Vous êtes sur la page 1sur 12

Published in Proceedings of the SPIE, Vol.

7309, 73090D (2009)

Suicide bomber detection


Naomi Alexander*a, Carlos Callejeroa, Franco Fioreb, Ignacio Gmeza, Ramn Gonzaloc, lvaro Enrquez de Lunad, Iigo Ederrac, Ins Palaciosc a Alfa Imaging S.A., C/General Pardias 91, 28006 Madrid, Spain; bNATO C3 Agency, Production Capability Area Team 5 ISR, Oude Waalsdorperweg 61, 2597AK, The Hague, Holland; cPublic University of Navarra, Electrical and Electronic Engineering Department, 31006 Pamplona, Navarra, Spain; dSpanish Navy Research and Development Centre, C/Arturo Soria 289, 28033 Madrid, Spain
ABSTRACT
The chance of suicide bomber attacks against troops in the Theatre of Operations is currently quite high. Most of the time checkpoints and compound gates are not equipped with the appropriate equipment to screen for potential suicide bombers. The ultimate solution would be to be able to perform stand-off screening under various weather conditions whilst avoiding contact between Force Protection personnel and potential suicide bombers. Radiation in the millimeterwave and the lower Terahertz range, having the useful property of being able to penetrate clothing in addition to fog and rain, makes it a clear candidate for imaging in this situation. A study has been made simulating real case scenarios to test practical detection performance and stand-off distances at a range of frequencies in this band, the results of which will be presented. Keywords: Millimeter-wave, imaging, explosives, IEDs, weapons

1. INTRODUCTION
Radiation in the millimeter-wave and the lower Terahertz range, having the useful property of being able to penetrate clothing in addition to fog and rain, makes it a clear candidate for imaging under various weather conditions whilst avoiding contact between Force Protection personnel and potential suicide bombers. The objectives of the study presented in this paper are: A comprehensive material study characterizing the transmission and reflection properties - and therefore the losses, dielectric constant and absorption - of the most common clothing, packaging and explosive materials at certain frequency ranges also making use of available trusted data. This study will determine the optimum operating frequency. Images simulating real case scenarios, developed in collaboration with the NATO C3 Agency, will be obtained at a specific range of frequencies in order to test practical detection performance and stand-off distances. Analysis of trial results to determine the ideal imager operating frequency for stand-off explosive and weapon detection and a comprehensive report on the stand-off distances and performance achievable.

1.1 Introduction to characterization measurements The measurements have been made using a ABmm MVNA- 8-350 Network Analyser at the Public University of Navarra. This instrument can use external mixers that make it capable of obtaining results up to the THz region. A transmitting antenna sends an electromagnetic signal to a quasi-optical bench that focuses the beam on the centre of the system with a spot size of about 0.7 times the diameter of the transmitting horn. The ray is received at another antenna of the same characteristics. A reference measurement is taken with no sample and then the samples are inserted in the centre of the quasi-optical system. The transmission and reflection measurements from each sample are compared with the reference measurement and therefore the materials electromagnetic properties can be calculated. These provide information about the transparency of the material as a function of frequency, thus enabling the determination of whether
*nalexander@alfaimaging.com; phone +34918830457; fax +34918827892; alfaimaging.com

Published in Proceedings of the SPIE, Vol.7309, 73090D (2009)

an imaging system will be able to see through these materials to detect an object hidden underneath, or in the case of the object to be detected, how opaque and therefore detectable it is. This measurement system produces ripple due to the reflections in the set-up and in order to eliminate it we need to average the data. The results are not 100% reliable. They usually show a deviation error up to 10% for the higher frequencies. This system also generates typically small resonant peaks. The solid and liquid IED and simulant samples and explosive samples characterized are presented in Table 1.

Table 1. Sample description. Sample no. Material SOLID IEDS & SIMULANTS 7 8 9 12 13 14 15 17 19 21 22 10 11 18 20 23 1 2 3 4 5 6 Charcoal Sugar Sulphur Baby food Magnesium Potassium chlorite Fertilizer pellets Peat (fertilizer) Empty sample holder Flour Plasticine LIQUID IEDS & SIMULANTS Oxygenated water Vinegar Iodine Water Empty holder EXPLOSIVES Goma 2 Eco PG2 TNT Bullet gunpowder Mine gunpowder Pentrite Plastic Plastic Granular powder Small pellets Large, irregular pellets Plastic 0,3 0,3 0,3 0,3 0,3 0,3 Liquid Liquid Liquid Liquid Empty 0,3 0,3 0,3 0,3 0,3 Powder Crystalline granules Powder Dense liquid Powder Powder granules Pellets Soil Empty Powder Malleable plastic 0,3 0,3 0,3 0,3 0,3 0,3 0,3 0,3 0,3 0,3 1,5 Macroscopic Structure Thickness (cm)

The characterization measurements will be presented in section 2. The objective of these measurements is to be able to compare the results with the images obtained. For this reason the approximate transmission value at the frequencies of interest (35, 94 and 220GHz) for each sample is presented. This value is the average within a window of 20GHz bandwidth centred at 35, 94 and 220GHz respectively. Initially the idea was to use the same explosives for the property characterization measurements and for the imaging trials, however, this was not possible in practice. The field trials were undertaken in Army facilities and they also prepared the explosive samples to be imaged. However the police corps in charge of National Security (Guardia Civil) provided and transported the explosive samples characterized at the Public University of Navarra, and therefore not all of the explosives/formats (macroscopic structures) used during the trials were available to characterize. The thickness of the characterized samples could also not be made the same as that of the imaged samples and therefore the characterization results must be considered as a guide to what can be expected in the images rather than compared directly to them.

Published in Proceedings of the SPIE, Vol.7309, 73090D (2009)

1.2 Imaging introduction The imager has an opto-mechanical scanning system that does not use any frequency selective components and can therefore operate with different frequency receivers, as well as having a theoretical 100% transmission. This system has been designed to have a 1K thermal sensitivity and a 6mrad spatial resolution at 94GHz. All images taken are passive, at 35 and 94GHz. The trial period was from 21st April 2008 to 9th May 2008 in the LQCA (Central Armaments Chemical Laboratory) facilities in La Maraosa. One of the project objectives was to take images in different scenarios, which are described below: Standard military shack: 18m long x 5m wide x 4m high. Metal structure with thick chipboard walls and corrugated metal roof. This building was used to simulate an indoor scenario. The shack had several windows along the walls which were covered with Aluminium foil (opaque to mm-waves). Blast walls with different dimensions. Two different concrete structures were tested at first, one 6m tall, 1m wide and 8m long; the other 5m tall, 2m wide and 10m long. The first structure has no roof and the second is covered with a concrete roof. Due to the great similarity of the indoor images to those taken in the covered blast wall scenario it was decided that all of the blast wall scenario mm-wave images should be taken in the first situation described above (uncovered blast walls). Outdoor, open field. This scenario is completely free of buildings, trees and other obstacles.

The threat objects imaged are described in Table 2.

Table 2. Threat objects imaged. Explosive/Simulant/Weapon TATP Tetralite Goma austrogel TNT Ammonium nitrate Hexogen (RDX) Amonitol Hexolite (HBX) PG2 Pentrite (PETN) Goma 2 Eco Ammonal Ammosal Electric primer Ammonium chloride Ammonium sulphate Timer Ceramic knife 300 300 Weight (g) 300 300 1500 200 300 300 300 300 1000 300 300 300 300 Size (width x height x thickness, cm) 15 x 25 x 3.5 12 x 20 x 3 8 x 40 x 4 18 x 8 x 5 15 x 20 x 2 15 x 20 x 3 average 12 x 18 x 3 12 x 18 x 3 27 x 5 x 5 10 x 20 x 3 average 21 x 15 x 2 20 x 13 x 4 20 x 13 x 4 4cm thick 12 x 17 x 3 12 x 17 x 3 9x9x3 27cm long x 5cm wide Granular Granular Plastic bag Plastic bag Macroscopic structure Powder Granular Plastic Granular Granular Solid tablets Granular Granular Plastic Solid tablets Plastic Powder Powder Presentation Plastic bag Plastic bag Plastic bag Rigid plastic cases 2 plastic bags Plastic bag Plastic bag Plastic bag Waxed paper Plastic bag Paper Plastic bag Plastic bag

Published in Proceedings of the SPIE, Vol.7309, 73090D (2009)

Metallic gun Metallic knife Bomber belt 1: water & liquid fertilizer Bomber belt 2: sugar & fertilizer pellets Bomber belt 3: sulphur & flour

22cm long x 4cm wide 32cm long x 5cm wide 34 x 20 x 6 40 x 20 x 6 40 x 20 x 6 Liquids Granular Powder

Plastic bag

PVC pipes, electric cable and paper tape PVC pipes, electric cable and paper tape PVC pipes, electric cable and paper tape

The stand-off imaging distances used for each of the above threats are shown in Table 3, related to the scenarios.

Table 3. Stand-off distances and scenarios. 2.65m Indoors Blast walls Open field 4.6m 6.9m 9.9m

No images are taken in the blast wall scenario at 9.9m due to the physical dimensions of the scenario; no images are taken at 9.9m in the open field scenario as it has been shown that the open sky produces such a large contrast in the image that the small differences between the threats and the subject cannot be seen. The environmental conditions such as the ambient temperature, humidity and weather conditions (cloud cover, sun, wind, rain...) were noted for each image. All the mm-wave images are of a subject, simulating a suicide bomber, with a threat hidden under one of two different fabrics: cotton and wool. These fabrics have been chosen as they are representative of the clothing materials used in the locations this study has been carried out for, it has also been seen in the characterization study that most common clothing materials have similar transmission properties at millimeter wavelengths. It has also been shown that, in general, the threats are detected under up to three layers of these materials. The indoor images have been processed with an automatic threat detection algorithm, which is constantly being updated, and the outdoor images are put though a deblurring and denoising process as no automatic threat detection algorithm has been found so far that achieves satisfactory results. Section 3 shows some example representative images at the closest and furthest stand-off distances taken during these trials.

2. CHARACTERIZATION
The characterization of clothing and packaging materials has been presented in a previous paper [1], the results are summarized in Table 4 for completion.

Table 4. Transmission and attenuation of clothing and packaging material samples around the possible imaging frequencies. Transmission/ Attenuation 1: Embroidered cotton 2: Jeans 40 GHz dB % Trans 0,1 0 103 99 90 GHz dB % Trans -0,4 -0,7 92 84 220 GHz dB % Trans -0,7 -0,7 85 85

Published in Proceedings of the SPIE, Vol.7309, 73090D (2009)

3: Polyester 4: Rain coat 5: Fake leather 6: Tracksuit 7: Cork 8: Cardboard 9: Paper 10: Bubble-wrap 11: Rubbish bag 12: Corrugated cardboard 13: Chipboard 14: Balsa wood 15: Packing foam 16: Wool

0,2 0,2 -0,2 0,2 -0,2 -0,3 0,3 0,4 0,4 -0,8 -1,7 -2,3 -0,1 -0.1

106 106 96 106 96 94 108 110 110 82 68 59 96 96

-0,4 -0,5 -0,7 -0,4 -1,3 -1,5 -0,4 -0,2 -0,3 -1,5 -3,0 -2.0 -0,3 -0.1

91 90 85 91 74 71 92 95 94 72 50 63 94 96

-0,5 -1,2 -1,0 -0,6 -2,2 -1,1 -0,6 -0,1 -0,1 -3,4 -5,0 -4,2 -0,5 -0.3

89 77 79 86 60 77 88 98 97 46 31 38 89 94

The characterization of the solid and liquid IED and simulant samples and explosive samples is summarized in Table 5.

Table 5. Transmission and attenuation of solid and liquid IED and simulant samples and explosive samples around the possible imaging frequencies. Transmission/ Attenuation
SOLID IEDS & SIMULANTS

40 GHz dB % Trans

90 GHz dB % Trans

220 GHz dB % Trans

7: Charcoal 8: Sugar 9: Sulphur 12: Baby food 13: Magnesium 14: Potassium chlorite 15: Fertilizer Pellets 17: Peat (fertilizer) 19: Empty sample holder 21: Flour 22: Plasticine
LIQUID IEDS & SIMULANTS

-8,2 -0,9 -1 -30 -2,8 -1,1 -14,4 -21,2 -0,1 -2,5 -11

15 81 79 0 52 78 4 1 98 56 8

-16,3 -1,5 -1,8 -43,3 -3,1 -2,8 -39,5 -38,3 -0,1 -7 -20

2 71 66 0 49 52 0 0 98 20 1

-40,8 -7,5 -2,4 <60 -4 -7.8 -58 <60 -0,2 -15 -43

0 18 58 0 40 0 0 0 95 3 0

10: Oxygenated water 11: Vinegar 18: Iodine 20: Water 23: Empty sample holder
EXPLOSIVES

-65 -65 -65 -65 -0,1

0 0 0 0 98

-84 -84 -84 -84 -0,1

0 0 0 0 98

-83 -83 -83 -83 -0,2

0 0 0 0 95

1: Goma 2 Eco 2: PG2 3: TNT 4: Bullet gunpowder

-8 -0,2 -0,5 -1

16 95 89 79

-16 -1,1 -1,3 -3

3 78 74 50

-39 -5,8 -5 -12

0 26 32 6

Published in Proceedings of the SPIE, Vol.7309, 73090D (2009)

5: Mine gunpowder 6: Pentrite

-31 -0,5

0 89

-58 -1,3

0 74

-58 -4,2

0 38

In general it is observed that, as is to be expected, the transmission decreases as the frequency is increased. This means that the obscuring materials become more opaque, however the materials to be detected do also, making it easier to detect any of the explosives measured at higher frequencies. Pentrite, TNT and PG2 are expected to be the hardest explosives to detect since they present the highest transmission in this set of samples.

3. IMAGES
3.1 Indoors Figure 1 shows the subject without a threat. This image can be used to compare with those in which the subject carries some sort of hidden threat.

a)

b)

c)

d)

e)

f)

Fig.1. Subject without threat: a) 35GHz, 2.65m, T=20.9C, H=40%, 05/05/08, 11:02; b) 94GHz, 2.65m, T=20.0C, H=46%, 25/04/08, 10:33; c) Visible image of subject at 2.65m; d) 35GHz, 9.9m, T=26.0C, H=42%, 05/05/08, 12:25; e) 94GHz, 9.9m, T=19.4C, H=47%, 23/04/08, 11:20; f) Visible image of subject at 9.9m

In view of figure 1a) and d) and the other 35GHz images taken throughout the trials it is concluded that the increase in spatial resolution at 94GHz is the most important difference between the two frequencies and therefore no further 35GHz images are presented in this subsection. In figures 2a) and c), the subject simulates a suicide bomber with Goma 2 Eco (plastic explosive) hidden under cotton located at 2.65m and 9.9m respectively from the millimeter-wave camera.

Published in Proceedings of the SPIE, Vol.7309, 73090D (2009)

a)

b)

c)

d)

Fig. 2. Subject with Goma 2 Eco covered with cotton: a) 94GHz, 2.65m, T=21.9C, H=44%, 25/04/08, 11:57; c) visible image showing hidden threat; c) 94GHz, 9.9m, T=21.5C, H=44%, 25/04/08, 11:39; d) visible image showing hidden threat

Figure 2 shows how Goma 2 Eco is detected automatically by the system. A high correlation can be observed between the size, shape and position of the threats in the mm-wave and visible images. It should be noted how well the automatic threat detection algorithm works, adjusting to the correct size and position of the threats. Figure 2c) shows the detection at 9.9m, especially clear when compared with the no-threat image at 94GHz (fig. 1e)).

a)

b)

c)

d)

Fig. 3. Subject with TNT covered with cotton: a) 94GHz, 2.65m, T=20.4C, H=46%, 25/04/08, 10:52; b) visible image showing hidden threat; c) 94GHz, 9.9m, T=21.6C, H=44%, 25/04/08, 11:42; d) visible image showing hidden threat

Figure 3a) shows that it is possible to detect TNT (granular explosive) hidden under cotton at 2.65m and 9.9m.

a)

b)

c)

d)

Fig. 4. Suicide bomber belt: fertilizer pellets and sugar covered with cotton: a) 94GHz, 2.65m, T=20.4C, H=46%, 25/04/08, 10:52; b) visible image showing hidden threat; c) 94GHz, 9.9m, T=23.5C, H=44%, 23/04/08, 17:10; d) visible image showing hidden threat

Published in Proceedings of the SPIE, Vol.7309, 73090D (2009)

a)

b)

c)

d)

e)

f)

Fig. 5. Suicide bomber belt: liquid fertilizer and water covered with cotton: a) 94GHz raw image, 2.65m, T=21.4C, H=33%, 28/04/08, 14:46; b) 94GHz processed image, 2.65m, T=21.4C, H=33%, 28/04/08, 14:46; c) visible image showing hidden threat; d) 94GHz raw image, 9.9m, T=24.0C, H=42%, 23/04/08, 17:22; e) 94GHz processed image, 9.9m, T=24.0C, H=42%, 23/04/08, 17:22; f) visible image showing hidden threat

It is interesting to observe that a difference can be seen in figure 4a) between the sugar and the fertilizer pellets (simulant and IED respectively) in the alternate pipes that make up the suicide bomber belt. Figures 4-5 show that the explosive belts are clearly detected at 2.65m and 9.9m. It should be noted that despite the stated detection, the threat is not automatically detected correctly by the algorithm in fig. 5b) due to the location of the threat: it does not fit completely in the field of view at this distance and, due to its large size, the automatic threat detection algorithm has boundary problems. From fig. 4 and 5 it can be seen that the belt containing liquids is whiter than the other, as liquids are much more opaque than the granules in the other belt (that let more of the subjects body radiation through). This has also been observed when comparing with the images of the belt containing powders. 3.2 Blast walls

a)

b)

c)

Published in Proceedings of the SPIE, Vol.7309, 73090D (2009)

d)

e)

f)

g)

h)

i)

Fig. 6. Suicide bomber belt: sulphur and flour covered with cotton: a) 35GHz, 2.65m, T=19.7C, H=36%, 07/05/08, 09:44, image of subject without threat; b) 35GHz, 2.65m, T=19.7C, H=36%, 07/05/08, 09:46; c) visible image showing hidden threat; d) 94GHz, 2.65m, T=23.3C, H=31%, 24/04/08, 16:55, image of subject without threat; e) 94GHz, 2.65m, T=23.1C, H=31%, 24/03/08, 17:53; f) visible image showing hidden threat; g) 94GHz, 6.9m, T=25.2C, H=30%, 24/04/08, 13:46, subject without threat; h) 94GHz, 6.9m, T=21.9C, H=33%, 24/04/08, 19:13; f) visible image showing hidden threat

Due to the large size of these kinds of threats, the detection in fig. 6. is clear and no comparison with the non-threat image (figs. 6a) and g)) is needed. It has, however, throughout the trials been seen to becomes necessary as the threats decrease in size and the stand-off distance is increased. A difference in the grey levels between two consecutive PVC pipes in the same belt can be seen in fig. 6b) and e). This is due to the fact that each belt is made up of two different substances in alternate pipes. The 35GHz image is included here as it has come out well, however this is not generally the case - especially as the stand-off distance is increased - and therefore, as in the previous scenario, the 94GHz images are considered more useful. 3.3 Open field

a)

b)

c)

Published in Proceedings of the SPIE, Vol.7309, 73090D (2009)

d)

e)

f)

Fig. 7. Suicide bomber belt: fertilizer pellets and sugar covered with cotton: a) 94GHz, 2.65m, T=32.1C, H=25%, 24/04/08, 12:53, subject without threat; b) 94GHz, 2.65m, T=22.2C, H=25%, 30/04/08, 13:58; c) visible image showing hidden threat; d) 94GHz, 6.9m, T=23.3C, H=43%, 24/04/08, 11:43; e) 94GHz, 6.9m, T=22.2C, H=25%, 30/04/08, 13:18; f) visible image showing hidden threat

Fig. 7b) shows that it is possible to detect the suicide bomber belt close-up under these conditions without need for comparison with the no-threat image, however as the distance increases a comparison becomes necessary (see fig. 7e)).

4. CONCLUSIONS
The following conclusions are those resulting from the complete study described in Section 1, some summaries/examples of which are presented in Sections 2 and 3. It has been shown both experimentally and theoretically that as the operating frequency is increased the detection capability increases. Thus, out of the two operating frequencies tested here, the optimum frequency is 94GHz. This is also explained by the following: The spatial resolution worsens as the frequency decreases. Thus the image quality increases with frequency and, in principle, the highest operating frequency will be the optimum. However, there is a limit to this rule as the transparency of the obscuring materials decreases with increasing frequency and a point will be reached when the hidden threats are no longer seen through the obscurants. As mentioned above, by observing the different material characterizations, it can be seen that the transmission increases with decreasing frequency. Considering the threats to be detected this means that they are more transparent at 35GHz than at 94GHz, thus making them more difficult to detect. The indoor detection capability is very high. A high detection probability is achieved for almost all of the imaged threats in indoor situations with little or no radiation incoming from the cold sky. Eight in every ten threats in these trials have been detected at every stand-off distance. In this situation, as has been noted previously, the detection limit is determined mainly by the imager resolution. The minimum object that can be detected with the imager used in these trials at a distance of d (metres) is approximately of size:

Size(m) = 6.5 10 3 d
Thus the optimum indoor stand-off distance will depend on the operating requirements, always taking into account the system spatial resolution. In outdoor situations the threat detection capability without comparing threat and no-threat images is low - less than 50% in the open field and less than 30% in the blast walls. These probabilities increase noticeably when a comparison between the threat and no-threat image is made. However, this solution is not very practical as it is not usual to have a no-threat mm-wave image of every subject. Several measures that could help in these situations are proposed: Extensive imager operator training.

Published in Proceedings of the SPIE, Vol.7309, 73090D (2009)

A standard set of no-threat images could be taken of subjects with different complexions which could be used for the comparison (choosing the standard complexion most similar to that of the subject in question) in the cases where the no-threat image for each subject is not available. Another measure that is currently being studied is that of implementing an automatic threat detection algorithm that can segment the subject from the background and then look for the threats on the subject only. Once real-time mm-wave images are available the reflections will appear and disappear as the subject moves. In the case of the subject hiding a threat, it will always be present in the image. In the blast wall situation the graphs show that the smaller the stand-off distance the better the results. This is mainly due to the scenario, in which the walls reflect both the cold sky and the subject, degrading the image and complicating the detection. The proposed solution is to cover the passage-way between the blast walls with an opaque material (for mm-waves) such as a sheet of metal. This structure does not need to be robust, its only job is to keep out the cold sky radiation. When the blast wall scenario was being chosen, the covered structure produced images that were very similar to those taken indoors. Thus, covering the blast wall scenario with an opaque material should increase the detection probability to the values obtained in the indoor scenario, with the consequent constant probability over distance. In the open field scenario it can be seen that the optimum stand-off distance is that closest to the imager. However, threats containing metal can be detected at larger stand-off distances in this situation. It is thus thought that explosives containing shrapnel will also be detected at larger stand-off distances. Despite this, the detection probabilities will be much lower than in indoor situations. One possible way to increase the detection probabilities is to work with horizontal polarization. It has been shown how this polarization increases the outdoor mm-wave image quality and therefore the detection capability will also increase. Other measures for increasing the detection probabilities have been described above. However, this study has not identified a solution that will increase the detection probabilities to those achieved indoors. In terms of the threats: the plastic and powder/granular explosives are detected with a high probability in nearly all of the scenarios, except for tetralite and hexolite, which have a low probability. The hexogen and pentrite samples, that were in tablet form, have completely different performances as hexogen is detected in most of the scenarios, whereas pentrite is detected in hardly any of them. In general it can be said that the explosives with the highest detection probability and with the most similar behaviours are the plastic explosives and the ones in powder form. It has been observed that there are three explosives that have a very low detection probability: pentrite, tetralite and hexolite. There is no apparent relationship between these explosives to be able to conclude which types cannot be detected. It has been seen that the detection capability is related to the material transmission, which depends on the macroscopic (scattering) and microscopic (absorption) structures. The macroscopic structure of the hexolite sample does not seem to be the cause as it is similar to that of the amonitol sample, which can be detected. The same can be deduced in the pentrite case, which has been characterized in powder/granule form but has been imaged in tablet form (and in both cases has been seen to have a high transmission). Based on these arguments it is reasoned that the no detection of the explosive (with detection probabilities below 50%) is due to some microscopic property, the study of which is beyond the scope of this project. As mentioned above, it has been seen that pentrite has a very high transmission at the frequencies studied, this means the explosive is transparent and cannot be detected. By studying the characterization in detail it seems that this explosive could be detected at a higher frequency. In the cases of hexolite and tetralite, as there is no characterization data, it is not known exactly why they are not detected in the mm-wave images, however the results seem to indicate they are in the same category as the pentrite sample. Another important conclusion is that the ceramic materials, such as the knife, cannot be detected. A detailed study of this has been undertaken, as it is generally stated that mm-waves can detect ceramic weapons. It is concluded that ceramic materials have a high transmission which makes them undetectable in most operating conditions at the frequencies studied here. However, on analysing the data there appears to be a

Published in Proceedings of the SPIE, Vol.7309, 73090D (2009)

tendency for the transmission to decrease with increasing frequency which implies that these threats could be detected at a higher frequency. It is also important to note that the hidden threats were generally small in size and thin. The larger the surface area of the threat, the more pixels that represent it in the image and thus the easier it is to detect. This can be seen in the case of the suicide bomber belts which, due to their large size, are always detected, irrespective of the stand-off distance and scenario. The experts at the L.Q.C.A. pointed out that suicide bombers use the largest amount of explosive possible, with a large amount of shrapnel, which means that in a real case the detection capability will be noticeably increased. With regards to the thickness, the thicker the threat the smaller its transmission and therefore the easier to detect. The majority of samples used throughout the trials were 2-4cm thick. The liquid threats have a high detection probability due to their low transmission and high absorption. The metallic threats have a high detection probability. This is due to their high reflectivity and near-zero transmission. Thus indoors the metallic threats reflect the background temperature, which is usually different to that of the subject and therefore the threats are detected, and outdoors, when the threats reflect the cold sky radiation, they are easy to detect. The experts at L.Q.C.A. pointed out that suicide bombers usually carry a large amount of metallic shrapnel in order to increase the damage caused by their attacks. This could increase the detection probability of the explosives as the shrapnel will be easier to detect. With the mm-wave imager used in this study and the current image processing techniques it is not possible to identify the threats. The identification of a threat consists in being able to detect it and determine what the substance is. It may be possible to identify threats with more powerful algorithms, at least to be able to distinguish metals from dielectrics. Comparing the results from the mm-wave image analysis with those from the threat characterization it is observed that there is a high correlation between the two. From the characterization, the explosives with the lowest transmission values (easier to detect) are Goma 2 Eco and Goma austrogel (this explosive was not characterized, however, the experts at the L.Q.C.A. pointed out that there is very little difference between the two), followed by PG2, then TNT and the most transmitting the pentrite sample. The only difference in the results is with the TNT and the PG2 samples. According to the characterization, TNT transmits slightly more than PG2. However, the detection probability of the TNT sample was slightly higher than that of the PG2 sample. This difference could be due to the packaging of the TNT in the imaging trials. The only atmospheric condition that severely affects the mm-wave images in these trials is the indoor room temperature, which decreases the temperature contrast and thus the detection probability. The humidity has not varied much throughout the course of these trials, however it has been observed in other studies that it has no effect over the stand-off ranges studied. The same has been observed for rain and fog.

5. REFERENCES
[1] Alexander,N.E., Callejero Andrs, C., Gonzalo,R., Multispectral mm-wave imaging: materials and images, Proc. SPIE 6948, 694803 (2008).

ACKNOWLEDGEMENTS
We would like to thank the LQCA for the preparation of the explosive samples, the use of their facilities and all their help throughout the trial period. We would also like to thank the Guardia Civil for providing and transporting the explosive samples for the characterization and Cristian Coman from the NC3A for his help and advice throughout the trials.

Vous aimerez peut-être aussi