Dual energy method of material recognition in high energy introscopy systems
Element analysis based on so-called dual energy method is widely used throughout the world in X-ray customs inspection systems for luggage control in airports. It facilitates the routine work of customs officer on identification of illegal drugs and explosives hidden in luggage. Due to the absorptio...
Збережено в:
| Дата: | 1999 |
|---|---|
| Автори: | , , |
| Формат: | Стаття |
| Мова: | English |
| Опубліковано: |
Національний науковий центр «Харківський фізико-технічний інститут» НАН України
1999
|
| Назва видання: | Вопросы атомной науки и техники |
| Онлайн доступ: | http://dspace.nbuv.gov.ua/handle/123456789/81505 |
| Теги: |
Додати тег
Немає тегів, Будьте першим, хто поставить тег для цього запису!
|
| Назва журналу: | Digital Library of Periodicals of National Academy of Sciences of Ukraine |
| Цитувати: | Dual energy method of material recognition in high energy introscopy systems/ V.L. Novikov, S.A. Ogorodnikov, V.I. Petrunin // Вопросы атомной науки и техники. — 1999. — № 4. — С. 93-95. — Бібліогр.: 2 назв. — англ. |
Репозитарії
Digital Library of Periodicals of National Academy of Sciences of Ukraine| id |
irk-123456789-81505 |
|---|---|
| record_format |
dspace |
| spelling |
irk-123456789-815052015-05-18T03:02:15Z Dual energy method of material recognition in high energy introscopy systems Novikov, V.L. Ogorodnikov, S.A. Petrunin, V.I. Element analysis based on so-called dual energy method is widely used throughout the world in X-ray customs inspection systems for luggage control in airports. It facilitates the routine work of customs officer on identification of illegal drugs and explosives hidden in luggage. Due to the absorption rate difference in material of X-rays generated by sources with different energies, discrimination of materials becomes possible. So the scanned image of inspected object can be represented in physical palette where materials are coloured according to their atomic number. 1999 Article Dual energy method of material recognition in high energy introscopy systems/ V.L. Novikov, S.A. Ogorodnikov, V.I. Petrunin // Вопросы атомной науки и техники. — 1999. — № 4. — С. 93-95. — Бібліогр.: 2 назв. — англ. 1562-6016 http://dspace.nbuv.gov.ua/handle/123456789/81505 en Вопросы атомной науки и техники Національний науковий центр «Харківський фізико-технічний інститут» НАН України |
| institution |
Digital Library of Periodicals of National Academy of Sciences of Ukraine |
| collection |
DSpace DC |
| language |
English |
| description |
Element analysis based on so-called dual energy method is widely used throughout the world in X-ray customs inspection systems for luggage control in airports. It facilitates the routine work of customs officer on identification of illegal drugs and explosives hidden in luggage. Due to the absorption rate difference in material of X-rays generated by sources with different energies, discrimination of materials becomes possible. So the scanned image of inspected object can be represented in physical palette where materials are coloured according to their atomic number. |
| format |
Article |
| author |
Novikov, V.L. Ogorodnikov, S.A. Petrunin, V.I. |
| spellingShingle |
Novikov, V.L. Ogorodnikov, S.A. Petrunin, V.I. Dual energy method of material recognition in high energy introscopy systems Вопросы атомной науки и техники |
| author_facet |
Novikov, V.L. Ogorodnikov, S.A. Petrunin, V.I. |
| author_sort |
Novikov, V.L. |
| title |
Dual energy method of material recognition in high energy introscopy systems |
| title_short |
Dual energy method of material recognition in high energy introscopy systems |
| title_full |
Dual energy method of material recognition in high energy introscopy systems |
| title_fullStr |
Dual energy method of material recognition in high energy introscopy systems |
| title_full_unstemmed |
Dual energy method of material recognition in high energy introscopy systems |
| title_sort |
dual energy method of material recognition in high energy introscopy systems |
| publisher |
Національний науковий центр «Харківський фізико-технічний інститут» НАН України |
| publishDate |
1999 |
| url |
http://dspace.nbuv.gov.ua/handle/123456789/81505 |
| citation_txt |
Dual energy method of material recognition in high energy introscopy systems/ V.L. Novikov, S.A. Ogorodnikov, V.I. Petrunin // Вопросы атомной науки и техники. — 1999. — № 4. — С. 93-95. — Бібліогр.: 2 назв. — англ. |
| series |
Вопросы атомной науки и техники |
| work_keys_str_mv |
AT novikovvl dualenergymethodofmaterialrecognitioninhighenergyintroscopysystems AT ogorodnikovsa dualenergymethodofmaterialrecognitioninhighenergyintroscopysystems AT petruninvi dualenergymethodofmaterialrecognitioninhighenergyintroscopysystems |
| first_indexed |
2025-07-06T06:29:12Z |
| last_indexed |
2025-07-06T06:29:12Z |
| _version_ |
1836877987254894592 |
| fulltext |
DUAL ENERGY METHOD OF MATERIAL RECOGNITION IN HIGH
ENERGY INTROSCOPY SYSTEMS
V. L.Novikov, S. A.Ogorodnikov, V. I.Petrunin
Scientific Production Complex of Linear Accelerators and Cyclotrons of the D.V.Efremov
Scientific Research Institute of Electrophysical Apparatus, St.-Petersburg, Russia
INTRODUCTION
Element analysis based on so-called dual energy
method is widely used throughout the world in X-ray
customs inspection systems for luggage control in
airports. It facilitates the routine work of customs
officer on identification of illegal drugs and explosives
hidden in luggage. Due to the absorption rate difference
in material of X-rays generated by sources with
different energies, discrimination of materials becomes
possible. So the scanned image of inspected object can
be represented in physical palette where materials are
coloured according to their atomic number.
METHOD
The physical principle of dual energy method is
based on the fact that due to the exponential law of
gamma radiation attenuation the ratio of logarithmic
transparencies at nominal and dual energy
( ) ( )
( )
( )
( )ZE
ZE
T
TZEE
tot
tot
,
,
ln
ln
,,
2
1
2
1
21
γ
γ
γγ µ
µ
δ == (1)
characterizes material of the barrier irrespective to its
thickness.
As far as effectiveness of dual energy method is
determined by Z-dependence of total attenuation
coefficient, it is evident that X-ray energy range <0.5
MeV, where the process of photoelectric interaction
dominates with its strong Z-dependence (στ∼Z5), is the
most preferable for element analysis. Opposite to the
photo-effect Compton scattering with µc∼Z/A has poor
Z-dependence. This ratio is approximately the same for
elements from, at least, top of periodic table,
compositions of those define the whole variety of
organic substances. However our goal is to investigate
the possibility of materials discrimination for the energy
range 1<E<10 MeV, where Compton scattering
dominates. One can see the little difference between
total attenuation coefficients of different elements for
energies from 1÷10 MeV range on Fig. 1.
Among many factors that stipulate its low
sensitivity the major one is the influence of scattered
radiation. Therefore we should carefully eliminate one
by collimating of the photon beam and all subsequent
formulas and calculations are fulfilled for the narrow
photon beam.
As far as we use flow of bremsstrahlung quanta
with continuous spectral distribution in our work, first
let us introduce new transparency T (inverse value of
absorption) of a barrier with atomic number Z and
thickness t (gr/cm2) for flow of bremsstrahlung with
boundary energy Eac (MeV).
( )
( ) ( )
( )∫
∫
⋅
⋅⋅
=
−
Eac
ac
Eac
tZE
ac
ac
dEEE
dE
dP
dEeEE
dE
dP
ZtET
0
0
,
,
,
,,
γγ
γ
γ
γµ
γ
γ , (2)
where energy distribution proportional to signal of
detector is a product of spectral distribution of
bremsstrahlung intensity according to Schiff formula [1]
and detector response factor.
Fig. 1. Total attenuation coefficient [2] as function of
energy for Carbon (Z=6), Aluminium (Z=13), Ferrum
(Z=26) and Plumbum (Z=82). Red curves are spectral
distributions of bremsstrahlung quanta (rel. units)
emerged from thick tungsten target for dual energies of
electron beam 8.8 MeV and 4.4 MeV.
As a first approach to the goal we estimated the
effectiveness of value introduced in (1) with newly
introduced transparencies for the purpose of material
discrimination. Unfortunately due to the continuous
character of bremsstrahlung spectrum δ depends on
thickness of barrier and therefore is not applicable for
the purpose of discrimination. Nevertheless, our first
approach has one important drawback. We restricted
ourselves with taking into account the ratio of
transparencies at nominal Eacc1 and dual Eacc2 energies
therefore a part of information has been lost. In order to
get rid of it we represent clusters on two-dimensional
absorption plane (inverse value of transparency), where
the most of materials can be discriminated irrespective
to their thickness since intersection of two data sets is
always lesser than their union.
Fig. 2 shows thickness curves for range from 0 to
120 gr/cm2 of a set of elements, where X-axis represents
inverted logarithmic transparency α1 (absorption) scaled
by factor Scale=1000 for nominal energy Eacc1=8 MeV
and Y-axis – (α2-α1) – difference of absorptions for
dual Eacc2=4 MeV and nominal energies:
( )
( ) ScaleTT
ScaleT
⋅−=−
⋅−=
2112
11
lnln
ln1
αα
α
(3)
ВОПРОСЫ АТОМНОЙ НАУКИ И ТЕХНИКИ. 1999. № 4.
Серия: Ядерно-физические исследования (35), с. 93-95.
93
Fig.2 Absorption curves of different elements for
thickness range from 0 to 120 gr/cm2. Energies: nominal
Eacc1= 8MeV and dual Eacc2=4MeV.
From figure one can see that absorption curves of
elements are separated for the whole thickness range,
except group of heavy metals (plumbum, tungsten) and
diverge as thickness increases. Now, therefore, our aim
is to formulate the recognition task in mathematical
terms for the common case.
Suppose we have two experimentally measured
transparencies at nominal and dual energies of a barrier
of unknown material and thickness: T1 exp, T2 exp. Our
goal is to associate them with material (Z) and
optionally with thickness t. So, using (2) we can write
system of two non-linear integral equations:
( )
( ) ( )
( )
( )
( ) ( )
( )
⋅
⋅⋅
==
⋅
⋅⋅
==
∫
∫
∫
∫
−
−
2
0
2
2
0
,
2
2exp2
1
0
1
1
0
,
1
1exp1
,
,
,,
,
,
,,
Eac
ac
Eac
tZE
ac
ac
Eac
ac
Eac
tZE
ac
ac
dEEE
dE
dP
dEeEE
dE
dP
ZtETT
dEEE
dE
dP
dEeEE
dE
dP
ZtETT
γγ
γ
γ
γµ
γ
γ
γγ
γ
γ
γµ
γ
γ
(4)
Thus we have two non-linear equations and two
unknown quantities: material and its thickness.
Formally this system must have solution as the
transparencies are obtained from experimental
measurements, but this solution might be not unique (as
we have seen that for heavy metals). This means that the
same pair of transparencies may correspond to a few
different materials with different thickness. Adding the
next measurements (for example, triple energy method
and so on) we may theoretically eliminate this
uncertainty, but practical realization of it is too
complicated. The next difficulty is how to solve this
system. If the first variable - thickness - is a continued
variable then second one - material - is discrete. The
most common method that has been used, is
minimization of functional. Let’s build positive form as
( )( ) ( )( )2
exp22
2
exp11 ,,,, TZtETTZtET
Functional
acac −+−
=
(5)
and start to minimize it as two-dimensional function.
Taking into account that Z variable is discrete and
number of materials is limited it is possible to search
minimum for thickness variable only and move along Z.
Minimum searching of one variable function can be
fulfilled either by Fibbonachi or Gold Cross-section
method. Graphically the minimization of functional (5)
can be represented as searching of the shortest distance
between experimental point (pixel on image) with
transparencies Texp1, Texp2 (logarithmic absorptions αexp1,
αexp2) and a curve of certain material from the set of
absorption curves. If procedure of minimization delivers
nil value functional with certain precision, this asserts
that material of barrier and its thickness are found and
that can be graphically represented for example by
colorizing of pixel on data image. This method
effectively works for the most of substances, but has
one major obstacle on the way of its practical
realization. The procedure of discrimination takes long
time and cannot be performed for each pixel in real time
mode because number of image pixels might reach
several millions. Therefore we can apply this method
locally only, for small zone of experimental image.
As we mentioned before solving of system of
non-linear equations (4) is a task that cannot be
performed in real time mode. The system is to be solved
for each data pixel and their number in image might
exceed million. Nevertheless the number of
mathematical operations fulfilled by computer processor
can be dramatically reduced and material recognition
procedure becomes possible for large data images. The
first main time eating procedure is calculation of
transparencies at high and low energies T1(Eac1,t,Z),
T2(Eac2,t,Z) in (5) can be substituted by their
interpolation algorithm. Indeed as far as we evaluated
high and low energy (we used the absorption method)
and also due to the fact that Z variable is discrete we can
build material recognition base for the limited number
of materials. The base represents itself as a set of frame
points:
( )( ) ( )( )( )jijiiij ZTLnZTLntR 21 ,,= , ni ...1= , mj ...1= (6)
where n – number of frame points, m – number of
materials. This set serves as interpolation base for
evaluation of high and low transparencies. As far as
transparency is exponential function of thickness, for
minimization of interpolation error its logarithmic
values were taken. Second, criterion of material
discrimination is minimization of Functional (5), which
also can be drastically simplified. Instead of searching
the shortest distance between experimental point
defined by a pair (T1exp,T2exp) and the closest thickness
curve from recognition set we use a new criterion.
Indeed, solving each of equations (4) relatively
thickness t with fixed Z, we can correspond high T1exp
and low T2exp experimental transparencies to the
thicknesses t1exp and t2exp. It is evident that if Z is chosen
right the both thicknesses must coincide. Therefore our
new criterion for material discrimination is evaluation
of the ratio:
ВОПРОСЫ АТОМНОЙ НАУКИ И ТЕХНИКИ. 1999. № 4.
Серия: Ядерно-физические исследования (35), с. 93-95.
93
( ) ( )
( ) ( )ii
i
ii ZTol
Zt
ZtZt
≤
−
exp1
exp2exp1 mi ,...,1= (7)
where m – number of chosen materials for recognition.
Though from the common point of view it looks
unusual to operate with two virtual thicknesses while
the physical value is unique. Nevertheless due to the
low sensitivity of method and noise of experimental
data the introduction of tolerance for each material in
recognition base is required.
So, now the discrimination procedure is
simplified considerably and requires only estimation of
ratio (7) for each material from recognition set. It does
not require numerous iterations and its duration is
defined by the number of materials m.
EXPERIMENTAL RESULTS
In order to get approval of our method a series of
experiments were carried out at NPK LUTS NIIEFA
stand "Inspector" equipped with:
• Industrial linear electron accelerator UELV-10-2D-
40 on travelling wave with energy at the nominal
mode up to 7-10 MeV and at dual 3-7 MeV. Pulse
duration – 5 µsec; repetition rate – 100 Hz
(interlaced mode).
• Collimating system of bremsstrahlung beam
consisting of two beam forming units and
collimator of detector line. Narrowed beam scans in
vertical plane.
• Detection line consists of 5 modules. Each module
contains 128 channels of detection; as a sensitive
element p-i-n diodes in combination with
scintillator crystals are used; a pitch of sensitive
elements in the array - 3,5 mm.
• System of transportation of tested samples across
radiation beam.
The procedure of experiment, data processing
and visualization were in the following order:
1. The electron accelerator operated at so called
interlaced mode, when its each even pulse
generates nominal energy and each odd – dual;
2. Raw data interlaced file is decomposed into high
energy and low energy image files;
3. Each file is processed separately: non-uniformity of
detectors and also angular anisotropy of
bremsstrahlung intensity are eliminated by means
of vertical correction of image;
4. Data counts are converted into transparencies by
means of normalization to the white field count
level;
5. Data of both files are taken logarithmic and one of
them is visualized in gray palette; each pixel on
image is characterized by two transparencies;
6. Both energies of accelerator are evaluated by
absorption method;
7. Material recognition is fulfilled by means of
processing of each pixel and its colorizing on
successful completion according to material
recognition palette.
In a series of experiments we used flat and
wedge samples made of polyethylene, duralumin, steel
and lead. Data processing according to above presented
method allowed us to discriminate and coloured
samples irrespective to their thickness (Fig. 3).
Fig. 3. Calculated absorption curves and spreads of
experimental data of polyethylene (50cm), duralumin
(40cm), steel (18cm) and lead (8cm) wedge samples at
Enom=9.04 MeV and Edual=4.65 MeV.
CONCLUSION
Theoretical consideration and experimental
results have shown that above represented algorithm
allows to carry out discrimination of materials
according to their atomic number and irrespective to
their thickness at dual energies of accelerator from 1÷10
MeV range. Low sensitivity of the method, noise level
and non-linearity of detectors restricted discrimination
in our experiments with Z-precision ~10. In order to
perform effective recognition of materials for custom
inspection purposes the installation has to satisfy the
following requirements:
• Accelerator must have high energy stability;
nominal and dual energy of accelerator must be
measured with proper precision;
• Accuracy of transparencies measurement must be
enough in order to ensure the third-fourth
significant digit, therefore the number of bits of
ADC must equals at least 16;
• Signal to noise ratio should be enough to assure the
correct transparency evaluation, therefore during
scanning averaging might be applied;
• The theory is based on the linear approach,
therefore the linearity of detectors has to be
provided;
• Speed of processor should be enough to fulfill the
necessary processing of large volumes of data;
• Discrimination ambiguity at small thickness of the
barrier requires the necessity of bremsstrahlung
filtering.
REFERENCES
[1] Shiff L. J. // Phis. Rev. 1951 v 83, Num 2.
[2] Storm E., Jsrael H. Photon Cross Section from 0.001
to 100 MeV for Elements 1 through 100. - Los Alamos
Scientific Laboratory, New Mexico, 1967.
ВОПРОСЫ АТОМНОЙ НАУКИ И ТЕХНИКИ. 1999. № 4.
Серия: Ядерно-физические исследования (35), с. 93-95.
93
|