (0) between 1 and 30 degrees. The results are shown in Fig. 1 where the median has been normalized to a common value for all curves. The close fit provided by the straight lines indicates that the data can be closely modeled by Weibull statistics. In addition, the decreasing skewness with increasing grazing angle indicates that the distribution is approaching a Rayleigh distribution
=

Manuscript received February 5, 1976. Copyright 1976 by The Institute of Electrical and Electronics Engineers, Inc.

From the detection standpoint, it can be said that the log-normal distribution represents a worst case distribution, the Rayleigh the best case, and the Weibull an intermediate model that may more accurately represent the real detection performance in clutter. While the detection performance is reasonably well established for log-normal and Rayleigh clutter, very little is known about the detection performance in Weibull clutter. An approximate analysis has been made [6] but the data available from this report have very limited scope. As in [ I ], this paper uses a Marcum-Swerling analysis to determnine the detection performance in Weibull clutter. This analysis is applicable where scan-to-scan integration
VOL AES-12, NO. 6 NOVEMBER 1976

2.0).

'736

IEEE TRANSACTIONS ON AEROSPACE AND ELECTRONIC SYSTEMS

.9999999 .9999995 La-I-L-L-Ll .9999990 F .9999950 .9999900 .9999500

6 '=

.-

2' O' 11 III

I 1 6 =' O.'5 .::i_L11

.9999000

v:
C V)

.9995000 .9990000 .9980000 .9950000 . 9900000

f. i 1 i z 11 lATtHfIf 1 1 1 1 1 1 1 tl, I 1

.9500000
30
50

><~~ 9000000 .

.98000000

'0
,
Cn

. 7000000 .6000000 . 5000000 . 4000000 .3000000 .2000000

A~~~~~~t
I

t:
G
G

.1000000 -. 0500000

.6000000 .5000000 .0050000

ZATION

lC

4-

.0000500
-3 0
3 6 9 12 15 CLUTTER AMPLITUDE/MEDIAN AMPLITUDE (dB)

.0001000

18

21

Fig. 1. Sea clutter distribution for several grazing angles.

.0000050 .0000010 .0000005 .0000001

-4-48

-36

-24

-12

0 12 24 36 SIGNAL/MEDIAN CLUTTER (dB)

48

60

72

Fig. 2. Weibull distribution function.

techniques are employed to obtain multiple independent looks at the clutter. Optimum performance is determined via a Chemoff bounding technique and practical receivers are identified that approach the optimum performance. As with log-normal clutter, it is determined that nonlinear receivers that tend to suppress, and hence normalize, high amplitude clutter returns are effective in Weibull clutter.
I. Weibull Clutter Characteristics

a where AC = A/Am, = 1/2, and Am is the median of the clutter cross-sectional area. The mean of the clutter cross section is given by

A=A/ln2=VU/ln2. m m

(3)

In a Weibull' clutter model the amplitude probability density function of the voltage out of an envelope detector, normalized to the median Vm is given by

fR(R) = a-Iln 2Ra- 1 exp [-ln 2-R],

R >O

0,

otherwise

(1)

If a = 2 (B = 1) and Vm = Am = 1, then (1) converts to the familiar Rayleigh amplitude probability density function of envelope detected noise and (2) converts to the exponential distribution of the corss-sectional area of Rayleigh clutter. The Weibull distribution function is shown in Fig. 2 on log-probability paper for various values of the parameter o. Table I gives values of a from [4] for various types of land clutter while sea clutter data are obtained by fitting Weibull distributions to the data from [3] and [5].
111. Detector Performance

where R = V/Vm and a is a parameter that relates to the skewness of the distribution. This is related to the ampliA. Linear Receiver Performance tude probability density function of the clutter cross section A by the transformation A = V2 resulting in The linear receiver (linear envelope detector with linear integrator) is known to approximate the optimum receiver in Rayleigh distributed clutter. Since many radars are deexp [-ln 2A], A >0 fA (Ac)= ,Bln signed with linear receivers, it is important to determine = 0, otherwise (2) how these receivers perform in Weibull clutter. The performance of the linear receiver is analyzed using ' Refer to Appendix A for characteristics of the Weibull distribua Marcum-Swerling [7] analysis where the Weibull distrition. bution of (1) and the Weibull-Rician distribution of

2.AO-l

SCHLEHER: RADAR DETECTION IN WEIBULL CLUTTER

737

TABLEI
Weibull Clutter Parameters for Distributed Clutter

Terrain/Sea State
Rocky mountains Wooded hills Forest Cultivated land
Sea state Sea state 3

Frequency
S

Beamwidth

L
x

(degrees) 1.5 1.7

0 (degrees)
z0.5 0.7 0.75.0

Pulsewidth

Cus)

3 0.17 0.17

0.512 0.626 0.506-0.531 0.606-2.0

0).5
K11

447
1.0-

0.02 0.1

1.452

1.160-1.783

30.0

(B )2 are used under the alternative hypotheses. The thresh- fi(V5 IHI)= 2 [a In 2 exp (2 V )/2ir] olds, as a function of Pfa (probability of a false alarm), are determined for n 1, 3, 10, and 30 (independent samples) (exp {- ln 2[exp (2V) -2A exp (V ) cos using a chlaracteristic function approach in conjunction with the FFT. The Pd (probability of detection) is determined +A2]112}/[exp (2V ) 2A exp (Vs) cos 0 in a similar manner using the FFT in conjunction with Gaus+A 2 1(2 -a)/2 ) d sian quadrature to evaluate the sum distributions. (5) The results of the linear receiver performance evaluation for aPfa - 10-6 are shown in Fig. 3 for o- 0.6, 0.8, 1.2, where the variables V, and V, extend from-co to +oo. and 2.0 and n = 1, 3, 1O, and 30 (independent samples). Fig. 4 shows the performance of the logarithmic receiver Table II gives the additional signal-to-median clutter ratio in Weibull clutter (a 0.5, 0.6, 0.8, 1.2, 2.0) for a Pfa = 1o-6 required for the performance (Pd 0.9' Pfa 10 6) in and n = 1, 3, 10, and 30 (independent samples). A comWeibull clutter (a = 0.6, 0.8, 1.2) to equal that in Rayleigh parison of the signal-to-median clutter differential for equal 10=6) in Weibull probabilities of detection (Pd 0.9,Pfa clutter (o= 2.0). and Rayleigh clutter is given in Table III. The performance Examination of Table 11 shows that performance degradation is severe for the higher skewed Weibull clutter distri- for a single pulse is almost identical to the linear receiver; butions. However, comparison with the linear receiver per- however, as the number of pulses increases to 30 the performance increases substantially. formance3 obtained in log-normal clutter [11 shows that clutter. less degradation occurs in Weibull C. Binary Integrator Performance B. Logarithmic Receiver Performance The binary integrator, depicted in Fig. 5, is a digital detection process that utilizes a double threshold. The enA logarithmic detector has an output voltage whose amvelope detected signal is compared to the first threshold. plitude is proportional to the logarithm of the input enThe number of threshold crossings in n repetitions of signal velope. This type of detector has found extensive use that is counted. When more than m crossings take place in the in Rayleigh distributed clutter since it has the property n trials, a target is assumed to be present. it provides a constant false alarm rate (CFAR) [8]. It was determined by Schwartz [9] that the second threshwith linear The logarithmic receiver (logarithmic detector old for a Rayleigh distributed envelope detected background manner to the linear reintegrator) is analyzed in a similar ceiver, except that the probability density under the clutter should be set at approximately 1.5V/n for optimum detection. This result is true for a broad range of false alarm hypothesis is given by probabilities (10-5 to 10-10) and probabilities of detection J(Vc IHo) a ln 2 exp (aV ) exp [- In 2-exp (aVe)] (4) (0.5 to 0.9). The performance of this suboptimum detector in Rayleigh clutter is within 1.5 dB of the optimum performance. and under the signal-in-clutter hypothesis by The performance of the binary integrator in Weibull clutter can be determined using the following procedure. First, using the Weibull distribution function (A2), given in Appendix A, find the threshold setting n, for all combinations of 2 See Appendix B. m and n of interest for a particular Pfa. Secondly, determine 'In [1] the curves and tables using the Rayleigh distribution for each m and n of interest the cumulative probability out should be adjusted by translating the Rayleigh curves 1.54 dB to the of the envelope detector that provides the desired Pd. Then, left and adding 1.54 dB to the tables where a comparison is made using a graphical plot of the single sample Weibull-Rician against Rayleigh clutter. This corrects an error made in adjusting from the mean to the median of the Rayleigh distribution. distribution function, determine the signal value that
=
-

738

IEEE TRANSACTIONS ON AEROSPACE AND ELECTRONIC SYSTEMS

NOVEMBER 1976

99. 99.
99.
'9

n= 1

PFA= 0o-6
1.2
a

RAYLEIGH
a

= 2.0

0.8

a = 0.6

99.8 99.S 99 98
95
i

99q9q

1.1

n= 3

PFA= io-6
aQ= 1.2
a= 0.8

RAYLEIGH a= 2,0

Q= 0.6

90

8
. ui

80
w

C)

PQ

7 6 5 4.0 3

70
50

Lo:- 60
=
cL

40
30

0-2

20

10

10

15

20

35 30 40 SIGNAL/MEDIAN CLUTTER (dB)

45

50

55

I~

12

SIGNAL/MEDIAN CLUTTER (dB)

16

20

24

28

l-

32

36

40

(A)
99. 99.
99. Q.
18

(B)
qq a ni.:rj

n= 10

PFA= io-6

RAYLEIGH
a

Sg,
09

2.0

a= 1.2

a-

0.8

a= 0.6

99.8 99.5 99 98
95
2

PFA = io-6
-

30
1.2
a =

RAYLE IGHI
a=

2.0l

0.8

a=0.6

90

w81
61 _51 <:4
3

80

,, 70
-60

cL

co330

50 40 X 30
-

20 10
5

18 15 21 12 SIGNAL/MEDIAN CLUTTER (dB)

I1

24

27

30

IE

-3

12

15

SIGNAUMEDIAN CLUTTER (dB)

18

21

24

(C)

(D) Fig. 3. Probability of detection, linear envelope detector.

TABLE 11 Linear Receiver Performance in Weibull Clutter Compared with Rayleigh Clutter (Additional Signal-to-Median Clutter Ratio in dB
forPd=O.9,Pfa =106)

1.2

0.8
1 4.4

0.6 28.3 24.7 20.0 16.0

1 3 10 30

7.5 6.1
4.4 3.9

1 7.7

1 1.5 9.3

interest and the second threshold setting for the best performance identified. The optimum values of m so found are given in Table IV and compared against the optimum setting in Rayleigh clutter (mr = 1.5 -4n). It is noted that the second threshold settings found for Weibull clutter (ci = 1.2) are almost identical to those found for log-normal clutter (p = 1 dB) [1] . The performance of the binary integrator, given in Section V, compares favorably with the optimum performance in Weibull clutter.
D. Median Detector Performance

corresponds to the cumulative probability for the determined threshold setting n,. The median detector, described in [1 I, can be implemented Using the preceding procedure, detection curves for the by employing the binary integrator with the second threshold binary integrator were formed for ox = 1.2 over the region of set at m = (n 1)/2. This effectively accomplishes the
-

SCHLEHER: RADAR DETECTION IN WEIBULL CLUTTER

739

99.9 '9,8 99,5 _

n= 1

99.8

99g,9 F
99
98 95

n= 3

RAYLEIGH 2

99 _

a= 2

a = 1,2

a= 8/

a = o. 6

99,5

RAYLEI GH
cz

PFA

1o-6
a
=

=o.si

909_
L 70

/ I

I
L
-

21

a=

L.2

a= 0.8

0.6

a= M

;m90
= 80
70

/1

50_ =Z 40 CD

30

20
99 8G

/
I

60 50

, 40 ,I 30

20

i
4

I
I

10

180
0
-

15
I

or 351,rrl 40 30 25' SIGNAL/MEDIAN CLUTTER (dB) on 20

'4I., 45

llr

cn 50

55

I __

12

SIGNAL/MEDIAN

16

20

24 CLUTTER

(dB)

28

32

36

40

-L-

99.98 F-99.8
99,i5 99 98
-

(A)
n= l

PFk = 10-6

99. 9 99,8
a

(B)
n= 30

PFA 10-6
/
I

H R,(AYLE IN

:1-

.8/

L0 I a= 5

99.5 99 98

RAYLEIGH

0.5

95

95 90
' 80 i 70 . 60 ; 50 40
/~~~~~~~~~~~~~~~~~~
30

goIL
'600 -50 t

20

100

10

II

21 18 15 SIGNAL/MEDIAN CLUTTEI (dB)

24

27

30

2 i

10 8 lO 8 6 SIGNAL/MEDIAN CLUTTER (dB)

12

14

1 16

18

(C)

(D)

Fig. 4. Probability of detection, log envelope detector.

TABLE Ill Logarithmic Receiver Performance in Weibull Clutter Compared with Rayleigh Clutter (Additional Signal-to-Median Clutter Ratio in dB forPd = 0.9,Pfa = 1-6)
n

1.2

0.8

1 3 10 30

7.5 6.0

3.6 2.4

17.8 12.6 7.8 3.8

0.6 28.3 20.1 12.2 6.1

0.5

37 26.0 15.9 7.9

examined. This might be expected, since all the information contained in the input distribution is not effectively utilized. The robustness of the median detector can be estimated by considering its performance over a range of log-normal to Weibull to Rayleigh clutter. This comparison is given in Table V for an integration of 10 pulses with a fa 10-6 and Pd values of 0.5 and 0.9. Examination of Table V shows that the performance is reasonably robust for high probabilities of detection when the small number of samples (ten) is considered.
=

IV. Optimum Performance: Chernoff Bound

of finding the median of the input distribution and then comparing it against a threshold. The median detector can be analyzed using a procedure identical to that used for the binary integrator. Detection curves were formed for at 1.2 and are presented in Section V. The performance is inferior to the other detectors
=

It is of interest to compare the performance of the receivers described in the previous sections with the optimum performance possible in Weibull clutter. In order to accomplish this comparison it is necessary to determine the optimum performance. This can be accomplished using a

740

IEEE TRANSACTIONS ON AEROSPACE AND ELECTRONIC SYSTEMS

NOVEMBER 1976

IF SIGNAL n PULSES

ENV DET

R> m

\<i

THRESHOLD

S~~~~~~m
H0

99.9 99.8 99.5 99 98

PFA = io-6
a= 1.2

n =

LIN DET MEDIAN DET

Fig. 5. Binary integrator.

TABLE IV Optimum Value of Second Threshold, Weibull Clutter (a = 1.2)

mopt
23
2 7

30
TABLE V

3 10

mr 2
4

z 80 , 70 60 n 50 X 40 t 30
-

20
10
5

II

18 16 10 14 12 SIGNAL/MEDIAN CLUTTER (dB)

20

22

24

Signal-to-Median Clutter in dB for Median Detector (Pfa n = 10)

10-6,
qq.I_q J' j

(A)
99.8 99.5
99
98

Pd
0.9 0.5

Rayleigh
7.5 5.7

Weibull (i=1.2)
11.6 10.9

Log-Normal
(a = 0.693) 11.3 10.1

PFA= io-6
a

1.2

n= 10

95

2
<D

90
80 v

w_

method, described in [1], that is derived from the Chernoff bounding technique [10]. It is shown in [1] that the semi-invariant, ,u(s), of the likelihood ratio for n independent samples is given by

>-

70 60 _: 50 A 40
30 20

p(s) = n In f

00

[(IHf['(IHo)

1-s

dv

(6)

10

where n is the number of independent samples, fv(v ) is the probability density function under the hypotheses Ho or H1, and s a variable between 0 and 1. It was further shown [1 ] that the probability of false alarm for the optimum receiver can be expressed in terms of,(s) and its derivatives as

12

14

16

8 14 10 6 12 SIGNAL/MEDIAN CLUTTER (dB)

16

18

20

YY Y

(B)

Pfa =

erfc [a(s)] * exp [a2 (s) + B(s)]

99.8 99.5 99 98
-_

95
90

-[p(s)16[A((s)]
+ (1Ia) [1

I exp [B(s)]

* {: erfc a(s). [s pp(s)] 3 exp [at2 (S)]

80
-

>

70 60

(7)

;
0

50 40 30
20

where

a(s) = s[(s)2] 1/2

(8)
(9)

10

An expression for Pm (Pm = 1 -Pd) can be found by substituting s - 1 for s in (7) through (9).
SCHLEHER: RADAR DETECTION IN WEIBULL CLUTTER

3 4 5 2 SIGNALMEDIAN CLUTTER (dB)

(C)

Fig. 6. Probability of detection, Weibull clutter.

741

TABLE VI
Integration Gain in dB for Rayleigh, Weibull, and Log-Normal Clutter

Appendix A Characteristics of Weibull Distributions [111

n Independent Samples 3 10 8.0 4.0 Rayleigh 12.3 6.2 Weibull (ca = 1.2) 18.2 11.8 Log-normal (av- 0.693)

30 11.2 17.6 22.6

The Weibuil probability density function with parameters exp [ -(px)a],

fx (x) = 4p(px)a-1
F (x)= I
-

(Al)

and is zero elsewhere. The distribution function is given by

The procedure of findingPfa andPd is provided in the following discussion. Equation (6) is evaluated by Gaussian quadrature for a particular signal value over a range of s from 0 to 1. Derivatives of u(s) are evaluated using a spline function numerical differentiation technique. A receiver operating curve is generated by plotting Pd versus Pfa for a particular signal, and Marcum-Swerling curves for the optimum receiver are generated from the receiver operating curves. Curves are presented in Section V that allow a comparison between practical and optimum performance in Weibull clutter and that also allow a comparison of the optimum performance in Rayleigh, Weibull, and log-normal [11 clutter.

exp

[-(px)O],

x > O.

(A2)

The rth moment is given by

r [(rla) +I ] /pr.

(A3)

The median is given by

(In 2)1/'p.

(A4)

The distribution function representing the envelope detected Weibull clutter can be solved for x = V(xm = Vm) as follows:
V=

Vm ( {ln [ I /(I - Fv(v))]}Iln 2)1/a.

(AS)

Letting VdB
V. Receiver Comparison
VdB
=

20 log V, then
+

VMdB

(3.184/a)

(20/a) log ln { 1/ [ I

Fv(v)]

}.

Curves showing the performance (Pfa = 106) of the linear receiver, logarithmic receiver, binary integrator, and median detector in Weibull clutter (a 1.2) are given in Fig. 6 for n equal to 3, 10, and 30 (independent samples). Also shown on these curves is the optimum performance in Weibull clutter (a = 1.2) and Rayleigh clutter. The performance of the binary integrator is the best of the receivers considered, approaching the optimum bound, while the performance of the median detector is the poorest. The performance of the logarithmic receiver is almost as good as the binary integrator and in many cases may be easier to implement than the binary integrator. Both the binary integrator and logarithmic receiver were also identified as being good receivers in log-normal clutter [1] and represent practical nonlinear receivers that provide good performance in clutter distributions that have long tails. Since both these receivers are almost optimum in Rayleigh clutter, they represent receivers that perform well over a wide range of clutter distributions. It is of interest to compare the integration gains available from an optimum detector in Rayleigh, Weibull (a 1.2), and log-normal (u, 0.693) clutter. Table VI shows the integration gain for a Pd of 0.9 and aPfa of 10-6. It is seen that for Weibull and log-normal clutter the situation exists whereby the integration gain is greater for noncoherent integration than it would be for coherent integration of a Gaussian distribution (1 0 log n).
=

(A6)
Using (A6) the Weibull distribution plots as a straight line log-Weibull paper. The straight line is easily determined by plotting the 0.50 probability point as
on
V
=

dB

VmdB

(A7)

and the 0.9999 probability point as

mdB +

22.470/a.

(A8)

The corresponding expression for the 0.9999 probability point of the cross-sectional area is (AdB = 10 logA):
A
dB

= A mdB +

11.235fl3.

(A9)

Conversely the parameters of the Weibull distribution can be easily determined by plotting the distribution on logWeibull paper and using the 0.5 probability point to determine Am dB and the 0.9999 probability point to determine A Weibull distributed variable x
[.

can

be generated by
(AlO)

x=xm [in (I /y S)/ln 2] '/'y

where Ys is generated by a random number generator uniformly distributed from 0 to 1.

742

IEEE TRANSACTIONS ON AEROSPACE AND ELECTRONIC SYSTEMS

Appendix B
Weibull-Rician Distribution

By applying the procedure used by Rice to determine the Rician distribution, it can be shown that the normalized (Vm = 1) density function of a steady signal (A) in Weibull clutter is given by:

f (v ) = r [(a ln 2 Vs)/2irl {exp [-In 2(V2 -2A V

cosO

1:

+A2 )(2-a)/21 d. (B1)

I 2

Note that this distribution reduces to the Weibull distribution (1) forA =0. The Weibull-Rician density function is shown in Fig. 7 for a = 1.2 and A equals 0 to 8. Note that the density function is highly concentrated in the vicinity of A.
References

Fig. 7. Weibull-Rician probability function.

[1] D.C. Schleher, "Radar detection in log-normal clutter," pre-

sented at the IEEE Int. Radar Conf., Apr. 1975 [2] M. Warden, "An experimental study of some clutter characteristics," in AGARD Conf Proc. 66-Advanced Radar Systems, May 1970. [3] G. Trunk and S. George, "Detection of targets in non-Gaussian sea clutter," IEEE Tranx Aerosp. Electron. Syst., vol. AES-6, Sept. 1970. [4] R. Boothe, "The Weibull distribution applied to the ground [8] J. Croney, "Clutter on radar displays," Wireless Eng., vol. 33, clutter backscatter coefficient," U.S. Army Missile Command, Apr. 1956. Tech. Rep. RE-TR-69-15, ASTIA Doc. AD691-109, June 1969. [9] M. Schwartz, "A coincidence procedure for signal detection," [5] Johns Hopkins Univ., Silver Spring, Md., "Sea clutter model, IEEE Trans. Inform. Theory, vol. IT-2, Dec. 1956. SCM-1," Nov. 25, 1970 (unpubl. rep.). [6] J. Ekstrom, "The detection of steady targets in Weibull [101 H. Van Trees, Detection, Estimation and Modulation Theory, clutter," presented at the IEE Conf., Radar-Present and pt. I. New York: Wiley, 1968. Future, London, Oct. 23-25, 1973, publ. 105. [11] E, Gumbel, Statistics ofExtremes. New York: Columbia Univ. Press, 1958. [7] M.I. Marcum, "Studies of target detection by pulsed radar," IEEE Trans Informi. Theory, vol. IT-6, Apr. 1960.

D. Curtis Schleher (S'53-A'55-M'60) received the B.E.E. degree in 1954, the M.S.E.E. degree in 1958, and the Ph.D. degree in 1975, all from the Polytechnic Institute of New York, Brooklyn, N.Y. He joined AIL, a division of Cutler-Hammer, Deer Park, N.Y., in 1954 and currently holds the position of Department Head in Radar Systems. At AIL he has been active in the design and development of both airborne and ground-based radar systems for surveillance and tracking applications. His principal interests are detection theory and radar signal processing.
SCHLEHER: RADAR DETECTION IN WEIBULL CLUTTER 743