System and method for radiant energy spectral classification

Abstract

A method of determining whether a radiation spectrum includes radiation corresponding to a naturally occurring radiation material source (NORM), the method comprising: collecting the radiation spectrum; normalizing the collected spectrum by using one or more predetermined normalization techniques; energy scaling the normalized spectrum by using one or more energy scaling techniques; and determining a material corresponding to each energy reading in the scaled energy spectrum by using one or more classification algorithms to classify the determined material as one or more predetermined categories.

Description

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS

[0002] This specification relies on U.S. Patent Provisional Application No. 63 / 374,272, entitled “Systems and Methods for Classification of Radiation Energy Spectra,” filed on September 1, 2022, for priority. The above application is incorporated herein by reference in its entirety. Technical Field

[0003] Embodiments of the present specification relate to the field of radiation-based inspection systems. Specifically, embodiments of the present specification relate to systems and methods for material discrimination based on real-time classification of detected energy spectra generated from radiation-based inspection systems. Background Art

[0004] Trade fraud, smuggling and terrorism have increased the demand for non-intrusive inspection systems (such as those based on X-ray, gamma ray and linear accelerator (LINAC)) for inspecting cargo, trucks, buses and other transportation systems, which effectively provide goods to move across borders. In addition, they also provide opportunities to include contraband such as weapons, explosives, illegal drugs and precious metals. In particular, non-intrusive inspection systems are used for applications ranging from roadside inspections of parked vehicles to scanning in crowded or high-traffic ports. The term port, although usually considered to refer to seaports, also applies to land border crossings or any port of entry.

[0005] Radiation Portal Monitors (RPMs) are passive systems used to detect the presence of a source of ionizing radiation passing through it and are typically used for security at borders or ports of entry. These systems are used to scan vehicles, cargo containers, rail cars, and pedestrians. In addition, similar systems are used to inspect scrap metal at steel recycling plants where radioactive sources may be melted and accidentally mixed with the scrap steel, representing a huge cost caused by millions of dollars of crucible contamination.

[0006] Naturally occurring radioactive materials (NORM) can cause these systems to alarm, requiring the operator of the system to stop traffic in order to investigate the cause of the alarm. NORM sources include isotopes such as Ra-226, Th-232 / Tl-208, K-40, and U-238. While the presence of NORM sources is generally acceptable for these applications, they represent "nuisance" alarms. The goal of these systems is to try to reduce these NORM nuisance alarms to a level below 1% as much as possible, as NORM sources are the cause of most nuisance alarms from radiation inlets.

[0007] Conventionally, the reduction of nuisance alarms due to the presence of NORM substances requires the use of expensive detectors that provide increased energy resolution in order to identify the specific NORM isotopes that cause the alarm. High energy resolution detectors are used to determine specific energy peaks within the detected spectrum due to NORM isotopes. Therefore, the method uses a very specific portion of the detected energy spectrum to determine the presence of a NORM source. Other known methods of reducing nuisance alarms due to the presence of NORM substances require the use of algorithms for performing clustering / binning of similar energy radiation, and then applying the decision to sound an alarm to the clustered radiation based on the ratio of counts in these energy bins to try and infer that the alarm is attributed to the NORM source. Even with automated systems, it becomes difficult to comply with the implicit requirement of keeping the number of false alarms low when the system is operated at high throughput.

[0008] Therefore, there is a need for a system and method for reducing nuisance alarms in radiation-based inspection portals. There is also a need for a system and method that uses a low-cost detector and performs analysis of the detected radiation spectrum as a whole in a machine learning framework to determine whether the alarm is caused by NORM materials. Summary of the invention

[0009] The following embodiments and aspects thereof are described and illustrated in conjunction with systems, tools, and methods, which are intended to be exemplary and illustrative, not limiting in scope.This application discloses many embodiments.

[0010] The present specification discloses a radiation portal monitor, which includes multiple detectors in data communication with a processing system, wherein the processing system is configured to determine whether a radiation spectrum includes radiation corresponding to a naturally occurring source of radiation material (NORM) by: generating a database of training spectra, each training spectrum in the database representing a NORM material; collecting radiation spectra; normalizing the collected spectra by using one or more predetermined normalization techniques; energy scaling the normalized spectra by using one or more energy scaling techniques; fitting the normalized and scaled spectra to each training spectrum in the database; determining a residual of fitting the normalized and scaled spectrum for each training spectrum; and identifying the normalized and scaled spectrum as a NORM material based on the residual and a threshold.

[0011] Optionally, if the residual is smaller than a threshold, the normalized and scaled spectrum is determined to be NORM material.

[0012] Optionally, the processor is further configured to identify the normalized scaled spectrum as belonging to an "other" category if the residual is greater than a threshold.

[0013] Optionally, the processor is configured to empirically determine the threshold based on a training spectrum and a residual of a scaled spectrum normalized by fitting a class of training spectra.

[0014] Optionally, collecting the radiation spectrum includes collecting an energy spectrum radiated by an inspected object in a radiation portal monitor by using one or more of a plurality of detectors. Optionally, the plurality of detectors are large area low resolution radiation detectors. Optionally, the radiation detector is a polyvinyl toluene (PVT) detector. Optionally, the radiation detector is a gamma radiation detector. Optionally, the radiation portal monitor is a passive system that does not include a radiation source. Optionally, the radiation portal monitor is a passive radiation portal monitor (RPM) for detecting the presence of an ionizing radiation source passing through the RPM. Optionally, the RPM includes a plurality of panels, wherein the panels are spaced apart to enable a cargo container to pass between two panels. Optionally, each panel includes a radiation detector.

[0015] Optionally, energy scaling the normalized spectrum comprises applying predetermined weights to the collected energy spectrum to obtain a weighting matrix of the spectrum.

[0016] Optionally, the one or more energy scaling techniques include a spectral dose index (SDI) technique. Optionally, the SDI technique adds more weight to higher energies in the normalized spectrum to help determine NORM materials.

[0017] Optionally, one or more training spectra used to train an Optimal Linear Associative Memory (OLAM) based artificial neural network are used to populate the database of training spectra. Optionally, the training spectra for the OLAM based artificial neural network include one spectrum for each isotope from the NORM material.

[0018] Optionally, an Optimal Linear Associative Memory (OLAM) neural network implemented within a Field Programmable Gate Array (FPGA) integrated circuit is used to determine whether the radiation spectrum includes radiation corresponding to a NORM material.

[0019] Optionally, the normalized and scaled spectrum is fitted to each training spectrum in the database in a sequential manner.

[0020] Optionally, the NORM material is one of the following: Ra-226, Th-232 / Tl-208, K-40 and U-238.

[0021] In some embodiments, the present specification discloses a method for determining whether a radiation spectrum includes radiation corresponding to a naturally occurring source of radiation material (NORM), the method comprising: generating a database of training spectra, each training spectrum in the database representing a NORM material; collecting radiation spectra; normalizing the collected spectra by using one or more predetermined normalization techniques; energy scaling the normalized spectra by using one or more energy scaling techniques; fitting the normalized and scaled spectra to each training spectrum in the database; determining the residual of fitting the normalized scaled spectrum for each class of training spectra; and identifying the normalized scaled spectrum as a NORM material based on the residual of fitting the normalized scaled spectrum with a class of training spectra and a threshold.

[0022] Optionally, if the residual of fitting the normalized scaled spectrum with a class of training spectra is less than a threshold, the normalized scaled spectrum is determined to be a NORM material.

[0023] Optionally, the method further comprises identifying the normalized scaled spectrum as belonging to the “other” category if a residual of fitting the normalized scaled spectrum with a category of training spectra is greater than a threshold.

[0024] Optionally, the threshold is determined empirically based on training spectra in a database and the residual of fitting a normalized scaled spectrum with a class of training spectra.

[0025] Optionally, collecting the radiation spectrum includes collecting an energy spectrum radiated by an object under inspection in a radiation inspection system by using one or more radiation detectors.

[0026] Optionally, the radiation detector is a large area low resolution detector. Still optionally, the radiation detector is a polyvinyl toluene (PVT) detector. Still optionally, the radiation detector is a gamma radiation detector.

[0027] Optionally, the radiation inspection system is a passive system that does not include a radiation source. Still optionally, the radiation inspection system is a passive radiation portal monitor (RPM) for detecting the presence of an ionizing radiation source passing through the RPM. Optionally, the RPM includes a plurality of panels, wherein the panels are spaced apart to enable a cargo container to pass between two panels. Optionally, each panel includes a radiation detector.

[0028] Optionally, energy scaling the normalized spectrum includes applying predetermined weights to the collected energy spectrum to obtain a weighting matrix of the spectrum.

[0029] Optionally, the one or more energy scaling techniques include a spectral dose index (SDI) technique. Still optionally, the SDI technique adds more weight to higher energies in the normalized spectrum to help determine NORM materials.

[0030] Optionally, one or more training sets used to train the OLAM-based artificial neural network are used to populate a database of training spectra.

[0031] Optionally, an Optimal Linear Associative Memory (OLAM) neural network implemented within a Field Programmable Gate Array (FPGA) integrated circuit is used to determine whether the radiation spectrum includes radiation corresponding to a NORM material.

[0032] Optionally, the training spectra for the OLAM-based artificial neural network include one spectrum for each isotope from the NORM material.

[0033] Optionally, the normalized and scaled spectrum is fit to each training spectrum in the database in a sequential manner.

[0034] Optionally, the NORM material is one of the following: Ra-226, Th-232 / Tl-208, K-40 and U-238.

[0035] The foregoing and other embodiments of the present specification will be described in more depth in the drawings and detailed description provided below. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] The accompanying drawings illustrate various embodiments of systems, methods, and embodiments of various other aspects of the present disclosure. It will be appreciated by those of ordinary skill in the art that the element boundaries (e.g., boxes, groups of boxes, or other shapes) shown in the figures represent an example of boundaries. It is possible that, in some examples, an element may be designed as multiple elements, or multiple elements may be designed as one element. In some examples, an element shown as an internal component of an element may be implemented as an external component in another element, and vice versa. In addition, the element may not be drawn to scale. A non-restrictive and non-exhaustive description is described with reference to the following drawings. The components in the drawings are not necessarily drawn to scale, but the emphasis is placed on the explanation principle.

[0037] Figure 1A An exemplary inspection system including a plurality of radiation portal monitors (RPMs) is illustrated in which the systems and methods of the present specification may be employed;

[0038] Figure 1B A radiation portal monitor (RPM) for inspecting cargo according to an embodiment of the present specification is illustrated;

[0039] Figure 1C A block diagram of a classification system according to an embodiment of the present specification is illustrated;

[0040] Figure 2 depicts a graph of a normalized and scaled energy-weighted spectrum according to an embodiment of the present specification;

[0041] Figure 3A is a flow chart illustrating a method of classifying an energy spectrum according to an embodiment of the present specification;

[0042] Figure 3B shows a normalized radiation spectrum of the radionuclide Ra-226 normalized with and without the use of an SDI constant according to an embodiment of the present specification;

[0043] Figure 3C illustrates a normalized radiation spectrum of the radionuclide K-40 normalized with and without the use of an SDI constant according to an embodiment of the present specification;

[0044] Figure 3D illustrates a normalized radiation spectrum of the radionuclide Th-232 normalized with and without the use of an SDI constant according to an embodiment of the present specification;

[0045] Figure 3E is a graph showing the difference in fitting of a background spectrum using a NORM material according to an embodiment of the present specification;

[0046] Figure 3F is a graph showing the difference in fit using the radionuclide Am-241 according to an embodiment of the present specification

[0047] Figure 3G is a graph showing the difference in fit using the radionuclide Co-57 according to an example of the present specification;

[0048] Figure 3H is a graph showing the difference in fit using the radionuclide Ba-133 according to an embodiment of the present specification;

[0049] Fig. 3I is a graph showing the difference in fit using the radionuclide Cs-137 according to an example of the present specification;

[0050] Figure 3J is a graph showing the difference in fit using the radionuclide Co-60 according to an example of the present specification;

[0051] Figure 4A illustrates normalized radiation spectra of various isotopes according to embodiments of the present specification; and

[0052] Figure 4B Pictured Figure 4A Spectra shown in FIG. 5 are weighted and classified into the categories “NORM” or “other”. DETAILED DESCRIPTION

[0053] In an embodiment, the present specification provides a system and method for reducing false alarms in a radiation-based inspection system. In an embodiment, the system and method of the present specification provides a rapid real-time classification that quickly classifies the detected energy spectrum from the radiation-based inspection system into naturally occurring radiation material sources (NORM) or "other" categories and the percentage of matches to each category. NORM sources include isotopes such as Ra-226, Th-232 / Tl-208, K-40, and U-238.

[0054] This specification relates to multiple embodiments. The following disclosure is provided to enable one of ordinary skill in the art to practice the present invention. The language used in this specification should not be interpreted as a general negation of any one specific embodiment or used to limit the meaning of the claims beyond the terms used therein. Without departing from the spirit and scope of the present invention, the general principles defined herein can be applied to other embodiments and applications. In addition, the terms and wordings used are for the purpose of describing exemplary embodiments and should not be considered restrictive. Therefore, the present invention will be given the widest scope including many substitutions, modifications and equivalents consistent with the disclosed principles and features. For the sake of clarity, details related to technical materials known in the technical field related to the present invention are not described in detail to avoid unnecessary obscurity of the present invention.

[0055] In the specification and claims of the present application, each of the words "includes," "comprising," "having," "containing," and forms thereof is not necessarily limited to the members of the list associated with the word. Therefore, they are intended to be equivalent in meaning and open-ended, in that one or more items following any of these words is not meant to be an exhaustive list of such one or more items, or to be limited to the listed one or more items. It should be noted herein that any feature or component described in association with a particular embodiment may be used and implemented with any other embodiment unless expressly indicated otherwise.

[0056] It must also be noted that, as used herein and in the appended claims, the singular forms "a," "an," and "the" include plural references unless the context dictates otherwise. Although any systems and methods similar or equivalent to those described herein can be used to practice or test embodiments of the present disclosure, the preferred systems and methods are presently described.

[0057] In various embodiments, the system includes at least one processor capable of processing program instructions, has a memory capable of storing the program instructions, and employs software consisting of a plurality of program instructions to perform the processes described herein.

[0058] In various embodiments, a computing device may be used to receive and process data signals and image data, and may include an input / output controller, at least one communication interface, and a system memory. The system memory includes at least one random access memory (RAM) and at least one read-only memory (ROM). These elements communicate with a central processing unit (CPU) to implement the operation of the computing device. In various embodiments, the computing device may be a conventional stand-alone computer, or alternatively, the functions of the computing device may be distributed over a network of multiple computer systems and architectures. In some embodiments, the execution of multiple program instructions or code sequences stored in one or more non-volatile memories enables or enables the CPU of the computing device to perform or enable various functions, processes, and algorithms, such as, for example, performing image reconstruction to display on a screen. In alternative embodiments, hard-wired circuits may be used instead of software instructions or in combination with software instructions to implement the processing of the systems and methods described in the present application. Therefore, the described systems and methods are not limited to any particular combination of hardware and software.

[0059] Figure 1A An inspection system including a plurality of radiation portal monitors (RPMs) according to an embodiment of the present description is shown. Figure 1A The radiation entrance monitor 1200 depicted in the embodiment includes a plurality of detector panels 1201, 1202, 1203, 1204, 1205 ... 120n, wherein each detector panel includes a plurality of radiation detectors. In an embodiment, the panels 1201, 1202, 1203, 1204, 1205 ... 120n are spaced apart at a distance to enable a cargo vehicle to pass between two panels. In an embodiment, the distance between the panels varies according to the size of the vehicle passing between the panels, and may be large enough to allow even the largest dump trucks used in steel recycling plants to pass. Thus, in an embodiment, the distance may vary from 1 meter to tens of meters. In other embodiments, the system may employ a single detector on either side without changing the operation / method described herein.

[0060] Figure 1B A radiation portal monitor (RPM) for inspecting cargo during use according to an embodiment of the present specification is illustrated. Figure 1BIn an embodiment, the RPM 1221 includes a first panel 1220 and a second panel 1222, the first panel 1220 and the second panel 1222 including one or more detectors for detecting radiation scattered by cargo 1224 traveling through and between the first panel 1220 and the second panel 1222, and a processor configured to process the detected radiation and provide information indicating whether the cargo includes any radiation source material. In an embodiment, the RPM 1221 does not include any radiation source, and the only radiation detected by the first panel 1220 and the second panel 1222 is radiation emitted from the cargo 1224 passing through the system. Therefore, in various embodiments, the inspection system of the present specification including multiple radiation portal monitors (RPMs) with multiple detector panels is a passive system that does not generate or emit X-rays for imaging purposes and operates only to detect radioactive materials traveling through the detector panels. The multiple panels can include gamma and neutron detector materials, such as polyvinyl toluene (PVT) plastic scintillator and LiZnS-PMMA, respectively. In an embodiment, the panel is capable of gamma and neutron detection and can detect and identify radionuclides.Such a modular drive-through inspection system can be configured to support the unique needs of the nuclear industry, law enforcement, private industry, research institutions, steel mills, scrap steel recycling plants, and federal agencies responsible for nuclear safety or public safety.

[0061] In an embodiment, the RPM 1221 can be used in conjunction with an active X-ray imaging system, particularly in conjunction with the application of blanking synchronization of the RPM system when emitting X-rays. In an embodiment, the method of the present specification for suppressing nuisance alarms due to NORM materials can be implemented in an RPM system included in an X-ray imaging system. U.S. Patent No. 8,963,094, entitled “Composite Gamma-Nutron Detection System” and filed on February 24, 2015, and U.S. Patent No. 10,393,915, entitled “Integrated primary and special nuclear material alarm resolution” and filed on August 27, 2019, assigned to the applicant of the present specification, are incorporated herein by reference in their entirety.

[0062] The classification systems and methods of the present specification enable the reduction of nuisance alarms in passive radiation portal monitors, wherein the classification information can be used to determine whether to notify an operator of an alarm based on the presence of a potential radiation-based threat. In embodiments, the classification systems and methods of the present specification can be used to reduce nuisance alarms in any passive radiation detection system, such as but not limited to portals, conveyors, handheld devices, scrap scanning systems, and area monitors, which passive radiation detection system is configured to generate an alarm when gamma radiation is detected and is used to measure the energy spectrum emitted from radioactive materials.

[0063] Figure 1C A block diagram of a classification system according to an embodiment of the present specification is illustrated. System 1100 includes a detection module 1102, a normalization module 1104, a scaling module 1106, and a classification module 1108. Detection module 1102 includes multiple radiation detectors for collecting energy spectra generated by an object being inspected in a radiation inspection system. In one embodiment, the radiation includes gamma rays. In an embodiment, multiple detectors are large-area low-resolution polyvinyltoluene (PVT) detectors, which have poor resolution and are not well suited for isotope identification. In an embodiment, conventional PVT detectors can be used without any physical modification to the detector. In an embodiment, detection module 1102 includes PVT detectors because they provide a large detection area and can be obtained at a low cost. In addition, PVT detectors can be programmed and calibrated / trained to more easily implement the requirements of classification module 1108 (described below) compared to high-resolution detectors. In a preferred embodiment, detection module 1102 includes one or more PVT detectors. In other embodiments, detection module 1102 includes any gamma ray detector with high or low resolution that can be used for isotope identification.

[0064] The normalization module 1104 normalizes the collected energy spectrum by using one or more predefined normalization techniques, as further described below.

[0065] The scaling module 1106 uses at least one predefined technique to energy scale the normalized energy spectrum. In one embodiment, but not limited to such an embodiment, the scaling module 1106 applies predetermined weights to the collected energy spectrum to form a weighting matrix of the spectrum by using a spectral dose index (SDI) technique, which helps to enhance high energy counts from isotopes that emit higher energy gamma radiation in the energy spectrum. In an embodiment, the SDI is a simple scaling to compensate for increases in detected gamma energy.

[0066] In one embodiment, the SDI technique includes determining the number of counts in a channel of a histogram of an energy spectrum, multiplying the determined counts by the number of channels, and summing all multiplied counts. As background, as used in the specification, a channel corresponds to a physically defined area of ​​a detector that generates a single electrical signal corresponding to the number of photons (also referred to as a "count") received at that physical location over a predetermined time period. In an embodiment, a scaling module 1106 enables amplification of higher energy isotopes from an object irradiated by gamma rays (which may correspond to a NORM source). In an embodiment, energy scaling causes the shape of the spectrum to be oriented toward higher energies to have more pronounced features. Energy scaling compensates for detector efficiency relative to detector energy. As is known, gamma ray detectors become less efficient as gamma ray energy increases. In an embodiment, the efficiency of the detector is measured for multiple gamma energies, and the resulting energy efficiency curve is used to improve the inefficient areas of the detector.

[0067] In an embodiment, the x-axis of the histogram of the energy spectrum depicts integer frequency bins 'x', where x = 1, 2, 3 ... n, representing the amplitude of the pulses collected from the gamma detector. The y-axis of the histogram of the energy spectrum represents the number of events occurring for each amplitude bin and is depicted as a function of x (i.e., F(x)), where x ranges from 1 to n, and F(x) ranges from 0 to the size of the variable selected to represent these values ​​(typically a 16-bit or 32-bit variable). In an embodiment, the SDI scaling technique is used to scale the function F(x) by multiplying each F(x) by its bin number x using the following formula:

[0068] SDI(x) = F(x)*x, where x = 1, 2, ... n. (Equation 1) Since the histogram of detector pulse amplitudes can be scaled so that the x-axis represents gamma energy, and the energy response of a gamma detector becomes less efficient as gamma energy increases, the SDI technique compensates for the reduced efficiency of the detector by increasing the number of events as the energy of the detected rays increases. Therefore, the use of the SDI technique adds more weight to higher energies in the spectrum, which helps determine NORM sources. For example, if the efficiency of the detector is 20% at 2 MeV, the counts received in the detector at that region are amplified by dividing the received counts by 0.2.

[0069] Figure 2 Graphs of normalized and scaled energy spectra according to embodiments of the present specification are illustrated. Graph 202 represents a normalized spectrum before energy scaling is applied to the spectrum. Graph 204 represents the energy spectrum of graph 202 after energy scaling the spectrum by applying one or more energy weighting techniques.

[0070] refer to Figure 1C, the classification module 1108 determines the material corresponding to each energy reading in the scaled energy spectrum and classifies each energy reading into one or more predefined categories. In one embodiment, the classification module classifies the energy spectrum into "NORM" or "other" categories and identifies the percentage that matches each category. In an embodiment, the "other" category marks all other materials / isotopes that are not NORM materials. Examples of NORM materials include, but are not limited to, radioactive isotopes of potassium (K-40), thorium (Th-232), uranium (U-238), and radium (Ra-226). In an embodiment, the classification module 1108 uses an artificial neural network based on optimal linear associative memory (OLAM) to classify the energy-scaled spectrum to determine energy sources belonging to the NORM category. In one embodiment, other machine learning techniques (such as but not limited to k-NN (k nearest neighbor) networks) can be used to classify the energy-scaled spectrum to determine energy sources belonging to the NORM category.

[0071] In various embodiments, in order to implement the OLAM-based implementation of the present specification, it is required that the collected energy spectrum is scaled to correspond to the training data and coefficients derived offline. Since the collected energy spectrum may contain more or less data than the predefined training data, the collected energy spectrum must be rescaled to a size corresponding to the training data in order to apply the training data coefficients to the collected energy spectrum. In an embodiment, the rescaling is achieved by using a mathematical operation such as multiplication or division. In an embodiment, the collected energy spectrum includes a plurality of energy bins, which include radiation of similar energy, wherein each energy bin is assigned a bin number. In an embodiment, the energy bin has a constant energy size, such as but not limited to 3keV per energy bin or "channel". The energy bin is formed by analog-to-digital conversion of the pulse height of the pulse received from the radiation detector. For example, if the pulse height has a 10-bit resolution, 1024 energy bins are formed. These correspond to the voltage of the pulse received from the radiation detector, which is proportional to the current from the detector, which in turn is proportional to the energy deposited in the detector by the gamma photons interacting within the detector. In one embodiment, after calibration, the energy bins may be energy scaled to 3 keV levels, for example, where the radiation detector is a high resolution detector. In an embodiment, for a PVT detector with a lower resolution, the pulse height of the pulses received from the detector may be 8-bit resolution, which results in 256 energy bins. If the bins are calibrated to cover 3 MeV (3000 keV) of energy, the energy bins may represent 3000 keV / 256 = 11.7 keV energy per bin.

[0072] In an embodiment, once the collected energy spectrum is rescaled to a size comparable to the training data by applying multiplication or division to each bin of the energy spectrum, the coefficient array for each category (NORM or other (OTHER)) is obtained by taking the dot product of the array of training data and the collected normalized energy spectrum. The dot product is the sum of the multiplication of the coefficient array and the bin-by-bin multiplication of the energy spectrum. This embodiment requires intensive mathematical processing and is therefore a less preferred embodiment.

[0073] In an embodiment, the normalization module is configured to enable an optimal linear associative memory (OLAM) to be implemented within a field programmable gate array (FPGA) integrated circuit. In a preferred embodiment, the normalization module 1104 is configured to normalize the collected energy spectrum by counting the number of counts in the energy spectrum collected into the energy histogram. In an embodiment, a predetermined number of counts need to be collected for the classification module 1108. For example, if 1000 counts are required to make a classification decision, when determining the coefficients for offline operation of the OLAM classifier, each category of the energy spectrum is normalized to contain a total of 1000 counts. Therefore, in the case where the energy spectrum lasts for a duration of 10 minutes and includes a million counts, each energy bin of the corresponding histogram is scaled so that the sum of all energy bins is equal to 1000.

[0074] In an embodiment, the coefficients for each energy category are determined from the normalized spectrum using the OLAM algorithm. In an embodiment, each count accumulates the coefficients for each category (NORM or other) until a total of 1000 counts have occurred, which is when it is determined whether the 1000 counts correspond to a radiation source of type "NORM" or "OTHER". When a count occurs, the amplitude of the count corresponds to the bin within the energy spectrum to which the count belongs, which is 1, while all other bins are 0. In one embodiment, the coefficient for that bin is added to an accumulator, and once the number of counts reaches a predefined number of counts (in this example, 1000), the process is complete, and the accumulator for each energy category is used to determine the winning energy category, which is used to further determine whether the energy category belongs to a radiation source of type NORM or OTHER.

[0075] In an embodiment, the SDI technique includes multiplying the number of counts in a particular energy bin by the bin number. For example, if there are two counts in an energy bin with a bin number of 200, the SDI scaling method will include multiplying the two counts by 200 (which is equal to 400). In an embodiment, a scaling technique that includes using the inverse of the energy efficiency curve of the detector is used to boost energy bins with lower energy. For example, if the energy represented by bin number 200 is only 20% efficient, the bin is scaled by dividing the counts within the bin by 0.2 to improve the energy efficiency of the energy bin. In other embodiments, various other suitable scaling techniques may be used to scale the collected energy spectrum without limiting the scope of this specification.

[0076] Classification module 1108 provides two categories to the OLAM algorithm training set, specifically "NORM" and "other". As is well known, when used for isotope identification, as an answer, the OLAM algorithm always presents only the results of one of the categories or isotopes present in the training set for training the artificial neural network based on OLAM. Therefore, the OLAM algorithm is able to identify only one or more isotopes present in its training set. If the OLAM network is trained by using each known NORM source and the fact that the NORM source generally has a higher energy signature than other radiation sources, and if the trained OLAM network is presented with a new isotope that is not present in the training set, the OLAM algorithm does not identify the new isotope, and uses the least squares residual technique to classify the isotope as "other" material by default. Therefore, if the OLAM algorithm does not recognize the specific energy spectrum obtained from the scaling module 1106, the energy spectrum is classified as belonging to "other" isotopes / materials. In other embodiments, the classification module 1108 uses various other classification techniques to classify the energy spectrum to identify the NORM isotope. In other embodiments, other neural networks, such as but not limited to k-NN (k-Nearest Neighbor) networks, may be used to classify the energy scaled spectrum to determine energy sources belonging to the NORM category.

[0077] In an embodiment, where an OLAM neural network is used to classify an unknown / collected energy-scaled spectrum into either a "NORM" or "other" category, the output of the OLAM algorithm is in the form of two numbers corresponding to the degree (how much) the unknown energy-scaled spectrum matches each of the two categories. For example, if the energy-scaled spectrum includes a "NORM" source, the output of the OLAM network is approximately 100% for the "NORM" category and approximately 0% for the "OTHER" category. In another example, if the scaled energy spectrum includes a mixture of isotopes including Am-241 and Ba-133 (which belong to the "other" category) and Ra-226 (which belongs to the "NORM" category), and each isotope contributes approximately 33% to the total counts in the scaled energy spectrum, the output of the OLAM network is approximately 66% for the "other" category and 33% for the "NORM" category.

[0078] Figure 3A 302 is a flow chart illustrating a method for classifying energy spectra according to an embodiment of the present specification. In step 302, a database of training spectra is generated. In an embodiment, one or more training sets for training an OLAM-based artificial neural network are used to populate the database of training spectra. In one embodiment, the training spectra of the OLAM-based artificial neural network include one spectrum from each isotope in the "NORM" category, which includes radioactive isotopes such as Ra-226, Tl-208 / Th-232, K-40, and U-238; and one spectrum from each isotope belonging to the "OTHER" category, which includes isotopes such as but not limited to Am-241, Co-57, Ba-133, Cs-137. It is well known that the isotopes belonging to the "other" category include more isotopes than are available, because some of these isotopes decay rapidly and are difficult to collect in an industrial environment, while others are special nuclear materials that are strictly controlled by government regulations. In various embodiments of the present specification, energy spectra of a large number of isotopes belonging to the "other" category are collected to fill the training set of an artificial neural network based on OLAM. In an embodiment, a single category of isotopes includes an energy histogram / scaled energy spectrum from each isotope belonging to that category. In an embodiment of the present specification, an OLAM-based neural network is trained to classify isotopes into two categories, namely "NORM" and "other". In other embodiments, the method of the present specification can be extended to identify multiple different isotope category pairs, such as but not limited to "medical" and "OTHER"; or special nuclear materials, namely "SNM" and "OTHER".

[0079] At step 304, an energy spectrum generated by an object under inspection in a radiation inspection system is collected by using one or more radiation detectors of the system. At step 306, the collected energy spectrum is normalized by using one or more predefined normalization techniques. At step 308, the normalized spectrum is energy scaled by using one or more predefined energy weighting techniques. In one embodiment, predetermined weights are applied to the collected energy spectrum to form a weighting matrix of the spectrum by using a spectral dose index (SDI) technique. At step 310, the normalized and scaled spectrum is fitted to each training spectrum in a database. In one embodiment, the normalized and scaled spectrum is fitted to each training spectrum in a database in a sequential manner. In one embodiment, in the case where an OLAM-based neural network is used for isotope classification, a set of coefficients for each category of the training spectrum is sequentially applied to the collected unknown scaled spectrum. In another embodiment, the scaled spectrum is fitted on all training sets simultaneously.

[0080] At step 312, the residuals of the fitted normalized scaled spectrum for each training spectrum category are determined. At step 314, the normalized scaled spectrum is identified as belonging to the category of the training spectrum that results in the minimum residual as determined at step 312. In an embodiment, the training spectrum belongs to the "NORM" category, and the normalized scaled spectrum that results in a residual ("r") value (as determined in step 312) less than a predefined threshold is determined to belong to the "NORM" category; and the normalized scaled spectrum that results in a residual value (as determined in step 312) greater than the predefined threshold is determined to belong to the "other" category. In one embodiment, after training the OLAM neural network and determining all residual values, the threshold is determined empirically. In an embodiment, the residual value of the isotope belonging to the NORM class must be lower than the residual values ​​of all other isotopes in the normalized scaled spectrum. For example, if the residual values ​​for the NORM source relative to the normalized scaled spectrum range from 2.4 to 20.5, and the residual values ​​for all other isotopes range from 27.2 to 115.3, then the threshold may be defined as 23.85, which is halfway between 20.5 and 27.2. In other embodiments, other suitable techniques may be used to predetermine the threshold.

[0081] In the training scenario, the radiation spectrum is acquired to generate experimental classes representing a predefined number of radionuclides belonging to the Naturally Occurring Radioactive Materials (NORM) group. Table 1 (below) illustrates all the training radionuclides acquired with each corresponding vector symbol.

[0082] Table 1

[0083] Radionuclides Vector Symbols Ra-226 Ra K-40 K Th-232 Th

[0084] To conduct training experiments, we first collect the background spectrum (Bkg∈R Nc ). Subsequently, the radiation spectrum is acquired for 60 seconds, with channels Nc=256 from the PVT detector.

[0085] Each acquired spectrum vector is represented as a vector v∈R Nc , where the ith individual element is v[i]. The preprocessed spectrum By using the following equation, taking into account Calculated in the case of:

[0086]

[0087] where normalization is performed using: and where "i" is the SDI constant.

[0088] Figure 3B The normalized radiation spectrum of the radionuclide Ra-226 normalized with and without the SDI constant as shown in Equation 2 above is illustrated. Figure 350 illustrates the Ra-226 spectrum normalized without the SDI constant, and Figure 352 illustrates the Ra-226 spectrum normalized by using the SDI constant. It can be seen that in Figure 352, the spectral features (especially at the higher channels) are more obvious for the enhanced spectrum comparison.

[0089] Figure 3C The normalized radiation spectrum of the radionuclide K-40 normalized with and without the SDI constant shown in Equation 2 above is shown. Figure 354 illustrates the K-40 spectrum normalized without the SDI constant, and Figure 356 illustrates the K-40 spectrum normalized by using the SDI constant. It can be seen that in Figure 356, the spectral features (especially at the higher channels) are more obvious for the enhanced spectrum comparison.

[0090] Figure 3D The normalized radiation spectrum of the radionuclide Th-232 normalized with and without the SDI constant is illustrated, as shown in Equation 1 above. Figure 358 illustrates the Th-232 spectrum normalized without the SDI constant, and Figure 360 ​​illustrates the Th-232 spectrum normalized by using the SDI constant. As can be seen, in Figure 360, the spectral features (especially at the higher channels) are more obvious for the enhanced spectrum comparison.

[0091] During training, the processed set of training spectra can be written more compactly as a matrix:

[0092]

[0093] It consists of a column vector of spectra, where Nt=3 is the number of training vectors.

[0094]

[0095] The measured spectrum used for classification is represented by:

[0096]

[0097] Also from having N c = 256 channels of the PVT detector for 60 seconds. Again, each measured spectrum is normalized in the same way and given by express.

[0098] Conceptually, if we reconstruct the measured spectrum using a linear combination of the training spectra Then the system can be represented by the matrix equation:

[0099]

[0100] in, is the weight of each training spectrum.

[0101] Since "A" is not full rank, there is no solution to this overdetermined problem. A matrix is ​​said to have full rank if its rank is equal to the maximum possible rank of a matrix of the same dimension, which is the smaller of the number of rows and columns. Therefore, the linear least squares solution is Used to solve for the coefficients using the following formula:

[0102]

[0103] These coefficients can be used to reconstruct the measured spectrum, where the reconstruction vector is given by It means that by converting the measured vector Project back into the space spanned by the columns of the training matrix 'A':

[0104]

[0105] As is well known, the residual is a measure of how well the line fits a single data point. Therefore, the residual from the fitted model is defined as the difference between the response data and the fit to the response data at each predicted value: residual = data - fit.

[0106] The resulting residual for each channel from the reconstruction consists of the vector: is the basic building block of classification and is defined by:

[0107]

[0108] The residual vector calculation can be simplified to a single matrix by calculating the projection matrix 'P', which transforms the measured spectrum vector Project into the null space of the training matrix 'A':

[0109]

[0110] where 'I' is the identity matrix, and

[0111]

[0112] In an embodiment, the classification method does not attempt to classify multiple classes simultaneously, since reconstruction may result in negative weights that have no physical meaning. In contrast, when measured spectra are not used for training, fitting typically results in large residuals, and therefore misclassifications. Therefore, in an embodiment, only the spectrum of a single class is trained at a time for binary decisions indicating whether or not a member of that class is a member. When a measured spectrum of a different class is encountered, the residuals are used as the basis for classification.

[0113] To obtain the size of the residuals compared to the total size of the measured spectrum, in an embodiment, the size of the residual vector is divided by the size of the measured spectrum using a simple decimal. This proportionally indicates how large the residuals of the fitted vector are compared to the original spectrum. The magnitude or size of each vector is calculated as the sum of the absolute values ​​of each element, or L in compact notation. 1 Norm function, for the measured vector v, L 1 The norm function is represented by ‖v‖1. Therefore, the relative residual scalar used for classification is It can be obtained by:

[0114]

[0115] Because, when using the measurement vector When , the relative residual is large. is not a member of the training class, and is small when it is a member of that class. In one embodiment, the present specification provides a simple threshold value that is created empirically using experimental data, as described above with respect to step 312 of FIG. 3 .

[0116] Figure 3E is a graph illustrating the fit difference of a background spectrum using a NORM material according to an embodiment of the present specification. The fit difference defines how well the measured spectrum fits the training spectrum of the NORM material, in this case the measured spectrum includes the majority of the NORM isotopes in concentration, since these naturally occurring isotopes produce the gamma radiation background. Graph 362 illustrates the accuracy of the fit resulting in a small relative residual:

[0117] Table 2 shows the values ​​of the measured spectra of multiple "other" (non-NORM) materials and their corresponding relative residual values. It can be seen from Table 2 that all the relative residual values ​​shown are significantly higher than the relative residual value of the background, i.e.

[0118] Table 2

[0119]

[0120] Figure 3F is a graph 364 showing the difference in fit using the radionuclide Am-241 in accordance with an embodiment of the present specification. Figure 3G is a graph 366 showing the difference in fit using the radionuclide Co-57 according to an embodiment of the present specification. Figure 3H is a graph 368 showing the difference in fit using the radionuclide Ba-133 according to an embodiment of the present specification. Fig. 3I is a graph 370 showing the difference in fit using the radionuclide Cs-137, in accordance with an embodiment of the present specification. Figure 3J is a graph 372 showing the difference in fits using the radionuclide Co-60 according to an embodiment of the present specification.

[0121] Figure 4A is a graph 402 illustrating normalized radiation spectra of various isotopes in accordance with an embodiment of the present specification. Figure 4B It is a graphic Figure 4A 404 of the spectrum shown in FIG, which is weighted and classified into the category of "NORM" or "other". Since NORM isotopes generally emit higher energy gamma radiation, in an embodiment, energy scaling of the radiation spectrum using a weighting matrix enables a classification algorithm (such as OLAM) to classify the spectrum into NORM or "other" category in real time by identifying high energy NORM isotopes.

[0122] In an exemplary test case, multiple detected radiation spectra as well as background spectra were analyzed using the classification method of the present specification. The background spectrum includes a large concentration of NORM isotopes because these naturally occurring isotopes produce a gamma radiation background. Using a classification algorithm such as OLAM, approximately 80% of the background was classified into the NORM category, while 20% was classified into the "other" category. When the same energy scaling and classification algorithm was applied to the Cs-137 spectrum collected from each of the four PVT detectors deployed in the radiation inspection system and the sum of the four PVT detectors, approximately 80% of the spectrum was classified into the "other" category and 20% was classified into the NORM category because the collected Cs-137 spectrum also includes background radiation. In various embodiments, the method of the present specification can be used to identify multiple different isotope category pairs, such as but not limited to "medical" and "OTHER"; or special nuclear materials, namely SNM and OTHER.

[0123] The above examples are merely illustrations of many applications of the systems, devices, and methods of this specification. Although only a few embodiments of the present invention are described herein, it should be understood that the present invention may be implemented in many other specific forms without departing from the spirit or scope of the present invention. Therefore, the present examples and embodiments are considered to be illustrative rather than restrictive, and the present invention may be modified within the scope of the appended claims.

Claims

1. A radiation portal monitor comprising a plurality of detectors in data communication with a processing system, wherein the processing system is configured to determine whether a radiation spectrum includes radiation corresponding to a naturally occurring source of radiation material (NORM) by: generating a database of training spectra, each training spectrum in the database representing a NORM material; collecting the radiation spectrum; normalizing the collected spectrum by using one or more predetermined normalization techniques; energy scaling the normalized spectrum using one or more energy scaling techniques; fitting the normalized and scaled spectrum to each training spectrum in the database; determining a residual of fitting the normalized and scaled spectrum to each of the training spectrum; as well as The normalized and scaled spectrum is identified as NORM material based on the residual and a threshold.

2. The radiation inlet monitor according to claim 1, wherein: If the residual is less than the threshold, the normalized and scaled spectrum is determined to be NORM material.

3. The radiation inlet monitor according to claim 1, wherein: The processor is further configured to identify the normalized scaled spectrum as belonging to an “other” category if the residual is greater than the threshold.

4. The radiation inlet monitor according to claim 1, wherein: The processor is configured to empirically determine the threshold based on the training spectrum and a residual of fitting the normalized scaled spectrum with a class of the training spectrum.

5. The radiation portal monitor of claim 1, wherein collecting the radiation spectrum comprises collecting an energy spectrum radiated by an object under inspection in the radiation portal monitor by using one or more of the plurality of detectors.

6. The radiation inlet monitor according to claim 5, wherein: The plurality of detectors are large area low resolution radiation detectors.

7. The radiation portal monitor of claim 5, wherein the radiation detector is a polyvinyl toluene (PVT) detector.

8. The radiation portal monitor of claim 5, wherein the radiation detector is a gamma radiation detector.

9. The radiation portal monitor of claim 5, which is a passive system that does not include a radiation source.

10. The radiation portal monitor of claim 5, the radiation portal monitor being a passive radiation portal monitor (RPM) for detecting the presence of an ionizing radiation source passing through the RPM.

11. The radiation portal monitor of claim 10, wherein the RPM comprises a plurality of panels, wherein the panels are spaced apart to enable a cargo container to pass between two panels.

12. The radiation portal monitor of claim 11, wherein each panel comprises a radiation detector.

13. The radiation portal monitor of claim 1, wherein energy scaling the normalized spectrum comprises applying predetermined weights to the collected energy spectrum to obtain a weighting matrix for the spectrum.

14. The radiation inlet monitor of claim 1, wherein: The one or more energy scaling techniques include a spectral dose index (SDI) technique.

15. The radiation inlet monitor of claim 14, wherein: The SDI technique adds more weight to higher energies in the normalized spectrum to help identify NORM materials.

16. The radiation portal monitor of claim 1, wherein one or more training spectra used to train an optimal linear associative memory (OLAM) based artificial neural network are used to populate a database of training spectra.

17. The radiation portal monitor of claim 16, wherein the training spectra for the OLAM-based artificial neural network include one spectrum for each isotope from the NORM material.

18. The radiation portal monitor of claim 1, wherein: An optimal linear associative memory (OLAM) neural network implemented within a field programmable gate array (FPGA) integrated circuit is used to determine whether the radiation spectrum includes radiation corresponding to NORM material.

19. The radiation portal monitor of claim 1, wherein: The normalized and scaled spectra are fitted to each training spectrum in the database in a sequential manner.

20. The radiation portal monitor of claim 1, wherein the NORM material is one of the following: Ra-226, Th-232 / Tl-208, K-40, and U-238.

Images