- Email: [email protected]
Comparison of neutron and high-energy X-ray dual-beam radiography for air cargo inspection
ARTICLE IN PRESS
Applied Radiation and Isotopes 66 (2008) 463–473 www.elsevier.com/locate/apradiso
Comparison of neutron and high-energy X-ray dual-beam radiography for air cargo inspection Y. Liu, B.D. Sowerby, J.R. Tickner CSIRO Minerals, Private Mail Bag 5, Menai, NSW 2234, Australia Received 24 September 2007; received in revised form 15 October 2007; accepted 16 October 2007
Abstract Dual-beam radiography techniques utilising various combinations of high-energy X-rays and neutrons are attractive for screening bulk cargo for contraband such as narcotics and explosives. Dual-beam radiography is an important enhancement to conventional single-beam X-ray radiography systems in that it provides additional information on the composition of the object being imaged. By comparing the attenuations of transmitted dual high-energy beams, it is possible to build a 2D image, colour coded to indicate material. Only high-energy X-rays, gamma-rays and neutrons have the required penetration to screen cargo containers. This paper reviews recent developments and applications of dual-beam radiography for air cargo inspection. These developments include dual high-energy X-ray techniques as well as fast neutron and gamma-ray (or X-ray) radiography systems. High-energy X-ray systems have the advantage of generally better penetration than neutron systems, depending on the material being interrogated. However, neutron systems have the advantage of much better sensitivity to material composition compared to dual high-energy X-ray techniques. In particular, fast neutron radiography offers the potential to discriminate between various classes of organic material, unlike dual energy X-ray techniques that realistically only offer the ability to discriminate between organic and metal objects. r 2007 Elsevier Ltd. All rights reserved. Keywords: Dual-beam radiography; Fast neutron radiography; X-ray radiography; Gamma-ray radiography; Air cargo; Contraband
1. Introduction There is a worldwide need for improved methods for the screening of air cargo for contraband. The main objectives are the detection of contraband such as illicit drugs, explosives, weapons and nuclear materials and the verification of declared manifests. Air cargo is usually packaged into lightweight aluminium containers called unit load devices (ULDs). The most common ULDs have a width of about 1.6 m and gross weight of up to 1.6 tonnes. However, pallet ULDs are significantly larger, with widths up to 2.4 m and maximum gross weights of up to 6.8 tonnes. For comparison, the most common sea container is the 40 ft container of dimensions 12.2 m (length) 2.4 m (width) 2.6 m (height) and maximum gross weight of 32.5 tonnes. The air cargo ULD content is commonly limited by its mass, not its Corresponding author. Tel.: +61 2 9710 6719; fax: +61 2 9710 6789.
E-mail address: [email protected] (B.D. Sowerby). 0969-8043/$ - see front matter r 2007 Elsevier Ltd. All rights reserved. doi:10.1016/j.apradiso.2007.10.005
volume, and the load is heterogeneously distributed. Material thicknesses (mass-per-unit area scanned) up to 100 g cm2 can be encountered. The requirements for a mass-screening system for air cargo are complex and demanding and include:
screening cargo containers without unpacking in 1–2 min; producing high quality, high-resolution (5–10 mm or better) images that can be readily interpreted by operators; differentiating between organic and inorganic materials and ideally discriminating between different classes of organic materials such as explosives, narcotics and benign substances; distinguishing ‘ordinary’ metals such as aluminium, iron, steel and copper from heavy metals such as lead, tungsten, uranium and plutonium that may be associated with smuggled nuclear materials;
ARTICLE IN PRESS 464
Y. Liu et al. / Applied Radiation and Isotopes 66 (2008) 463–473
compliance with strict radiation safety requirements for both operating staff and cargo; suitability for continuous operation in an airport environment; and availability at reasonable cost.
The majority of cargo inspection systems installed today use single-beam high-energy X-ray or gamma-ray radiography to form a 2D (scan depth integrated) transmission image of the cargo’s contents. The image is analysed by a human operator, who is required to make a decision in a few minutes as to whether the container should be opened for manual inspection. Image quality is crucial in assisting the operator to reach the correct decision and recent X-ray scanner developments have focussed on improving the resolution and contrast of the collected images. Highenergy X-ray or gamma-ray radiography is primarily sensitive to density and is particularly effective in detecting hidden objects with readily identifiable shapes such as firearms and other weapons. However, single-energy radiography provides no material discrimination and is less well suited to finding low-density organic threats that lack a definite shape. The requirement to determine both shape and material composition points to imaging cargo using two different types of radiation. From the two measurements, it may then be possible to calculate an image showing both massper-unit area and composition. For this approach to be effective both types of radiation must be sufficiently penetrating to be detectable after passing through the cargo and the attenuation of both types of radiation in different materials must be distinguishable. Only fast (high energy) neutrons and high-energy X-rays or gamma-rays have the required penetration for imaging the contents of most air cargo containers. The ranges of even the most energetic charged particles that could be obtained from a practical accelerator suitable for industrial use are at most a few g cm2. For example, the ranges of 10 MeV electrons and protons in water are 5.0 and 0.12 cm, respectively. The first aim of any dual-beam radiography technique is to determine separately the mass-per-unit area and a measure of material composition for each pixel in the final 2D image. Ideally, these two quantities should be independent, so that the mass-per-unit area determination does not depend on the material composition and vice versa. The types of radiation used need to comply with relevant regulations, particularly with regard to the irradiation of foods commonly present in air cargo. For example, in the USA and UK, to be exempt from these regulations the dose imparted to the cargo must not exceed 0.01 Gy in the case of inspection devices that utilise neutrons and 0.5 Gy otherwise. In these countries, the maximum irradiation energy is 10 MeV in the case of X-rays and 14 MeV in the case of neutrons. Note that these regulations differ between countries.
In the present paper, two dual-beam radiography systems, one based on X-rays alone and the other on fast neutrons and gamma-rays (or X-rays), are assessed, discussed and compared for application to the screening of consolidated airfreight. The big advantage of dual-beam systems is that they provide additional information on the composition of the object being imaged. By comparing the attenuations of transmitted dual beams, it is possible to build a 2D image, colour coded to indicate average composition integrated along the beams of each image pixel. Some results from a specific implementation of a fast neutron and gamma-ray radiography (FNGR) system for air cargo scanning at Brisbane International Airport in Australia are first analysed to determine the likely penetration depth requirements for a dual-beam radiography system of air cargo. 2. Characterisation of air cargo at Brisbane International Airport Air cargo container sizes vary widely and contents range from single commodity loads through to consolidations containing many hundreds of different items. Air cargo containers are generally more cluttered than sea cargo containers. In late 2006, a detailed statistical survey was carried out to determine the distribution of loads in cargo passing through the commercial prototype FNGR scanner that was installed at an Australian Customs Service facility at Brisbane International Airport (Eberhardt et al., 2006). The scanner facility incorporated an extensive automated cargo handling system, making possible the rapid scanning of large numbers of containers. The scanner facility was trialled by Australian Customs over 12–18 months at Brisbane International Airport. Per-pixel attenuation histograms for 14 MeV neutron and 60Co gamma-ray (1.17 and 1.33 MeV) transmissions were calculated for 940 cargo images collected over a period of 3 weeks. A more detailed manual analysis was made of 202 cargo images collected over a 4-day window inside this period. This detailed analysis included additional information on the ULD type and manifest details (where available). The breakdown by ULD type is shown in Fig. 1. Detailed descriptions and drawings of the different ULD types are available from a number of sources on the web (e.g., Air Freight Council of NSW, 2007). The commonest observed ULDs are AKEs with dimensions of 1530 (width) 2010 (length) 1630 (height) mm. The commonest observed pallets are the PMCs and PAGs with dimensions of 2440(2240 for PAG) 3180 1630 mm. Cargos are scanned through the narrower (width) dimension. Overall, about 50% of cargo was loaded into ULDs and 50% on pallets. From the per-pixel gamma-ray attenuation histograms, it is possible to calculate the distribution in mass-per-unit area for different classes of cargo. In this analysis, measured mass-per-unit areas of less than 2 g cm2 are ignored and a material independent mass-attenuation
ARTICLE IN PRESS Y. Liu et al. / Applied Radiation and Isotopes 66 (2008) 463–473
coefficient of 0.058 cm2 g1 is assumed. Fig. 2 shows the simple and cumulative distributions in mass-per-unit area estimated for cargos classified as ‘‘light’’ and ‘‘heavy’’ in the manual analysis, along with the average distributions for all cargos. The limited number of bits used to record the attenuation histograms is responsible for the cut-off in the curves around 106 g cm2, but the fraction of pixels above this level is very small. A scanner using radiation capable of penetrating through 100 g cm2 would be able to image through 99.6% of the area of an average air cargo ULD at Brisbane airport. A scanner using radiation capable of
Fig. 1. Breakdown of 202 incoming air cargo containers by ULD and pallet types at Brisbane International Airport over a 4-day period in late 2006. The ULD identifiers commencing with A and D indicate aluminium containers and those commencing with P are larger pallet cargo. The most common AKE containers are half-width lower deck containers used on wide body aircraft.
465
penetrating through only 50 g cm2 would be able to image through 87% of the area of an average air cargo ULD at Brisbane airport. It is a reasonable hypothesis that similar penetration depths would apply to imaging of air cargo ULDs at airports around the world. 3. Dual-energy X-ray methods 3.1. Background Dual-energy X-ray radiography is an important enhancement to X-ray systems that provides additional information on the composition of the object being imaged. By comparing the attenuations of transmitted high- and low-energy X-ray beams, it is possible to build a 2D image, colour coded to indicate metal and organic materials. Dual energy X-ray systems can utilise the Z dependence of the mass-attenuation coefficients at either low energies (due to the photoelectric effect) or high energies (due to pair production) (Fig. 3(a)). In applications such as examining passenger luggage, dual-energy X-ray radiography using low-energy X-rays has become the usual screening method. However, the limited penetration of the lower energy X-rays (around 50–100 keV) used to provide composition information prevents the method being used effectively on consolidated airfreight. To illustrate this, Fig. 3(b) shows the thickness of various materials for 0.1% X-ray transmission as a function of X-ray energy, calculated using mass-attenuation coefficients from NIST (2007). For energies above about 1 MeV, X-rays are sufficiently penetrating to pass through
Fig. 2. Cumulative (log scale, left-hand axis) and simple (linear scale, right-hand axis) distributions of image pixels as a function of mass-per-unit area for incoming air cargo at Brisbane International Airport. The ‘‘All Cargo’’ data shown is for all 940 incoming ULD/pallet types scanned over a 3-week period. Pixels with measured mass-per-unit area less than 2.0 g cm2 were excluded from calculation.
ARTICLE IN PRESS 466
Y. Liu et al. / Applied Radiation and Isotopes 66 (2008) 463–473
3.2. Calculations for the determination of composition For the ideal case of narrow-beam transmission of monoenergetic radiation, the transmissions t1,2 of two X-ray beams 1,2 of different energy can be written as ti ¼
Ii ¼ expðmi X Þ ði ¼ 1; 2Þ, I 0i
(1)
where Ii is the measured strength of the transmitted radiation with the cargo in position, I 0i is the measured strength with no cargo in the beam, mi is the massattenuation coefficient, and X is the mass-per-unit area. If the energy of the first radiation beam is selected such that the mass-attenuation coefficient is a weak function of material, as is the case for X-rays with energies in the range 1–3 MeV, then the two transmission equations can be solved to give: X ¼
R¼
Fig. 3. (a) X-ray and gamma-ray mass-attenuation coefficients (R-values) (cm2 g1) for various materials over the energy region 50 keV to 10 MeV. (b) Thickness of the same materials for 0.1% transmission plotted as a function of gamma-ray energy.
100 g cm2 of all materials. The mass-attenuation coefficients for the heaviest elements (Pb and U) begin to differ significantly at energies below about 500 keV and the lighter metals (Al and Fe) can be distinguished at energies below about 100–150 keV. At these low energies, however, penetration is limited to 30–50 g cm2 for the lightest materials and less than 10–20 g cm2 for the high-atomic number elements. Therefore, the application of conventional dual energy X-ray scanners based on X-ray tube sources (up to about 450 keV) is limited to cargo with low mass-per-unit area primarily because of the very limited penetration of the low-energy X-rays (100–200 keV) required for material discrimination. At higher energies above about 4–5 MeV, it again becomes possible to distinguish different materials. Penetrations at these high energies are excellent. Dual highenergy X-ray radiography for material discrimination is discussed below.
1 I1 log 0 , m1 I1
m2 logðI 2 =I 02 Þ ¼ . m1 logðI 1 =I 01 Þ
(2)
(3)
Here R is just the ratio of mass-attenuation coefficients that, for the appropriate choice of X-ray energies, is diagnostic of the material present. Unfortunately, convenient monoenergetic X-ray or gamma-ray sources are not available for energies above 1–2 MeV. Instead, high-energy Bremsstrahlung sources using electron accelerators with accelerating voltages typically up to 9 MV must be used, which produce broadspectrum radiation with energies up to the accelerating potential. For broad-spectrum radiation produced by such a source, Eq. (1) needs to be replaced with R E MAX fðEÞ exp mi ðEÞX Þ rðEÞ dE Ii 0 ti ¼ 0 ¼ ði ¼ 1; 2Þ, R E MAX Ii fðEÞrðEÞ dE 0 (4) where the integrals extend over all radiation energies produced by the source, f(E) is the energy spectrum of X-rays emitted by the source and r(E) is the relative detector response as a function of X-ray energy. Because X-ray mass-attenuation coefficients vary strongly with energy, the attenuation of broad-spectrum radiation is no longer strictly exponential and the simple solutions for X and R given in Eqs. (2) and (3) do not strictly apply. Nevertheless, with appropriate choices of X-ray energies these relations provide a good starting point for the analysis of dual high-energy X-ray transmission images.1 1 Other authors have taken and composition information measurements, for example, (1997). The problem reduces
different approaches to extracting density from dual high-energy X-ray transmission the ‘banana-plots’ of Rushbrooke et al. to finding functions fX and fR such that
ARTICLE IN PRESS Y. Liu et al. / Applied Radiation and Isotopes 66 (2008) 463–473
467
photon depends on both the probability of the X-ray interacting in the detector element and the energy that it deposits if it does interact. We have estimated these factors for a small CsI scintillator element using a simple Monte Carlo simulation based on the EGSnrc (Kawrakov and Rogers, 2006) code. We find that the energy dependence of the detector response r(E) appearing in Eq. (4) can be well represented above energies of 0.2 MeV by the functional form rðEÞ / 1 þ 0:535E 1:5 .
Fig. 4. Bremsstrahlung X-ray spectra with accelerating voltages and endpoint energies of 4, 5, 6, 7, 8 and 9 MeV. Spectra have been normalised to have the same peak intensity.
(5)
Utilising this expression for r(E) in Eq. (4), the transmissions of 5 and 9 MeV Bremsstrahlung through polythene, carbon, aluminium, iron lead and uranium have been calculated. R-values calculated using Eq. (3) are plotted as a function of material thickness in Fig. 5 for two cases, namely (a) no pre-filtration of the Bremsstrahlung beam and (b) pre-filtration with 30 g cm2 of lead to remove low-energy X-rays.
Fig. 4 plots the energy spectrum of Bremsstrahlung sources with accelerating voltages of 4, 5, 6, 7, 8 and 9 MeV, calculated using the simple analytical model of Findlay (1989). A 3 g cm2 tantalum target is assumed and the spectra have been normalised to have the same peak intensity. To determine the X-ray transmission at two different energies, or more practically over two different energy ranges, it is necessary to either (i) discriminate the energies of individual X-ray quanta hitting the detector system, (ii) use two detector arrays, filtered so that one preferentially responds to higher energy and one to lower energy quanta or (iii) collect separate measurements at different accelerating voltages, using interlaced energies on alternate accelerator pulses. Approach (i) is technically very difficult, due to the high intensity and low-duty cycle (of order 103; Mishin, 2005) of the RF linear accelerators typically used as X-ray sources. A group at Cambridge Imaging implemented this approach using multiple scintillation crystals for each detector element (Rushbrooke et al., 1997). Approach (ii) would be difficult to implement because of the relatively small differences in mass-attenuation coefficients above a few MeV. Approach (iii) is the most widely used (Ogorodnikov and Petrunin, 2002; Bjorkholm, 2004; Chen and Wang, 2006) and is discussed in more detail below. Generally, with the interlaced dual energy pulsed-source technique the first end-point energy is chosen to be in the range 4–6 MeV and the second 8–9 MeV. The detector arrays measure the total energy deposited in each detector element per pulse. The average energy deposited per X-ray (footnote continued) X ¼ fX(t1, t2) and R ¼ fR(t1, t2) such that X depends only on mass-per-unit area and R depends only on (average) material composition.
Fig. 5. R-values based on ratios of the transmission of 5 and 9 MV Bremsstrahlung radiation, plotted as a function of material thickness for various materials for (a) no pre-filtration of the Bremsstrahlung beam and (b) pre-filtration with 30 g cm2 of lead.
ARTICLE IN PRESS 468
Y. Liu et al. / Applied Radiation and Isotopes 66 (2008) 463–473
It can be seen at once that, as expected, the calculated R-values are not independent of material thickness. More seriously, for no pre-filtration and material thicknesses below about 20 g cm2 the R-values are not unique, nor monotonic with increasing atomic number. This is due to the different attenuation mechanisms of low- and highenergy X-rays in materials of different atomic number. At relatively low energies, where Compton scattering dominates the total cross-section, the low-atomic number materials have the highest mass attenuations. At high energies, where pair production becomes increasingly important, the high-atomic number materials have the highest mass attenuations. On the basis of this analysis, unique solutions for dual X-ray R-values (and associated material composition) at low-material thickness appears to require additional information. These results point to the importance of pre-filtering the Bremsstrahlung radiation to remove the lower energy component. When the Bremsstrahlung beams are pre-filtered with 30 g cm2 of lead the R-values are still not constant with thickness but a relatively simple empirical function could be used to correct the measured R- and thickness-values back to a standard mass-per-unit area, say 30 g cm2. 3.3. Dual high-energy X-ray cargo inspection systems The development and application of dual energy highenergy X-ray scanners has been reported by a number of authors. A summary of some of the features of these scanners is given in Table 1. Unfortunately, only very limited technical information is available on the Varian and Nuctech systems. The scanners developed by the Efremov Scientific Institute, Varian and Nuctech all image a cargo container using successive pulses of high- and low-energy polychromatic X-ray beams. Typical X-ray Bremsstrahlung energy spectra from the X-ray sources used are shown in Fig. 4. In this situation, non-unique compositional information is a possibility unless pre-filtering of low-energy X-rays is used. The high energy is selected so that the pair production cross-section accounts for a significant proportion of the attenuation. The high energy is usually in the 8–9 MeV
range. The lower energy is selected to provide reduced pair production attenuation and at the same time to give a sufficiently high flux. Generally, energies in the range 4–6 MeV are selected. These three systems use a linear array of detectors, typically with a pitch or spacing of 3.5–5 mm. The system developed by Cambridge Imaging Ltd. uses a different approach. The X-ray source used was a 10 MeV linac operated at a single energy. The energy discrimination was provided by the detector system although the details are not disclosed (Rushbrooke et al., 1997). A number of optional detector systems are discussed by Neale et al. (1996). These detection systems are complex but they do offer the possibility of preferentially measuring the transmitted X-rays that produce annihilation radiation, therefore improving signal-to-noise compared to the interlaced pulsing methods used by the other authors. 4. Neutron radiography techniques 4.1. Background The limitations of X-ray systems for the detection of organics-based explosives and drugs have stimulated the development of alternative methods for cargo inspection including those based on neutrons. Neutron radiography can be performed using either fast neutrons, with energies of a few hundred keV or larger, or using thermal neutrons with energies around 0.025 eV. Thermal neutron crosssections vary by more than 7 orders of magnitude, from about 104 b for oxygen through to 46,617 b for gadolinium. Other elements with significant thermal crosssections include lithium, boron, cadmium and erbium. The wide variation in cross-section, which correlates poorly with the threat detection requirements outlined in Section 1, coupled with the lack of practical sources, makes thermal neutron radiography ill-suited to cargo screening. Fast neutron techniques are attractive for cargo inspection as they have the required penetration, they can be used to determine elemental composition and they interact with matter in a manner complementary to X-rays. X-rays and gamma-rays scatter primarily from
Table 1 Summary of published information on some dual high-energy X-ray radiography systems for cargo inspection Cambridge Imaging Ltd.
Efremov Scientific Institute
Varian
Nuctech
References
Rushbrooke et al. (1997)
X-ray source
10 MeV linac; pulsed mode 200 Hz 100# elements; 5 5 mm2 Optical fibre/CCD Laboratory trial conducted
Ogorodnikov and Petrunin (2002) Pulsed linac; 8/4 MeV interlaced mode 100 Hz, 5 msec CdWO4; 3.5 mm pitch PIN diodes Laboratory trial conducted
Bjorkholm (2004), Budner (2006) Pulsed linac; 9/5 MeV interlaced mode 200 Hz 5 mm typical
Chen and Wang (2006), Tang et al. (2006) Pulsed linac; 9/6 MeV interlaced mode Pulse separation o4 ms
Laboratory trial conducted
Commercial unit
Pulse rate and duration Detector type and size Read-out system Status
ARTICLE IN PRESS Y. Liu et al. / Applied Radiation and Isotopes 66 (2008) 463–473
electrons whereas fast neutrons scatter from the protons and neutrons in the nucleus. As a result neutrons are scattered and slowed more by low-atomic number (Z) elements than by the high-Z elements. Convenient sources of neutrons include sealed tube generators using the deuterium–deuterium (DD), deuterium–tritium (DT) and tritium–tritium (TT) fusion reactions, higher energy accelerators with beam energies generally above 1 MeV using a variety of deuterium and proton induced reactions, radioisotope sources and photoneutron sources using Bremsstrahlung radiation from high-energy electron accelerators. Table 2 summarises the main properties of these sources. All of these sources produce fast neutrons with energies of a few hundred keV or greater. Fig. 6(a) plots the neutron mass-attenuation coefficients as a function of energy for the same materials as previously investigated for dual-energy X-ray radiography. There is significant variation from material to material and the mass-attenuation coefficients for some materials also show pronounced resonances. Fig. 6(b) shows the thicknesses of various materials through which 0.1% of neutrons are transmitted from DD or DT sources. Results for highenergy accelerator-based sources and 241Am-Be, 252Cf sources will be broadly similar, depending on the exact neutron energy spectrum produced. Fast neutrons have similar penetrations to 1–2 MeV X-rays for aluminium; they are significantly more penetrating for heavier elements and less penetrating for organic materials.
469
uranium calculated for DD and DT neutrons and (average energy of 1.25 MeV) gamma-rays.
60
Co
4.3. Fast neutron radiography cargo inspection systems As for dual energy X-ray techniques, material discrimination using fast neutron radiography relies on determining
4.2. Determination of composition For a monoenergetic neutron source (DD or DT generators) and a monoenergetic gamma-ray source, Eqs. (2) and (3) (Section 3.2) can be used to determine a cross-section ratio R and a material density X independent of material thickness. In this case, the subscript 1 refers to the gamma-ray measurement and 2 to the neutron measurement. If either the neutron or X-ray source is not monoenergetic, then Eq. (4) has to be used to calculate the transmission and the R-value determined using Eq. (3) can be expected to vary somewhat with material thickness. For the simpler monoenergetic case, Table 3 shows the R-values for polythene (also known as polyethylene, formula (C2H4)n), graphite, aluminium, iron, lead and
Fig. 6. (a) Neutron mass-attenuation coefficients versus energy for various materials. (b) Thicknesses of the same materials through which 0.1% of DD neutrons, or DT neutrons, 60Co gamma-rays (1.17 and 1.32 MeV) and X-rays from a 5 MV Bremsstrahlung source are transmitted.
Table 2 Properties of some available fast neutron sources Source
Beam energy
Reaction
Neutron energy (MeV)
Maximum output (s1)
Sealed tube, DD Sealed tube, DT Sealed tube, TT Accelerator 241 Am-Be 252 Cf Photoneutron
o200 keV o200 keV o200 keV 1–6 MeV Radioisotope Radioisotope 44 MeV
2
2.5 14 0–9 Varies 0–12 0–10 Varies
108 1010 2 108 1010 5 107 109 1010–11
H(d,n)3He H(d,n)4He 3 H(t,2n)4He Various 9 Be(a,n)12C Fission (g,n) 3
ARTICLE IN PRESS Y. Liu et al. / Applied Radiation and Isotopes 66 (2008) 463–473
470
Table 3 Calculated R-values for DD/DT neutrons and
5. Comparison of dual-beam radiography techniques 60
Co gamma-rays
Material
R-value (DD/60Co)
R-value (DT/60Co)
Polythene Graphite Aluminium Iron Lead Uranium
4.496 1.404 0.852 0.725 0.295 0.305
1.781 1.153 0.712 0.526 0.267 0.237
mass-attenuation coefficients of fast neutrons and gammarays or X-rays. Various alternative neutron radiography systems have been reviewed (Gozani, 2002; Buffler, 2004; Sowerby and Tickner, 2007). Some systems (Miller et al., 1996) are attempting to use radiography with broadenergy spectrum or ‘white’ neutron sources, usually coupled with some means of determining the energy of the transmitted neutrons, to identify specific elements through their characteristic resonances. Whilst potentially providing powerful material discrimination, these approaches require complex sources and detector systems and currently remain at the laboratory trial stage. The alternative approach is to perform ‘total-beam’ radiography, performing a single measurement of the attenuation of the entire spectrum of neutrons emitted by the source. Either a monoenergetic DD or DT source (Eberhardt et al., 2005, 2006) or a broad-energy radioisotope source (Bartle, 1995) may be used. The lower penetration (particularly through organic materials) and the lower source strengths realistically limit the application of radioisotope and DD sources to scanning smaller items such as suitcases and parcels. High intensity DT neutron sources, however, are well suited to scanning air cargo and are discussed below. CSIRO has developed a FNGR technique that utilises a 14 MeV DT generator and a 60Co gamma-ray source (Eberhardt et al., 2005, 2006). The first commercial-scale FNGR scanner was developed and installed in an Australian Customs Service facility at Brisbane International Airport (Eberhardt et al., 2006) and the technology and associated business processes have been trialled in a real time operational environment. Comparative tests against two commercial single-beam X-ray scanners on a range of air cargo showed that, with improved spatial resolution (5 mm detectors or smaller) and multi-view capability, the CSIRO Air Cargo Scanner has the potential to significantly outperform current commercial X-ray air cargo scanners. It should be noted that FNGR can also be performed using a DT neutron source in combination with continuous X-ray source in place of the 60Co source (Sowerby and Tickner, 2003; Kang et al., 2007). The advantages of this approach include higher penetration, higher intensity source, improved image quality and improved safety as the source can be switched off. Necessary corrections for beam hardening effects can be based on measured X-ray transmission.
To simplify the following discussion, we limit ourselves to considering two specific instances of dual-beam radiography systems, namely:
a dual-energy 5 and 9 MeV interlaced high-energy X-ray scanner; and a FNGR scanner using DT neutrons and 60Co gammarays.
Table 4 compares the penetration and R-values of the two systems. The penetration figures show the thickness of each material through which 0.1% of the least penetrating of the two types of radiation associated with each system is transmitted. The R-value figures show the ratio of the logarithms of the transmissions of the two types of radiation, at 30 g cm2 material depth for the dual-energy X-ray scanner. The R-value for the neutron/gamma-ray scanner is independent of material thickness. As can be seen, the penetration of the dual-energy X-ray scanner is larger than that of the neutron/gamma-ray scanner for all materials, due to the high-energy radiation used. The penetration of the neutron/gamma-ray scanner is limited by the neutron transmission for organic materials and by the 60Co gamma-ray transmission for the metals. Replacing the 60Co source with a 4–5 MeV high-energy X-ray source would increase the penetration of the neutron/ gamma-ray scanner for all except organic materials. The most striking advantage of the FNGR technique is the wide range of R-values between different materials (Fig. 7). There is a factor of 7.5 variation in R-values between polythene and the heaviest metals, compared to a factor of 1.2 variation for the dual-energy X-ray method. The variation in R is also more uniformly spread, with good separation between different classes of organic material and light and heavy metals. In contrast, different organic materials are virtually indistinguishable using the Table 4 Comparison of penetration and R-value discrimination of dual-energy X-ray and neutron/gamma-ray scanners Material
Polythene C Al Fe Pb U
Dual X-ray scanner
Neutron/gamma-ray scanner
Penetration (g cm2)
R-value
Penetration (g cm2)
R-value
172 193 191 183 154 147
0.789 0.797 0.826 0.873 0.960 0.954
60 106 129 132 121 112
1.781 1.153 0.712 0.526 0.267 0.237
Note: By replacing the 60Co source in the neutron/gamma-ray scanner with a 4–5 MeV, high-energy X-ray source would significantly increase penetration for all except organic materials.
ARTICLE IN PRESS Y. Liu et al. / Applied Radiation and Isotopes 66 (2008) 463–473
Fig. 7. R-values for various materials for the dual-energy 9 and 5 MeV X-ray scanner and a fast neutron/gamma-ray scanner using DT neutrons and 60Co gamma-rays.
dual-energy X-ray approach and most of the variation is between light (Zo15) and heavy (Z425) elements. To investigate the material discrimination abilities of the two methods further, it is necessary to estimate the accuracies with which R-values can be determined. Contributions to the uncertainty in R-value determination arise from three main sources (Chen et al., 2007):
Counting statistics, due to the finite number of radiation quanta received at the detectors. Deviations from narrow-beam transmission due to radiation scattering. Instability in the beam energies of the Bremsstrahlung X-ray source and variations in the energy spectrum with angle.
The magnitudes of these uncertainties are not straightforward to estimate accurately. However, some approximations can be made. The I0 neutron count rate for the neutron/gamma-ray scanner described in Eberhardt et al. (2006) is approximately 1000 counts per pixel. The I0 count rates for highenergy X-ray-based systems can be estimated if assumptions are made about the source strength and detector geometry. Assuming a 5 MeV source producing 1 Gy min1 at 1 m, an 8 m source/detector distance, 5 5 mm detectors and a 10 m min1 scan rate, an I0 rate of approximately 100,000 counts per pixel is obtained. In our FNGR scanners (Eberhardt et al., 2006), we observe deviations from narrow-beam transmission due to neutron and gamma-ray scattering. These amount to approximately 10% and 1% for the two radiations; that is, 10% of detected neutrons and 1% of detected gammarays are scattered rather than directly transmitted. A similar figure for X-ray scattering is reported in Chen
471
et al. (2007). Fortunately, most of this scattering is highly local, between closely neighbouring pixels, and can be corrected rather accurately. In most parts of the image (except near object edges where the radiation attenuation is varying extremely rapidly), we estimate a residual uncertainty after scattering correction of 1% of I/I0 for neutrons and 0.1% for gamma-rays. Variations in electron beam energy used in the Bremsstrahlung source produce significant shifts in the measured R-values. For example, a shift of 0.5 MeV in the higher beam energy changes the R-value for iron by 0.01 and the R-value for organic materials by 0.02. In the absence of manufacturers’ information on beam energy stability, we do not include beam instability effects in the total uncertainties discussed below, but they represent a potentially significant source of uncertainty in material identification using a dual-energy X-ray system. Combining the first two sources of uncertainty in quadrature, we can estimate the total R-value uncertainty per pixel for the two scanner systems as a function of material thickness. These estimated R-value uncertainties are used to calculate a figure of merit as shown in Fig. 8, determined as the ratio of R-value uncertainties divided by R-value differences for graphite and polythene (top graph) and iron and polythene (lower graph). Graphite and polythene have been selected to illustrate the capability of the techniques to distinguish various classes of organic materials and iron and polythene have been selected to illustrate the capability of the techniques to distinguish organic materials from metal. The figures of merit are plotted as a function of material (iron) thickness for dualenergy X-ray and FNGR scanners. The figures of merit for a single pixel and also averaged over an area of 100 pixels are shown. Note that the higher the ratio, the greater the ability of the scanner to distinguish the two materials; a ratio less than unity means that the scanner is essentially unable to distinguish between the two materials. The material discrimination abilities of the FNGR scanner, expressed as the ratio of the R-value uncertainties to the R-value difference between two materials, are generally superior to the dual-energy X-ray scanner. The ability of the dual-energy X-ray scanner to discriminate between different organic materials is very limited, even averaged over 100 pixels. In contrast, the uncertainty for the FNGR scanner at the scan thickness is only 4–7% of the R-value difference between polythene and carbon, showing that the FNGR system can distinguish between a wide range of organic materials. Both the dual-energy X-ray and FNGR scanners are able to effectively distinguish between organic materials and steel and also to distinguish heavy metals (lead, uranium) from common metals such as steel and copper. In cluttered environments such as air cargo, it is relatively unusual to find regions of scan images comprising a single material. Rather, the integration along the radiation beams that is inherent in 2D radiography means that many overlapping materials are often present. In these
ARTICLE IN PRESS 472
Y. Liu et al. / Applied Radiation and Isotopes 66 (2008) 463–473
High-energy X-ray systems have the advantage of better penetration and higher count rates than neutron systems. However, neutron systems have the advantage of much better sensitivity of R-value measurement compared to dual high-energy X-ray techniques. Both dual high-energy X-ray and neutron scanning techniques are suitable for widespread deployment in an airport environment using currently available technology and can meet essential requirements such as rapidly scanning consolidated cargo and radiation safety. Both technologies produce high-resolution images, can readily identify common metal items by shape and have the potential to detect nuclear materials such as uranium, plutonium and common radiation shielding materials such as lead and tungsten. FNGR offers the additional significant advantage of discriminating between a wide range of organic materials. This improved discrimination assists operators in interpreting the complex, highly cluttered cargo images that result from scanning air cargo and have the potential to facilitate in the detection of narcotics, explosives and other organic threat materials.
References
Fig. 8. Figure of merit, calculated as R-value errors divided by the difference in R-value between (a) graphite and polythene and (b) iron and polythene. Both graphs are plotted versus material thickness for dualenergy X-ray and FNGR scanners. Results for single and 100 pixel average measurements are shown.
circumstances, it is important that the material discrimination figure-of-merit is significantly larger than unity, to allow regions of varying material composition to be picked out by image analysts. 6. Discussion Recent developments and applications of dual-beam radiography for cargo inspection have been reviewed. These recent developments include dual high-energy X-ray techniques as well as fast neutron and gamma-ray (or X-ray) radiography systems. Both techniques can readily discriminate between metals and organics, and between high-atomic number metals such as lead and uranium and common metals such as aluminium and steel.
Air Freight Council of NSW, 2007. /http://www.airfreightnsw.com.au/html/ s02_article/article_view.asp?id=131&nav_cat_id=148&nav_top_id=60S. Bartle, C.M., 1995. Method and apparatus for detecting concealed substances. US Patent No. 5,479,023, 26 December 1995. Bjorkholm, P.J., 2004. Dual energy detection of weapons of mass destruction. Port Technol. Int. 22, 155–158. Budner, G.J., 2006. Dual energy detection of weapons of mass destruction. Proc. SPIE 6178, G1–G9. Buffler, A., 2004. Contraband detection with fast neutrons. Radiat. Phys. Chem. 71, 853–861. Chen, Z., Wang, X., 2006. Cargo X-ray imaging technology for material discrimination. Port Technol. Int. 30, 163–165. Chen, G., Bennet, G., Perticone, D., 2007. Dual-energy X-ray radiography for automatic high-Z material detection. Nucl. Instrum. Methods Phys. Res. B 261, 356–359. Eberhardt, J.E., Rainey, S., Stevens, R.J., Sowerby, B.D., Tickner, J.R., 2005. Fast neutron radiography scanner for the detection of contraband in air cargo containers. Appl. Radiat. Isot. 63, 179–188. Eberhardt, J.E., Liu, Y., Rainey, S., Roach, G.J., Stevens, R.J., Sowerby, B.D., Tickner, J.R., 2006. Fast neutron and gamma-ray interrogation of air cargo containers. International Workshop on Fast Neutron Detectors, Cape Town, 3–6 April 2006, Proceedings of Science (FNDA2006) 092. /http://pos.sissa.itS. Findlay, D.J.S., 1989. Analytic representation of Bremsstrahlung spectra from thick radiators as a function of photon energy and angle. Nucl. Instrum. Methods A 276, 598–601. Gozani, T., 2002. A review of neutron based non-intrusive inspection technique. In: Conference on Technology for Preventing Terrorism, Hoover Institution National Security Forum, Stanford University, 12–13 March 2002, p. 12. Kang, K., et al., 2007. Method and equipment for discriminating materials by employing fast neutron and continuous spectral X-ray. US Patent Application Publication No. US 2007/0096036, 3 May 2007. Kawrakov, I., Rogers, D.W.O., 2006. The EGSnrc Code System, Monte Carlo Simulation of Electron and Photon Transport, NRCC Report PIRS-701.
ARTICLE IN PRESS Y. Liu et al. / Applied Radiation and Isotopes 66 (2008) 463–473 Miller, T.G., Van Staagen, P.K., Gibson, B.C., Krauss, R.A., 1996. Contraband identification in sealed containers using neutron transmission. SPIE Proc. 2867, 215–218. Mishin, A.V., 2005. Advances in X-band and S-band linear accelerators for security, NDT and other applications. In: Proceedings of the 2005 Particle Accelerator Conference, Knoxville, TN, IEEE, 240244. Neale, W.W., Rushbrooke, J.G., Ansorge, 1996. Material identification using X-rays. United States Patent No. 5,524,133, 4 June 1996. NIST, 2007. /http://physics.nist.gov/hysRefData/X-rayMassCoeffS (Table 1). Ogorodnikov, S., Petrunin, V., 2002. Processing of interlaced images in 4–10 MeV dual energy customs system for material recognition. Phys. Rev. Special Top.—Accelerators Beams 5, 104701.
473
Rushbrooke, J.G., Neale, W.W., Ansorge, R.E., Hooper, C.E., Lacey, R.J., Head, L., 1997. High energy material discrimination (HEMD) in the X-ray imaging of cargo containers. In: Proceedings—Harnessing Technology to Support the National Drug Control Strategy, Chicago, August 1997, pp. 6–7, 6–19. Sowerby, B.D., Tickner, J.R., 2003. Radiographic equipment. International Patent Application No. WO2004053472PCT/AU03/01641, 10 December 2003. Sowerby, B.D., Tickner, J.R., 2007. Recent advances in fast neutron radiography for cargo inspection. Nucl. Instrum. Methods A 580, 799–802. Tang, Ch., Chen, H., Liu, Y., Wang, X., 2006. Low energy linacs and their application in Tsinghua University, In: Proceedings of LINAC 2006, Knoxville, TN, pp. 256–258.
Applied Radiation and Isotopes 66 (2008) 463–473 www.elsevier.com/locate/apradiso
Comparison of neutron and high-energy X-ray dual-beam radiography for air cargo inspection Y. Liu, B.D. Sowerby, J.R. Tickner CSIRO Minerals, Private Mail Bag 5, Menai, NSW 2234, Australia Received 24 September 2007; received in revised form 15 October 2007; accepted 16 October 2007
Abstract Dual-beam radiography techniques utilising various combinations of high-energy X-rays and neutrons are attractive for screening bulk cargo for contraband such as narcotics and explosives. Dual-beam radiography is an important enhancement to conventional single-beam X-ray radiography systems in that it provides additional information on the composition of the object being imaged. By comparing the attenuations of transmitted dual high-energy beams, it is possible to build a 2D image, colour coded to indicate material. Only high-energy X-rays, gamma-rays and neutrons have the required penetration to screen cargo containers. This paper reviews recent developments and applications of dual-beam radiography for air cargo inspection. These developments include dual high-energy X-ray techniques as well as fast neutron and gamma-ray (or X-ray) radiography systems. High-energy X-ray systems have the advantage of generally better penetration than neutron systems, depending on the material being interrogated. However, neutron systems have the advantage of much better sensitivity to material composition compared to dual high-energy X-ray techniques. In particular, fast neutron radiography offers the potential to discriminate between various classes of organic material, unlike dual energy X-ray techniques that realistically only offer the ability to discriminate between organic and metal objects. r 2007 Elsevier Ltd. All rights reserved. Keywords: Dual-beam radiography; Fast neutron radiography; X-ray radiography; Gamma-ray radiography; Air cargo; Contraband
1. Introduction There is a worldwide need for improved methods for the screening of air cargo for contraband. The main objectives are the detection of contraband such as illicit drugs, explosives, weapons and nuclear materials and the verification of declared manifests. Air cargo is usually packaged into lightweight aluminium containers called unit load devices (ULDs). The most common ULDs have a width of about 1.6 m and gross weight of up to 1.6 tonnes. However, pallet ULDs are significantly larger, with widths up to 2.4 m and maximum gross weights of up to 6.8 tonnes. For comparison, the most common sea container is the 40 ft container of dimensions 12.2 m (length) 2.4 m (width) 2.6 m (height) and maximum gross weight of 32.5 tonnes. The air cargo ULD content is commonly limited by its mass, not its Corresponding author. Tel.: +61 2 9710 6719; fax: +61 2 9710 6789.
E-mail address: [email protected] (B.D. Sowerby). 0969-8043/$ - see front matter r 2007 Elsevier Ltd. All rights reserved. doi:10.1016/j.apradiso.2007.10.005
volume, and the load is heterogeneously distributed. Material thicknesses (mass-per-unit area scanned) up to 100 g cm2 can be encountered. The requirements for a mass-screening system for air cargo are complex and demanding and include:
screening cargo containers without unpacking in 1–2 min; producing high quality, high-resolution (5–10 mm or better) images that can be readily interpreted by operators; differentiating between organic and inorganic materials and ideally discriminating between different classes of organic materials such as explosives, narcotics and benign substances; distinguishing ‘ordinary’ metals such as aluminium, iron, steel and copper from heavy metals such as lead, tungsten, uranium and plutonium that may be associated with smuggled nuclear materials;
ARTICLE IN PRESS 464
Y. Liu et al. / Applied Radiation and Isotopes 66 (2008) 463–473
compliance with strict radiation safety requirements for both operating staff and cargo; suitability for continuous operation in an airport environment; and availability at reasonable cost.
The majority of cargo inspection systems installed today use single-beam high-energy X-ray or gamma-ray radiography to form a 2D (scan depth integrated) transmission image of the cargo’s contents. The image is analysed by a human operator, who is required to make a decision in a few minutes as to whether the container should be opened for manual inspection. Image quality is crucial in assisting the operator to reach the correct decision and recent X-ray scanner developments have focussed on improving the resolution and contrast of the collected images. Highenergy X-ray or gamma-ray radiography is primarily sensitive to density and is particularly effective in detecting hidden objects with readily identifiable shapes such as firearms and other weapons. However, single-energy radiography provides no material discrimination and is less well suited to finding low-density organic threats that lack a definite shape. The requirement to determine both shape and material composition points to imaging cargo using two different types of radiation. From the two measurements, it may then be possible to calculate an image showing both massper-unit area and composition. For this approach to be effective both types of radiation must be sufficiently penetrating to be detectable after passing through the cargo and the attenuation of both types of radiation in different materials must be distinguishable. Only fast (high energy) neutrons and high-energy X-rays or gamma-rays have the required penetration for imaging the contents of most air cargo containers. The ranges of even the most energetic charged particles that could be obtained from a practical accelerator suitable for industrial use are at most a few g cm2. For example, the ranges of 10 MeV electrons and protons in water are 5.0 and 0.12 cm, respectively. The first aim of any dual-beam radiography technique is to determine separately the mass-per-unit area and a measure of material composition for each pixel in the final 2D image. Ideally, these two quantities should be independent, so that the mass-per-unit area determination does not depend on the material composition and vice versa. The types of radiation used need to comply with relevant regulations, particularly with regard to the irradiation of foods commonly present in air cargo. For example, in the USA and UK, to be exempt from these regulations the dose imparted to the cargo must not exceed 0.01 Gy in the case of inspection devices that utilise neutrons and 0.5 Gy otherwise. In these countries, the maximum irradiation energy is 10 MeV in the case of X-rays and 14 MeV in the case of neutrons. Note that these regulations differ between countries.
In the present paper, two dual-beam radiography systems, one based on X-rays alone and the other on fast neutrons and gamma-rays (or X-rays), are assessed, discussed and compared for application to the screening of consolidated airfreight. The big advantage of dual-beam systems is that they provide additional information on the composition of the object being imaged. By comparing the attenuations of transmitted dual beams, it is possible to build a 2D image, colour coded to indicate average composition integrated along the beams of each image pixel. Some results from a specific implementation of a fast neutron and gamma-ray radiography (FNGR) system for air cargo scanning at Brisbane International Airport in Australia are first analysed to determine the likely penetration depth requirements for a dual-beam radiography system of air cargo. 2. Characterisation of air cargo at Brisbane International Airport Air cargo container sizes vary widely and contents range from single commodity loads through to consolidations containing many hundreds of different items. Air cargo containers are generally more cluttered than sea cargo containers. In late 2006, a detailed statistical survey was carried out to determine the distribution of loads in cargo passing through the commercial prototype FNGR scanner that was installed at an Australian Customs Service facility at Brisbane International Airport (Eberhardt et al., 2006). The scanner facility incorporated an extensive automated cargo handling system, making possible the rapid scanning of large numbers of containers. The scanner facility was trialled by Australian Customs over 12–18 months at Brisbane International Airport. Per-pixel attenuation histograms for 14 MeV neutron and 60Co gamma-ray (1.17 and 1.33 MeV) transmissions were calculated for 940 cargo images collected over a period of 3 weeks. A more detailed manual analysis was made of 202 cargo images collected over a 4-day window inside this period. This detailed analysis included additional information on the ULD type and manifest details (where available). The breakdown by ULD type is shown in Fig. 1. Detailed descriptions and drawings of the different ULD types are available from a number of sources on the web (e.g., Air Freight Council of NSW, 2007). The commonest observed ULDs are AKEs with dimensions of 1530 (width) 2010 (length) 1630 (height) mm. The commonest observed pallets are the PMCs and PAGs with dimensions of 2440(2240 for PAG) 3180 1630 mm. Cargos are scanned through the narrower (width) dimension. Overall, about 50% of cargo was loaded into ULDs and 50% on pallets. From the per-pixel gamma-ray attenuation histograms, it is possible to calculate the distribution in mass-per-unit area for different classes of cargo. In this analysis, measured mass-per-unit areas of less than 2 g cm2 are ignored and a material independent mass-attenuation
ARTICLE IN PRESS Y. Liu et al. / Applied Radiation and Isotopes 66 (2008) 463–473
coefficient of 0.058 cm2 g1 is assumed. Fig. 2 shows the simple and cumulative distributions in mass-per-unit area estimated for cargos classified as ‘‘light’’ and ‘‘heavy’’ in the manual analysis, along with the average distributions for all cargos. The limited number of bits used to record the attenuation histograms is responsible for the cut-off in the curves around 106 g cm2, but the fraction of pixels above this level is very small. A scanner using radiation capable of penetrating through 100 g cm2 would be able to image through 99.6% of the area of an average air cargo ULD at Brisbane airport. A scanner using radiation capable of
Fig. 1. Breakdown of 202 incoming air cargo containers by ULD and pallet types at Brisbane International Airport over a 4-day period in late 2006. The ULD identifiers commencing with A and D indicate aluminium containers and those commencing with P are larger pallet cargo. The most common AKE containers are half-width lower deck containers used on wide body aircraft.
465
penetrating through only 50 g cm2 would be able to image through 87% of the area of an average air cargo ULD at Brisbane airport. It is a reasonable hypothesis that similar penetration depths would apply to imaging of air cargo ULDs at airports around the world. 3. Dual-energy X-ray methods 3.1. Background Dual-energy X-ray radiography is an important enhancement to X-ray systems that provides additional information on the composition of the object being imaged. By comparing the attenuations of transmitted high- and low-energy X-ray beams, it is possible to build a 2D image, colour coded to indicate metal and organic materials. Dual energy X-ray systems can utilise the Z dependence of the mass-attenuation coefficients at either low energies (due to the photoelectric effect) or high energies (due to pair production) (Fig. 3(a)). In applications such as examining passenger luggage, dual-energy X-ray radiography using low-energy X-rays has become the usual screening method. However, the limited penetration of the lower energy X-rays (around 50–100 keV) used to provide composition information prevents the method being used effectively on consolidated airfreight. To illustrate this, Fig. 3(b) shows the thickness of various materials for 0.1% X-ray transmission as a function of X-ray energy, calculated using mass-attenuation coefficients from NIST (2007). For energies above about 1 MeV, X-rays are sufficiently penetrating to pass through
Fig. 2. Cumulative (log scale, left-hand axis) and simple (linear scale, right-hand axis) distributions of image pixels as a function of mass-per-unit area for incoming air cargo at Brisbane International Airport. The ‘‘All Cargo’’ data shown is for all 940 incoming ULD/pallet types scanned over a 3-week period. Pixels with measured mass-per-unit area less than 2.0 g cm2 were excluded from calculation.
ARTICLE IN PRESS 466
Y. Liu et al. / Applied Radiation and Isotopes 66 (2008) 463–473
3.2. Calculations for the determination of composition For the ideal case of narrow-beam transmission of monoenergetic radiation, the transmissions t1,2 of two X-ray beams 1,2 of different energy can be written as ti ¼
Ii ¼ expðmi X Þ ði ¼ 1; 2Þ, I 0i
(1)
where Ii is the measured strength of the transmitted radiation with the cargo in position, I 0i is the measured strength with no cargo in the beam, mi is the massattenuation coefficient, and X is the mass-per-unit area. If the energy of the first radiation beam is selected such that the mass-attenuation coefficient is a weak function of material, as is the case for X-rays with energies in the range 1–3 MeV, then the two transmission equations can be solved to give: X ¼
R¼
Fig. 3. (a) X-ray and gamma-ray mass-attenuation coefficients (R-values) (cm2 g1) for various materials over the energy region 50 keV to 10 MeV. (b) Thickness of the same materials for 0.1% transmission plotted as a function of gamma-ray energy.
100 g cm2 of all materials. The mass-attenuation coefficients for the heaviest elements (Pb and U) begin to differ significantly at energies below about 500 keV and the lighter metals (Al and Fe) can be distinguished at energies below about 100–150 keV. At these low energies, however, penetration is limited to 30–50 g cm2 for the lightest materials and less than 10–20 g cm2 for the high-atomic number elements. Therefore, the application of conventional dual energy X-ray scanners based on X-ray tube sources (up to about 450 keV) is limited to cargo with low mass-per-unit area primarily because of the very limited penetration of the low-energy X-rays (100–200 keV) required for material discrimination. At higher energies above about 4–5 MeV, it again becomes possible to distinguish different materials. Penetrations at these high energies are excellent. Dual highenergy X-ray radiography for material discrimination is discussed below.
1 I1 log 0 , m1 I1
m2 logðI 2 =I 02 Þ ¼ . m1 logðI 1 =I 01 Þ
(2)
(3)
Here R is just the ratio of mass-attenuation coefficients that, for the appropriate choice of X-ray energies, is diagnostic of the material present. Unfortunately, convenient monoenergetic X-ray or gamma-ray sources are not available for energies above 1–2 MeV. Instead, high-energy Bremsstrahlung sources using electron accelerators with accelerating voltages typically up to 9 MV must be used, which produce broadspectrum radiation with energies up to the accelerating potential. For broad-spectrum radiation produced by such a source, Eq. (1) needs to be replaced with R E MAX fðEÞ exp mi ðEÞX Þ rðEÞ dE Ii 0 ti ¼ 0 ¼ ði ¼ 1; 2Þ, R E MAX Ii fðEÞrðEÞ dE 0 (4) where the integrals extend over all radiation energies produced by the source, f(E) is the energy spectrum of X-rays emitted by the source and r(E) is the relative detector response as a function of X-ray energy. Because X-ray mass-attenuation coefficients vary strongly with energy, the attenuation of broad-spectrum radiation is no longer strictly exponential and the simple solutions for X and R given in Eqs. (2) and (3) do not strictly apply. Nevertheless, with appropriate choices of X-ray energies these relations provide a good starting point for the analysis of dual high-energy X-ray transmission images.1 1 Other authors have taken and composition information measurements, for example, (1997). The problem reduces
different approaches to extracting density from dual high-energy X-ray transmission the ‘banana-plots’ of Rushbrooke et al. to finding functions fX and fR such that
ARTICLE IN PRESS Y. Liu et al. / Applied Radiation and Isotopes 66 (2008) 463–473
467
photon depends on both the probability of the X-ray interacting in the detector element and the energy that it deposits if it does interact. We have estimated these factors for a small CsI scintillator element using a simple Monte Carlo simulation based on the EGSnrc (Kawrakov and Rogers, 2006) code. We find that the energy dependence of the detector response r(E) appearing in Eq. (4) can be well represented above energies of 0.2 MeV by the functional form rðEÞ / 1 þ 0:535E 1:5 .
Fig. 4. Bremsstrahlung X-ray spectra with accelerating voltages and endpoint energies of 4, 5, 6, 7, 8 and 9 MeV. Spectra have been normalised to have the same peak intensity.
(5)
Utilising this expression for r(E) in Eq. (4), the transmissions of 5 and 9 MeV Bremsstrahlung through polythene, carbon, aluminium, iron lead and uranium have been calculated. R-values calculated using Eq. (3) are plotted as a function of material thickness in Fig. 5 for two cases, namely (a) no pre-filtration of the Bremsstrahlung beam and (b) pre-filtration with 30 g cm2 of lead to remove low-energy X-rays.
Fig. 4 plots the energy spectrum of Bremsstrahlung sources with accelerating voltages of 4, 5, 6, 7, 8 and 9 MeV, calculated using the simple analytical model of Findlay (1989). A 3 g cm2 tantalum target is assumed and the spectra have been normalised to have the same peak intensity. To determine the X-ray transmission at two different energies, or more practically over two different energy ranges, it is necessary to either (i) discriminate the energies of individual X-ray quanta hitting the detector system, (ii) use two detector arrays, filtered so that one preferentially responds to higher energy and one to lower energy quanta or (iii) collect separate measurements at different accelerating voltages, using interlaced energies on alternate accelerator pulses. Approach (i) is technically very difficult, due to the high intensity and low-duty cycle (of order 103; Mishin, 2005) of the RF linear accelerators typically used as X-ray sources. A group at Cambridge Imaging implemented this approach using multiple scintillation crystals for each detector element (Rushbrooke et al., 1997). Approach (ii) would be difficult to implement because of the relatively small differences in mass-attenuation coefficients above a few MeV. Approach (iii) is the most widely used (Ogorodnikov and Petrunin, 2002; Bjorkholm, 2004; Chen and Wang, 2006) and is discussed in more detail below. Generally, with the interlaced dual energy pulsed-source technique the first end-point energy is chosen to be in the range 4–6 MeV and the second 8–9 MeV. The detector arrays measure the total energy deposited in each detector element per pulse. The average energy deposited per X-ray (footnote continued) X ¼ fX(t1, t2) and R ¼ fR(t1, t2) such that X depends only on mass-per-unit area and R depends only on (average) material composition.
Fig. 5. R-values based on ratios of the transmission of 5 and 9 MV Bremsstrahlung radiation, plotted as a function of material thickness for various materials for (a) no pre-filtration of the Bremsstrahlung beam and (b) pre-filtration with 30 g cm2 of lead.
ARTICLE IN PRESS 468
Y. Liu et al. / Applied Radiation and Isotopes 66 (2008) 463–473
It can be seen at once that, as expected, the calculated R-values are not independent of material thickness. More seriously, for no pre-filtration and material thicknesses below about 20 g cm2 the R-values are not unique, nor monotonic with increasing atomic number. This is due to the different attenuation mechanisms of low- and highenergy X-rays in materials of different atomic number. At relatively low energies, where Compton scattering dominates the total cross-section, the low-atomic number materials have the highest mass attenuations. At high energies, where pair production becomes increasingly important, the high-atomic number materials have the highest mass attenuations. On the basis of this analysis, unique solutions for dual X-ray R-values (and associated material composition) at low-material thickness appears to require additional information. These results point to the importance of pre-filtering the Bremsstrahlung radiation to remove the lower energy component. When the Bremsstrahlung beams are pre-filtered with 30 g cm2 of lead the R-values are still not constant with thickness but a relatively simple empirical function could be used to correct the measured R- and thickness-values back to a standard mass-per-unit area, say 30 g cm2. 3.3. Dual high-energy X-ray cargo inspection systems The development and application of dual energy highenergy X-ray scanners has been reported by a number of authors. A summary of some of the features of these scanners is given in Table 1. Unfortunately, only very limited technical information is available on the Varian and Nuctech systems. The scanners developed by the Efremov Scientific Institute, Varian and Nuctech all image a cargo container using successive pulses of high- and low-energy polychromatic X-ray beams. Typical X-ray Bremsstrahlung energy spectra from the X-ray sources used are shown in Fig. 4. In this situation, non-unique compositional information is a possibility unless pre-filtering of low-energy X-rays is used. The high energy is selected so that the pair production cross-section accounts for a significant proportion of the attenuation. The high energy is usually in the 8–9 MeV
range. The lower energy is selected to provide reduced pair production attenuation and at the same time to give a sufficiently high flux. Generally, energies in the range 4–6 MeV are selected. These three systems use a linear array of detectors, typically with a pitch or spacing of 3.5–5 mm. The system developed by Cambridge Imaging Ltd. uses a different approach. The X-ray source used was a 10 MeV linac operated at a single energy. The energy discrimination was provided by the detector system although the details are not disclosed (Rushbrooke et al., 1997). A number of optional detector systems are discussed by Neale et al. (1996). These detection systems are complex but they do offer the possibility of preferentially measuring the transmitted X-rays that produce annihilation radiation, therefore improving signal-to-noise compared to the interlaced pulsing methods used by the other authors. 4. Neutron radiography techniques 4.1. Background The limitations of X-ray systems for the detection of organics-based explosives and drugs have stimulated the development of alternative methods for cargo inspection including those based on neutrons. Neutron radiography can be performed using either fast neutrons, with energies of a few hundred keV or larger, or using thermal neutrons with energies around 0.025 eV. Thermal neutron crosssections vary by more than 7 orders of magnitude, from about 104 b for oxygen through to 46,617 b for gadolinium. Other elements with significant thermal crosssections include lithium, boron, cadmium and erbium. The wide variation in cross-section, which correlates poorly with the threat detection requirements outlined in Section 1, coupled with the lack of practical sources, makes thermal neutron radiography ill-suited to cargo screening. Fast neutron techniques are attractive for cargo inspection as they have the required penetration, they can be used to determine elemental composition and they interact with matter in a manner complementary to X-rays. X-rays and gamma-rays scatter primarily from
Table 1 Summary of published information on some dual high-energy X-ray radiography systems for cargo inspection Cambridge Imaging Ltd.
Efremov Scientific Institute
Varian
Nuctech
References
Rushbrooke et al. (1997)
X-ray source
10 MeV linac; pulsed mode 200 Hz 100# elements; 5 5 mm2 Optical fibre/CCD Laboratory trial conducted
Ogorodnikov and Petrunin (2002) Pulsed linac; 8/4 MeV interlaced mode 100 Hz, 5 msec CdWO4; 3.5 mm pitch PIN diodes Laboratory trial conducted
Bjorkholm (2004), Budner (2006) Pulsed linac; 9/5 MeV interlaced mode 200 Hz 5 mm typical
Chen and Wang (2006), Tang et al. (2006) Pulsed linac; 9/6 MeV interlaced mode Pulse separation o4 ms
Laboratory trial conducted
Commercial unit
Pulse rate and duration Detector type and size Read-out system Status
ARTICLE IN PRESS Y. Liu et al. / Applied Radiation and Isotopes 66 (2008) 463–473
electrons whereas fast neutrons scatter from the protons and neutrons in the nucleus. As a result neutrons are scattered and slowed more by low-atomic number (Z) elements than by the high-Z elements. Convenient sources of neutrons include sealed tube generators using the deuterium–deuterium (DD), deuterium–tritium (DT) and tritium–tritium (TT) fusion reactions, higher energy accelerators with beam energies generally above 1 MeV using a variety of deuterium and proton induced reactions, radioisotope sources and photoneutron sources using Bremsstrahlung radiation from high-energy electron accelerators. Table 2 summarises the main properties of these sources. All of these sources produce fast neutrons with energies of a few hundred keV or greater. Fig. 6(a) plots the neutron mass-attenuation coefficients as a function of energy for the same materials as previously investigated for dual-energy X-ray radiography. There is significant variation from material to material and the mass-attenuation coefficients for some materials also show pronounced resonances. Fig. 6(b) shows the thicknesses of various materials through which 0.1% of neutrons are transmitted from DD or DT sources. Results for highenergy accelerator-based sources and 241Am-Be, 252Cf sources will be broadly similar, depending on the exact neutron energy spectrum produced. Fast neutrons have similar penetrations to 1–2 MeV X-rays for aluminium; they are significantly more penetrating for heavier elements and less penetrating for organic materials.
469
uranium calculated for DD and DT neutrons and (average energy of 1.25 MeV) gamma-rays.
60
Co
4.3. Fast neutron radiography cargo inspection systems As for dual energy X-ray techniques, material discrimination using fast neutron radiography relies on determining
4.2. Determination of composition For a monoenergetic neutron source (DD or DT generators) and a monoenergetic gamma-ray source, Eqs. (2) and (3) (Section 3.2) can be used to determine a cross-section ratio R and a material density X independent of material thickness. In this case, the subscript 1 refers to the gamma-ray measurement and 2 to the neutron measurement. If either the neutron or X-ray source is not monoenergetic, then Eq. (4) has to be used to calculate the transmission and the R-value determined using Eq. (3) can be expected to vary somewhat with material thickness. For the simpler monoenergetic case, Table 3 shows the R-values for polythene (also known as polyethylene, formula (C2H4)n), graphite, aluminium, iron, lead and
Fig. 6. (a) Neutron mass-attenuation coefficients versus energy for various materials. (b) Thicknesses of the same materials through which 0.1% of DD neutrons, or DT neutrons, 60Co gamma-rays (1.17 and 1.32 MeV) and X-rays from a 5 MV Bremsstrahlung source are transmitted.
Table 2 Properties of some available fast neutron sources Source
Beam energy
Reaction
Neutron energy (MeV)
Maximum output (s1)
Sealed tube, DD Sealed tube, DT Sealed tube, TT Accelerator 241 Am-Be 252 Cf Photoneutron
o200 keV o200 keV o200 keV 1–6 MeV Radioisotope Radioisotope 44 MeV
2
2.5 14 0–9 Varies 0–12 0–10 Varies
108 1010 2 108 1010 5 107 109 1010–11
H(d,n)3He H(d,n)4He 3 H(t,2n)4He Various 9 Be(a,n)12C Fission (g,n) 3
ARTICLE IN PRESS Y. Liu et al. / Applied Radiation and Isotopes 66 (2008) 463–473
470
Table 3 Calculated R-values for DD/DT neutrons and
5. Comparison of dual-beam radiography techniques 60
Co gamma-rays
Material
R-value (DD/60Co)
R-value (DT/60Co)
Polythene Graphite Aluminium Iron Lead Uranium
4.496 1.404 0.852 0.725 0.295 0.305
1.781 1.153 0.712 0.526 0.267 0.237
mass-attenuation coefficients of fast neutrons and gammarays or X-rays. Various alternative neutron radiography systems have been reviewed (Gozani, 2002; Buffler, 2004; Sowerby and Tickner, 2007). Some systems (Miller et al., 1996) are attempting to use radiography with broadenergy spectrum or ‘white’ neutron sources, usually coupled with some means of determining the energy of the transmitted neutrons, to identify specific elements through their characteristic resonances. Whilst potentially providing powerful material discrimination, these approaches require complex sources and detector systems and currently remain at the laboratory trial stage. The alternative approach is to perform ‘total-beam’ radiography, performing a single measurement of the attenuation of the entire spectrum of neutrons emitted by the source. Either a monoenergetic DD or DT source (Eberhardt et al., 2005, 2006) or a broad-energy radioisotope source (Bartle, 1995) may be used. The lower penetration (particularly through organic materials) and the lower source strengths realistically limit the application of radioisotope and DD sources to scanning smaller items such as suitcases and parcels. High intensity DT neutron sources, however, are well suited to scanning air cargo and are discussed below. CSIRO has developed a FNGR technique that utilises a 14 MeV DT generator and a 60Co gamma-ray source (Eberhardt et al., 2005, 2006). The first commercial-scale FNGR scanner was developed and installed in an Australian Customs Service facility at Brisbane International Airport (Eberhardt et al., 2006) and the technology and associated business processes have been trialled in a real time operational environment. Comparative tests against two commercial single-beam X-ray scanners on a range of air cargo showed that, with improved spatial resolution (5 mm detectors or smaller) and multi-view capability, the CSIRO Air Cargo Scanner has the potential to significantly outperform current commercial X-ray air cargo scanners. It should be noted that FNGR can also be performed using a DT neutron source in combination with continuous X-ray source in place of the 60Co source (Sowerby and Tickner, 2003; Kang et al., 2007). The advantages of this approach include higher penetration, higher intensity source, improved image quality and improved safety as the source can be switched off. Necessary corrections for beam hardening effects can be based on measured X-ray transmission.
To simplify the following discussion, we limit ourselves to considering two specific instances of dual-beam radiography systems, namely:
a dual-energy 5 and 9 MeV interlaced high-energy X-ray scanner; and a FNGR scanner using DT neutrons and 60Co gammarays.
Table 4 compares the penetration and R-values of the two systems. The penetration figures show the thickness of each material through which 0.1% of the least penetrating of the two types of radiation associated with each system is transmitted. The R-value figures show the ratio of the logarithms of the transmissions of the two types of radiation, at 30 g cm2 material depth for the dual-energy X-ray scanner. The R-value for the neutron/gamma-ray scanner is independent of material thickness. As can be seen, the penetration of the dual-energy X-ray scanner is larger than that of the neutron/gamma-ray scanner for all materials, due to the high-energy radiation used. The penetration of the neutron/gamma-ray scanner is limited by the neutron transmission for organic materials and by the 60Co gamma-ray transmission for the metals. Replacing the 60Co source with a 4–5 MeV high-energy X-ray source would increase the penetration of the neutron/ gamma-ray scanner for all except organic materials. The most striking advantage of the FNGR technique is the wide range of R-values between different materials (Fig. 7). There is a factor of 7.5 variation in R-values between polythene and the heaviest metals, compared to a factor of 1.2 variation for the dual-energy X-ray method. The variation in R is also more uniformly spread, with good separation between different classes of organic material and light and heavy metals. In contrast, different organic materials are virtually indistinguishable using the Table 4 Comparison of penetration and R-value discrimination of dual-energy X-ray and neutron/gamma-ray scanners Material
Polythene C Al Fe Pb U
Dual X-ray scanner
Neutron/gamma-ray scanner
Penetration (g cm2)
R-value
Penetration (g cm2)
R-value
172 193 191 183 154 147
0.789 0.797 0.826 0.873 0.960 0.954
60 106 129 132 121 112
1.781 1.153 0.712 0.526 0.267 0.237
Note: By replacing the 60Co source in the neutron/gamma-ray scanner with a 4–5 MeV, high-energy X-ray source would significantly increase penetration for all except organic materials.
ARTICLE IN PRESS Y. Liu et al. / Applied Radiation and Isotopes 66 (2008) 463–473
Fig. 7. R-values for various materials for the dual-energy 9 and 5 MeV X-ray scanner and a fast neutron/gamma-ray scanner using DT neutrons and 60Co gamma-rays.
dual-energy X-ray approach and most of the variation is between light (Zo15) and heavy (Z425) elements. To investigate the material discrimination abilities of the two methods further, it is necessary to estimate the accuracies with which R-values can be determined. Contributions to the uncertainty in R-value determination arise from three main sources (Chen et al., 2007):
Counting statistics, due to the finite number of radiation quanta received at the detectors. Deviations from narrow-beam transmission due to radiation scattering. Instability in the beam energies of the Bremsstrahlung X-ray source and variations in the energy spectrum with angle.
The magnitudes of these uncertainties are not straightforward to estimate accurately. However, some approximations can be made. The I0 neutron count rate for the neutron/gamma-ray scanner described in Eberhardt et al. (2006) is approximately 1000 counts per pixel. The I0 count rates for highenergy X-ray-based systems can be estimated if assumptions are made about the source strength and detector geometry. Assuming a 5 MeV source producing 1 Gy min1 at 1 m, an 8 m source/detector distance, 5 5 mm detectors and a 10 m min1 scan rate, an I0 rate of approximately 100,000 counts per pixel is obtained. In our FNGR scanners (Eberhardt et al., 2006), we observe deviations from narrow-beam transmission due to neutron and gamma-ray scattering. These amount to approximately 10% and 1% for the two radiations; that is, 10% of detected neutrons and 1% of detected gammarays are scattered rather than directly transmitted. A similar figure for X-ray scattering is reported in Chen
471
et al. (2007). Fortunately, most of this scattering is highly local, between closely neighbouring pixels, and can be corrected rather accurately. In most parts of the image (except near object edges where the radiation attenuation is varying extremely rapidly), we estimate a residual uncertainty after scattering correction of 1% of I/I0 for neutrons and 0.1% for gamma-rays. Variations in electron beam energy used in the Bremsstrahlung source produce significant shifts in the measured R-values. For example, a shift of 0.5 MeV in the higher beam energy changes the R-value for iron by 0.01 and the R-value for organic materials by 0.02. In the absence of manufacturers’ information on beam energy stability, we do not include beam instability effects in the total uncertainties discussed below, but they represent a potentially significant source of uncertainty in material identification using a dual-energy X-ray system. Combining the first two sources of uncertainty in quadrature, we can estimate the total R-value uncertainty per pixel for the two scanner systems as a function of material thickness. These estimated R-value uncertainties are used to calculate a figure of merit as shown in Fig. 8, determined as the ratio of R-value uncertainties divided by R-value differences for graphite and polythene (top graph) and iron and polythene (lower graph). Graphite and polythene have been selected to illustrate the capability of the techniques to distinguish various classes of organic materials and iron and polythene have been selected to illustrate the capability of the techniques to distinguish organic materials from metal. The figures of merit are plotted as a function of material (iron) thickness for dualenergy X-ray and FNGR scanners. The figures of merit for a single pixel and also averaged over an area of 100 pixels are shown. Note that the higher the ratio, the greater the ability of the scanner to distinguish the two materials; a ratio less than unity means that the scanner is essentially unable to distinguish between the two materials. The material discrimination abilities of the FNGR scanner, expressed as the ratio of the R-value uncertainties to the R-value difference between two materials, are generally superior to the dual-energy X-ray scanner. The ability of the dual-energy X-ray scanner to discriminate between different organic materials is very limited, even averaged over 100 pixels. In contrast, the uncertainty for the FNGR scanner at the scan thickness is only 4–7% of the R-value difference between polythene and carbon, showing that the FNGR system can distinguish between a wide range of organic materials. Both the dual-energy X-ray and FNGR scanners are able to effectively distinguish between organic materials and steel and also to distinguish heavy metals (lead, uranium) from common metals such as steel and copper. In cluttered environments such as air cargo, it is relatively unusual to find regions of scan images comprising a single material. Rather, the integration along the radiation beams that is inherent in 2D radiography means that many overlapping materials are often present. In these
ARTICLE IN PRESS 472
Y. Liu et al. / Applied Radiation and Isotopes 66 (2008) 463–473
High-energy X-ray systems have the advantage of better penetration and higher count rates than neutron systems. However, neutron systems have the advantage of much better sensitivity of R-value measurement compared to dual high-energy X-ray techniques. Both dual high-energy X-ray and neutron scanning techniques are suitable for widespread deployment in an airport environment using currently available technology and can meet essential requirements such as rapidly scanning consolidated cargo and radiation safety. Both technologies produce high-resolution images, can readily identify common metal items by shape and have the potential to detect nuclear materials such as uranium, plutonium and common radiation shielding materials such as lead and tungsten. FNGR offers the additional significant advantage of discriminating between a wide range of organic materials. This improved discrimination assists operators in interpreting the complex, highly cluttered cargo images that result from scanning air cargo and have the potential to facilitate in the detection of narcotics, explosives and other organic threat materials.
References
Fig. 8. Figure of merit, calculated as R-value errors divided by the difference in R-value between (a) graphite and polythene and (b) iron and polythene. Both graphs are plotted versus material thickness for dualenergy X-ray and FNGR scanners. Results for single and 100 pixel average measurements are shown.
circumstances, it is important that the material discrimination figure-of-merit is significantly larger than unity, to allow regions of varying material composition to be picked out by image analysts. 6. Discussion Recent developments and applications of dual-beam radiography for cargo inspection have been reviewed. These recent developments include dual high-energy X-ray techniques as well as fast neutron and gamma-ray (or X-ray) radiography systems. Both techniques can readily discriminate between metals and organics, and between high-atomic number metals such as lead and uranium and common metals such as aluminium and steel.
Air Freight Council of NSW, 2007. /http://www.airfreightnsw.com.au/html/ s02_article/article_view.asp?id=131&nav_cat_id=148&nav_top_id=60S. Bartle, C.M., 1995. Method and apparatus for detecting concealed substances. US Patent No. 5,479,023, 26 December 1995. Bjorkholm, P.J., 2004. Dual energy detection of weapons of mass destruction. Port Technol. Int. 22, 155–158. Budner, G.J., 2006. Dual energy detection of weapons of mass destruction. Proc. SPIE 6178, G1–G9. Buffler, A., 2004. Contraband detection with fast neutrons. Radiat. Phys. Chem. 71, 853–861. Chen, Z., Wang, X., 2006. Cargo X-ray imaging technology for material discrimination. Port Technol. Int. 30, 163–165. Chen, G., Bennet, G., Perticone, D., 2007. Dual-energy X-ray radiography for automatic high-Z material detection. Nucl. Instrum. Methods Phys. Res. B 261, 356–359. Eberhardt, J.E., Rainey, S., Stevens, R.J., Sowerby, B.D., Tickner, J.R., 2005. Fast neutron radiography scanner for the detection of contraband in air cargo containers. Appl. Radiat. Isot. 63, 179–188. Eberhardt, J.E., Liu, Y., Rainey, S., Roach, G.J., Stevens, R.J., Sowerby, B.D., Tickner, J.R., 2006. Fast neutron and gamma-ray interrogation of air cargo containers. International Workshop on Fast Neutron Detectors, Cape Town, 3–6 April 2006, Proceedings of Science (FNDA2006) 092. /http://pos.sissa.itS. Findlay, D.J.S., 1989. Analytic representation of Bremsstrahlung spectra from thick radiators as a function of photon energy and angle. Nucl. Instrum. Methods A 276, 598–601. Gozani, T., 2002. A review of neutron based non-intrusive inspection technique. In: Conference on Technology for Preventing Terrorism, Hoover Institution National Security Forum, Stanford University, 12–13 March 2002, p. 12. Kang, K., et al., 2007. Method and equipment for discriminating materials by employing fast neutron and continuous spectral X-ray. US Patent Application Publication No. US 2007/0096036, 3 May 2007. Kawrakov, I., Rogers, D.W.O., 2006. The EGSnrc Code System, Monte Carlo Simulation of Electron and Photon Transport, NRCC Report PIRS-701.
ARTICLE IN PRESS Y. Liu et al. / Applied Radiation and Isotopes 66 (2008) 463–473 Miller, T.G., Van Staagen, P.K., Gibson, B.C., Krauss, R.A., 1996. Contraband identification in sealed containers using neutron transmission. SPIE Proc. 2867, 215–218. Mishin, A.V., 2005. Advances in X-band and S-band linear accelerators for security, NDT and other applications. In: Proceedings of the 2005 Particle Accelerator Conference, Knoxville, TN, IEEE, 240244. Neale, W.W., Rushbrooke, J.G., Ansorge, 1996. Material identification using X-rays. United States Patent No. 5,524,133, 4 June 1996. NIST, 2007. /http://physics.nist.gov/hysRefData/X-rayMassCoeffS (Table 1). Ogorodnikov, S., Petrunin, V., 2002. Processing of interlaced images in 4–10 MeV dual energy customs system for material recognition. Phys. Rev. Special Top.—Accelerators Beams 5, 104701.
473
Rushbrooke, J.G., Neale, W.W., Ansorge, R.E., Hooper, C.E., Lacey, R.J., Head, L., 1997. High energy material discrimination (HEMD) in the X-ray imaging of cargo containers. In: Proceedings—Harnessing Technology to Support the National Drug Control Strategy, Chicago, August 1997, pp. 6–7, 6–19. Sowerby, B.D., Tickner, J.R., 2003. Radiographic equipment. International Patent Application No. WO2004053472PCT/AU03/01641, 10 December 2003. Sowerby, B.D., Tickner, J.R., 2007. Recent advances in fast neutron radiography for cargo inspection. Nucl. Instrum. Methods A 580, 799–802. Tang, Ch., Chen, H., Liu, Y., Wang, X., 2006. Low energy linacs and their application in Tsinghua University, In: Proceedings of LINAC 2006, Knoxville, TN, pp. 256–258.