Ultrawideband (UWB) high-resolution noiseradar for ... · DRDC Ottawa TM 2009-190 . October 2009. Ultrawideband (UWB) high-resolution noise ... an ultra wideband (UWB) noise radar

Ultrawideband (UWB) high-resolution noiseradar for concealed weapon detection: electromagnetic simulation - phase 1 T. Thayaparan and N. Nikolova

Defence R&D Canada – Ottawa Technical Memorandum

DRDC Ottawa TM 2009-190 October 2009

Ultrawideband (UWB) high-resolution noiseradar for concealed weapon detection:electromagnetic simulation - phase 1T. ThayaparanDefence R&D Canada – Ottawa

N. NikolovaMcMaster University

Defence R&D Canada – OttawaTechnical MemorandumDRDC Ottawa TM 2009-190October 2009

Principal Author

Original signed by T. Thayaparan

T. Thayaparan

Approved by

Original signed by Caroline Wilcox

Caroline WilcoxRAST Section

Approved for release by

Original signed by Brian Eatock

Brian EatockChair/Document Review Panel

c° Her Majesty the Queen in Right of Canada as represented by the Minister ofNational Defence, 2009

c° Sa Majesté la Reine (en droit du Canada), telle que représentée par le ministrede la Défense nationale, 2009

AbstractThis report describes the preliminary simulation results of ultra wideband polarizedscattering from various types of concealed weapons carried by a human. Results areshown for both isolated weapons as well as weapons in close proximity to a human,simulating a human carrying the weapon. By selecting the frequency range to coverthe region of resonance, i.e., when the weapon size is a wavelength, resonant scatteringmechanism provides unique spectral features that can be used for detecting theseweapons. Our preliminary simulations show that it is indeed possible to use thewideband polarized backscatter to identify concealed weapons carried by humans.The report consists of five parts which address the following topics: 1) literaturesurvey, 2) capabilities of full-wave electromagnetic models: feasibility and preliminarycase studies, 3) parametric studies in terms of distance, weapon size, mutual positionand polarization, 4) complete system models and their performance in frequency andtime domain analyses, and 5) recommendations for system parameters.

RésuméLe présent rapport décrit les résultats des simulations préliminaires de dispersionpolarisée à bande ultralarge de divers types d’armes dissimulées sur des individus.Les résultats portent sur des armes isolées ainsi que des armes à proximité d’unindividu, qui simule un individu transportant une arme. Grâce à la sélection d’unegamme de fréquences qui permet de couvrir la région de résonance, c. à d. lorsquela taille de l’arme est d’une longueur d’onde, le mécanisme de diffusion résonantefournit des caractéristiques spectrales uniques qui peuvent être utilisées pour détecterces armes. Selon nos simulations préliminaires, il est en effet possible d’utiliser larétrodiffusion polarisée à large bande pour déceler des armes dissimulées sur desindividus. Le rapport comprend cinq parties qui portent sur les sujets suivants :1) recherches bibliographiques ; 2) capacités des modèles électromagnétiques à ondepleine : faisabilité - certaines études de cas préliminaires ; 3) études paramétriquesen fonction de la distance, de la taille des armes, de la position mutuelle et de lapolarisation ; 4) modèles du système complet et leurs performances dans les analysesde la fréquence et du domaine temporel ; 5) recommandations pour les paramètres desystème.

DRDC Ottawa TM 2009-190 i

This page intentionally left blank.

ii DRDC Ottawa TM 2009-190

Executive summary

Ultrawideband (UWB) high-resolution noise radar forconcealed weapon detection:electromagnetic simulation - phase 1

T. Thayaparan, N. Nikolova; DRDC Ottawa TM 2009-190; Defence R&D Canada– Ottawa; October 2009.

Background: Concealed weapons pose a significant threat to military, security, andlaw enforcement personnel. Wideband radar waveforms can excite natural electro-magnetic resonances for better characterization of the size and shape of an object.Resonance techniques have been used for many years to characterize the propertiesof physical, mechanical, and electrical systems. The size, shape, and physical com-position of the system determine the resonant response. An incident wideband radarsignal induces resonant or "standing wave" currents on an object thereby producingradar return. Since the resonances are determined by the physical characteristics ofthe object, the radar return becomes an electromagnetic "fingerprint" that can beused to recognize the object. Prior experimental results under noiseless conditionsusing impulse waveforms confirmed the existence of a unique, aspect independent,radar signature that can be used as a "fingerprint". Additional tests were conductedwith both weapons and nuisances being carried by a human. These tests confirmedthat it is possible to distinguish the armed from the unarmed case, even when theweapon was concealed behind the back.

Illuminating a weapon over a range of frequencies maps its resonant structure andprovides a signature that is uniquely determined by its size, shape, and materialcomposition. In order to induce a resonant response in an object, it is necessaryto illuminate it in the frequency band of the natural resonances. In the frequencydomain, this can be easily accomplished by sweeping the signal over the appropriateband one frequency at a time using a commercial network analyzer. Since concealedweapons may range in size from 10 cm (wavelength corresponding to 3 GHz) to 75cm (wavelength corresponding to 400 MHz), an ultra wideband (UWB) noise radaroperating over the 400-3000 MHz band can be made to excite resonant scattering fromweapons of different sizes. Since these frequencies easily penetrate thick clothing aswell as walls and foliage, concealed weapons can be easily detected at a distancewithout false alarms caused by a dense RF environment. The proposed techniquethus significantly alters the landscape in favour of the friendly forces by using a high-resolution waveform that is also immune from interference.

Results: This report describes the preliminary simulation results of ultra widebandpolarized scattering from various types of concealed weapons carried by a human.

DRDC Ottawa TM 2009-190 iii

Results are shown for both isolated weapons as well as weapons in close proximityto a human, simulating a human carrying the weapon. By selecting the frequencyrange to cover the region of resonance, i.e., when the weapon size is a wavelength,resonant scattering mechanism provides unique spectral features that can be used fordetecting these weapons. Our preliminary simulations show that it is indeed possibleto use the wideband cross-polarized backscatter to identify concealed weapons carriedby humans. The report consists of five parts which address the following topics: 1)literature survey, 2) capabilities of full-wave electromagnetic models: feasibility andpreliminary case studies, 3) parametric studies in terms of distance, weapon size,mutual position and polarization, 4) complete systemmodels and their performance infrequency and time domain analyses, and 5) recommendations for system parameters.

Significance: The primary objective of this project is to: 1) develop theoretical andnumerical models of electromagnetic wideband back-scattering from various targettypes, such as concealed weapons using noise waveforms and 2) develop and demon-strate a radar sensor to detect concealed weapons carried by an individual underconditions of multipath, clutter, and foliage obscuration, while remaining immune todetection, jamming, and interference.

Potential niche area spin-offs are covert communications, ground penetration ima-ging, SAR imaging, polarimetric ISAR imaging, foliage penetration (FOPEN) SARimaging, anti-jamming imaging performance, through wall imaging, detection of im-provised explosive devises (IEDs), radar tags, MIMO (multiple input multiple output)noise radar, compact ID (identification), police and firefighting applications such assearch-and-rescue or hostage-taking scenarios, etc.

The anticipated benefits for DRDC and the client are: 1) to develop expertise innoise radar technology, especially in the area of concealed weapon detection andcovert communications to provide advice, support, consultation on IEDs/concealedweapons on a small ship or FIAC (fast inshore attack craft) threats, IEDs buried insand/soil/clay, human target detection, and covert networks; 2) to enhance situationalawareness, intent prediction and decision making for achieving operational superior-ity, for example, reliable and timely situational awareness information involving thepresence of suicide bombers, concealed weapons, or IEDs; and 3) asymmetric advant-age for defeating terrorist groups and tactics, for example, monitoring and detectionof potential suicide bombers, concealed weapons, and IED threats.

iv DRDC Ottawa TM 2009-190

Sommaire

Ultrawideband (UWB) high-resolution noise radar forconcealed weapon detection :electromagnetic simulation - phase 1

T. Thayaparan, N. Nikolova ; DRDC Ottawa TM 2009-190 ; R & D pour la défenseCanada – Ottawa ; octobre 2009.

Contexte : Les armes dissimulées constituent une menace importante pour le per-sonnel militaire, de sécurité et d’application de la loi. Les formes d’onde des radarsà large bande permettent d’exciter des résonances électromagnétiques naturelles afinde mieux caractériser la taille et la forme d’un objet. Des techniques de résonanceont été utilisées pendant nombre d’années pour caractériser les propriétés de systèmesphysiques, mécaniques et électriques. La taille, la forme et la composition physique dusystème déterminent la réponse résonante. Un signal incident du radar à large bandeinduit des courants résonants ou à "onde stationnaire" sur un objet, ce qui produit unécho radar. Comme les résonances sont déterminées par les caractéristiques physiquesde l’objet, l’écho radar devient une " empreinte " électromagnétique qui peut servirà reconnaître l’objet. Des résultats d’expériences précédentes réalisées en conditionssilencieuses au moyen de formes d’onde à impulsions ont confirmé l’existence d’unesignature radar unique et indépendante d’aspect qui peut être utilisée comme une" empreinte ". D’autres essais ont été réalisés à l’aide d’armes et d’objets nuisiblestransportés par un individu. Ces essais ont confirmé qu’il est possible de distinguerdes personnes armées de personnes non armées, et ce, même si l’arme est dissimuléederrière le dos.

L’illumination d’une arme sur une gamme de fréquences permet de tracer la structurerésonante de l’objet et de fournir une signature déterminée uniquement en fonctionde sa taille, de sa forme et de sa composition matérielle. Afin d’induire une réponserésonante dans un objet, il faut illuminer celui ci dans la bande de fréquences des ré-sonances naturelles. Dans le domaine de fréquences, on peut alors facilement balayerle signal sur la bande appropriée, une fréquence à la fois, au moyen d’un analyseurde réseau commercial. Étant donné que la taille des armes dissimulées peut varierentre 10 cm (longueur d’onde correspondante de 3 GHz) et 75 cm (longueur d’ondecorrespondante de 400 MHz), un radar à bruit à bande ultralarge (UWB) fonction-nant sur la bande de 400-3000 MHz peut être fabriqué pour exciter la dispersionrésonante d’armes de tailles variées. Comme ces fréquences pénètrent facilement lesvêtements épais ainsi que les murs et le feuillage, les armes dissimulées sont facilementdétectables à distance sans fausses alarmes causées par un environnement dense enRF. La technique proposée modifie donc considérablement le paysage au profit des

DRDC Ottawa TM 2009-190 v

forces amies, grâce à une forme d’onde à haute résolution qui est aussi insensible aubrouillage.

Résultats : Le présent rapport décrit les résultats des simulations préliminaires dedispersion polarisée à bande ultralarge de divers types d’armes dissimulées sur desindividus. Les résultats portent sur des armes isolées ainsi que des armes à proximitéd’un individu, qui simule un individu transportant une arme. Grâce à la sélectiond’une gamme de fréquences qui permet de couvrir la région de résonance, c.-à-d.lorsque la taille de l’arme est d’une longueur d’onde, le mécanisme de diffusion réso-nante fournit des caractéristiques spectrales uniques qui peuvent être utilisées pourdétecter ces armes. Selon nos simulations préliminaires, il est en effet possible d’uti-liser la rétrodiffusion à polarisation croisée à large bande pour déceler des armesdissimulées sur des individus. Le rapport comprend cinq parties qui portent sur lessujets suivants : 1) recherches bibliographiques ; 2) capacités des modèles électroma-gnétiques à onde pleine : faisabilité - certaines études de cas préliminaires ; 3) étudesparamétriques en fonction de la distance, de la taille des armes, de la position mu-tuelle et de la polarisation ; 4) modèles du système complet et leurs performancesdans les analyses de la fréquence et du domaine temporel ; 5) recommandations pourles paramètres de système.

Portée : Le présent projet vise principalement : 1) à développer des modèles théo-riques et numériques de rétrodiffusion à large bande électromagnétique à partir dedivers types de cibles, comme des armes dissimulées, au moyen de formes d’onde debruit ; 2) à mettre au point et à faire la démonstration d’un capteur radar visant àdétecter des armes dissimulées sur un individu dans des conditions de brouillage partrajet multiple, clutter et feuillage, tout en restant immunisé contre la détection, lebrouillage intentionnel et les interférences.

Les domaines spécialisés dérivés potentiels comprennent notamment les communica-tions secrètes, l’imagerie par pénétration du sol, l’imagerie SAR, l’imagerie ISAR po-larimétrique, l’imagerie SAR de pénétration du feuillage (FOPEN), les performancesd’imagerie anti brouillage, l’imagerie à travers les murs, la détection de dispositifsexplosifs de circonstance (IED), les étiquettes radar, le radar à bruit à entrées et àsorties multiples (MIMO), l’identification compacte et les applications policières etde lutte contre l’incendie, comme des scénarios de recherche et de sauvetage et deprise d’otages.

Les avantages prévus pour RDDC et le client sont les suivants : 1) le développementd’une expertise dans la technologie des radars à bruit, notamment dans le domainede la détection d’armes dissimulées et des communications secrètes, en vue d’offrirdes informations et de l’appui et de tenir des consultations sur les IED et les armesdissimulées sur un petit navire ou les menaces liées aux embarcations d’assaut rapide(FIAC), aux IED enfouis dans le sable, le sol ou l’argile, ainsi que la détection de cibles

vi DRDC Ottawa TM 2009-190

humaines et de réseaux secrets ; 2) l’accroissement de la sensibilisation à la situation,de l’intention de prédiction et de prise de décision en vue d’obtenir la supérioritéopérationnelle, comme des informations de sensibilisation à la situation fiables et àtemps concernant la présence de bombes humaines, d’armes dissimulées ou d’IED ; 3)l’avantage asymétrique en vue de venir à bout des groupes et des tactiques terroristes,comme la surveillance et la détection de menaces liées aux bombes humaines, auxarmes dissimulées et aux IED.

DRDC Ottawa TM 2009-190 vii


viii DRDC Ottawa TM 2009-190

Table of contentsAbstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . i

Résumé . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . i

Executive summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iii

Sommaire . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . v

Table of contents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ix

List of figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xi

1 introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

2 Literature Survey . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

2.1 Concealed Weapon Detection Using Late Time Response (LTR) . . 5

2.2 Microwave/Millimeterwave Holography . . . . . . . . . . . . . . . . 10

2.2.1 Single frequency MMW Holography Inversion . . . . . . . . 12

2.2.2 Wideband MMW Holography Inversion . . . . . . . . . . . . 13

2.3 Millimeterwave (MMW) Radiometry . . . . . . . . . . . . . . . . . . 16

2.4 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18

3 Capabilities of Fullwave Electromagnetic Models: Feasibility andpreliminary Case Studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19

3.1 Solid Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19

3.2 Summary of Project Setup and Requirements . . . . . . . . . . . . . 23

3.3 Some Weapon Signatures . . . . . . . . . . . . . . . . . . . . . . . . 25

4 Parametric Studies in Terms of Distance, weapon, Size, Mutual Positionand Polarization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33

4.1 Defining the Weapon Signature . . . . . . . . . . . . . . . . . . . . . 33

4.1.1 Plane Wave Excitation (PWE) . . . . . . . . . . . . . . . . 33

4.1.2 Excitation with an UWB Antenna . . . . . . . . . . . . . . 34

DRDC Ottawa TM 2009-190 ix

4.2 Weapon Size . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35

4.3 Weapon Shape . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37

4.4 Weapon Orientation and Incident Wave Polarization (Beretta inOpen Space) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42

4.4.1 Plane Wave Along Z (Broadside Illumination) . . . . . . . . 43

4.4.2 Plane Wave Along X and Y (Narrow-side Illuminations) . . 52

4.5 Weapon Position on the Human Body (Beretta and Man) . . . . . . 55

5 Complete System Models and Their Performance in Frequency and TimeDomain Analyses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63

6 Recommendations Regarding Real System Parameters . . . . . . . . . . . 65

6.1 Radar Range Equation . . . . . . . . . . . . . . . . . . . . . . . . . 65

6.2 Estimated Receive-to-Transmit Power Ratios . . . . . . . . . . . . . 68

6.2.1 Results in the Fresnel Region . . . . . . . . . . . . . . . . . 70

6.2.2 Results in the Fraunhofer Region . . . . . . . . . . . . . . . 75

6.3 Antenna Gain and the Half-power Beamwidths (HPBW) . . . . . . . 79

6.4 Maximum Transmitted Power . . . . . . . . . . . . . . . . . . . . . 82

6.4.1 Safety Code 6 Canada [42] . . . . . . . . . . . . . . . . . . . 82

7 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86

x DRDC Ottawa TM 2009-190

List of figuresFigure 1: Comparison of different types of high-resolution waveforms from

which the featureless characteristics of the noise waveform areclearly evident: impulse (top), linear frequency modulated(middle), and random noise (bottom). The reflected signal fromthe target is cross-correlated with a time-delayed replica of thetransmit waveform. . . . . . . . . . . . . . . . . . . . . . . . . . 2

Figure 2: Normalized RCS of a sphere of radius a . . . . . . . . . . . . . . 6

Figure 3: Output of the EST detection system [7-8]. . . . . . . . . . . . . . 6

Figure 4: AKELA’s swept frequency experimental setup. . . . . . . . . . . . 8

Figure 5: AKELA’s time domain experimental setup. . . . . . . . . . . . . . 8

Figure 6: The block diagram of the RF/IF section of the interrupted CWradar of EST [7-8]. . . . . . . . . . . . . . . . . . . . . . . . . . . 9

Figure 7: Experimental setup for LTR studies in [9]. . . . . . . . . . . . . . 10

Figure 8: Coordinate setup in the single-frequency holography theory [17]. . 11

Figure 9: Coordinate setup in the wide-band holography theory [17]. . . . . 14

Figure 10: (a) 35-GHz hologram, and (b) reconstructed image of amannequin with a concealed weapon [17]. . . . . . . . . . . . . . . 16

Figure 11: Man carrying metallic (top) and ceramic (bottom) handguns: (a)optical image; (b) MMW radiometry image [2]. . . . . . . . . . . . 17

Figure 12: Three CAD models for the study of the BerettaM9FS handgunwhere illumination is done through an UWB Vivaldi antenna: (a)gun only, (b) man202 only, (c) man202 plus gun suspended at hisabdomen. Illumination direction is along +z. Polarization isvertical. Distance from antenna to target is approximately 1 m. . 21

Figure 13: Geometry and size of objects whose back-scatter signatures arecompared: (a) sphere, (b) hand grenade, (c) BerettaM9FShandgun, and (d) automatic rifle AK47. . . . . . . . . . . . . . . . 26

DRDC Ottawa TM 2009-190 xi

Figure 14: Normalized magnitude spectra of the co-pol (y-polarized)backscattered far-zone E-field of four objects: sphere of radius a =3 cm, a hand grenade, the BerettaM9FS handgun, and theautomatic rifle AK47. . . . . . . . . . . . . . . . . . . . . . . . . . 27

Figure 15: Co-pol (y-polarized) RCS of four objects: sphere of radius a = 3cm, a hand grenade, the BerettaM9FS handgun, and theautomatic rifle AK47. . . . . . . . . . . . . . . . . . . . . . . . . . 28

Figure 16: Same as FIG. 14 without the co-pol RCS of AK47. . . . . . . . . 29

Figure 17: RCS plots for AK47 showing a major low-frequency peak in theco-pol RCS at 245 MHz. . . . . . . . . . . . . . . . . . . . . . . . 29

Figure 18: Normalized magnitude spectra of the cross-pol (x-polarized)backscattered far-zone E-field of four objects: sphere of radius a =3 cm, a hand grenade, the BerettaM9FS handgun, and theautomatic rifle AK47. . . . . . . . . . . . . . . . . . . . . . . . . . 30

Figure 19: Cross-pol (x-polarized) RCS of four objects: sphere of radius a =3 cm, a hand grenade, the BerettaM9FS handgun, and theautomatic rifle AK47. . . . . . . . . . . . . . . . . . . . . . . . . . 31

Figure 20: Same as FIG. 19 without the cross-pol RCS of AK47. . . . . . . . 32

Figure 21: Three simulations representing three positions of theBerettaM9FS handgun on the body: (a) front, (b) side, and (c)back. In all cases the plane wave is incident from the front and isvertically polarized. . . . . . . . . . . . . . . . . . . . . . . . . . . 34

Figure 22: The magnitude RCS spectrum of the Beretta handgun in a widerfrequency range (from 50 MHz to 9 GHz). The plane wave isincident along +z (see pointing vector P in inset and field ispolarized along y). . . . . . . . . . . . . . . . . . . . . . . . . . . 36

Figure 23: The simple I and Γ shapes and their characteristic dimensions :(a) I-cylinder, (b) I-plate, (c) Γ-plate. The plane wave excitationis along z and its field is x-polarized. . . . . . . . . . . . . . . . . 37

Figure 24: Co-pol (X-pol) RCS of I-cylinders from 0.3 GHz to 3 GHz. . . . . 38

Figure 25: Co-pol (X-pol) RCS of I-plates from 0.3 GHz to 3 GHz. . . . . . . 39

Figure 26: Co-pol RCS of the G plates from 0.3 GHz to 3 GHz. . . . . . . . . 40

xii DRDC Ottawa TM 2009-190

Figure 27: Cross-pol RCS of the G plates from 0.3 GHz to 3 GHz. . . . . . . 41

Figure 28: The orientation of the handgun with respect to the plane waveincident along z and the observation points at which the RCS isrecorded. The Poynting vector P shows the direction of incidenceand the vector shows the E-field polarization. The point P1records a monostatic return while the points P2 and P3 are at 90degrees along x and —y, respectively. The distances from the threeobservation points to the weapon are not to scale — they are onlyrepresenting directions. All three points are in the far zone. . . . . 43

Figure 29: RCS of the Beretta handgun at P1 (monostatic returns, along —z).The ‘Y-pol’ RCS is co-polarized while the ‘X-pol’ RCS iscross-polarized with respect to the incident wave (z-incidentpolarized along y). . . . . . . . . . . . . . . . . . . . . . . . . . . 44

Figure 30: RCS of the Beretta handgun at P2 (along x). The ‘Y-pol’ RCS isco-polarized while the ‘X-pol’ RCS is cross-polarized with respectto the incident wave (z-incident polarized along y). . . . . . . . . 45

Figure 31: RCS of the Beretta handgun at P3 (along —y). The ‘Y-pol’ RCS isco-polarized while the ‘X-pol’ RCS is cross-polarized with respectto the incident wave (z-incident polarized along y). . . . . . . . . 46

Figure 32: RCS of the Beretta handgun at P1 (monostatic returns, along —z).The ‘Y-pol’ RCS is co-polarized while the ‘X-pol’ RCS iscross-polarized with respect to the incident wave (z-incidentpolarized along y). . . . . . . . . . . . . . . . . . . . . . . . . . . 47

Figure 33: Co-pol (Y-pol) RCS of the Beretta handgun at the threeobservation points P1, P2 and P3 for a z-incident wave polarizedalong y. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48

Figure 34: Z-pol RCS of the Beretta handgun at the three observation pointsP1, P2 and P3 for a z-incident wave polarized along y. . . . . . . 49

Figure 35: Co-pol (X-pol) RCS of the Beretta handgun at the threeobservation points P1, P2 and P3 for a z-incident wave polarizedalong x. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50

Figure 36: Cross-pol (Y-pol) RCS of the Beretta handgun at the threeobservation points P1, P2 and P3 for a z-incident wave polarizedalong x. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51

DRDC Ottawa TM 2009-190 xiii

Figure 37: Orientations of the incident waves in two narrow-sideilluminations of the Beretta handgun. The plane waveilluminations are: (a) PxEy, (b) PyEx. . . . . . . . . . . . . . . . 52

Figure 38: Comparison of the weapon co-pol RCS when it is illuminatedfrom the front (PzEy) and from the side (PxEy). . . . . . . . . . . 53

Figure 39: Comparison of the weapon co-pol RCS when it is illuminatedfrom the front (PzEx) and from the side (PyEx). . . . . . . . . . . 54

Figure 40: Comparison between the co-pol (Y pol) back-scattered far fieldrecorded at P1 in the case of the man with a weapon IN FRONTand in the case of a man only (no weapon). . . . . . . . . . . . . . 56

Figure 41: The signature of the weapon in open space Sw and its signatureextracted in the case of the weapon IN FRONT of the man. . . . 57

Figure 42: Comparison between the co-pol (Y pol) back-scattered far fieldrecorded at P1 in the case of the man with a weapon AT THESIDE and in the case of a man only (no weapon.) . . . . . . . . . 58

Figure 43: The signature of the weapon in open space Sw and its signatureextracted in the case of the weapon AT THE SIDE of the man. . . 59

Figure 44: Comparison between the co-pol (Y pol) back-scattered far fieldrecorded at P1 in the case of the man with a weapon AT THEBACK and in the case of a man only (no weapon.) . . . . . . . . 60

Figure 45: The signature of the weapon in open space Sw and its signatureextracted in the case of the weapon AT THE BACK of the man. . 61

Figure 46: Monostatic radar scenario illustrating the significance of themutual position and orientation of the antennas and the target.The radar return depends on the antenna directivity D whichsubstantially depends on the angular position (θt, φt) of the targetrelative to the antenna boresight. The directivity of the receivingantenna depends on the angle (θr, φr), at which the backscatterarrives. Assuming that its boresight is oriented the same way asthat of the transmitting antenna, (θr, φr) = (θt, φt). Thebackscattered signal strength depends substantially on the angleof incidence (θi, φi) at which the target is illuminated. The angleof incidence is taken in the local coordinate system of the target. . 67

xiv DRDC Ottawa TM 2009-190

Figure 47: Receive-to-transmit co- and cross-pol power ratios at 400 MHzversus range within the Fresnel region; the antenna gain is 1.5(1.8 dB). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71

Figure 48: Receive-to-transmit co- and cross-pol power ratios at 1 GHz versusrange within the Fresnel region; the antenna gain is 3.0 (4.7 dB). . 72

Figure 49: Co-pol return in dB versus range within the Fresnel region forfour gain values. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73

Figure 50: Cross-pol return in dB versus range within the Fresnel region forfour gain values. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74

Figure 51: Receive-to-transmit co- and cross-pol power ratios at 400 MHzversus range within the far zone; the antenna gain is 1.5 (1.8 dB). 75

Figure 52: Receive-to-transmit co- and cross-pol power ratios at 1 GHzversus range within the far zone; the antenna gain is 3.0 (4.7 dB). 76

Figure 53: Co-pol return in dB versus range within the far zone for four gainvalues. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77

Figure 54: Cross-pol return in dB versus range within the far zone for fourgain values. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78

Figure 55: Illustration of the antenna HPBW θV in the vertical direction andits relation to the illuminated metric width WV at a range R. . . . 81

DRDC Ottawa TM 2009-190 xv


xvi DRDC Ottawa TM 2009-190

1 introductionConcealed weapons pose a significant threat to military, security, and law enforce-ment personnel. Terrorists and law-breakers routinely carry concealed weapons, suchas revolvers, guns, rigged cell-phones, and rifles in their clothing. The uncontrolledenvironments associated with peacekeeping and the move toward relaxation of con-cealed weapons laws provide a strong motivation for developing weapons detectiontechnologies, which are non-invasive and can function non-cooperatively. Canadianand allied forces routinely deal with the detection of suicide bombers, concealedweapons, improvised explosive devices (IEDs), and terrorists crouching under foliageor behind building walls. Existing weapons detection systems are primarily orientedto detecting metal and require the cooperation of the person being searched. Newgeneration of detectors under development focusing primarily on imaging methodsface problems related to privacy. The need exists for a weapons detector, which isportable, detects weapons remotely from a distance, avoids privacy issues, differen-tiates between car keys and a knife, and is affordable enough to be issued to everypeacekeeper or police officer.

Current technologies to detect concealed weapons, such as X-Rays and passive ra-diometry, are expensive, cumbersome, slow, and prone to false alarms [1]. Since aconcealed weapon comes in different sizes and shapes, ultra wideband (UWB) radartechniques can be used to excite natural electromagnetic resonances that characterizethe size, shape, and material composition of an object. Neural network processingcan then be employed to classify the difference between weapons and nuisance ob-jects, and even to recognize the weapon type. Current radar detection systems arebased on short-pulse or linear frequency-modulated waveforms, which are more proneto radio frequency, electromagnetic and noise interference sources, and clutter envir-onments. Recently there has been a notable increase in conflicts between emergingmilitary radar technologies and existing users of the spectrum. Thus, probing wave-forms must spread the energy over a very wide bandwidth, be non-redundant, appearas random noise, and possess the spectral efficiency features [2]. Radar technology ismuch more reliable compared to newly-emerging terahertz technologies, which havevery low sensitivity and can detect only a few vapours using spectroscopy, but cannotdetect metals having large radar signature concealed behind clothing.

Random noise radar is an attractive and viable option for use in these applications.Random noise radar refers to techniques and applications that use incoherent noise asthe probing transmit waveform. Research work in this area has been conducted sincethe late 1950s [3]. However, only in the last few years have commercial chips reachedto GHz frequencies and made such systems practical. In such a system, the transmitsignal is pure thermally generated noise. A major advantage of using noise as thetransmit signal is its inherent immunity from radio frequency and electromagnetic

DRDC Ottawa TM 2009-190 1

0 5 00 10 00 15 00-0 .5

0

0 .5

1Im pu ls e waveform

Am

plit

ud

e

0 5 00 10 00 15 00-1

-0 .5

0

0 .5

1L FM waveform

Am

plit

ud

e

0 5 00 10 00 15 00

-2

0

2

Ran do m n oise wa ve fo rm

Tim e

Am

plit

ud

e

Figure 1: Comparison of different types of high-resolution waveforms from which thefeatureless characteristics of the noise waveform are clearly evident: impulse (top),linear frequency modulated (middle), and random noise (bottom). The reflectedsignal from the target is cross-correlated with a time-delayed replica of the transmitwaveform.

interferences and improved spectrum efficiency. Figure 1 compares different typesof high-resolution waveforms from which the featureless characteristics of the noisewaveform are clearly evident: impulse (top), linear frequency modulated (middle),and random noise (bottom). The reflected signal from the target is cross-correlatedwith a time-delayed replica of the transmit waveform. When the internal delay exactlyequals the round-trip time to the target, a peak is obtained in the correlation signal;otherwise, the correlator output is zero. The range resolution is inversely proportionalto the transmit bandwidth, and a 1-GHz bandwidth yields a range resolution of 15cm. In 2004, a Task Group (TG) on Noise Radar Technology was established byNATO to further develop this technology for military applications with active DRDCparticipation.

A conventional noise signal is not phase-coherent since the transmit waveform is inco-herent. Thus, the correlation process captures only the amplitude but not the phaseof the return signal. However, a technique called heterodyne correlation overcomesthis problem in a unique and novel manner [4]. In this method, the return signalfrom the target is cross-correlated with a time-delayed and frequency-offset replica ofthe transmit signal. The frequency-offset waveform is achieved by beating the noisesignal with a phase-locked oscillator at the offset frequency. When this is done, the

2 DRDC Ottawa TM 2009-190

correlator output at the correct time-delay is always exactly at the offset frequency,superimposed with the Doppler frequency shift caused by target motion. In addition,the correlator output is zero at time delays that do not match the round trip timeto the target. Since the correlator output is at the offset frequency, its relative phasecan be obtained by comparing its waveform with the phase-locked offset frequencyoscillator in a phase detector. Thus, the phase of the reflected signal can be extracteddespite the fact that the illuminating waveform is totally incoherent. By measuringphase, even low micro-Doppler frequency modulations can be analyzed and smalltarget movements estimated, as shown in many applications [5-8].

Noise radars have tremendous advantages over conventional radar systems, which areespecially relevant for defence and homeland security applications. These include: (1)Clutter rejection: Since the noise waveform is spread over a wide frequency range, ithas the necessary diversity to reduce clutter and multipath effects; (2) Electromag-netic compatibility: Many noise radars can occupy the same spectral band, with negli-gible cross-interference as the signal from one will not correlate with the others’ trans-mit replica [9]; (3) Spectrum efficiency: Due to the uncorrelated and non-interferencenature of the waveform, these systems possess enhanced spectral occupancy [9]; (4)Ease of signal processing: Thermal noise is easy to generate, and modulators withgood linearity or antennas with good impulse response are not needed; (5) Frequencyshaping: The noise spectrum can be shaped to enhance detection of specific targettypes, prevent signal leakage into adjacent bands, or prevent in-band spectral fratri-cide with friendly systems; (6) Thumbtack range-Doppler ambiguity function: Bothrange and Doppler can be simultaneously estimated and independently controlled byvarying bandwidth and integration time; (7) Immunity from interference and jam-ming: External signals caused by jammers or other interfering transmitters will notcorrelate with the time-delayed transmit replica and hence will yield zero output; (8)Immunity from detection: Since the waveform is not repeatable, it does not seem asan intentional signal on the adversary’s receiver.

Although noise radars can be applied to related areas such as concealed weapon,IED detection, through wall imaging, etc., we are specifically focusing on concealedweapon in this study. High-resolution ultrawideband (UWB) noise radar systems areideal sensors for detecting a variety of concealed weapons of different sizes. Resonancetechniques have been used for many years to characterize the properties of physical,mechanical, and electrical systems. The size, shape, and physical composition ofthe system determine the resonant response. An incident wideband radar signalinduces resonant or “standing wave’ currents on an object thereby producing radarreturn. Since the resonances are determined by the physical characteristics of theobject, the radar return becomes an electromagnetic “fingerprint’ that can be usedto recognize the object. Prior experimental results under noiseless conditions usingimpulse waveforms confirmed the existence of a unique, aspect independent, radarsignature that can be used as a “fingerprint”. Additional tests were conducted with


both weapons and nuisances being carried by a human. These tests confirmed thatit is possible to distinguish the armed from the unarmed case, even when the weaponwas concealed behind the back [10].

Illuminating a weapon over a range of frequencies maps its resonant structure andprovides a signature that is uniquely determined by its size, shape, and materialcomposition. In order to induce a resonant response in an object, it is necessaryto illuminate it in the frequency band of the natural resonances. In the frequencydomain, this can be easily accomplished by sweeping the signal over the appropriateband one frequency at a time using a commercial network analyzer. Since concealedweapons may range in size from 10 cm (wavelength corresponding to 3 GHz) to 75 cm(wavelength corresponding to 400 MHz), an UWB noise radar operating over the 400-3000 MHz band can be made to excite resonant scattering from weapons of differentsizes. Since these frequencies easily penetrate thick clothing as well as walls andfoliage, concealed weapons can be easily detected at a distance without false alarmscaused by a dense RF environment. The proposed technique thus significantly altersthe landscape in favour of the friendly forces by using a high-resolution waveform thatis also immune from interference. This approach therefore uses ‘matched illumination’noise waveforms whose properties are matched to the specific types of targets desiredto be detected.

The report is organized as follows. The literature survey about the detection ofconcealed weapons based on radar technology in given in Section 2. Capabilitiesof full-wave electromagnetic models in relation to concealed weapon carried by ahuman is presented in Section 3. Section 4 presents the parametric studies in termsof distance, weapon size, mutual position and polarization. The complete systemmodels and their performance in frequency and time domain analyses are given inSection 5. Recommendations for system parameters are given in Section 6.

It should be noted that this project is a multiyear study in three phases. In thisreport, we present the results from phase 1 of the project.


2 Literature Survey2.1 Concealed Weapon Detection Using Late Time

Response (LTR)There are several approaches to the detection of concealed weapons based on radartechnology: 1) identification of the resonant signatures of weapons in the microwaverange in the late-time response, 2) microwave and millimeter—wave holography com-bined with synthetic-aperture radar (SAR), and 3) passive millimeter—wave radiometry.We review these approaches and summarize their advantages and shortcomings. Tech-nologies not considered here are: 1) detection of the distortion of the Earth’s magneticfield, 2) magnetic resonance imaging (MRI), 3) inductive magnetic field methods, 4)acoustic and ultrasonic detection, 5) infrared imagers, etc. Comprehensive reviews inthis area can be found in [11, 12, 13].

Every structure has intrinsic (or natural) resonant frequencies, which depend on itssize, shape and constitutive parameters. Hunt et al. (associated with AKELA Inc.)seem to be the first to develop a practical system which uses wideband radar to excitethe natural resonances of concealed weapons [14, 15, 16]. Another practical systemwas reported by Hausner and West [17, 18] (of Electro Science Technologies, EST).EST personnel have several patents for their system — hardware and signal processingmethods. Similar investigations in an ultra-wide band are reported in [19]. Theenhancement of the target radar cross section (RCS) at resonant frequencies can beillustrated by the normalized (with respect to the geometrical cross—section) RCS ofa sphere [see Figure 2].

The advantage of the resonant radar return is that it is aspect independent, i.e., itdoes not depend on the direction of illumination. The polarization, however, mayplay an important role in the ability of the incident wave to excite the dominantnatural modes [19, 20]. Likewise, the polarization of the LTR signal carries importantinformation and both co-polarized and cross-polarized returns should be recorded andprocessed [17, 19].

The LTR radar technique is a non-imaging detection method for real—time warningof possible threat [see Figure 3]. It does not produce an image of the scanned object.It analyzes the radar return looking to identify objects whose resonant signaturesare known beforehand. In particular, it looks at the LTR of the signal. It is wellknown that the early—time (specular) part of the radar return signal carries inform-ation about the location of the scattering centers. In contrast, the late—time partof the signal consists essentially of slowly decaying oscillations carrying the resonant“signatures" of the scatterers.

The efficiency of the LTR approach depends crucially on the ability of the detection


Figure 2: Normalized RCS of a sphere of radius a

Figure 3: Output of the EST detection system [7-8].


algorithm to identify the resonant signatures of interest and to distinguish them fromthe signatures of innocuous items (pocket clutter, nuisance objects). It is reportedthat the resonant signatures of handguns lie between 500 MHz and 2 GHz. Theused detection algorithms are: 1) genetic—algorithm classifiers [14, 15, 16]; 2) neuralnetworks [17, 18]; 3) singularity expansion method [19]. The singularity expansionmethod (SEM) was introduced by Baum; see, for example, [21, 22]. The LTR of atarget is represented in the s-plane with its poles and zeros. Known targets are firstanalyzed to obtain their “natural modes" in terms of their poles and zeros in thes-plane. Then, the LTR of the interrogated target is processed and compared withthe database for the objects of interest. In [19], the poles and zeros of the LTR areextracted using the generalized pencil—of—function (PoF) method [23, 24].

AKELA [14, 15, 16] has developed a swept-frequency continuous wave (SF-CW) radarand a pulsed radar; see Figure 4 and Figure 5. The SF-CW radar covers the frequencyrange from 450 MHz to 2 GHz. It is digitally synthesized and controlled. It hasrange resolution of 4 inches and maximum range of 10 to 15 feet. The SF-CWsystem provides precise waveform control. They have used mostly Vivaldi antennasbut have also developed more compact spiral antennas. In 2002, AKELA reportsprobability of detection Pd = 64 % and probability of false alarm Pfa = 36 % in thecase of a classifier trained to classify “any individual". In the case of a classifier fora particular individual, the results are somewhat better. Their classifiers seem to bebased on genetic algorithms.

EST [17, 18] developed a low—cost system which emulates an impulse radar. Theimpulse radar, which can transmit pulses as short as several hundreds of picoseconds,is too expensive, large and heavy in order to be deployed in practical systems. It alsorequires high—performance antennas. Instead, EST uses a sequence of measurementswith a relatively wide pulse (of duration 10 ns) transmitted with a fixed frequency.This frequency is stepped through the desired band from 9.50 GHz to 10.55 GHz.There are 256 such steps in a complete measurement. The receiver has a bandwidth ofseveral hundred Mega-hertz. One complete measurement takes several milliseconds,which is in effect real—time. The system is termed “interrupted CW" radar, whichreflects the fact that the time duration of the transmitted pulse is long compared tothe transit time of the signal to the target and back. A horizontally polarized signal istransmitted through one TX antenna but both horizontal (co-polarized) and vertical(cross-polarized) signals are detected through two linearly polarized RX antennas. Atthe end of the 256 measurements, a frequency—domain response is formed, which isthen transformed into a time—domain waveform using Chirp-Z transform. The time-domain response is processed using a trained neural network. Excellent performanceis reported with Pd ≥ 95 % and Pfa ≤ 10 %. The operating range is 9 to 15 feet. Theblock diagram of the RF/IF section of the interrupted CW radar of EST is shown inFigure 6. The antennas are patch arrays.


Figure 4: AKELA’s swept frequency experimental setup.

Figure 5: AKELA’s time domain experimental setup.


Figure 6: The block diagram of the RF/IF section of the interrupted CW radar ofEST [7-8].


Figure 7: Experimental setup for LTR studies in [9].

LTR target identification via the SEM has been studied in the context of concealedweapon detection in [19]. The measurements use test instrumentation; see Figure 7.Vector network analyzer (VNA) drives a transmitting antenna (port 1) and collectsthe back—scattered signal from the receiving antenna (port 2) emulating a swept—frequency radar. The antennas are wideband Vivaldi elements. The time—domainbackscattered waveform is obtained from the frequency—sweep measurement throughinverse Fourier transform. The LTR is extracted from the time waveform. Thepoles and zeros of the LTR are extracted and compared to those of known targets.The measurements are done in the frequency band from 0.7 GHz to 6.7 GHz. Thescattered signal is extracted using a calibration measurement in which there is notarget within the antenna range.

The major drawback of the LTR methods seems to be in the noise due to the person’ssignature. The signature of an individual with a weapon is very similar to one withouta weapon. Besides, the signature of a person is strong and highly unpredictable. Thissignature has to be efficiently filtered out from the total response. This seems to bea problem, hence the high false-alarm rate (in AKELA’s implementation).

2.2 Microwave/Millimeterwave HolographyActive microwave and millimeterwave (MMW) sensors have the capability to detectconcealed weapons because they can achieve resolution well below 1/10th of the size


Figure 8: Coordinate setup in the single-frequency holography theory [17].

of a typical hand weapon. Also, at these frequencies the reflectivity contrast betweenthe body and a metallic weapon is fairly high. The proposed approaches can beclassified as: 1) real aperture radar (RAR), and 2) radar holography.

RAR has a number of advantages such as relative simplicity and low cost. Its majordrawback is that it has relatively low resolution in order to reliably image smallerweapons such as knives or small guns (e.g., sizes of about 15 cm or smaller).

In radar holography, a transmitter and a receiver are scanned linearly over an aperturewhile being swept over a very wide frequency band. This wide frequency bandwidthallows for a good range resolution despite the fact that the target is located very closeto the scanned aperture.

Several practical MMW holographic systems were built by the Pacific North WestNational Laboratory (PNNL) [25, 26, 27, 28] operating at: 22.0 to 47.5 GHz, 40 to 60GHz, and 90 to 120 GHz. Holographic imaging relies on amplitude and phase meas-urements of the scattered signal across a 2-D aperture [see Figure 8]. The techniquedeveloped at PNNL essentially merges the characteristics of microwave holographyand synthetic aperture radar (SAR).


2.2.1 Single frequency MMW Holography Inversion

The received waveforms are discretely sampled over the frequency band and over theaperture (x0, y0, z0 = z0 ). They are expressed as

s(x0, y0) =

ZZSo

f(x, y, z)e−j2k√(x−x0)2+(y−y0)2+z20dxdy , (1)

where So is the object surface, f(x, y, z) is the object’s reflectivity function, and k isthe wave number. The reflectivity function is the ratio of reflected to incident fieldand it is the imaged quantity. In Equation (1), it is assumed that the target’s surfaceis flat and parallel to the scan plane, i.e., z is constant, z = 0 .

The exponential term in Equation (1) can be viewed as a spherical wave emanatingfrom (x0, y0) and with a wave number 2k. It can be represented by a superposition ofplane waves at z = 0 :

e−j2k√(x−x0)2+(y−y0)2+z20 =

Z 2k

−2k

Z 2k

−2ke−j[kx(x−x

0)+ky(y−y0)−kzz0]dkxdky . (2)

We can now substitute Equation (2) in Equation (1) to obtain

s(x0, y0) =

ZZSo

f(x, y, 0)

∙Z 2k

−2k

Z 2k

−2ke−j[kx(x−x

0)+ky(y−y0)−kzz0]dkxdky

¸dxdy . (3)

Re-arranging,

s(x0, y0) =

Z 2k

−2k

Z 2k

−2k

∙ZZSo

f(x, y, 0)e−jkxx−jkyydxdy

¸| {z }

F2D (kx,ky)

ejkxx0+jkyy0+jkzz0dkxdky . (4)

Denoting the 2-D spatial Fourier transform of f(x, y, 0) as F2D(kx, ky), Equation (4)can be written as

s(x0, y0) =

Z 2k

−2k

Z 2k

−2k

£F2D(kx, ky)e

jkzz0¤ejkxx

0+jkyy0dkxdky = (2π)2F−12D

©F2D(kx, ky)e

jkzz0ª,

(5)where F−12D denotes the 2-D inverse Fourier transform. It follows that

F2D(kx, ky)ejkzz0 = F2D {s(x0, y0)} . (6)

The reflectivity function can now be determined as [see Equation (4)]

f(x, y, 0) = F−12D [F (kx, ky)] = F−12D©e−jkzz0F2D[s(x0, y0)]

ª. (7)


From the dispersion relation,

k2x + k2y + k2z = (2k)2 , (8)

we determinekz =

q4k2 − k2x − k2y , (9)

which is substituted in Equation (7) to yield the reconstruction formula

f(x, y, 0) = F−12D [F (kx, ky)] = F−12Dne−jz0√4k2−k2x−k2yF2D[s(x0, y0)]

o. (10)

The above formula summarizes the single—frequency holographic image reconstruc-tion.

To allow for near—field 3-D imaging, wide frequency band is used and an algorithmwhich does not use the far—field approximation. It allows for the reconstructionof the 3-D dependence of the reflectivity function f(x, y, z) where z is no longer aconstant [see Figure 9]. The algorithm, however, assumes that the data representsa single reflection from the target (no multiple reflections) and that there are nodispersion and polarization changes due to the target. The wideband reconstructionalgorithm can be considered as an extension of the single—frequency “backward—wave"holographic reconstruction algorithm.

2.2.2 Wideband MMW Holography Inversion

In the 3-D (wideband) case, the measured response at the radar aperture is definedas the triple integral

s(x0, y0, ω) =

ZZZVo

f(x, y, z)e−j2k√(x−x0)2+(y−y0)2+z20dv . (11)

Following the same steps as in the case of a 2-D target, we express Equation (11) as

s(x0, y0, ω) =

Z 2k

−2k

Z 2k

−2k

£F (kx, ky, kz)e

jkzz0¤ejkxx

0+jkyy0dkxdky

= (2π)2F−12D©F (kx, ky, kz)e

jkzz0ª,

(12)

where (compare with Equation (4))

F (kx, ky, kz) =

ZZZVo

f(x, y, z)e−jkxx−jkyy−jkzzdxdydz . (13)

From Equation (12), we obtain

F (kx, ky, kz)ejkzz0 = F2D {s(x0, y0, ω)} . (14)


Figure 9: Coordinate setup in the wide-band holography theory [17].


We can make use of the wide—band frequency information through the dispersionrelation,

k2x + k2y + k2z = (2k)2 = 4(ω/c)2 , (15)

where c is the speed of light. We can obtain the 3-D reflectivity function by applying3-D inverse Fourier transform to Equation (14) and expressing ω as a function of kzusing Equation (15):

f(x, y, z) = F−13D {F (kx, ky, kz)} = F−13D©e−jkzz0F2D {s(x0, y0, ω)}

ª, (16)

whereω =

³cqk2x + k2y + k2z

´/2 . (17)

The reconstruction formula Equation (16) can produce the reflectivity of the targetexactly, provided that the assumptions for single reflection and lack of de-polarizationand dispersion hold. Of course, exact reconstruction assumes also continuous fre-quency and spatial (over the radar aperture) scans. The latter is not possible inpractice and the limitations of the sampling rates must be observed. The lateralsampling along the aperture must satisfy

∆ < λ/4 , (18)

which ensures that the phase shift from one sampling point to the next is smaller thanπ rad (Nyquist limit) for signals propagating from and back to the aperture. Thefrequency sampling is determined from the requirement that the phase shift 2kRmaxdue to a change in the wave number∆k is less than π rad. Here, Rmax is the maximumrange. We obtain that

∆f <c

4Rmax. (19)

Alternatively, the number of frequency samples for a bandwidth B must be

Nf >2Rmaxc/(2B)

. (20)

The cross—range resolution is

δx ≈λ0

4 sin(θb/2), (21)

where λ0 is the wavelength at the central frequency and θb is the lesser of the antennabeamwidth or the angle subtended by the aperture. If the range R is much greaterthan the aperture D,

δx ≈λ02· RD

. (22)

The same principles apply for the other cross—range resolution δy.


Figure 10: (a) 35-GHz hologram, and (b) reconstructed image of a mannequin witha concealed weapon [17].

The range resolution depends on the frequency bandwidth:

δz ≈c

2B. (23)

A full—body 3-D wideband holographic scan requires less than 2 seconds. Resolutionbetter than 5 mm is obtained. Distances to the targets are between 10 and 25 feet[25]. Cylindrical imaging apertures have also been used [26].

The result of the holographic reconstruction is an image where the concealed weaponis hopefully visible against the background of the body [see Figure 10].

2.3 Millimeterwave (MMW) RadiometryThese are passive millimeter—wave systems typically operating in the frequency rangebetween 35 GHz and 94 GHz. They respond to the thermal energy in this frequencyband emitted or reflected by the target scene. The principles are similar to those ininfra—red imaging; however, the cross—range resolution is poorer because of the largerwavelength. On the other hand, millimeter waves have much better penetration thaninfra—red waves through obscurants such as smoke, fog, rain or clothing. An excellentreview can be found in [29]. Since this technology seems to depart from the objectives


Figure 11: Man carrying metallic (top) and ceramic (bottom) handguns: (a) opticalimage; (b) MMW radiometry image [2].

of this project, the current review is limited to the principles and not the technicaldetails.

In the case of metallic objects, the image contrast is due to the differences in theemissivities of objects. For example, while ground appears relatively warm (∼ 290 ◦K),metallic objects appear relatively cold (∼ 80 ◦K in the W band, which is from 75 GHzto 111 GHz). The human body, which has very good emissivity close to unity, is usu-ally at a higher temperature (∼ 37 ◦C ≈ 310 ◦K) than metallic weapons whose typicalemissivity is about 0.2. This makes for a very good temperature contrast betweenthe body and the weapon even if the weapon is somewhat warmed up by the body.

MMW systems are capable of detecting body—worn non-metallic objects as well suchas plastic or ceramic guns, explosives, or dangerous chemicals. In this case, the tem-perature contrast is due to the concealed object blocking partially the body emissionand substituting it partially with its own lower emission. See Figure 11 (from [12]).

Encouraging results have been reported of prototype systems for concealed weapondetection via MMW radiometry [30, 31, 32, 33]. The challenge is to reduce the scan


time below a second. The time needed to obtain the image relates to the systemnoise temperature Tsys, its bandwidth B, its integration time τ , and the number n ofsimultaneous beams (receivers).

2.4 ConclusionsIt is clear from the preliminary studies, that in the microwave range most of theinformation about the weapon is contained in the late—time response (LTR) of theradar signal. The following conclusions and recommendations can then be made:

1. Sophisticated algorithms will be needed to identify the LTR and separate itfrom the specular reflection.

2. The LTR will then have to be post—processed and compared to the LTRs ofknown weapons.

3. The radar signals can be simulated provided the full—wave EM models prove tobe feasible and sufficiently accurate.

4. Item 2 indicates that the signatures of all weapons of interest will have to befirst acquired and stored in a database for comparison with the measured radarresponse.

5. The process of comparison may include neural networks and genetic algorithms(as suggested in the literature) or some novel feature extraction approachesdeveloped specifically for this project. Items 1, 2 and 5 require expertise intarget identification and signal processing.

6. It is not clear at this point whether the specular part of the back—scatter canbe of any use. It carries information mostly about the distance to the target.


3 Capabilities of Fullwave ElectromagneticModels: Feasibility and preliminary CaseStudies

3.1 Solid ModelsThe development of the full—wave electromagnetic models requires the following CAD1 models:

• Weapons: handguns (pistols), knives, rifles, grenades, explosives

• Humans: standing male (arms spread), standing male (arms down), sittingmale, respective female models, standing child model

• Radar antennas: ultra-wideband (UWB) Vivaldi, horns, respective arrays

• Furniture and standard indoor scatterers: metallic desks and bookshelves, desktopcomputers and monitors, brick walls, etc.

We have acquired the following solid models, which are downloaded from website(http://www.3dcadbrowser.com/default.aspx):

WEAPONS

1. 404_Beretta_M9FS: DXF, SAT, STL

2. 3078_Kalasnikov_AK74: DXF, STL

3. 5893_AK47_Assault_Rifle: DXF, STL

4. 6619_Ingram_M10: DXF, STL

5. 8073_Winchester_Rifle: DXF, STL

6. 9199_Machine_Gun: DXF, STL

7. 9782_80_Articulated_Segments_Hand_Grenade: DXF, STL

8. 9795_Sniper_Rifle_Shooting_Bullets: DXF, STL

9. 11974 Glock 21: STL

10. 11976_40MM_Grenade_Launcher: DXF, STL

1Here, CAD (computer-aided design) models refers to the shape of 3D solids. The CAD modelin its most common use implies that only geometrical information is provided.


11. 13514_Smith_and_Wesson_Revolver: DXF, STL

HUMANS

1. 202_Man: STL (male, arms spread)

2. 2567_Man: DXF, STL (male, arms down)

3. 17987_Seated_Human_Dummy: DXF, STL

ROOMS

1. 10507_Room_Interior: DXF, STL

2. 14611_Room_Interior: DXF, STL

FURNITURE

1. 2481_Lamp: DXF, STL

2. 4512_Bookshelf: DXF, STL

3. 5921_File_Cabinet: DXF, STL

COMPUTERS AND ELECTRONICS

1. 12274_Dell_Laptop_Computer: DXF, STL

The STL models can be successfully imported in computer simulation technology(CST) Microwave Studio (CST MWS) [34] and Ansoft high-frequency structure sim-ulator (HFSS) [35]. The DXF files can be imported into the XFDTD (commer-cial finite-difference time-domain) package [36] and can be further exported to SATformat. The latter is compatible with both HFSS and CST MWS. Caution must beexercised with regard to units and size. The units of most of the models are notspecified. It was found that many of the weapon CAD models are in mils while thehuman models must be properly scaled to fit into any of the standard units for length.

Some representative full—wave models were built in CSTMWS. These models are laterused to examine the signature of weapons with and without a human, with a plane—wave illumination and with illumination through a UWB antenna, for polarizationstudies, etc. For example, the initial study of the BerettaM9FS handgun required7 full—wave models as listed in Table 1. Three of these models (first column) useplane—wave excitation (PWE). The other four models use the UWB antenna as anexcitation.


(a)

(b)

(c)

(a)

(b)

(c)

Figure 12: Three CAD models for the study of the BerettaM9FS handgun whereillumination is done through an UWB Vivaldi antenna: (a) gun only, (b) man202only, (c) man202 plus gun suspended at his abdomen. Illumination direction is along+z. Polarization is vertical. Distance from antenna to target is approximately 1 m.


In all CAD files, the absolute and mutual position of objects and excitation is exactlypreserved. This is illustrated in Figure 12. This is essential in deriving the weapon’ssignature in two scenarios: (1) weapon in open space and (2) weapon at the humanbody. The second scenario actually requires two simulations — one with the humanonly and one with the human and the gun. By subtracting the response of theformer from that of the latter simulation, one can obtain the gun response only. It isimportant to find out whether the so obtained response is going to differ substantiallyfrom the response of the weapon in case (1).

The full—wave EM models are very demanding in terms of memory requirements andcomputation time. They are analyzed on a dual—core Intelr Xeonr CPU5160 @ 3GHz, 3 GHz, with 32 GB RAM. Run time is on the order of hours when the memoryrequirements are on the order of 2 to 3 GB.

Some preliminary conclusions have been made:

1. The EMmodels can accommodate problems where the UWB antenna is no morethan 1.5 m away from the target. Problems larger than that require excessivecomputational time.

2. Models where the UWB antenna is replaced by a plane—wave excitation havesmaller computational requirements and can be made more accurate by increas-ing the spatial sampling.

3. It has been confirmed that the resonant frequencies of the handguns, whoseCAD models are available, indeed lie between 300 MHz and 2 GHz as suggestedin the literature.

4. It has been confirmed that the weapon signatures are indeed affected by thepresence of the human body.

Table 1: Models used in the initial study of BERETTAM9FS (mutual position, ori-entation, and wave polarization are constant)

Plane Wave Excitation UWB Antenna ExcitationAntenna only

Gun only Gun + AntennaMan202 only Man202 + AntennaMan202 + Gun Man202 + Gun + Antenna


3.2 Summary of Project Setup and RequirementsThe frequency range of all simulations is set from 0.3 GHz to 3 GHz.

In all projects, metallic weapons (or metallic parts of weapons) are modeled as perfectelectric conductors (PEC) because conduction loss can be neglected in the frequencyrange of interest. The metals used for guns are alloys with very high content ofcopper, as well as tin, zinc, and steel, all very good conductors.

The human body is modeled as a non-dispersive material of constitutive parametersr = 40 and σ = 0.8 S/m. These values are close to those of muscle in the low-GHz frequency range and, therefore, the results are representative of a human ofmuscular built. It should be noted that the constitutive parameters of fat are verydifferent ( r ∼ 5, σ ∼ 0.07 S/m). Further studies must address: (1) the effectof frequency dispersion in order to evaluate its importance for the accuracy of theobtained results; (2) the impact of the body type (mostly fatty or mostly muscular)on the weapon signature. Frequency dispersion requires additional computationalresources and should be avoided if possible.

• Project Description: BerettaM9FS (Plane Wave Excitation)

Table 2: Summary of project “BERETTAM9FS”

Mesh Solver RequirementsProperties Setup

min step max step number energy max number energy time memorymm mm mesh cells convergence pulses balance MB5.25512 5.66283 1,277,760 -80 dB 100 1e-12 0h 12m 16s 750

• Project Description: Man202 (Plane Wave Excitation)

Table 3: Summary of project “MAN202”


min step max step number energy max number energy time memorymm mm mesh cells convergence pulses balance MB6.2272 8.46233 9,363,250 -80 dB 100 1e-12 5h 0m 32s 3,000


Table 4: Summary of project “MAN202 plus BERETTAM9FS”


min step max step number energy max number energy time memorymm mm mesh cells convergence pulses balance MB6.2272 8.46233 9,363,250 -80 dB 100 1e-12 5h 59m 25s 3,000

Table 5: List of representative simulation projects with brief description

Name Description1. BerettaM9FS Gun in open space; plane-wave excitation (PWE)2. Man202 Human (arms spread) in open space; PWE3. BerettaM9FS plus Man202 Human (arms spread) with gun at abdomen

in open space; PWE4. BerettaM9FS plus Man202 (side) Human (arms spread) with gun under armpit

in open space; PWE5. BerettaM9FS plus Man202 (back) Human (arms spread) with gun at back

in open space; PWE6. Grenade Grenade in open space; PWE7. AK47 AK assault rifle in open space; PWE8. BerettaM9FS plus Antenna Gun in open space; UWB antenna excitation9. Man202 plus Antenna Human (arms spread) in open space; UWB

antenna

• Project Description: Man202 + BerettaM9FS front (Plane Wave Excitation)

In the following tables (Table 2 to Table 4), representative simulation projects, run inCSTMicrowave Studio [34], are described in terms of the parameters of the simulationsetup as well as their time and memory requirements.

Table 5 lists the types of projects that have been simulated. The computationalrequirements of projects 1 to 7 are comparable or smaller than those listed in Table4.

The most demanding projects are those including a realistic antenna model [see mod-els 8 and 9 in Table 5]. The antenna features small geometrical details which requirefine local sampling. Without shape simplification, the number of mesh cells is onthe order of 109 and the computational time exceeds two days (on a dual-core Intelr

Xeonr CPU 5160 @ 3 GHz, 3 GHz, with 32 GB of RAM).


For future studies, even more complicated scenarios will have to be considered, whichcan properly include additional sources of clutter such as walls, furniture, pocketobjects (e.g., keys), etc. It is advisable to consider the simplification of some of thefeatures of the human body. The degree of simplification must be carefully chosen sothat it does not affect the accuracy of the results for the signature of the weapon. Thefollowing defintions are used for the co-polarized and cross-polarized backscatteredfields. The co-polarized backscattered field is the backscattered field whose polariz-ation is the same as that of the incident field and the cross-polarized backscatteredfield is the backscattered field whose polarization is orthogonal to that of the incidentfield. Note that the same definition applies to the radar cross-section of a target sincethe radar cross-section is calculated from the backscattered field.

3.3 Some Weapon SignaturesThe signatures of four objects have been studied: a sphere of radius a = 3 cm, a handgrenade, the BerettaM9FS handgun, and the automatic rifle AK47. The structuresand their approximate dimensions are shown in Figure 13. All objects are consideredperfectly conducting.

Figure 14 shows the normalized amplitude spectra of the y-polarized (co-pol) far—zone E-field for the four objects. It is clear that the relative spectral content of thesignatures of the objects is very different. Figure 15 shows the respective co-pol radarcross sections (RCS). The RCS of the AK47 is much stronger than the other threeobjects, which is due to its larger size. Also, AK47 features very strong resonance at2.7327 GHz. Since the AK47 RCS is much stronger than the RCS of the other threeobjects, for better visibility, the RCS of the sphere, the grenade, and the handgun areplotted again in Figure 16. We also note that the AK47 has another major resonancebelow 300 MHz. This is shown in Figure 17 where the co-pol and cross-pol RCSof AK47 are plotted from 100 MHz to 3 GHz. A major peak of the co-pol RCS isobserved at 245 MHz.

The cross-pol (x-polarized) far—field normalized spectra of the four objects are plottedin Figure 18. It is clear that the cross-pol weapon signatures are different from theco-pol signatures. This is an indication that polarization diversity should be exploitedfor the purpose of target identification. The cross-pol RCS of the four objects areplotted in Figure 19 and Figure 20. The peak cross-pol RCS of AK47 is at 0.3999GHz which is at the opposite end of the frequency band as compared to the peak ofits co-pol RCS.


radius a = 3 cm length: 13 cmwidth: 9.2 cm

length: 16.6 cmwidth: 10.6 cm

thickness: 4 cm

length: 50 cmwidth: 21 cm

thickness: 3 cm

(a) (b)

(c) (d)

radius a = 3 cm length: 13 cmwidth: 9.2 cm

length: 16.6 cmwidth: 10.6 cm

thickness: 4 cm

length: 50 cmwidth: 21 cm

thickness: 3 cm

(a) (b)

(c) (d)

Figure 13: Geometry and size of objects whose back-scatter signatures are compared:(a) sphere, (b) hand grenade, (c) BerettaM9FS handgun, and (d) automatic rifleAK47.


0.5 1 1.5 2 2.5 3x 10

9

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

frequency (Hz)

norm

aliz

ed m

ag s

pect

rum

spheregrenadeberettaAK47

Figure 14: Normalized magnitude spectra of the co-pol (y-polarized) backscatteredfar-zone E-field of four objects: sphere of radius a = 3 cm, a hand grenade, theBerettaM9FS handgun, and the automatic rifle AK47.


Figure 15: Co-pol (y-polarized) RCS of four objects: sphere of radius a = 3 cm, ahand grenade, the BerettaM9FS handgun, and the automatic rifle AK47.


Figure 16: Same as FIG. 14 without the co-pol RCS of AK47.

Figure 17: RCS plots for AK47 showing a major low-frequency peak in the co-polRCS at 245 MHz.


0.5 1 1.5 2 2.5 3x 10

9

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

frequency (Hz)

norm

aliz

ed m

ag s

pect

rum

spheregrenadeberettaAK47

Figure 18: Normalized magnitude spectra of the cross-pol (x-polarized) backscatteredfar-zone E-field of four objects: sphere of radius a = 3 cm, a hand grenade, theBerettaM9FS handgun, and the automatic rifle AK47.


Figure 19: Cross-pol (x-polarized) RCS of four objects: sphere of radius a = 3 cm, ahand grenade, the BerettaM9FS handgun, and the automatic rifle AK47.


Figure 20: Same as FIG. 19 without the cross-pol RCS of AK47.


4 Parametric Studies in Terms of Distance,weapon, Size, Mutual Position andPolarization

4.1 Defining the Weapon SignatureThe parametric studies have for now been limited to the case where the target is inopen space or suspended on the human body, i.e., there are no other sources of clutterbut the human.

The target signatures can be extracted using two different responses for the twodifferent types of excitations (the PWE and the UWB antenna).

4.1.1 Plane Wave Excitation (PWE)

The weapon signature is expressed in terms of the complex—valued monostatic far—zone field, i.e., the scattered field is evaluated in the direction from which the planewave is incident. The far field is acquired for the weapon in open space. This responseis denoted as Sw.

Another weapon signature can be obtained when the weapon is on the human body.Then, two simulations are performed — one with the human alone and one with thehuman (in exactly the same position) and the weapon on the body. In the firstsimulation, the far field of the human is denoted as the response Sh. In the secondsimulation, the response Shw is that of the human with the weapon. We can thenobtain the weapon signature (when suspended on the human body) as

S(h)w = Shw − Sh . (24)

It is important to compare S(h)w with Sw in order to evaluate the changes in the

weapon signature due to the close proximity of the human body.

In this report, the signature of the BerettaM9FS handgun is studied in three positions:

• handgun at the abdomen (waist height),

• handgun on the side under the armpit,

• handgun at the back (above waist).

In all cases, the plane wave is incident from the front and is vertically polarized. Thethree cases are shown in Figure 21.


(a) (b) (c) (a) (b) (c)

Figure 21: Three simulations representing three positions of the BerettaM9FS hand-gun on the body: (a) front, (b) side, and (c) back. In all cases the plane wave isincident from the front and is vertically polarized.

4.1.2 Excitation with an UWB Antenna

In this case, the radar return can be expressed in terms of the reflection coefficientS11 measured at the antenna terminals. The weapon signature is then obtained as

S(a)w = (S11)wa − (S11)a (25)

where S(a)w is the weapon signature in the presence of the antenna alone (no clutter),(S11)wa is the reflection at the antenna terminals in the presence of the weapon,and (S11)a is the reflection coefficient of the antenna in open space. The obtainedsignature can then be compared to the signature of the weapon in open space aswell as the weapon signature in the presence of the human, both obtained by theplane—wave illumination [refer to the previous subsection].

Further, the weapon signature received by the UWB antenna must be studied forthe cases where it is body worn. Similarly to the case of a plane—wave excitation,we acquire the antenna return S11 in the case of a human alone and in the case of ahuman with the weapon. The weapon signature is then determined as

S(ah)w = (S11)hwa − (S11)ha (26)

where (S11)hwa is the return in the model where the weapon is on the human body and(S11)ha is the return at the antenna terminals where the human is the only scatterer(no weapon).


The weapon signature as captured by the UWB antenna is to be studied in thesecond phase of the project when the UWB antenna design will be completed. Thiswill also enable us to determine accurately the level of return with respect to thedistance between the antenna and the target. We note that the interaction betweenthe human body as a scatterer and the UWB antenna is likely to be in their mutualreactive—near—field or radiating—near—field zones (separation distances of couple ofmeters). To determine the maximum range of the system, we also need to know thereceiver signal—to—noise ratio.

4.2 Weapon SizeThe weapon size affects its radar return through two mechanisms. First, the sizedetermines its intrinsic resonance frequencies. The radar return is enhanced at res-onance frequencies exhibiting a characteristic peak in magnitude. The smaller theweapon is, the higher its resonance frequencies. The intrinsic frequencies determinethe major content of the late—time response (LTR) of a target. If the observed RCSin the given frequency band has no peaks but is increasing with frequency, the targetresonances must be lying in a higher frequency band. Such monotonically increasingRCS is seen in Figure 2 for the case of a metallic sphere [see Rayleigh Region]. TheRayleigh Region corresponds to target sizes of approximately λ/6 or less. If in thefrequency band of interest, the RCS exhibits one or several peaks of comparable mag-nitudes, then this frequency band falls within the target’s Resonance Region. This isa desirable scenario for a weapon detection system based on resonant signatures.

Second, the size is crucial to the strength of the radar return, especially its specular(early-time) part. Thus, even if the incident pulse does not excite resonances inthe structure (i.e., its frequency band does not contain any of the weapon resonancefrequencies), the RCS may still be large due to the size of the target. As the targetbecomes several wavelength and larger, its RCS approaches its cross—section in theplane perpendicular to the direction of incidence [see Figure 2, Optical Region].

This is illustrated with the RCS of the BerettaM9FS handgun which was calculatedin three frequency bands: from 50 MHz to 300 MHz, from 300 MHz to 3 GHz, andfrom 3 GHz to 9 GHz [see Figure 22]. It is clear that the resonance region for thisweapon lies approximately between 250 MHz and 3 GHz.

Further discussion related to the importance of the object size on its RCS is givenbelow for some simple shapes.


Figure 22: The magnitude RCS spectrum of the Beretta handgun in a wider frequencyrange (from 50 MHz to 9 GHz). The plane wave is incident along +z (see pointingvector P in inset and field is polarized along y).


L

D

PE L

W

T

PE

L

H

W

T

P E

(a) (b) (c)

L

D

PE L

W

T

PE

L

H

W

T

P E

(a) (b) (c)

Figure 23: The simple I and Γ shapes and their characteristic dimensions : (a) I-cylinder, (b) I-plate, (c) Γ-plate. The plane wave excitation is along z and its field isx-polarized.

4.3 Weapon Shape

Table 6: Dimensions of the simple shapes studied at frequencies from 0.3 to 3 GHz

I-cylinder (in mm) I-plate (in mm) Γ-cylinder (in mm)# L D L W T L H W T1. 100 10 100 10 3 100 50 25 32. 200 20 200 20 3 200 100 50 33. 300 30 300 30 3 300 150 75 34. 400 40 400 40 3 400 200 100 3

In order to study the effect of the shape on the weapon signature, we consider threetypes of simple shapes: the ‘I’ shape (e.g., the shape of a knife or a rifle) and the‘Γ’ shape (e.g., the shape of a handgun). The investigation is done for the shapein conjunction with the size of the object so that a recommendation for the mostsuitable frequency band can be made. The simulated structures are shown in Figure23 and their dimensions are summarized in Table 6.

The co-pol (X pol) RCS of the I-cylinders are plotted in Figure 24. It is obvious that asthe cylinder increases in size, its major resonance shifts toward the lower frequencies.All four shapes, however, exhibit at least one major resonance in the frequency bandbetween 0.3 GHz and 3 GHz. The cross-pol RCS are negligible (smaller than theco-pol RCS by a factor of 10−14) and are not plotted here. Very similar results areobserved in the case of the I-plates whose co-pol RCS are plotted in Figure 25. Thecross-pol RCS are again negligible. At the low frequency end, the I-plate responsesare almost indistinguishable from those of the I-cylinders. This is due to the fact


Figure 24: Co-pol (X-pol) RCS of I-cylinders from 0.3 GHz to 3 GHz.


Figure 25: Co-pol (X-pol) RCS of I-plates from 0.3 GHz to 3 GHz.


Figure 26: Co-pol RCS of the G plates from 0.3 GHz to 3 GHz.


Figure 27: Cross-pol RCS of the G plates from 0.3 GHz to 3 GHz.


that the lengths of the two types of objects are the decisive parameter while theircross—sectional shape parameters are less important. The decisive role of the lengthis specific to this frequency band where it becomes comparable to the wavelengths ofthe incident pulse. The cross—sectional parameters are too small to have a significantimpact.

The co-pol and cross-pol RCS of the Γ-plates are plotted in Figure 26 and Figure 27.This time there is significant cross-pol (Y-pol) return (only an order smaller than theco-pol return), which is due to the presence of a long metallic object along the y-axis.The major resonance of a Γ-plate of the same length L as that of an I-cylinder is atlower frequency than that of the I-cylinder, which is due to the overall increase inthe length of the metallic object (i.e., L+H > L). Thus, the lowest resonance of thelongest Γ-plate sinks below 0.3 GHz.

For quick approximations, the following rule of thumb can be used: the overall meanlength of the object (e.g., for a Γ-shape, this is L + H −W ) when multiplied by afactor of 2 will give an approximate value of the wavelength at the lowest—frequencymajor resonance of the structure. For a thin I-shaped object, the overall mean lengthis simply L. If the object is thicker (e.g., for an I-cylinder, D > L/10), the resonancecurve is likely to broaden and shift toward lower frequencies. In general, when theaspect ratio of an object approaches 1 (e.g., sphere, cube, grenade), the cross—sectionallength of the object (e.g., the sphere diameter) should be multiplied by a factor of πin order to obtain the approximate value of the wavelength at the lowest—frequencymajor resonance. Also, for such “round” or “cubical” objects, the resonances are lesspronounced.

Thus, a frequency bandwidth from 300 MHz to 3 GHz is suitable for detecting res-onance signatures of thin I-shaped objects of length anywhere between 5 and 50 cm.With more complicated shapes such as handguns, the mean length is the most im-portant factor. The above frequency band is obviously well suited for practically allhandguns, small and large.

4.4 Weapon Orientation and Incident WavePolarization (Beretta in Open Space)

The effect of the position and orientation of a representative weapon (the Ber-ettaM9FS handgun) with respect to the plane wave is summarized here. We considerthree cases where the handgun is illuminated with a plane wave incident along z,along y, and along −x. We first consider the case of the plane wave along z.


x

y

2P

3P

incEP

x

z

2P

1P

incE

Pz

y

3P

1PP

incE

(a) (b) (c)

x

y

2P

3P

incEP

x

z

2P

1P

incE

Pz

y

3P

1PP

incE

(a) (b) (c)

Figure 28: The orientation of the handgun with respect to the plane wave incidentalong z and the observation points at which the RCS is recorded. The Poynting vectorP shows the direction of incidence and the vector shows the E-field polarization. Thepoint P1 records a monostatic return while the points P2 and P3 are at 90 degreesalong x and —y, respectively. The distances from the three observation points to theweapon are not to scale — they are only representing directions. All three points arein the far zone.

4.4.1 Plane Wave Along Z (Broadside Illumination)

The polarization of the plane wave and the gun orientation in this case are illustratedin Figure 28. The three directions, at which the weapon RCS is investigated, areshown by the observation points P1, P2, and P3. Note that the RCS is a far—zonequantity and the distances from the three observation points to the weapon in Figure28 are not to scale – they are only representing the directions. All three points arein the far zone.

The x-, y- and z-polarized RCS are plotted at points P1, P2 and P3, in Figure 29,Figure 30, and Figure 31, respectively. At P1 (monostatic returns), both the co- andcross-polarized RCS exhibit clear resonance at 645 MHz. This resonance appears alsoin the RCS at P2 and P3. Additional strong resonance appears in the co-pol (Y pol)RCS at P1 and (much weaker) at P2.

It is clear from the plots in Figure 29, Figure 30, and Figure 31, that a given RCScarries similar features regardless of whether it is acquired as a monostatic return or areturn at near—perpendicular directions. To verify this conclusion, each polarization ofthe RCS is plotted for the three observation points in a separate plot [Figure 32, Figure33, and Figure 34]. Both the Y- and X-pol RCS are of comparable magnitudes andvery similar frequency dependence in the monostatic and the orthogonal directions.

Due to the reciprocity of the problem, we expect similar results if the z-incident wave


Figure 29: RCS of the Beretta handgun at P1 (monostatic returns, along —z). The‘Y-pol’ RCS is co-polarized while the ‘X-pol’ RCS is cross-polarized with respect tothe incident wave (z-incident polarized along y).


Figure 30: RCS of the Beretta handgun at P2 (along x). The ‘Y-pol’ RCS is co-polarized while the ‘X-pol’ RCS is cross-polarized with respect to the incident wave(z-incident polarized along y).


Figure 31: RCS of the Beretta handgun at P3 (along —y). The ‘Y-pol’ RCS is co-polarized while the ‘X-pol’ RCS is cross-polarized with respect to the incident wave(z-incident polarized along y).


Figure 32: RCS of the Beretta handgun at P1 (monostatic returns, along —z). The‘Y-pol’ RCS is co-polarized while the ‘X-pol’ RCS is cross-polarized with respect tothe incident wave (z-incident polarized along y).


Figure 33: Co-pol (Y-pol) RCS of the Beretta handgun at the three observationpoints P1, P2 and P3 for a z-incident wave polarized along y.


Figure 34: Z-pol RCS of the Beretta handgun at the three observation points P1, P2and P3 for a z-incident wave polarized along y.


Figure 35: Co-pol (X-pol) RCS of the Beretta handgun at the three observationpoints P1, P2 and P3 for a z-incident wave polarized along x.


Figure 36: Cross-pol (Y-pol) RCS of the Beretta handgun at the three observationpoints P1, P2 and P3 for a z-incident wave polarized along x.


x

y

2P

3P

incE

P x

y

2P

3P

incEP

(a) (b)

x

y

2P

3P

incE

P x

y

2P

3P

incEP

(a) (b)

Figure 37: Orientations of the incident waves in two narrow-side illuminations of theBeretta handgun. The plane wave illuminations are: (a) PxEy, (b) PyEx.

is x-polarized. Indeed, both the co-pol (X pol) and the cross-pol (Y-pol) RCS retaintheir major low—frequency peak in the monostatic and the orthogonal directions [seeFigure 35 and Figure 36].

In summary, we can conclude that a bistatic scenario is not going to bring significantadvantages to the system performance. In other words, for a given position of thetransmitting antenna, there will be no significant advantage in acquiring the scatteredfield at different angular positions. It would suffice to measure the backscatter only.On the other hand, recording both co- and cross-polarized signatures at the trans-ceiver location could almost double the information about the target in comparisonwith a single linearly polarized antenna. A dual—polarization system is desirable.

4.4.2 Plane Wave Along X and Y (Narrow-side Illuminations)

In these scenarios, the orientation of the handgun with respect to the global co-ordinate system remains the same as described in the previous subsection, only thedirection of the plane wave is changed. This is illustrated in Figure 37. This study isimportant in order to evaluate the impact of the mutual position of the weapon andthe transmitting antenna.

The case in Figure 37a is referred to as ‘PxEy’ in view of the Poynting vector orienta-tion along x and the E-field polarization along y. The case in Figure 37b is referred toas ‘PyEx ’. Following this nomenclature, the cases studied in the previous subsection(incidence along z) are denoted as ‘PzEy’ and ‘PzEx ’.

The co-pol (Y pol) RCS in the PxEy case is compared to that of the PzEy case.


0.5 1 1.5 2 2.5 3x 10

9

0.02

0.04

0.06

0.08

0.1

0.12

0.14

0.16

0.18

0.2

0.22

frequency (Hz)

RC

S am

plitu

de s

pect

rum

Y pol, PzEyY pol, PxEy

Figure 38: Comparison of the weapon co-pol RCS when it is illuminated from thefront (PzEy) and from the side (PxEy).


0.5 1 1.5 2 2.5 3x 10

9

0.05

0.1

0.15

0.2

frequency (Hz)

RC

S am

plitu

de s

pect

rum

X pol, PzExX pol, PyEx

Figure 39: Comparison of the weapon co-pol RCS when it is illuminated from thefront (PzEx) and from the side (PyEx).


In both cases, the E-field is polarized along y. The difference is in the angle ofillumination, which results in narrow—side and broadside illumination, respectively.The weapon RCS in both cases are compared in Figure 38. The low—frequency res-onance at about 635 MHz is clearly unaffected by the angle of illumination. Thehigh—frequency resonances, however, change significantly.

Similarly, we compare the co-pol (X pol) RCS in the PyEx case (narrow—side) with thePzEx (broadside) case. Figure 39 shows that the two returns differ most significantlyin the high—frequency end of the band.

In view of these observations, we can conclude that the direction of illumination doesnot influence strongly the weapon’s response; yet, some resonance shifts appear athigher frequencies. This can be explained with the fact that the RCS depends notonly on the intrinsic resonances but also on the weapon cross—section relative to thewavelength. At broadside illumination, the weapon electrical cross—section is largerand growing with frequency. This explains the much greater difference between thebroadside and the narrow—side illumination RCS at higher frequencies than at thelow—frequency end where the intrinsic resonance dominates the return.

It may be beneficial to exploit two or three transceiver stations which illuminate thesubject from different angles. The stations can operate independently as monostaticradars.

4.5 Weapon Position on the Human Body (Beretta andMan)

Here, weapon position refers to the position and orientation of the weapon withrespect to the human body for a given incident wave. So far, 3 positions are studiedand compared: in front of the body, on the side of the body and behind the body[see Figure 21]. The three positions are referred to as ‘Front’, ‘Side’, and ‘Back ’.

Using Equation 24, the signature of the Beretta handgun is calculated for the threepositions in the case of a z-incident wave polarized along y (PzEy plane wave) interms of its RCS. These signatures are compared with the respective weapon RCSwhen the weapon is in open space (no human). Here, all far—field results are recordedat P1 (monostatic return). The weapon, when in front of the human, influencesstrongly both the magnitude and the phase of the far—field return. This is illustratedin Figure 40. The weapon return, when suspended in front of the man, is comparedwith its return in open space in Figure 41. The two returns differ significantly. It isclear that strong coupling exists between the man and the weapon, which affects theweapon signature.

The influence of the weapon on the overall man—plus—weapon return visibly decreases


0.5 1 1.5 2 2.5 3x 10

9

0.05

0.1

0.15

0.2

0.25

0.3

frequency (Hz)

Far F

ield

Mag

Spe

ctru

m (V

/m)

man+weaponman only

Figure 40: Comparison between the co-pol (Y pol) back-scattered far field recordedat P1 in the case of the man with a weapon IN FRONT and in the case of a manonly (no weapon).


0.5 1 1.5 2 2.5 3x 10

9

0.01

0.02

0.03

0.04

0.05

0.06

frequency (Hz)

Far F

ield

Mag

Spe

ctru

m (

V/m

)

Sw

Sw(h)

Figure 41: The signature of the weapon in open space Sw and its signature extractedin the case of the weapon IN FRONT of the man.


0.5 1 1.5 2 2.5 3x 10

9

0.05

0.1

0.15

0.2

0.25

frequency (Hz)

Far F

ield

Mag

Spe

ctru

m (V

/m)

man+weaponman only

Figure 42: Comparison between the co-pol (Y pol) back-scattered far field recordedat P1 in the case of the man with a weapon AT THE SIDE and in the case of a manonly (no weapon.)


0.5 1 1.5 2 2.5 3x 10

9

0.01

0.02

0.03

0.04

0.05

0.06

frequency (Hz)

Far F

ield

Mag

Spe

ctru

m (V

/m)

Sw

Sw(h)

Figure 43: The signature of the weapon in open space Sw and its signature extractedin the case of the weapon AT THE SIDE of the man.


0.5 1 1.5 2 2.5 3x 10

9

0.05

0.1

0.15

0.2

0.25

frequency (Hz)

Far F

ield

Mag

Spe

ctru

m (V

/m)

man+weaponman only

Figure 44: Comparison between the co-pol (Y pol) back-scattered far field recordedat P1 in the case of the man with a weapon AT THE BACK and in the case of aman only (no weapon.)


0.5 1 1.5 2 2.5 3x 10

9

0.01

0.02

0.03

0.04

0.05

0.06

frequency (Hz)

Far F

ield

Mag

Spe

ctru

m (

V/m

)

Sw

Sw(h)

Figure 45: The signature of the weapon in open space Sw and its signature extractedin the case of the weapon AT THE BACK of the man.


in the cases of a handgun at the side and at the back of the man [see Figure 42 andFigure 44]. However, the weapon signature is still significant, in the sense that it isof the same order of magnitude as the weapon signature recorded in open space (nohuman). Moreover, as seen in Figure 43 and Figure 45, the weapon signature is lessaffected by the presence of the human in these two cases.

So far, these studies are done with the plane—wave excitation model. In the secondphase of the project, these studies must be extended to the excitation with a UWBantenna so that absolute estimates can be given about the impact of the mutualposition and polarization on the range of the system.


5 Complete System Models and TheirPerformance in Frequency and TimeDomain Analyses

This is a numerical study of the accuracy of the simulation models of the wholesystem. These models are very complex and the simulation results may be erroneousand thus misleading. Convergence analysis has been done with regard to

• mesh size

• distance to absorbing boundaries and absorbing—boundary setup

• convergence of the time waveforms

• shape details of the human body and the weapon.

Some recommendations are given in Table 7 regarding setting up full—wave simula-tions. These recommendations are to be followed in both time— and frequency—domainsimulations. In addition, in time—domain simulations, the recommended frequencybandwidth of the excitation pulse should not exceed 1:10 ratio.

Establishing the importance of the shape details for the accuracy of the weapon re-sponse has been crucial to the computational efficiency of the electromagnetic models.The shapes of the weapon and the human body should be simplified as much as pos-sible in order to reduce memory requirements, allow for larger cell sizes and, therefore,reduce the computational time. In turn, this allows accommodating larger problemswhere the antenna can be moved several meters away from the target.

The antenna is the major source of mesh—size increase due to its fine shape details.The typical number of mesh cells in a project involving the antenna, the human andthe weapon is 2 × 109 when the antenna is about 1.5 m away from the human. Atthe same time, the same project without the antenna (using plane—wave excitation)requires about 9 × 106 cells. For now, the maximum allowable distance betweenthe simulated antenna and human cannot exceed 1.2 m. This problem must beaddressed in the second phase of this project by the introduction of proper modelsimplification. Such simplifications will also allow considering more complex scenariosincluding indoor furniture, several humans and multiple weapons in the future stagesof this project. This study is only preliminary and must become a focal point of theproject in phase 2.


Table 7: Summary of recommendations for setting up simulation

mesh cell size optimal: λmin/25maximum: λmin/16

adaptive mesh refinement recommendeddistance to absorbing boundaries at least λmax/6convergence criteria S-parameter convergence error: = 0.005

energy (time-domain): at least -80 dBshape details ignore details smaller than λmin/12dispersion of human tissue to be determined


6 Recommendations Regarding Real SystemParameters

The recommendations given below address the frequency range, polarization, angleof illumination, and bistatic vs. monostatic radar.

A. For the reliable detection of handguns, knives and hand grenades, it is recom-mended to cover the frequency band from 300 MHz all the way up to 3 GHz.While prior publications suggest that frequencies up to 2 GHz are sufficient,the simulations performed here clearly show that there is significant resonancecontent in the weapon signature beyond 2 GHz. Even a large weapon such asAK 47 exhibits a characteristic resonance at a frequency as high as 2.74 GHz.The recommended frequency band will cover most of the intrinsic resonancesdue to weapons of sizes roughly from 5 to 50 cm.

B. There is no doubt that a successful system will have to exploit polarization di-versity, i.e., each transceiver will have to be equipped with two cross-polarizedantennas. The electromagnetic models show that the cross-polarized radar re-turns of weapons are distinct from the co-polarized return. They carry specificinformation for parts of the weapon oriented at right angles to the E-field of theincident wave. In contrast, the co-polarized radar return carries informationabout those metallic parts, which are oriented along the E-field of the incidentwave.

C. At the same time, the studies show that a bistatic radar system will not havesignificant advantages over a monostatic one. This is to say that it will sufficeto have a receiver at the location of the transmitter.

D. On the other hand, it would seem that the angle at which a target is illuminatedmay have a significant impact on its response. This is mostly due to the factthat the radar cross-section strongly depends on the cross-section of the weaponin a plane perpendicular to the direction of incidence. It is thus recommendedto have several transceiver stations, each operating as a separate monostaticradar.

Further recommendations given hereafter address the range, antenna gain, and max-imum transmitted power.

6.1 Radar Range EquationThe radar range equation gives the ratio of received to transmitted power valuesprovided the radar cross-section (RCS) of the target is known. In a monostatic


scenario (transmitting and receiving antennas are co-located), the equation is

Pr

Pt= eAteAr ·(1−|S11t|2)(1−|S11r|2)·

µλ

4πR2

¶2·Dt(θt, ϕt) ·Dr(θr, ϕr)

4π·σ(θi, ϕi), (27)

where Pt and Pr are the transmitted and received power values, respectively; eAt andeAr are the transmitting and receiving antenna efficiencies,2 respectively; |S11t| and|S11r| are the reflection coefficients at the transmitting and receiving antenna termin-als, respectively; Dt and Dr are the transmitting and receiving antenna directivities,3

respectively; λ is the wavelength, R is the distance between the antenna and thetarget, and σ is the RCS of the target.4 The directivities and the RCS are evaluatedat the respective mutual position and orientation of the antenna and the target. Fig-ure 46 illustrates the angles which influence the strength of the backscattered signal.These are: the target angle (θt, φt), the angle of reception (θr, φr), and the angle ofincidence at the target (θi, φi).

When the radar exploits polarization diversity, there are two receiving antennas.Typically, one has polarization identical to that of the transmitting antenna (co-polRX antenna) and the other has polarization orthogonal to that of the transmittingantenna (cross-pol RX antenna). In general, the backscattered signal carries botha co-pol and a cross-pol component. The two backscatter components are math-ematically described by the co-pol and the cross-pol RCS of the target, σ|| and σ⊥.The backscatter is split into its components by the orthogonally polarized antennas.Thus, two received power values have to be calculated, the co-pol power Pr|| and thecross-pol power Pr⊥, resulting from a single transmission signal of power Pt whose po-larization is the reference. Note that two antennas are orthogonal if reception is zeroin open space and in the absence of a target, with both antennas in each other’s dir-ection of maximum radiation. We say that the antennas are polarization-decoupled.This may happen when the two antennas are linearly polarized and are geometricallyorthogonal to each other; e.g., a TX dipole is along x, an RX dipole is along y, andthe propagation is along z. However, circularly polarized antennas can be polariza-tionally decoupled as well; e.g., a TX antenna is right-hand circularly polarized andan RX antenna is left-hand circularly polarized, or vice versa.

2The antenna efficiency is the ratio of the power transmitted by the antenna to the power fed toits terminals. It is a measure of the losses in the antenna structure.

3The directivity D is the same as the antenna gain G if the antenna efficiency eA is 1 (or 100%),i.e., G = eAD. By definition, the directivity shows the radiation intensity in the given directiongenerated by the antenna as compared to the averaged radiation intensity. The latter is obtainedas the total radiated power divided by 4π steradians (the full solid angle of a sphere).

4The RCS σ of a target is its effective scattering area, which, when multiplied by the powerdensity of the incident wave, produces an effective total scattered power such that its averagedscattered intensity (total scattered power divided by 4π) is the same as that produced by the realtarget at the receiving antenna.


tR R=

rR R=

( , )t tθ ϕ

TX

RX

( , ) ( , )r r t tθ ϕ θ ϕ=

( , )i iθ ϕ

Figure 46: Monostatic radar scenario illustrating the significance of the mutual po-sition and orientation of the antennas and the target. The radar return depends onthe antenna directivity D which substantially depends on the angular position (θt, φt)of the target relative to the antenna boresight. The directivity of the receiving an-tenna depends on the angle (θr, φr), at which the backscatter arrives. Assumingthat its boresight is oriented the same way as that of the transmitting antenna,(θr, φr) = (θt, φt). The backscattered signal strength depends substantially on theangle of incidence (θi, φi) at which the target is illuminated. The angle of incidenceis taken in the local coordinate system of the target.


Here, we assume that the transmitting and receiving antennas are identical, i.e., thedirectivities are the same for both the co-pol and the cross-pol receiving antennasas well as for the transmitting antenna. We also assume that the transmitting andreceiving antennas are oriented so that their boresight directions are co-parallel, i.e.,(θr, φr) = (θt, φt), as illustrated in Figure 46. With these assumptions, Equation (27)becomes

Pr||,⊥

Pt= e2A · (1− |S11|2)2 ·

µλ

4πR2

¶2· D

2(θt, ϕt)

4π· σ||,⊥(θi, ϕi). (28)

The subscripts ||,⊥ emphasize that two power ratios are calculated when co- andcross-pol returns are measured.

6.2 Estimated Receive-to-Transmit Power RatiosThe following constant parameter values are assumed when generating the receive-to-transmit power ratios versus the range R:

• the co-pol RCS σ|| = 0.1 m2 corresponds to that of a small handgun (Beretta)

at its dominant resonance (at about 600 MHz); for comparison, the maximumco-pol RCS of a hand grenade is σ|| = 0.06 m2 (at about 800 MHz);

• the cross-pol RCS σ⊥ = 0.025 m2 corresponds to that of a small handgun(Beretta) at its dominant resonance (at about 600 MHz); for comparison, themaximum cross-pol RCS of a hand grenade is σ⊥ = 0.002 m2 (at about 800MHz);

• the antenna efficiency is assumed eA = 0.9, which is typical for a high-qualityprinted Vivaldi antenna;

• the reflection coefficient is set at the worst allowable value of |S11| = 0.316,which corresponds to −10 dB.

Note that larger weapons may have larger RCS by a factor of 2 to 3. However, muchlower RCS (by as much as an order of magnitude) may also occur depending on theweapon size and the angle of illumination.

The radar range equation holds when the target is in the far zone (Fraunhofer region)of the radar antennas where their directivity is a meaningful quantity. In the farzone, the transmitted power density decreases steadily as 1/R2 and maintains thesame angular distribution.

In the intermediate (radiation near, or Fresnel) region of the antenna, the powerdensity decreases nearly as 1/R2 (with some fluctuations) but its angular distribution


depends on R and is generally unpredictable. One could, however, assume that theradar range equation would produce reasonable estimates in the intermediate regionas well.

In the reactive near zone, the radar range equation will produce erroneous results.This is because in this zone, the transmitted field is strongly reactive and has stronglynonlinear dependence on both the distance R to the antenna and the angular positionof the observation point. Most importantly, if the target is within the near zone ofthe antenna, it will couple electromagnetically to the antenna, thus, becoming partof the radiating system itself, influencing strongly the antenna. The antenna gain ismeaningless in the reactive near zone.

In this study, we assume that the antennas will follow the currently available ultra-wideband Vivaldi design. This planar design features a size of 300× 350 mm2. Thus,the largest size of the antenna (along the PCB (printed circuit board) diagonal andneglecting the PCB height) is Dmax =

√3002 + 3502 ≈ 461 mm. Using the equation

for the far-zone lower limit,

Rfar =2D2

max

λ, (29)

with the shortest wavelength in our frequency range,

λmin =c

fmax≈ 3× 10

8

3× 109 = 0.1, (30)

we calculate the minimum distance required for the observation point to be in the farzone as Rfar ≈ 4.25 m. Thus, strictly speaking, the radar range equation can be usedfor reliable estimates only for distances R ≥ 4.25 m.

We next calculate the minimum distance required for the observation point to be inthe intermediate (Fresnel) region using

Rint = 0.62

sD3max

λmin≈ 0.614. (31)

Thus, the Fresnel region of the UWB Vivaldi is in the range 0.614 ≤ R ≤ 4.25 m.

Since the radar range equation can still produce reasonable results in the Fresnelregion, we will compute the receive-to-transmit power ratio for two intervals: 1) from1 m to 4 m (Fresnel region where results are only estimates), and 2) from 4 m to 10m (Fraunhofer region where results are reasonably accurate).

We note that the border between the Fresnel and Fraunhofer regions is not distinctand the field properties change very gradually at the transition between the two.


6.2.1 Results in the Fresnel Region

Since the Vivaldi has a gain which changes with frequency significantly, we considertwo frequency cases:

1) At 400 MHz, the Vivaldi gain (or directivity) is 1.8 dB or 1.5 on a linear scale.Essentially, the Vivaldi behaves as a dipole at 400 MHz where its directivity isthe same as that of a small dipole.

2) At 1 GHz, the Vivaldi gain (or directivity) is 4.7 dB or about 3.0 on a linearscale.

The co-pol and cross-pol power ratios at 400 MHz and 1 GHz are plotted in Figure 47and Figure 48, respectively. We observe that the power ratio is actually higher at400 MHz than at 1 GHz despite the higher antenna gain at 1 GHz. This is due tothe mutual cancellation of the directivity term D2 and the wavelength term λ2 inEquation (28). While D increases 2 times from 400 MHz to 1 GHz, the wavelengthdecreases 2.5 times. The cross-pol returns at 400 MHz and at 1 GHz are 4 timesweaker than the co-pol returns, which is due to the fact that σ⊥ = σ||/4.

Next, we consider the worst-case scenario in terms of wavelength when λ = λmin = 0.1m (fmax = 3 GHz). We will use the same co- and cross-pol RCS as before. This time,we plot a family of curves for directivity values ranging from 2 dB to 11 dB with astep of 3 dB. The co-pol curves are plotted in Figure 49 and the cross-pol curves areshown in Figure 50. Since the power ratio depends on the directivity squared, thecurves in each family are separated exactly by 6 dB.


1 1.5 2 2.5 3 3.5 4

0.5

1

1.5

2

2.5

3

3.5

4x 10

-5

range (m)

P r / P

t

Power ratio for co-pol 400 MHzPower ratio for cross-pol 400 MHz

Figure 47: Receive-to-transmit co- and cross-pol power ratios at 400 MHz versusrange within the Fresnel region; the antenna gain is 1.5 (1.8 dB).


1 1.5 2 2.5 3 3.5 4

0.5

1

1.5

2

2.5x 10

-5

range (m)

P r / P

t

Power ratio for co-pol 1 GHzPower ratio for cross-pol 1 GHz

Figure 48: Receive-to-transmit co- and cross-pol power ratios at 1 GHz versus rangewithin the Fresnel region; the antenna gain is 3.0 (4.7 dB).


Figure 49: Co-pol return in dB versus range within the Fresnel region for four gainvalues.


Figure 50: Cross-pol return in dB versus range within the Fresnel region for four gainvalues.


4 5 6 7 8 9 10

2

4

6

8

10

12

14

16x 10

-8

range (m)

P r / P t

Power ratio for co-pol 400 MHzPower ratio for cross-pol 400 MHz

Figure 51: Receive-to-transmit co- and cross-pol power ratios at 400 MHz versusrange within the far zone; the antenna gain is 1.5 (1.8 dB).

6.2.2 Results in the Fraunhofer Region

The same type of results as those in Subsection 6.2.1 are now presented for the rangevalues 4 ≤ R ≤ 10 m, which fall within the far zone (Fraunhofer region). Figure 51and Figure 52 show the receive-to-transmit power ratios at 400 MHz and at 1 GHzwith the respective antenna gains.

Figure 53 and Figure 54 show the radar co- and cross-pol returns in dB for four gainvalues. Again, an antenna gain increase of 3 dB leads to a 6 dB increase in the radarreturn.


4 5 6 7 8 9 10

2

4

6

8

10x 10

-8

range (m)

P r / P t

Power ratio for co-pol 1 GHzPower ratio for cross-pol 1 GHz

Figure 52: Receive-to-transmit co- and cross-pol power ratios at 1 GHz versus rangewithin the far zone; the antenna gain is 3.0 (4.7 dB).


Figure 53: Co-pol return in dB versus range within the far zone for four gain values.


Figure 54: Cross-pol return in dB versus range within the far zone for four gainvalues.


6.3 Antenna Gain and the Half-power Beamwidths(HPBW)

For antennas of well-defined pencil beams (highly directive antennas), the relationbetween antenna directivity and its half-power beamwidths (HPBW) in the two or-thogonal principal planes can be found using Kraus’ formula [37],

D ≈ 41000θ◦Eθ

◦H

(32)

where θ◦E is the E-plane HPBW and θ◦H is the H-plane HPBW, both in degrees. Thisformula overestimates the directivity of our UWB Vivaldi design at 2 GHz wherethe antenna has its highest directivity and the beamwidths are reported as θ◦E = 40,θ◦H = 70. It gives a value of 11.6 dB when the actual gain is about 8.2 dB.

More realistic prediction is given by Elliot’s formula [38]

D ≈ 32400θ◦Eθ

◦H

(33)

used for planar arrays. It yields an estimate of 10.6 dB. Roughly the same estimateis obtained using the Tai and Perreira formula [39]:

D ≈ 72815

θ◦2E + θ◦2H. (34)

The Vivaldi gain is not well estimated by these approximate expressions because itis not a very directive antenna and its pattern is not well approximated by a singlepencil beam. A good approximation is given by the following expression

D ≈ 19000θ◦Eθ

◦H

(35)

where the coefficient in the numerator was adjusted to fit the gain and HPBW dataavailable for the Vivaldi antenna.

The antenna beamwidths have two-fold importance to the system performance. First,narrow beamwidths mean higher gain, thus, higher receive-to-transmit power ratios.Second, narrow beamwidths mean less multipath interference as the beam will benormally focused on the scanned person.

On the other hand, narrow beamwidths may provide insufficient area coverage whenthe man is very close to the antenna. Figure 55 illustrates the simple geometrical


relation between the antenna beamwidth in the vertical direction θV , the range Rand the covered metric width WV :

tan

µθV2

¶=

WV

2R. (36)

The Vivaldi design features averaged HPBW in the vertical direction θV = 77◦. At

a range of RV min ≈ 1.26 m, this angle corresponds to a metric width in the verticaldirection of WV = 2 m (the height of a very tall man). For any distance greater thanRV min, the metric width will be greater than the man and significant reflections fromsurrounding objects can be expected. For distances below RV min, the antenna is notgoing to illuminate entirely a target which is taller than 2 meters.

The same arguments apply in the horizontal direction where a maximummetric widthof a man could be assumed as WH = 1 m. Our current antenna design features anaverage horizontal beamwidth of θH = 65◦. This corresponds to a minimum distanceof RHmin ≈ 0.8 m at which the antenna beam will exactly match a metric width of 1m. If R > RHmin, the beam will be wider than the target.

Here, we have to consider the fact that the far zone of the Vivaldi antenna is limitedfrom below by the distanceRfar ≈ 4.25m. As we discussed previously, it is desirable toperform the measurements in the far zone where the target will not couple reactivelyto the antenna. Since Rfar > RV min and Rfar > RHmin, it is clear that the antennabeam will cover much larger metric area than the geometrical cross-section of a manand there always will be reflections from the surrounding environment. Moreover,at low frequencies, our Vivaldi design radiates practically as a dipole, i.e., it has noazimuth directivity and a very broad beamwidth in the elevation plane (≈ 100◦).Thus, at the low frequency end, clutter and multipath are inevitable.

One may ask whether we could improve the antenna directivity, or, equivalently,reduce its beamwidths. The answer is that this can be done only at the expense ofmaking the antenna larger. The current design is already fairly large for a mobile sta-tion which is to be quickly deployed at any location. We can say with great certaintythat it is probably close to the smallest possible design for the bandwidth between400 MHz and 3 GHz by simply observing that at 400 MHz it behaves essentially asa small dipole and is omnidirectional.

Note also that increasing the size of the antenna will increase Rfar. A distance of 4 to5 m is deemed adequate for scanning in an uncontrolled environment in a high-risksituation as it puts good distance between the security forces and the scanned person.In a controlled environment, fixed stations can be installed with larger antennas oreven antenna arrays.


R

Vθ VW

Figure 55: Illustration of the antenna HPBW θV in the vertical direction and itsrelation to the illuminated metric width WV at a range R.

In a controlled environment with a known range, it could be beneficial to constrainthe beamwidths so that they fit as much as possible the maximum target extents inthe vertical and horizontal directions. This would reduce the reflections and the elec-tromagnetic interference. In addition, this would improve the received-to-transmittedpower ratio. As seen from Equation (28) and Equation (35), this ratio depends onthe two HPBW as

Pr

Pt∼ 1

(θ◦Eθ◦H)

2 . (37)

However, beamwidth limitations will impose severe requirements on the antenna size.To appreciate this problem, consider the following scenario. The range is fixed atR = 5 m and the desired vertical metric width of the beam coverage is 2 m. UsingEquation (36), we calculate the desired vertical beamwidth as θV ≈ 23◦. Analogously,if the desired horizontal metric width of the beam coverage is 1 m, the horizontalantenna beamwidth is obtained as θH ≈ 11.4◦. These angles would correspond to again of about 69 or 18 dBi as per Equation (35) (which is very likely to underestimatethe gain). A practical equation relating the maximum achievable linear gain to theantenna size is [40][41]

Gmax ≈ (ka)2 + 2ka (38)

where k = 2π/λ is the wave number and a is the radius of the smallest sphere whichcircumscribes the antenna. For λmax = 1m (at 300 MHz) andGmax = 69, we calculatethat ka ≈ 7.3666. At λmax, this requires a minimum radius of amin = 1.172 m. Notethat this implies an antenna whose largest dimension is greater than 2 m. Also notethat Equation (38) is a theoretical limit which is practically impossible to achieve, i.e.,the antenna size will have to be well above 2amin. As an example, a high-quality hornantenna achieves a gain of 16 dBi when its aperture is about 2.5λ in both directionswhere λ in our case is λmax = 1 m.


Table 8: Maximum allowed power density for human exposure according to SafetyCode 6

Frequency (MHz) Power Density W/m2 Averaging Time (min)300 to 1500 f/100∗ 61500 to 15000 10 6f∗ is in MHz

In summary, it is not realistic to impose design specifications for the antenna designbased on gain or beamwidths if the antenna is to be used in a portable station. Thespecifications should address phase linearity and low VSWR (voltage standing waveratio) in the bandwidth of interest in order to ensure waveform fidelity. The effect ofclutter and multipath will have to be addressed by the receiver design.

6.4 Maximum Transmitted PowerDuring the system design, the following limitations may need to be considered.

6.4.1 Safety Code 6 Canada [42]

Safety Code 6, Section 2.2, specifies exposure limits for persons NOT classified as RFand microwave workers (the general public included) in terms of electric field strengthand power density for frequencies greater than 100 MHz. The values in terms of powerdensity are summarized in TABLE IX in the relevant frequency bands.

The relation between the maximum power density pmax and the maximum total trans-mitted power is

Ptmax =pmax · 4πR2

G0(39)

where G0 is the antenna gain in the direction of maximum radiation and R is therange. For a small dipole (G0 = 1.5), for example, a maximum power density ofpmax = 10 W/m2 at R = 5 m translates into a significant total radiated power of

Ptmax =10 · 4π · 25

1.5= 2094.4W. (40)

We must bear in mind that the antenna gain depends on the frequency. For example,the UWBVivaldi element has maximum gain at 2 GHz and it isG0max = 6.6 (8.2 dBi).At 2 GHz, we calculate the maximum allowable transmitted power as Ptmax ≈ 476W(56.8 dBm). Notice that at 400 MHz, the antenna has a gain of 1.5 and the maximumpower is given by Equation (40).


In a noise radar, we can assume that the power level is distributed fairly uniformlyin the band of interest. Thus, if a limit is to be specified over 1 MHz of bandwidth,it becomes

P(1 MHz)tmax =

Ptmax

∆fMHz=

476

(3000− 300) ≈ 0.1763W/MHz (22.46 dBm/MHz). (41)

The above limit is significantly lower than the limits imposed by the Safety Code 6Canada.


7 ConclusionThe electromagnetic simulations are sufficiently accurate with a plane-wave excitation(no antenna). With such excitation, they can represent shapes in great detail. CADfiles of weapons and human bodies can be used directly without simplification.

Electromagnetic models including the UWB Vivaldi antenna are extremely demand-ing due to the huge number of mesh cells. Trade-offs have been attempted in termsof spatial discretization in violation of the recommendations given in Section 5. Theresults are so far inaccurate since they do not exhibit sufficient convergence.

The recommendations regarding future work are outlined below.

1) Further studies must address the impact of the body on the weapon signature.At this stage, it is clear that the body can significantly alter the weapon signa-ture, especially when the weapon is "in front" or "at the side" of the human, i.e.,when it is directly illuminated by the incident wave. One conclusive effect hasbeen observed, namely, the presence of the body causes the weapon resonancesto shift toward lower frequencies. This can be explained with the presence ofhigh-permittivity tissue right next to the weapon. It is worthwhile to try andextract a mathematical mapping between the weapon resonant frequencies inopen space and those when it is suspended at the body.

2) In this study, only one body type was considered — a "muscular" male. In Year2 of the project, a parametric study must be performed regarding the type ofbody at which the weapon resides. As a minimum, three body types shouldbe considered: muscular, fatty, and transitional. The human body need not bemodeled accurately. The level of simplification must be carefully determined. Inaddition to the body type, the size and shape of the human body are parametersof interest.

3) It is recommended to further investigate the importance of the frequency disper-sion of the constitutive parameters of the human tissues. Taking into accountthe frequency dispersion in time-domain simulations increases the computationtime. If dispersion has to be taken into account, a careful comparison betweenthe time-domain and the frequency-domain models must be made. While thetime-domain models have the advantage of generating responses in a very widefrequency bands, the frequency-domain simulations may turn out to be moreaccurate and computationally efficient.

4) Further studies must also address the inclusion of additional sources of clutter,especially pocket objects, belts, chains, necklaces, bracelets, etc.


5) In view of the problem observed in simulations including the UWB antenna, itis recommended that one of the focal points of the work in Year 2 becomes thedevelopment of reliable electromagnetic models, which can properly include theUWB antenna. The following simplifications must be investigated, which mayeventually decrease the overall computational load of the simulations:

• shape simplification in the human body and the weapon,• equivalent antenna models exploiting equivalent far-field excitation andcalculation of the antenna return from the back-scattered far field and theequivalent antenna aperture,

• investigation of models based on integral-equation computations.

6) These models have to include both the UWB antenna and the human. Sincethe human and the antenna are likely to be in each other’s near zones, the effectof the distance from the human to the antenna must be carefully examined forboth co-pol and cross-pol returns. In the latter case, the model will need toinclude two antennas, which is going to be particularly challenging.


References[1] N.G. Paulter, Guide to the Technologies of Concealed Weapon and Contraband

Imaging and Detection, NIJ Guide 602—00, U.S. Department of JusticeNational Law Enforcement and Corrections Technology Center, Rockville, MD,USA, 72 p.

[2] L. Turner, “The evolution of featureless waveforms for LPI communications,’Proceedings of the 1991 National Aerospace and Electronics Conference(NAECON), Dayton, OH, pp. 1325-1331, May 1991.

[3] B.M. Horton, “Noise-modulated distance measuring systems,’ Proceedings ofthe IRE, vol. 47, pp. 821-828, May 1959.

[4] R.M. Narayanan, Y. Xu, P.D. Hoffmeyer, and J.O. Curtis, “Design,performance, and applications of a coherent ultrawideband random noiseradar,’ Optical Engineering, vol. 37, pp. 1855-1869, June 1998.

[5] R.M. Narayanan and M. Dawood, “Doppler estimation using a coherentultrawideband random noise radar,’ IEEE Trans. on Antennas andPropagation, vol. 48, pp. 868-878, June 2000.

[6] R.M. Narayanan, R.D. Mueller, and R.D. Palmer, “Random noise radarinterferometry,’ Proc. SPIE Conf. on Radar Processing, Technology, andApplications, Denver, CO, vol. 2845, pp. 75-82, August 1996.

[7] Y. Xu, R.M. Narayanan, X. Xu, and J.O. Curtis, “Polarimetric processing ofcoherent random noise radar data for buried object detection,’ IEEE Trans. onGeoscience and Remote Sensing, vol. 39, pp. 467-478, March 2001.

[8] Y. Zhang and R.M. Narayanan, “Monopulse radar based on spatiotemporalcorrelation of stochastic signals,’ IEEE Trans. on Aerospace and ElectronicSystems, vol. 42, pp. 160-173, January 2006.

[9] T. Thayaparan and C. Wernik, Noise Radar Technology Basics, DRDC TM2006-266, Ottawa, December 2006, 46 p.

[10] A.R. Hunt and R.D. Hogg, “Stepped-frequency CW radar for concealedweapon detection and through-the-wall surveillance,’ Proc. SPIE Conf. onSensors, and Command, Control, Communications, and Intelligence (C3I)Technologies for Homeland Defence and Law Enforcement, Orlando, FL, vol.4708, pp. 99-105, April 2002.

[11] N. G. Paulter, "Guide to the technologies of concealed weapon imaging anddetection," NIJ Guide 602—00, 2001, on line,


http://www.ncjrs.gov/pdffiles1/nij/184432.pdf (Access date: October 20,2009).

[12] H.-M. Chen, S. Lee, R. M. Rao, M. A. Slamani, and P. K. Varshney, "Imagingfor concealed weapon detection," IEEE Signal Processing Mag., Mar. 2005, pp.52—61.

[13] A. Agurto, Y. Li, G. Y. Tian, N. Bowring, and S. Lockwood, "A review ofconcealed weapon detection and research in perspective," Proc. of the 2007IEEE Int. Conf. on Networking, Sensing and Control, Apr. 2007, pp. 443—448.

[14] A. R. Hunt, R. D. Hogg, and W. Foreman, "Concealed weapons detection usingelectro-magnetic resonances," SPIE Conference on Enforcement and SecurityTechnologies, Boston, MA, Nov. 1998, pp. 62—67.

[15] A. R. Hunt and R. D. Hogg, "A stepped-frequency CW radar for concealedweapon detection and through the wall surveillance," in SPIE Proc. on Sensorsand Command, Control, Communications and Intelligence (C3I) Technologiesfor Homeland Defense and Law Enforcement, SPIE Proc. vol. 4708, 2002, pp.99—105.

[16] AKELA, "Demonstration of a concealed weapons detection system usingelectromagnetic resonances, final report", US Dept. of Justice, 2001, on line,http://www.ncjrs.gov/pdffiles1/nij/grants/190134.pdf (Access date: October20, 2009).

[17] J. Hausner and N. West, "Radar based concealed threat detector," IEEEMTT-S Int. Microwave Symp., June 2007, pp. 765—768.

[18] J. Hausner and N. West, "Radar based concealed threat detector," IEEE Int.Conference on Microwaves, Communications, Antennas and ElectronicSystems, 2008 (COMCAS 2008), pp. 1—8.

[19] M. Gashinova, M. Cherniakov, and A. Vasalos, "UWB signature analysis fordetection of body-worn weapons," Int. Conference on Radar (CIE 2006), pp.1—4.

[20] N. Shuley and D. Longstaff, "Role of polarization in automatic targetrecognition using resonance descriptions," Electronic Lett., vol. 40, no. 4, Feb.2004, online no: 20040170.

[21] C. E. Baum, "The singularity expansion method: background anddevelopments," IEEE Antennas and Propagation Society Newsletter, Aug.1986, pp. 15—23.


[22] C. E. Baum, "The singularity expansion method," Chapter 3 in TransientElectromagnetic Fields, L. Felsen ed., Springer-Verlag, Berlin, 1976.

[23] Y. Hua and T. K. Sarkar, "Generalized pencil-of-function method forextracting poles of an EM system from its transient response," IEEE Trans.Antennas Propagat., vol. 37, no. 2, Feb. 1989, pp. 229—234.

[24] T. K. Sarkar and O. Pereira, "Using the matrix pencil method to estimate theparameters of a sum of complex exponentials," IEEE Antennas Propagat.Mag., vol. 37, No. 1, Feb. 1995, pp. 48—55.

[25] D. L. McMakin, D. M. Sheen, and H. D. Collins, "Remote concealed weaponsand explosive detection on people using millimeter-wave holography," 30thAnnual 1996 Int. Carnahan Conference on Security Technology, Oct. 1996, pp.19—25.

[26] P. E. Keller, D. L. McMakin, D. M. Sheen, A. D. McKinnon, and J. W.Summet, "Privacy algorithm for cylindrical holographic weapons surveillancesystem," IEEE AES Systems Mag., Feb. 2000, pp. 17—23.

[27] D. M. Sheen, D. L. McMakin, and T. E. Hall, "Three-dimensionalmillimeter-wave imaging for concealed weapon detection," IEEE Trans.Microw. Theory Tech., vol. 49, no. 9, Sep. 2001, pp. 1581—1592.

[28] D. M. Sheen, D. L. McMakin, and T. E. Hall, "Near field imaging at microwaveand millimeter wave frequencies," IEEE MTT-S Int. Microwave Symp., June2007, pp. 1693—1696.

[29] L. Yujiri, M. Shoucri, and P. Moffa, "Passive millimeter-wave imaging," IEEEMicrowave Mag., Sep. 2003, pp. 39—50.

[30] K. St. J. Murphy, R. Appleby, G. Sinclair, A. McClumpha, K. Tatlock, R.Doney, and I. Hutcheson, "Millimetre wave aviation security scanner," IEEE36th Annual Int. Conf. on Security Technology, 2002, pp. 162—166.

[31] G. Sinclair, R. N. Anderton, and R. Appleby, "Outdoor passive millimetrewave security screening," IEEE 35th Carnahan Conf. on Security Technologies,2001, pp. 172—179.

[32] A. Pergande and L. Anderson, "Video rate millimetre-wave camera forconcealed weapon detection," Proc. SPIE, vol. 4373, 2001, pp. 35—39.

[33] C. A. Martin, S. E. Clark, J. A. Lovberg, and J. A. Galliano, "Real-timewide-field-of-view passive millimetre-wave imaging," Proc. SPIE, vol. 4719,2002, pp. 341—349.


[34] CST Studio Suite, CST Computer Simulation Technology AG,http://www.cst.com/ (Access date: October 20, 2009).

[35] High-Frequency Structure Simulator, Ansoft,http://www.ansoft.com/products/hf/hfss/ (Access date: October 20, 2009).

[36] XFDTD, Remcom, http://www.remcom.com/ (Access date: October 20, 2009).

[37] J. D. Kraus, Antennas, McGraw-Hill, 1988.

[38] R. S. Elliot, "Beamwidth and directivity of large scanning arrays, Part II," TheMicrowave J., pp. 74—82, Jan. 1964.

[39] C. Tai and C. Pereira, "An approximate formula for calculating the directivityof an antenna," IEEE Trans. Antennas Propagat., vol. AP-24, No. 2, March1976, pp. 235-236.

[40] R. F. Harrington, "Effect of antenna size on gain, bandwidth, and efficiency,"J. of Research of the National Bureau of Standards—D, Radio Propagation, vol.64D, No. 1, January-February 1960, pp. 1-12.

[41] A. Skrivernik and J. F. Zürcher, "Electrically small antenna design," Tutorialfrom Proceedings the ACE School on Small Antennas, online,www.its.bldrdoc.gov/isart/art08/slides08/tutorial_skr_a-08.pdf (Access date:October 20, 2009).

[42] Health Canada, Limits of Human Exposure to Radiofrequency ElectromagneticFields in the Frequency Range from 3 kHz to 300 GHz (Safety Code 6).

[43] FCC Regulations Section 15 — Radio Frequency Devices, Subpart F —Ultra-Wideband Operation.




DOCUMENT CONTROL DATA(Security classification of title, body of abstract and indexing annotation must be entered when document is classified)

1. ORIGINATOR (The name and address of the organization preparing thedocument. Organizations for whom the document was prepared, e.g. Centresponsoring a contractor’s report, or tasking agency, are entered in section 8.)

Defence R&D Canada – Ottawa3701 Carling Avenue, Ottawa ON K1A 0Z4, Canada

2. SECURITY CLASSIFICATION (Overallsecurity classification of the documentincluding special warning terms if applicable.)

UNCLASSIFIED

3. TITLE (The complete document title as indicated on the title page. Its classification should be indicated by the appropriateabbreviation (S, C or U) in parentheses after the title.)

Ultrawideband (UWB) high-resolution noise radar for concealed weapon detection:electromagnetic simulation - phase 1

4. AUTHORS (Last name, followed by initials – ranks, titles, etc. not to be used.)

Thayaparan, T.; Nikolova, N.

5. DATE OF PUBLICATION (Month and year of publication ofdocument.)

October 2009

6a. NO. OF PAGES (Totalcontaining information.Include Annexes,Appendices, etc.)

112

6b. NO. OF REFS (Totalcited in document.)

43

7. DESCRIPTIVE NOTES (The category of the document, e.g. technical report, technical note or memorandum. If appropriate, enterthe type of report, e.g. interim, progress, summary, annual or final. Give the inclusive dates when a specific reporting period iscovered.)

Technical Memorandum

8. SPONSORING ACTIVITY (The name of the department project office or laboratory sponsoring the research and development –include address.)

Defence R&D Canada – Ottawa3701 Carling Avenue, Ottawa ON K1A 0Z4, Canada

9a. PROJECT OR GRANT NO. (If appropriate, the applicableresearch and development project or grant number underwhich the document was written. Please specify whetherproject or grant.)

12pz16

9b. CONTRACT NO. (If appropriate, the applicable number underwhich the document was written.)

10a. ORIGINATOR’S DOCUMENT NUMBER (The officialdocument number by which the document is identified by theoriginating activity. This number must be unique to thisdocument.)

DRDC Ottawa TM 2009-190

10b. OTHER DOCUMENT NO(s). (Any other numbers which maybe assigned this document either by the originator or by thesponsor.)

11. DOCUMENT AVAILABILITY (Any limitations on further dissemination of the document, other than those imposed by securityclassification.)( X ) Unlimited distribution( ) Defence departments and defence contractors; further distribution only as approved( ) Defence departments and Canadian defence contractors; further distribution only as approved( ) Government departments and agencies; further distribution only as approved( ) Defence departments; further distribution only as approved( ) Other (please specify):

12. DOCUMENT ANNOUNCEMENT (Any limitation to the bibliographic announcement of this document. This will normally correspondto the Document Availability (11). However, where further distribution (beyond the audience specified in (11)) is possible, a widerannouncement audience may be selected.)

13. ABSTRACT (A brief and factual summary of the document. It may also appear elsewhere in the body of the document itself. It is highlydesirable that the abstract of classified documents be unclassified. Each paragraph of the abstract shall begin with an indication of thesecurity classification of the information in the paragraph (unless the document itself is unclassified) represented as (S), (C), or (U). It isnot necessary to include here abstracts in both official languages unless the text is bilingual.)

This report describes the preliminary simulation results of ultra wideband polarized scatteringfrom various types of concealed weapons carried by a human. Results are shown for both isol-ated weapons as well as weapons in close proximity to a human, simulating a human carryingthe weapon. By selecting the frequency range to cover the region of resonance, i.e., when theweapon size is a wavelength, resonant scattering mechanism provides unique spectral featuresthat can be used for detecting these weapons. Our preliminary simulations show that it is in-deed possible to use the wideband polarized backscatter to identify concealed weapons carriedby humans. The report consists of five parts which address the following topics: 1) literaturesurvey, 2) capabilities of full-wave electromagnetic models: feasibility and preliminary case stud-ies, 3) parametric studies in terms of distance, weapon size, mutual position and polarization, 4)complete system models and their performance in frequency and time domain analyses, and 5)recommendations for system parameters.

14. KEYWORDS, DESCRIPTORS or IDENTIFIERS (Technically meaningful terms or short phrases that characterize a document and couldbe helpful in cataloguing the document. They should be selected so that no security classification is required. Identifiers, such asequipment model designation, trade name, military project code name, geographic location may also be included. If possible keywordsshould be selected from a published thesaurus. e.g. Thesaurus of Engineering and Scientific Terms (TEST) and that thesaurus identified.If it is not possible to select indexing terms which are Unclassified, the classification of each should be indicated as with the title.)

Noise RadarRadar Cross-sectionConcealed WeaponUltra WidebandLPIElectromagnetic ModelCovert detection

Ultrawideband (UWB) high-resolution noiseradar for ... · DRDC Ottawa TM 2009-190 . October 2009. Ultrawideband (UWB) high-resolution noise ... an ultra wideband (UWB) noise radar

Documents