Multifaceted radiation detection and classification system

Abstract

To provide a system of identifying a source of radiation.SOLUTION: The system includes a radiation source detector and a radiation source identifier. The radiation source detector receives measurements of radiation; for one or more sources, generates detection metrics indicating whether the sources are present in the measurements; and evaluates the detection metrics to detect whether a source is present in the measurements. When the presence of a source in the measurements is detected, the radiation source identifier generates, for one or more sources, identification metrics indicating whether the sources are present in the measurements, generates a null-hypothesis metric indicating whether no source is present in the measurements; and evaluates the one or more identification metrics and the one or more null-hypothesis metrics to identify the source, if any, which is present in the measurements.SELECTED DRAWING: Figure 9

Description

Technical Field

[0001] [Cross - Reference to Related Applications] This patent document claims the benefit of the priority of U.S. Provisional Patent Application No. 62 / 673,750, entitled "Radiation Detection Analysis Kit (RDAK) (Radiation Detection Analysis Kit (RDAK))", filed on May 18, 2018, and U.S. Patent Application No. 62 / 805,825, entitled "Radiation Detection Analysis Kit (RDAK) (Radiation Detection Analysis Kit (RDAK))", filed on February 14, 2019, both of which are incorporated herein by reference.

[0002] [Description of Federally Sponsored Research or Development] The United States Government has rights in this invention pursuant to Contract No. DE - AC52 - 07NA27344 between the United States Department of Energy and Lawrence Livermore National Security, LLC for the operation of Lawrence Livermore National Laboratory.

Background Art

[0003] The detection of radiation sources is particularly important for protecting the safety of the general public, military personnel, and first responders, etc., while safeguarding the safety of nuclear materials. Radiation sources are classified into radiation source classes based on the use of radiation. For example, radiological dispersion weapons (e.g., dirty bombs) and radiopharmaceuticals (e.g., within medical patients) 99mIt is considered that the radiation source classes are different between (a) and (Tc). The radiation source classes of many radiation sources (e.g., medical patients) correspond to approved uses because all safety risks are considered acceptable, while the radiation source classes of other radiation sources (e.g., contaminated bombs) correspond to unapproved uses because the safety risks are considered unacceptable. Since there are approved and unapproved uses depending on the application, simply detecting the presence of a radiation source is of little use. Therefore, one purpose of radiation source detection is to distinguish between radiation source classes that may pose unacceptable safety risks and those that do not pose such safety risks.

[0004] To assist in radiation detection, a radiation detector, also called a detector (i.e., a physical device that detects photons), is used to collect the number of photons having energy within the radiation spectrum. The radiation detector can only identify the energy of photons up to a certain level of accuracy. As a result, the radiation detector can classify the number of photons into energy ranges. The radiation detector can collect counts over a certain period (e.g., 1 second) and indicate the counts within each energy range as a measurement of the radiation (i.e., the radiation spectrum).

[0005] To detect radiation sources over a wide area (e.g., a city), a drone equipped with a radiation detector can be flown in a measurement collection pattern over the airspace of that area. The detector can also be placed in a vehicle and driven, carried by hand, or placed at strategic fixed locations. These measurements can be provided to an analysis system to identify whether there are radiation sources of interest that can be considered radiation sources indicating acceptable or unacceptable safety risks. Since there can be significantly different background radiations in different areas (e.g., due to different terrains or building materials), measurements representing the same radiation source can vary greatly depending on the area. As a result, such an analysis system needs to consider the background radiation present in the area where the measurements were collected. SUMMARY OF THE INVENTION

Problems to be Solved by the Invention

[0006] The speed and accuracy of detecting a radiation source and classifying it (e.g., as an acceptable safety risk or an unacceptable safety risk) are important. Also, it is important to identify the radionuclide that is the radiation source (e.g., U or 137 Cs). However, some state-of-the-art analysis systems require a huge amount of computing resources for processing measurement values, identifying background radiation, identifying the radiation source, and classifying the radiation source class of the radiation source. Also, many of the currently deployed analysis systems employ detection algorithms that are not very accurate. Therefore, using such analysis systems is not practical due to the required computing resources or low accuracy. There is a strong need for an algorithm that combines high computational efficiency and high accuracy.

Brief Description of the Drawings

[0007]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Figure 9

[0008] A method and system for detecting and classifying radiation sources are provided. A radiation detection and analysis ("RDA") system processes a series of radiation measurements (i.e., gamma ray spectrum data and neutron measurements), analyzes these measurements in real time to detect the presence of a radiation source, and identifies a radiation source (e.g., U or 137 Cs) and determines the radiation source class of the radiation source. The RDA system provides a radiation source detector component ("radiation source detector") that can operate on a wide range of detector types including different materials and detector sizes using multiple detection algorithms. Each detection algorithm generates one or more detection metrics updated with each new measurement within a series of measurements. The detection metrics optimize the signal-to-noise ratio under various conditions. The radiation source detector employs various background estimation algorithms to estimate background radiation ("background") so as to be able to determine the contribution of the radiation source to the measurements. The radiation source detector continuously retrains the background estimation algorithm based on background measurements (i.e., measurements where no radiation source is detected). The RDA system also includes a radiation source identifier component ("radiation source identifier") that identifies a radiation source (e.g., 137 Cs) and its radiation source class when the radiation source is detected. When a radiation source class of interest (which can be for approved or unapproved uses) is identified, the RDA system can output an alarm indicating the radiation source class.

[0009] The RDA system can be used in fixed or mobile detection applications. Not all algorithms provided by the RDA system are suitable for all detection applications. The RDA system provides an architecture that allows the selection of different combinations of a detection algorithm and a background estimation algorithm to optimize performance. The various combinations of algorithms can be used for (actual or simulated) sample measurements. For each application, the combination of algorithms most suitable for that application can be selected. Then, the most suitable combination can be used within the RDA system to operate in real time with high precision. Further, the RDA system can be used to process measurements collected in the past much faster than when processing them in real time.

[0010] FIG. 1 is a block diagram showing the overall architecture of the RDA system in some embodiments. The RDA system 100 includes a radiation source detector 110 and a radiation source identifier 120.

[0011] The radiation source detector inputs measurement values and outputs a source detection flag indicating whether a source exists within each measurement value (or group of measurement values). The radiation source detector includes a calibrate measurements component 111, a calculate metrics component 112, an evaluate metric component 113, an estimate background component 114, and a detection source definitions store 115. The calibrate measurements component calibrates the measurement values to account for detector sensor drift and non-linearity. The calculate metrics component calculates a metric of the measurement values to indicate whether any source as defined by the source definition exists within the measurement values. The source definition of a source includes its source signature (e.g., spectral shape) considering specified shielding such as shielding type and shielding thickness. Thus, a radiation source can be associated with the source definition of each specified shielding. The source definition can include multiple sources. Some algorithms operate without a source definition that detects any anomalies different from the background. The evaluate metric component sets the source detection flag individually or overall based on the evaluation of the metric to determine whether a radiation source may exist within the measurement values. The estimate background component updates the estimated background based on measurement values in which the presence of a radiation source was not detected. The calculate metrics component considers the estimated background when calculating the metric. The detection source definitions store includes source definitions that define the source signatures of the radiation sources to be detected.

[0012] When the radiation source detector detects the presence of a radiation source, the radiation source identifier identifies the source type and its radiation source class. The radiation source identifier 120 includes a metric calculation component 121, a source identification component 122, and an identified source definition store 123. The metric calculation component calculates metrics in the same way as the metric calculation component of the radiation source detector, but can use algorithms that are particularly effective for identifying the source. The source identification component identifies the source based on an evaluation of the metrics. The source identification component can output the probability of each radiation source class. The identified source definition store includes the source definitions of the sources to be identified. The identified source definition store includes a more comprehensive set of source definitions than that of the detected source definition store to improve the accuracy of source identification.

[0013] The RDA system inputs spectral measurements represented as a spectral histogram that includes a fixed number of energy bins, each representing an energy range within the detection energy range. For example, if the energy range is 100 keV to 1 MeV, the histogram can have 50 energy bins, each representing an energy range of 20 keV. The first energy bin has an energy range of 100 keV to 120 keV, the second energy bin has an energy range of 120 keV to 140 keV, and so on up to 980 keV to 1 MeV. The histogram has energy edges at 100 keV, 120 keV, 140 keV, and so on. Therefore, there is one more energy edge than the number of energy bins. Each energy bin contains the number of detected photons that were present within the energy range of that energy bin over one measurement interval (e.g., one second).

[0014] The following table shows the explanations of the terms used to describe the RDA system.

Table 1

[0015] Figure 2 is a block diagram showing in more detail the components of the radiation source detector of the RDA system in some embodiments than 110 of FIG. 1. The radiation source detector 200 inputs measurement values, detects whether a source may exist within the measurement, and outputs a source detection flag. However, the radiation source detector does not perform the computationally expensive process of actually identifying the source. The radiation source detector includes a measurement value calibration component 201, generate metric components 202, an aggregate metrics component 203, a metric evaluation component 204, train metric components 205, a veto component 206, a background estimation component 207, and a generate calibration component 208. The radiation source detector also includes a detected source definition store 211 and a history store 212. Each component may also include a store for locally storing data generated by that component or data received from other components.

[0016] The measurement value calibration component generates calibrated measurement values to account for detector measurement value drift and non-linearity using a re-binning process. The measurement value calibration component inputs measurement values and calibration data and outputs calibrated measurement values. The calibration data indicates how to re-bin the measurement values based on detector drift.

[0017] The metric generation component can include metric generation components for various types of detection algorithms. The metric generation component takes as input a calibration measurement, optionally navigation information (e.g., GPS information for identifying the presence of a building), and a set of coefficients for a line source signature, and generates a metric for each line source signature indicating whether the presence of the source of that line source signature, or the presence of any source other than background, is represented by the measurement (i.e., the calibration measurement). Each set of coefficients is associated with a combination of a line source signature and a detection algorithm. For example, the first set of coefficients can be for a line source signature of Cs that does not include shielding and (as described later) an orthonormal subspace projection ("OSP") matched filter 137 and can be for a line source signature of Cs that includes 10 cm of lead shielding and an OSP matched filter 137 and the third set of coefficients can be for a line source signature of U that does not include shielding and (as described later) a likelihood ratio test ("LRT") matched filter. The metric generation component for the associated detection algorithm uses these coefficients to generate, for each set of coefficients, a metric as a semi - definitive indication that a source represented by the line source signature is present. The radiation source identifier provides a more definitive indication for each source. Continuing with this example, the detection algorithm for the matched filter applies the set of coefficients for the line source signature of Cs without shielding to generate a first metric and applies the set of coefficients for the line source signature of Cs that includes 10 cm of lead shielding and an OSP matched filter to generate a second metric. Each metric can be represented as a pair of a numerator and a denominator called metric partials. For example, for a particular detection algorithm without shielding 137 applies the set of coefficients for the line source signature of Cs to generate a first metric and for a line source signature of Cs that includes 10 cm of lead shielding and an OSP matched filter 137 applies the set of coefficients for the line source signature of Cs to generate a second metric. Each metric can be represented as a pair of a numerator and a denominator called metric partials. For example, for a particular detection algorithm without shielding 137The metric of the Cs line source signature can be (1,5). Regarding the detection algorithm, it will mainly be described as a matched filter, but the detection algorithm can also be an anomaly detector. The anomaly detector generates a metric indicating whether the deviation of the measurement from the estimated background is sufficient to indicate the presence of a source regardless of what the source is.

[0018] The metric aggregation component generates an aggregated metric for each combination of the line source signature, the detection algorithm, and the window of measurements (e.g., a rolling window). Each window includes the number of most recent measurements such as 1, 2, and 4. For example, without occlusion 137 The metric partial for a specific detection algorithm of the Cs line source signature and four consecutive measurements can be (1,4), (3,5), (1,2), and (3,5). The metric aggregation component sums the numerators of the metric partials and sums the denominators to set the aggregated metric to the value obtained by dividing the sum of the numerators by the square root of the sum of the denominators in order to aggregate the metric of the window for the line source signature. Continuing with this example, the individual metric partial (1,5) results in a time window of 1, and dividing 1 by the square root of 4 gives 0.5. When aggregating the first two metric partials for a time window of 2, the sum of the numerators is 4, the sum of the denominators is 9, and multiplying 4 by the square root of 9 gives 1.33. When aggregating all four metric partials for a time window of 4, the sum of the numerators is 8, the sum of the denominators is 16, and dividing 8 by the square root of 16 gives an aggregated metric of 2. Thus, without occlusion 137 For Cs and a specific detection algorithm with windows containing 1, 2, and 4 measurements, the aggregated metrics are 0.5, 1.33, and 2. Aggregation can also be performed based on measurements obtained at the same location even at different times.

[0019] The metric evaluation component takes an aggregated metric as input and outputs a single decision metric indicating whether a source has been detected. First, the metric evaluation component normalizes each aggregated metric by subtracting the measured mean and dividing by the measured standard deviation. Next, the decision metric is determined from the evaluation of all the normalized aggregated metrics including different time windows, different algorithms, and different sources. Usually, this evaluation uses the maximum value of each normalized aggregated metric and can also use weighting factors determined based on the performance analysis of different methods or using machine learning as described later.

[0020] Next, the metric evaluation component compares this decision metric with a plurality of thresholds set to control the operation of the radiation source detector. One such threshold is the source present threshold. If the decision metric exceeds the source present threshold, the metric evaluation component sets a source detection flag (i.e., true or false). This source detection flag triggers the radiation source identifier. The source present threshold can be set so that the failure of source detection (false positive or false alarm) when there is no source, such as a false alarm occurring every 8 hours, occurs at a certain rate. This certain rate can be based on the period (e.g., 8 hours), measurement rate, number of rolling windows, and number of metrics. As another alternative, the metric evaluation component can calculate the decision metric using a classifier generated via a machine learning algorithm (e.g., support vector machine, Bayesian classifier, and neural network). The classifier can be trained using training data including feature vectors and labels. The features of the feature vector can include the detection algorithm, source, window size, and aggregated metric. The feature vector can be generated based on data collected and generated during actual detection regarding different sources. The feature vector includes a 10 cm iron shield 137Cs, and different angular velocities between the radiation source and the detector (e.g., identified based on GPS readings). 137 It can also be generated using a mathematical model that generates simulated measurement values of different types of radiation sources, such as Cs. The label can be an indication (e.g., true or false) as to whether the feature vector represents the presence or likelihood of a radiation source. The label can be generated manually or based on data generated when the radiation source identifier identifies the radiation source and the radiation source class.

[0021] The metric evaluation component also compares the decision metric with an extended aggregation threshold. If this threshold is exceeded, the radiation source identifier is triggered to aggregate until the decision metric falls below an end aggregation threshold. The decision metric can also be compared with a veto threshold, which is usually set very low. When the decision metric exceeds the veto threshold, a lockout flag that controls whether to use the measurement value for calibration or background estimation is sent to the rejection component 206.

[0022] The background estimation component can include a background estimation component for each background algorithm. The background estimation component inputs measurement values and generates background statistics for background estimation. Each background algorithm can generate different statistics, such as the average value (referred to as the ratio) within the measurement values representing the background, the average measurement value of the background, or the basis of the background. The detection algorithm can employ different background statistics. For example, one detection algorithm can employ a rate statistic, and another detection algorithm can employ a rate statistic and a basis statistic. The background estimation component can generate a statistic that is a weighted sum of the new statistic and the previous statistic. For example, if the new ratio is 160, the previous ratio is 200, and the weight is 0.25, the weighted ratio to be used as the current ratio is 190 (i.e., 200 * 0.75 + 160 * 0.25). Thus, the weight controls how quickly the estimated value is adjusted based on the new statistic. The background estimation component can generate background statistics periodically (e.g., every 5 minutes) or based on an analysis that the current background is significantly different from the previous background.

[0023] The rejection component provides the uncalibrated measurement values to the calibrated measurement value generation component and the calibrated measurement values to the background estimation component. On the other hand, the rejection component filters out measurement values that should not be used for background estimation. For example, each measurement value used for generating an aggregation metric exceeding a rejection threshold can be filtered out. The measurement values before and after these measurement values can contribute to the line source. The rejection component can filter out the measurement values within a (fixed or variable) extended window before and after the measurement values used for generating the aggregation metric to remove this contribution from the background. Therefore, using the extended window, it is possible to filter out measurement values that may be affected by the line source but are not included in the aggregation metric. If these measurement values are not filtered out, they will be considered in the background calculation, tending to make the estimated background less susceptible to the influence of the line source. The rejection threshold can be set so that a rejection miss occurs at a certain rate, such as once every 10 minutes. This certain rate can be based on a period (e.g., 10 minutes) and the measurement rate.

[0024] The metric training component inputs the line source signature and background statistics and generates a set of coefficients for a combination of the line source signature and the detection algorithm. However, the set of coefficients may not be generated for each combination. For example, a certain detection algorithm 137 may be particularly effective for the identification of Cs but may not be effective for the identification of U. In such a case, this detection algorithm 137 generates a set of coefficients for Cs and does not generate for U.

[0025] The history store includes each measurement value and calibrated measurement value. The detected line source definition store includes the line source definition for each line source to be detected (e.g., one for each shield).

[0026] FIG. 3 is a block diagram showing components of a radiation source identifier of an RDA system in some embodiments. When the radiation source detector detects the possibility of a source, the radiation source identifier is executed to identify the source (e.g., 137 Cs) and the radiation source class, which is the main contributing factor to the measured non-background photons. The radiation source identifier inputs the measurement values and background statistics used by the radiation source detector to detect the presence of the source. The measurement values can be the sum of the measurement values included in each window. For example, if a window contains 512 measurement values, the radiation source identifier sums these measurement values to give a single total measurement value. The radiation source identifier outputs an indication of the sources present. The radiation source identifier can output the probability that each source is present. For example, the radiation source identifier can output a probability of 0.67 for 137 Cs and a probability of 0.11 for U. The radiation source identifier also applies radiation source class rules for identifying the radiation source class of the source. For example, the radiation source class of the source can be medical, industrial, fission, etc. The radiation source class can also be represented by a probability. The radiation source identifier can include a measurement value calibration component 313, a metric generation component 301, a metric training component 302, a select source metrics component 303, a generate null-hypothesis metric component 304, a generate source scores component 305, and an identify radiation source class component 306. The radiation source identifier can also include an identification source definitions store 311 and a radiation source class rules store 312.

[0027] The metric calibration component of the line source classifier inputs the measurement values used when the radiation source detector detects the line source. The metric calibration component aligns the bin range of the measurement values with the bin range of the line source signature. The metric calibration component can shift either the bin range of the measurement values or the bin range of the line source signature. Shifting the bin range of the measurement values induces the correlations between bins that need to be tracked to guarantee statistical accuracy. Shifting the line source definition increases statistical accuracy but requires more computational resources as each line source signature needs to be shifted.

[0028] The identification line source definition store is similar to the detection line source definition store but has a more comprehensive set of line source definitions. This set is made more comprehensive to account for the detection efficiency of the line source definitions. Detection efficiency means the likelihood that a line source signature is effective for the detection of the line source regardless of shielding. For example, a line source signature with 5 cm of lead shielding can have a detection efficiency of 0.25, and a line source signature with 7 cm of lead shielding can have a detection efficiency of 0.15. The goal is to have line source signatures for sufficient shielding configurations such that shielding configurations with detection efficiencies below the maximum possible detection efficiency are minimized. To maximize overall performance, a trade-off is needed between the detection efficiency across all the shielding configurations of interest and the false alarm opportunities, which increase with the number of signatures in use.

[0029] In some embodiments, it is useful to capture signatures observed in the field. Using a semi-supervised learning modality, detections with spectra that have some similarity to the source signatures but do not match as well as expected are extracted from the measurements to create new source signatures, which can be added to the source signature store. To determine whether the measured spectra are well represented by the source signatures in the source signature store, the detection efficiency at the extracted spectra is compared to the detection efficiency of the existing source signatures in the source signature store. This method can increase the sensitivity to sources under test that are intermediate between the shielding configurations and unexpected source combinations used in the source signature store or have some differences from the source signatures in the source signature store. This new source signature can be added locally to the source signature store of the detector in use and, if desired, distributed to the source signature stores used by other detectors of the same type.

[0030] For identification, the source is assumed to be already detected and the false alarm rate is no longer considered. Thus, the limit on the number of source signatures for identification is not as critical and is limited only by the computational cost. Figure 4 is a graph showing the source identification efficiency. Graph 400 includes an x-axis representing the thickness and a y-axis representing the detection efficiency. The solid line 401 represents the identification efficiency of the source signature without shielding. The identification efficiency has a maximum value of 0.5 at a thickness of 0 cm and decreases to 0.0 at a thickness of 5 cm, i.e., a source shielded by 5 cm is not identified with a source signature without shielding. The solid lines 402 and 403 show the identification efficiencies of source signatures of 5 cm and 7 cm, respectively. The dotted line 404 connects the detection efficiencies at these thicknesses for the source signatures. Regions 405 and 406 represent the gap between the maximum identification efficiency and the actual identification efficiency. Such gaps can cause the source to be misidentified. The area of the gap can be reduced by adding source definitions with larger shielding thicknesses. For example, gap 405 is reduced by adding a source definition with a thickness of 3 cm.

[0031] The metric generation component includes metric generation components for each identification algorithm used by the radiation source identifier. In some embodiments, only one identification algorithm is used, such as one similar to the OSP matching filter but with a different normalization term that takes into account the difference between detection efficiency and identification efficiency. The metric training component can be similar to the metric trainer component of the radiation source detector.

[0032] The source metric selection component takes the metrics as input, identifies the best metric for each source, and outputs these metrics as source metrics. For example, for 137 unshielded 137 Cs, and 137 Cs with 1 cm and 2 cm of shielding, the metrics can be 1.0, 2.0, and 0.5 respectively. In such a case, the source metric selection component

[0033] The null hypothesis metric generation component takes the measurements, background statistics, and source metrics as input and generates a null hypothesis metric representing the statistically worst scenario for the background. The null hypothesis metric generation component calculates the variance of each channel of the measurements and the expected background. The null hypothesis metric generation component also calculates the variance and covariance of each metric and the covariance of the source metrics. The null hypothesis metric generation component calculates a null hypothesis score based on the variance and covariance of the source metrics and outputs the variance of the null hypothesis (e.g., the average of the variances of the metrics) and the null hypothesis score.

[0034] The line source score generation component inputs the target metrics (i.e., the line source metric and the null hypothesis metric) and the initial probabilities (prior probabilities) of each target (i.e., the line source and the background). The score generation component calculates the probability of each target given a measurement value that is the maximum of the probabilities that the metric of the target is greater than the target metrics of each other target. The maximum probability can be based on a multivariate normal distribution centered on that measurement value, given the measurement value and the variances of each target metric. The line source score generation component generates the target probabilities (posterior probabilities) of each target using, for example, a Bayesian estimator that generates target probabilities.

[0035] The radiation source class identification component inputs the target probabilities and applies radiation source class rules to generate the radiation source class probabilities of each radiation source class. The radiation source class rule store contains the radiation source class rules. Alternatively, the radiation source class identification component can select, for each radiation source class, the maximum radiation source class probability that the target is that radiation source class. The maximum radiation source class probability can be based on the multivariate normal distribution and variance of that radiation source class, similar to the target probabilities.

[0036] Detection algorithm The RDA system supports the use of various combinations of detection algorithms, metric trainers, and aggregation methods. By supporting such combinations, the RDA system can be used in a wide range of applications and can be adjusted to suit these applications.

[0037] The RDA system can be used with projection-based detection algorithms and quadratic-based detection algorithms. The projection-based detection algorithm can be decomposed into a set of linear operations that each act on the measurements individually. For example, the detection algorithm can generate a metric by applying a linear transformation to the sum of the measurements over a period of time to calculate the numerator and denominator as represented by the following equation,

Number

Number

Number

[0038] In the quadratic-based detection algorithm, the numerator or denominator is generated as the product of the measurements through a transformation matrix as represented by the following equation.

Number

Number

[0039] Orthogonal subspace projection (「OSP」) matching filter The metric trainer of the OSP matched filter constructs an optimal multiplicative spectral filter that best excludes the background. The metric trainer uses the estimated background basis function B, which is an orthonormal subspace projection. This projection is maximized towards the source signature S with the weight matrix W based on the estimated background. The metric generated by the OSP matched filter has a Gaussian distribution and can be normalized to have zero mean and unit variance. The metric of the OSP matched filter can be represented by the following equation.

Number

[0040] The metric trainer is for the source signature S i , the estimated background Input JPEG2025111467000009.jpg6150 and the background basis vector. The estimated background can be calculated using an exponential smoothing function. The weight matrix can be calculated by dividing the previous background measurement B. The metric trainer can calculate the optimal projection vector T as follows. 1) The estimated background where each energy bin variance spans 1 + scalar times the square root of the expected background Calculate the reciprocal of the diagonal of the weight matrix W from JPEG2025111467000010.jpg6150. 2) Calculate the weighted background. B W =(B t WB) -1 B t W 3) Weight the source signature by the weight matrix. S t W 4) Calculate the weighted projection of the source signature. S t WB 5) Calculate the background weighted projection. S t WB(B W ) 6) Set the optimal projection vector T to the difference between the source signature and the background weighted projection. 7) Set the variance scalar V to the sum of the square of the optimal projection vector T times the estimated background. JPEG2025111467000011.jpg14150 8) Divide the terms of the optimal projection vector T by the square root of the expected variance so that the expected variance becomes 1. JPEG2025111467000012.jpg6150

[0041] The performance characteristics of the OSP matching filter can be modeled using the following equation,

Equation

[0042] Hybrid matching filter The hybrid matched filter combines the characteristics of the matched filter and the characteristics of gross count type algorithms. This combination can enable more effective detection when the systematic noise and the statistical noise are similar to each other in terms of the timescale of the maximum integration. Such a similarity between the systematic noise and the statistical noise is often seen in small detectors (e.g., walkie-talkies).

[0043] The hybrid matched filter does not remove all the shapes related to the background, but removes the parts that do not correspond to the source signature. Since a part of the background is retained, the hybrid matched filter estimates and removes the bias. The metric trainer of the hybrid matched filter generates a transformation T as represented by the following equation,

Equation

Equation

[0044] The performance characteristics of the alignment filter can be modeled using the following equation, [Equation] where D here is the performance characteristic, ε is the detection efficiency of the line source signature, S Mc is the number of line source counts within the measurement, B Mc is the number of background counts within the measurement, and σ 2 is a position variable drift term that depends on how well the current background matches the previous background used for estimation. Generally, since the position-dependent drift is smaller than the drift of the gross count metric, the hybrid alignment filter can perform better than the standard gross count metric when properly adjusted.

[0045] Bidimensional (「BD」) matching filter The BD integration filter helps address problems that occur when the detection algorithm uses the background rate. When the background rate suddenly increases, the detection algorithm may underestimate the background. As a result, this underestimated background is subtracted from the measurement value, causing the average of the metric to increase. Also, as the background increases, there is more Poisson noise, causing the variance of the metric to increase. Therefore, in measurements with an increasing background, the metric is likely to exceed the threshold and produce false positives.

[0046] Similar to the hybrid integration filter, the BD integration filter enables the intensity of the measurement over the background (i.e., the count number) to contribute to changes in the metric and the measurement shape. The BD integration filter introduces non-linear elements to limit the effect of this contribution. The BD integration filter balances these non-linear elements so that an increase in the background reduces the metric and the chance of false positives remains constant regardless of the conditions.

[0047] The BD integration filter algorithm does not use the oblique OSP. Instead, the BD integration filter algorithm calculates the total count in the line source signature and each background basis vector representing the background. The decomposition uses both the measurement value and the expected background measurement value. The BD integration filter is two-dimensional in the sense that it produces both an estimated line source and an estimated background for the measurement value. <##

[0048] The estimated background of the measurement values is based on both the measurement values and the estimated background, so the estimated background will be lower than the actual background. Therefore, the BD consistency filter algorithm calculates the penalty by applying a penalty function based on the difference between the estimated background and the expected background. The penalty tends to balance the increase in the mean and variance of the metric. However, when the estimated background is lower than the expected background, the metric is less likely to produce false positives, and thus the penalty is unnecessary. When using a linear penalty function, the noise within the estimated background is always amplified when the estimated background is low. The non-linear penalty function allows for a certain false positive rate at the estimated background. Furthermore, unlike the OSP consistency filter, the estimated variance is calculated using the expected background. Therefore, the metric is Rather than JPEG2025111467000021.jpg10150 Scaled as JPEG2025111467000022.jpg10150. As a result, the BD consistency filter has a higher detection count per unit, and thus higher sensitivity.

[0049] The BD consistency filter can be based on a partitioned matrix problem as represented by the following equation,

Equation

[0050] Directly solving this partitioned matrix problem can be computationally expensive as it requires calculating the vector projection of each of the source and background components for each time step. This partitioned matrix problem can be solved by partitioning it into problems having solutions represented by the following equations.

Equation

[0051] After the estimated source count and the estimated background count are generated, the metric of the DB consistency filter can be calculated as represented by the following equation.

Equation

[0052] Similar to other consistency filter algorithms, the numerator and denominator terms can be summed and integrated over multiple time segments. The penalty can be calculated such that the total integrated distribution remains constant for different background intensities and conditions of the background estimation term. Since the distribution function is not linear or fixed as a function of a threshold, the slope can be selected such that the distribution integral never exceeds the expected value. The slight over penalty generated thereby may sometimes reduce the overall sensitivity, but this sensitivity loss can be no worse than the sensitivity obtained from the use of the estimated background in at least the denominator term.

[0053] Likelihood ratio test (「LRT」) matching filter The LRT integration filter can employ the same metric evaluator as the hybrid integration filter. When the background shape is known, the LRT integration filter uses Poisson probability statistics for the measurement to detect the line source signature. The LRT integration filter deletes only one line source signature at a time. The LRT integration filter has excellent sensitivity in static detection scenarios and can be suitable for small detectors where statistical noise is the main factor. The LRT integration filter may not be suitable for large moving detectors.

[0054] The LRT integration filter constructs two hypotheses as represented by the following equations: H0 = B s B r Δt

Number

Number

Number

[0055] The resulting LRT is based on a two-sided test, while other matched filters are based on one hypothesis test at normal statistical values. Therefore, the resulting LRT can be transformed by subtracting the mean from the expected background and dividing by the expected noise of the estimated background. This result represents one hypothesis test of the background. The expected background rate is required to generate the metric. As a result, the metric can be represented by the following equation, [Number] where DM here represents the metric, and κ represents the bias term (or penalty) of any difference between the estimated total count in the background and the expected background total count in the background. The bias term is an adjustment that can be made observationally based on the background estimate. When performing simulations using a known statistical distribution (truth) and using random sampling to form the estimates, a bias term can exist. If this bias term is in the negative direction, sources will be detected more frequently than expected. Therefore, a bias term is added to account for the difference between the truth (actual expected background) and the estimated background. The LRT matched filter is optimal only for the target time window and a single intensity. These target values are fed into the metric trainer.

[0056] The LRT matched filter follows the behavioral model given by,[[]] [Number] where ε here represents the detection efficiency of source S, S Mc represents the number of source counts in the measurement, B Mc represents the number of background counts in the measurement, σ 2represents a position-variable drift term that depends on how well the current expected background and the previous expected background match. Since the position-dependent drift is always smaller than the drift of the gross count metric, the hybrid alignment filter can be a better performer than the standard gross count metric when properly adjusted.

[0057] Target NSCRAD algorithm NSCRAD algorithm The Nuisance-Rejection Spectral Comparison Ratio Anomaly Detection (「NSCRAD」) algorithm was developed by the Pacific Northwest National Laboratory. The NSCRAD algorithm first converts the measurements into a set of regions of interest and applies the transformation to remove the expected background vector. Since the regions of interest may overlap, the NSCRAD algorithm takes into account the correlation matrix estimated from the background samples.

[0058] Feature vector of the region of interest Given JPEG2025111467000033.jpg6150 and the background vector B = [b i , the NSCRAD transform α is JPEG2025111467000034.jpg6150 and can be represented by the following equation.

Equation

[0059] In addition to removing the background components, the NSCRAD algorithm can also employ a partial subspace projection matrix γ. The NSCRAD algorithm is based on the matrix of background sources Perform calculations starting from JPEG2025111467000036.jpg6150. The NSCRAD algorithm calculates the partial subspace projection transformation as represented by the following equation. γ(N)=(I - N(N t N) -1 N t )

[0060] Since the eigenvectors are projected and not independent, the NSCRAD algorithm can use oblique subspace projection. The transformed covariance matrix can be represented by the following equation, β=(αΣα t ) -1 α where Σ here represents covariance. The NSCRAD algorithm adjusts the subspace projection to remove the background as represented by the following equation. γ(N;α,β)=(I - αN(N t β t αN) -1 N t β t )

[0061] Many different algorithms can be developed from these transformations. The NSCRAD algorithm can be represented by the following equation. DM = X t (β t γα)X This equation is a classical quadratic form where Q = β t γα. This metric has degrees of freedom given by subtracting 1 from the number of energy features for background removal and further subtracting the number of background sources to be removed, and has a χ 2 statistic. The implementation of the NSCRAD algorithm can be as described below.

Table 2

[0062] Since the NSCRAD algorithm has a quadratic form, the aggregation is performed not simply on the numerator and denominator as in other algorithms, but on the spectral comparison ratio ("SCR") vector and the projected SCR vector. As a result, the amount of memory required to spatially aggregate the measured values increases. Therefore, the NSCRAD algorithm may be appropriate when using temporal aggregation, but may not be appropriate when using spatial aggregation.

[0063] The NSCRAD algorithm essentially has the statistics of χ 2 . Therefore, as will be described later, this algorithm may not be associated with metric aggregation. The NSCRAD algorithm only removes a single dynamic degree of freedom and a specified number of fixed background dimensions. With this method, the complexity of the background evaluator decreases, but the systematic variability attenuation may decrease. Therefore, the maximum aggregation time is limited for large detectors. The NSCRAD algorithm may produce an insufficient statistical distribution when the total number is small. This tendency sometimes requires increasing the threshold regarding the level expected by the χ 2 statistics.

[0064] NSCRAD matching filter Unlike the NSCRAD algorithm, the NSCRAD integrated filter has statistics with unit variance and normal distribution. As a result, the tendency to have outliers is reduced, there is an increase in systematic variance rejection, and the operating range becomes wider. The minus side of this method is that the region of interest for optimizing performance is different for each target line source. Therefore, it is necessary to pair all detection metric evaluators with corresponding metric trainers. The implementation of the NSCRAD integrated filter is the same as that of the NSCRAD algorithm, except that the integrated filter coefficients are calculated for each region of interest as represented by the following equation, G = S t β t γ(N; α, β)α The decision metric is represented by the following equation.

Equation

[0065] Unlike the NSCRAD algorithm, the NSCRAD integrated filter generates a metric for one line source signature. To use the filter to cover the line source range, a description of the line source signature covering the line source to be detected is required.

[0066] Metric aggregation The metric aggregation component can aggregate metrics spatially or temporally. Many spectrum detection algorithms perform aggregation within a fixed time window. The spatial aggregation method generates an aggregated metric from measurements collected at the same location but at different times (e.g., the detector makes multiple sweeps across one area). The temporal aggregation method generates an aggregated metric for each time interval based on the previous time interval. The metric aggregation component can adopt various aggregation methods to generate an aggregated metric. When the RDA system adopts multiple aggregation methods, the metric aggregation component generates an aggregated metric for each aggregation method. The metric evaluation component can detect whether a line source exists using the maximum value of the aggregated metric.

[0067] The temporal aggregation method uses a rolling window of measurements when generating an aggregated metric. The aggregation component generates aggregated metrics for multiple rolling windows, each containing a different number of measurements.

[0068] The spatial aggregation method can apply a weighted back projection method assuming a simple 1 / R 2 kernel to generate an aggregated metric.

[0069] Metric standardization The metric generation component can adopt a metric normalization method to correct the statistical distribution of a detection algorithm that generates a standard normal distribution on individual metrics. The metric normalization method maintains a running estimate of the mean and variance of each metric. The metric normalization method calculates the deviation (bias) from zero of the mean and the excess variance from this running estimate. Next, the metric normalization method corrects the detection metric to remove this bias and reduce the variance.

[0070] The metric normalization method can employ a bias forgetting factor and a variance forgetting factor. The metric normalization method uses these factors within a first order infinite impulse response filter to maintain bias and variance estimates. The smaller this number is, the slower the learning of the algorithm with respect to changes in conditions becomes. For example, the forgetting factor can be set to approximately 0.01, which means that the metric normalization method learns the changes in approximately 100 measurements. These forgetting factors become smaller when the background is completely stationary and larger when the background is changing significantly. The bias forgetting factor and the variance forgetting factor can have the same value or different values. The metric normalization method can exclude metrics that clearly include line sources when calculating these bias and variance estimates. For example, metrics that exceed a threshold can be excluded.

[0071] Measurement calibration The RDA system calibrates the measurement values to account for the drift of the detector sensors during photon collection. As a result of the drift, photons at one energy level may be identified as being at different energy levels, and the spectrum may be scaled wider or narrower. To account for this drift, the measurement calibration component reassigns the counts from the measurement bins to different bins, which is referred to as a change or rebinning of the measurement binning structure. The radiation source detector can periodically collect calibration measurements of known sources and / or measurements from the background from the detector to identify the characteristics of the drift. The measurement calibration component identifies the necessary calibration by comparing the calibration measurements to the expected measurements of the known source to determine the drift. For example, if the source signature and calibration measurements of a known source have corresponding peaks in different bins, it may be necessary to reassign the counts of the measurement values based on the bin differences. For example, in a simple case, the measurement values can have exactly the same shape as the source signature, except that they are shifted to a lower energy level by one bin due to drift. The calibration component can also correct for the non-linearity of the detector response, including the non-linearity inherent in the photon absorber and the non-linearity from the amplifier. Generally, these non-linearities are captured in the characterization measurements performed prior to unfolding and are provided in the form of polynomial coefficients or spline coefficients, or energy / channel pairs.

[0072] In some embodiments, monitoring can be performed using a strong measurement of a line source having well-separated features, and in some cases further correction for non-linearity can be performed. To do this, each line source signature is enhanced using a look-up table that includes the effective energy of each feature within the line source signature useful for calibration (e.g., an emission line of gamma rays or a Compton edge), where the effective energy is the energy seen for a perfectly measured line source signature that takes into account the shift from the true energy due to the effects of shielding and scattering on the feature position extraction procedure (e.g., centroiding or line fitting). This table also lists, in addition to the effective energy, the expected range of the effective energy (e.g., standard deviation) along with the range of the ratio of peak counts to total counts over the spectrum. The effective energy, the range of the effective energy, and the peak-to-total count ratio can all be functions of the signal-to-noise ratio of the measurement and the total number of counts in the measurement. These dependencies are included in the look-up table of line source signatures. This look-up table also includes settings for the feature extraction procedure (e.g., region of interest). Some line source signatures, such as those with a large amount of shielding or those of nuclides having no well-defined features, are not suitable for calibration monitoring and correction, have no effective energy, and do not include relevant values.

[0073] When a source is detected with a sufficiently high decision metric and source identification probability, the effective energy of each feature is calculated and compared to the energy in the lookup table of the source signature. If the extracted energy and the effective energy listed in the lookup table differ significantly from the expected range of the measured conditions faced, a flag indicating a calibration problem can be set (calibration monitor), and / or this information can be used to update the energy scale (calibration correction). Also, the measured peak-to-total count ratio is compared to the appropriate range in the lookup table to confirm that the correct features and source signature have been found regardless of the energy scale. Multiple measurements of calibration errors are required to trigger a correction. In this case, a constrained spline is used that maintains the overall shape of the non-linearity from more detailed controlled pre-deployment calibration measurements of a particular detector or general detector type, but also adjusts to match exactly the measurements at the available energies, to generate a corrected energy scale.

[0074] The measurement calibration component reassigns the count of each bin of the measurement values to the next higher energy level to account for changes in the energy scale due to drift and non-linearity. As another example, the drift in the energy range can vary across the entire energy range. For example, the counts in the low energy bins can be compressed into fewer bins, and the counts in the high energy bins can be expanded into more bins. As another example, when it is necessary to assign the counts of the bins of the measurement values to multiple bins, there can be a correlation in the statistics of the data. The statistical analysis method may introduce a bias that can affect the detection metric due to this correlation. To prevent such statistical bias, the measurement calibration component employs a statistical rebinning method. For example, the measurement calibration component can randomly select a binomial variable of the count number of the bins of the measurement values that need to be reassigned to other bins. By introducing this randomness, the Poisson statistics of the rebinning are maintained. The non-linearity correction can consider environmental factors such as temperature or operational factors such as the count rate. Therefore, the coefficients can be functions of these factors.

[0075] Figure 5 shows the calibration of measurement values based on the drift of the detector. The measurement value 510 includes an extended bin 511 with extended edges 512 and 513. The calibrated measurement value 520 includes bins 521 - 523 with edges 524 - 527 specified by the binning structure. The calibrated measurement value represents the compression of the scale of the measurement value. Bin 511 overlaps bins 521 - 523. The measurement calibration component assigns the fraction of the count of bin 511 to bin 521 to distribute the count of bin 511 among bins 521 - 523. This fraction is based on the ratio of the energy range 528 to the energy range of bin 511. Similarly, the fraction of the count corresponding to the energy range 529 is assigned to bin 523. The remaining counts are assigned to bin 522. The measurement calibration component can select the count number based on the binomial distribution of the fraction of the count instead of using the exact fraction of the count.

[0076] Dynamic background estimation The RDA system provides a dynamic estimation of the measurement background. When the detector is moving, the measured background can change significantly (e.g., by a factor of 5). Gross count algorithms, such as the gross counts k-sigma algorithm, are dominated by these variations and have little sensitivity to weak sources. The spectral detection algorithm detects sources by looking for changes in the spectral signature rather than changes in the overall intensity. Therefore, the spectral detection algorithm is trained using a set of representative backgrounds for the area to be measured. Typically, such representative backgrounds are developed based on collecting background data over several hours for use as a fixed background training set.

[0077] In some cases, it may be difficult to generate representative backgrounds. For example, when the detector is planned to be deployed in a new area, it may not be possible to collect background data. Also, the detector response may change over time. These changes can include gain drifts, changes in the non-linear response, and changes in energy resolution and efficiency. Although the measurement calibration component corrects for gain drifts, the correction of other changes in the detector response is limited. As a result, a representative background generated from background data collected on one day may not represent the background detected on another day. The RDA system can use dynamic background estimation, which is typically trained "on the fly" during normal operation, to make the detection sensitivity and false positive rate more reliable, robust, and predictable.

[0078] The RDA system receives measurement values from detectors that perform both background measurements and source measurements. Therefore, the background estimation component excludes the contribution of the source. These sources can include intentional sources, unknown sources found as part of source exploration, and accidental encounters with sources commonly found in the (e.g., medical) environment. The background estimation component recognizes when the measurement values represent accidental sources in order to improve the estimated background and excludes these measurement values when generating the estimated background. The RDA system employs a rejection component that analyzes metrics to determine which measurement values should be excluded. It is also possible not to exclude the measurement values of the source. For example, if a source is present when the detector starts collecting measurement values, or if the radiation from the source increases gradually, the source may not be detected and thus may not be excluded. As a result, a small amount of the source signature may be incorporated into the background, which may reduce the sensitivity of the background estimation for this source. In such cases, the measurement values of the source based on slightly higher intensities are not recognized as sources and are not excluded further.

[0079] To prevent such sources from not being detected and thus not being excluded, the RDA system employs a sensitivity tester. The sensitivity tester periodically checks the sensitivity to many sources by injecting sources of various intensities into the estimated background to generate test measurement values. Then, the sensitivity tester uses the estimated background generated by the background estimation component to execute a detection algorithm on the test measurement values to determine the minimum intensity of the source required for detection. The sensitivity tester compares this minimum intensity with the previous minimum intensity to determine whether the sensitivity has declined. If so, it restarts the background estimation component to generate a new estimated background.

[0080] Rank average The rank average algorithm generates the background rate for all received measurements. The RDA system can use the rank average algorithm together with the gross count detection algorithm when the operating conditions cause the background of the measurements to temporarily increase or decrease. For example, when an object (e.g., a truck) passes the detector, the background rate may decrease, and when the detector passes a structure containing a radiation source (e.g., a building), the background rate may increase.

[0081] The rank average algorithm sorts the total counts of each measurement collected over a certain period to account for a temporary decrease or increase in the background. Next, the rank average algorithm generates the average of the top (or bottom) counts. This average is a biased estimate of the background. The rank average algorithm calculates a bias factor using standard statistical assumptions. Then, the rank average algorithm corrects the biased estimate based on the bias factor. For example, the rank average algorithm can calculate the expected offset of the Gaussian distribution within the average total count and subtract this offset from the estimated background.

[0082] The rank average algorithm can be particularly useful when the detector is stationary, when the background is obscured by nearby objects, and when an estimated background is needed in the absence of these objects.

[0083] Progressive Projection The RDA system employs a progressive projection algorithm that estimates a set of background vectors over all possible linear combinations of background shapes. This spanning set is called a basis, but the background vectors are not orthogonal. The progressive projection algorithm queues background measurements (i.e., background where the source is not detected) until there is a change in the spectral shape of the background measurements or until a certain minimum time has elapsed. Next, the progressive projection algorithm compares the background measurements with the changed spectral shape to the current basis vectors to determine which basis vector the background measurement best represents. If the background measurement is found to be not strongly related or separated from any of the existing basis vectors, this background measurement does not represent anything new regarding the background. Conversely, if the background measurement is found to be strongly related or separated from a basis vector, the progression projection algorithm averages this background measurement with the existing basis vectors. Next, the progressive projection algorithm attempts to maximize the eigenvalues associated with the correlation matrix formed by the basis vectors so that the basis vectors have the maximum diversity. The progressive projection algorithm employs a forgetting function so that the progressive projection algorithm does not update the basis vectors so that they become collinear when the background measurements are the same over a long period of time.

[0084] The progressive projection algorithm can initialize the basis vectors from background measurements collected after the start or after clearing of the algorithm. If the detector is exposed to the source during initialization, this source is trained into the basis vectors and thus the sensitivity tester can become unresponsive until it is activated. Progressive projection can have three starting methods. The hot start method uses a predetermined set of basis vectors specified within the configuration data. The warm start method generates the average of background measurements over a specified period and mixes this average with predetermined background measurements from the configuration unit (or previous runs). The cold start method injects known background measurements after taking the average over a specified period.

[0085] Each start method has different advantages and disadvantages. The cold start method is robust in the sense that it operates even if the source is present at the same location as the detector, as long as the source is not removed. However, this robustness comes at the cost that it can take a long time for the method to be fully trained and reach full sensitivity. If the known background measurements of the cold start method do not accurately reflect the variability of the background measurements, the RDA system may not detect the source correctly and may need to be restarted. The warm start method uses previous background measurements and is thus trained quickly. However, if the background measurements are significantly different from the previous background measurements (e.g., there is a source present in the background measurements), the progressive projection algorithm may become unresponsive to the source and may need to be restarted. The hot start method can generate the basis vectors immediately. However, if the background measurements are not represented by the predetermined basis vectors, the RDA system generates false positives, i.e., the background measurements are shown as if a source is present. As a result, the progressive projection algorithm does not update the basis vectors, the false positives continue, and it is necessary to restart using the cold start method or the warm start method.

[0086] Classification In some embodiments, the metric generation component employs an identification algorithm that is similar to the OSP alignment filter detection algorithm but has a different normalization term. The metric generation component uses the expected background calculated before measurement to calculate the expected variance Σ within each channel B When considering the mixing of the line source signatures of the line sources, the metric generation component calculates the mixing of the line source signatures of each line source that optimally solves Σ B X = Σ B Ax and creates a mixed line source signature S = Ax. When not considering mixing, the metric generation component uses each line source signature S i .

[0087] For each line source signature (mixed or unmixed), the metric generation component calculates the transformation as represented by the following equation

Equation

[0088] The null hypothesis generation component calculates the variance Σ of the measured values using the expected background Y . The null hypothesis generation component generates the variance of each metric and the covariance of the metrics using the following equation σ i 2 = Q i Σ X Q i t c ij = Q i Σ X Q j t where σ herei 2 represents the variance of S i and c ij represents the covariance of S i and S j The null hypothesis generation component selects the maximum of the variances of the null hypotheses. Next, the null hypothesis generation component calculates the metric of the null hypothesis based on the maximum value of a random walk near zero using a fixed number of draws as represented by the following equation,

Equation

[0089] Computer system A computer system (e.g., a network node or a collection of network nodes) capable of implementing the RDA system and other described systems can include a central processing unit, an input device, an output device (e.g., a display device and a speaker), a storage device (e.g., a memory and a disk drive), a network interface, a graphics processing unit, a cellular or other wireless link interface, a global positioning system device, and inertial navigation, etc. The input device can include a keyboard, a pointing device, a touch screen, a gesture recognition device (e.g., for air gestures), a head and eye tracking device, and a microphone for voice recognition, etc. The computer system can include a high-performance computer system, a cloud-based server, a desktop computer, a laptop, a tablet, an e-reader, a personal digital assistant, a smartphone, a game console, and a server, etc. For example, simulation and training can be executed using a high-performance computer system, and classification can be executed by a mobile device that is part of a network node. The computer system can access computer-readable media including a computer-readable storage medium and a data transmission medium. The computer-readable storage medium is a tangible storage means that does not include a transient propagated signal. Examples of the computer-readable storage medium include a primary memory, a cache memory, a secondary memory (e.g., a DVD), and other storage devices. The computer-readable storage medium can store or encode computer-executable instructions or logic for implementing the RDA system and other described systems. The data transmission medium is used to transmit data via a wired or wireless connection through a transient propagated signal or a carrier wave (e.g., electromagnetic). The computer system can include a secure encryption processor as part of the central processing unit that generates keys and securely stores them and uses these keys to encrypt and decrypt data.

[0090] The RDA system and other described systems can be described in the general context of computer-executable instructions, such as program modules and components executed by one or more computers, processors, or other devices. Generally, a program module or component includes routines, programs, objects, and data structures that perform the tasks of the RDA system and other described systems or implement data types. Usually, the functions of program modules can be combined or distributed as desired in various examples. Aspects of the RDA system and other described systems can be implemented in hardware, for example, using application-specific integrated circuits (ASICs) or field-programmable gate arrays (FPGAs).

[0091] Flow chart FIG. 6 is a flowchart showing the high-level processing of the OSP alignment filter component of the RDA system in some embodiments. Measurements X over time t t and a component 600 that transfers the indication of the line source definition S are called. In block 601, the component calculates a transform projection T that complements the background B. In block 602, the component calculates a partial projection of the background-removed X t onto the line source definition S. In block 603, the component estimates the partial noise n t at time t. In block 604, the component updates the projection of the moving window so as to add the partial projection over time t and subtract the partial projection over time t - Δ. In block 605, the component updates the noise of the moving window so as to add the partial noise over time t and subtract the partial noise over time t - Δ. In block 606, the component calculates a detection metric as the projection divided by the noise. Then, the component terminates.

[0092] Figure 7 is a flowchart showing the process of calibration components of the RDA system in some embodiments. A component 700 that modifies existing bins is called based on the drift of the sensor of the radiation detector in signal energy detection. A component that transfers the indication of the current bin and the edges and measured values of scale s is called. In blocks 701 to 705, the component initializes the number of overflow bins (#OFbins) and the number of underflow bins (#UFbins). In block 701, the component sets the number of overflow bins to 0. In decision block 702, if the scale is less than 1.0, the component proceeds to block 703, otherwise the component proceeds to block 704. In block 703, the component sets the number of underflow bins to the value obtained by subtracting the integer part of the number of bins (#bin) multiplied by the scale from the number of bins. In block 704, the component sets the number of underflow bins to 0. In decision block 705, if the scale is equal to 1.0, since rebinning is not required, the component ends, otherwise the component proceeds to block 706. In block 706, the component initializes variables b0 and b1 to track the energy of the bin edge and the index i of the bin. In blocks 707 to 713, the component loops through the execution of rebinning. In block 707, the component increments the index i to the next bin. In decision block 708, if all bins have already been indexed, the component ends, otherwise the component proceeds to block 709. In block 709, the component sets b0 to the current edge and b1 to the next edge. In block 710, the component sets index i0 and i1 to the integer parts of index b0 and index b11 respectively. In decision block 711, if index i0 is equal to index i1, the component proceeds to block 713, otherwise the component proceeds to block 712.In block 712, after the component calls the bin splitting component to split the bins and then loops back to block 707 to select the next bin, otherwise the component proceeds to block 713. In block 713, the component increments the count out[i0] of the new bin by the count X[i] of the measured values and then loops back to block 707 to select the next bin.

[0093] Figure 8 is a flowchart showing the processing of the bin splitting component of the RDA system in some embodiments. A component 800 for splitting bins is called. In block 801, the component sets the variable n to the count of bin[i], sets the variable split to a fraction of these counts, and assigns it to out[i0]. In block 802, the component selects the number of counts x1 to be added to out[i0] by applying a binomial distribution based on the variables n and split. In block 803, the component adds the count x1 to out[i0] and decrements the remaining count of the variable n by the count x1. In block 804, the component sets the index j to index i0 plus 1. In decision block 805, if the variable j is greater than the minimum value of the variable x1 and the number of bins #bin, the component ends; otherwise, the component increments the variable j and proceeds to block 806. In block 806, the component sets the variable x1 to a value based on the binomial distribution of the variable n and 1.0 divided by the scale s. In block 807, the component decrements the variable n by the variable x1. In block 808, the component increments the count of out[i1] by the variable x1. In decision block 809, if the index i1 is less than the number of bins, the component proceeds to block 801; otherwise, the component proceeds to block 811. In block 810, the component increments out[i1] by the variable n and loops back to block 805. In block 811, the component sets the overflow count to the variable n and loops back to block 805.

[0094] FIG. 9 is a flowchart showing the processing of an estimated background based on the rank average component of the RDA system in some embodiments. A component 900 for adjusting the estimated value of the background is called. In block 901, the component initializes a variable i to index the entire bin. In decision block 901, the component increments the variable i, and if it is greater than the number of bins, the component ends; otherwise, the component proceeds to block 903. In block 903, the component creates a sort of the counts within the moving window of the indexed bin. In block 904, the component calculates the average of the top counts. In block 905, the component calculates the expected offset of the top counts based on the average. In block 906, the component adjusts the count of the indexed bin by the offset and loops back to block 902 to select the next bin.

[0095] In the following paragraphs, various embodiments of aspects of the RDA system and other systems will be described. The implementation of the system can also adopt any combination of the embodiments. The processes described below can be executed by one or more computer systems comprising a processor that executes computer-executable instructions stored on a computer-readable storage medium implementing the system.

[0096] In some embodiments, one or more computer systems are provided for identifying a radiation source. The one or more computer systems include one or more computer-readable storage media storing computer-executable instructions, and one or more processors executing the computer-executable instructions stored on the one or more computer-readable storage media. The instructions implement a radiation source detector that receives measurements of radiation, generates a detection metric for one or more sources indicating whether the source is present within the measurements, and evaluates the detection metric to detect whether a source is present within the measurements. The instructions implement a radiation source identifier that, when a source is detected within the measurements, generates an identification metric for one or more sources indicating whether the source is present within the measurements, generates a null hypothesis metric indicating whether no source is present within the measurements, and evaluates the one or more identification metrics and the one or more null hypothesis metrics to identify the source if a source is present within the measurements. In some embodiments, the radiation source detector generates the detection metric for a source based on a detection source signature of the source and generates the identification metric for the source based on a plurality of identification source signatures of the source. In some embodiments, the identification source signatures of the source represent sources with different shielding. In some embodiments, the radiation source identifier identifies a radiation source class of the identified source. In some embodiments, the measurements are calibrated to account for drift within the detector that collected the measurements. In some embodiments, the radiation source identifier generates an aggregate metric for each source. In some embodiments, the measurements within a window are aggregated and radiation is detected and identified based on the aggregated measurements.

[0097] In some embodiments, a method for detecting the presence of a radiation source from radiation measurement values, executed by a computer system. The method uses, for one or more sources and for one or more detection algorithms, a detection algorithm that takes into account an estimated value of background radiation generated based on previous measurement values of the current measurement value, to generate a metric indicating the similarity between the current measurement value and the source signature of the source, and generates an aggregated metric of the source from one or more metrics of the source. The method analyzes the aggregated metric of one or more sources to determine whether a source exists within the current measurement value. The method indicates that the presence of the source has been detected when it is determined that there is a high likelihood that a source exists within the current measurement value. In some embodiments, the method further generates an estimated value of background radiation based on previous measurement values excluding the previous measurement values in which the presence of the source has been detected. In some embodiments, the estimated value of background radiation is a weighted average based on the measured value of background radiation and a priori estimate. In some embodiments, the source signature is related to the identification of the source and the shielding of the source. In some embodiments, the method generates a metric for each source signature related to the source for each detection algorithm. In some embodiments, the measurement value is represented by a histogram of energy ranges including the number of photons in each energy range. In some embodiments, a metric is generated for each of a plurality of window measurement values, each window including the current measurement value and many adjacent previous measurement values. In some embodiments, the detection algorithm is related to a coefficient, and the method dynamically adjusts the coefficient based on the estimated value of background radiation. In some embodiments, the analysis of the aggregated metric includes applying a classifier trained using training data including a feature vector having features including the detection algorithm, window size, and aggregated metric, and a label for each feature vector indicating whether a source exists. In some embodiments, the feature vector is generated from the collected radiation measurement values. In some embodiments, the feature vector is generated from simulated radiation measurement values.In some embodiments, the method further calibrates the current measurement to account for drift associated with the detector used to collect the measurement values.

[0098] In some embodiments, a method is provided for identifying a radiation source from radiation measurement values, executed by a computer system. The method receives current measurement values in which the presence of a source has been detected based on a detection source signature. The method generates a metric for each of a plurality of identification source signatures using an identification algorithm. The metric indicates a similarity between the current measurement values and the source signature of that source. The identification algorithm takes into account an estimated value of background radiation generated when detecting the presence of the source. The identification source signature is more comprehensive than the detection source signature. The method generates a source metric for each source based on one or more metrics generated using the source signature of that source. The method generates a null hypothesis metric indicating the similarity between the current measurement values and the estimated value of background radiation. The source and the estimated value of background radiation are targets. The source metric and the null hypothesis metric are target metrics. The method generates a target probability representing the presence of that target in the current measurement values for each target. The target probability is based on the target metric of that target. In some embodiments, the method generates a radiation source class probability for each of a plurality of radiation source classes, in which the current measurement values represent a source of that radiation source class. The radiation source class probability is generated based on the target probability. In some embodiments, the generation of the target probability takes into account the prior probability of each target. In some embodiments, the identification algorithm is based on an orthonormal subspace projection matched filter algorithm.

[0099] Orthogonal subspace projection matching filter algorithm In some embodiments, a method is provided for generating a metric related to a radiation source in radiation measurements, which is executed by one or more computer systems. The method accesses a source signature, an estimated background, and a background basis vector of the source. The method generates a projection vector based on the source signature, the estimated background, and the background basis vector. The method accesses the measurements. The method generates a metric based on the source signature, the estimated background, and the background basis vector. In some embodiments, the source signature is a histogram representing an energy range divided into energy bins, and each energy bin has a value representing the number of photons radiated by the source over a certain time interval. In some embodiments, generating the projection vector includes generating a weight matrix from the estimated background, where the variance of each energy bin is based on the expected background. In some embodiments, generating the projection vector further includes generating a source-signature weighted projection and a background weighted projection, and setting the projection vector to the difference between the source-signature weighted projection and the background weighted projection. In some embodiments, the source signature represents the shielding of the source. In some embodiments, the method further generates an aggregated measurement of a different number of measurements, and for each aggregated measurement, generates a metric based on the aggregated measurement, the projection vector, and the expected variance of the aggregated measurement. In some embodiments, the method indicates that the presence of the source is detected when the metric meets a source presence threshold. In some embodiments, the method indicates that the presence of the source is semi-definitively detected when the metric meets a source presence threshold. In some embodiments, generating the detection metric includes dividing the product of the projection vector and the measurement by the square root of the expected variance.

[0100] In some embodiments, one or more computer systems are provided for generating a metric related to a radiation source in a measured value of radiation. The one or more computer systems include one or more computer-readable storage media storing computer-executable instructions and one or more processors executing the computer-executable instructions stored on the one or more computer-readable storage media. The instructions, when executed, access a source signature, an estimated background, and a background basis vector of the source. The instructions generate a projection vector based on the source signature, the estimated background, and the background basis vector. The instructions access the measured value. The instructions generate a metric based on the measured value, the projection vector, and an expected variance of the measured value. In some embodiments, the source signature is a histogram representing an energy range divided into energy bins, and each energy bin has a value representing the number of photons emitted by the source over a certain time interval. In some embodiments, the instructions for generating the projection vector generate a weight matrix from the estimated background, and the variance of each energy bin is based on the expected background. In some embodiments, the instructions for generating the projection vector further generate a source-signature weighted projection and a background weighted projection, and set the projection vector to the difference between the source-signature weighted projection and the background weighted projection. In some embodiments, the source signature represents shielding of the source. In some embodiments, the instructions further generate an aggregated measurement of a different number of measured values, and for each aggregated measurement, generate a metric based on the aggregated measurement, the projection vector, and an expected variance of the aggregated measurement. In some embodiments, the instructions further indicate that the presence of the source has been detected when the metric meets a source presence threshold. In some embodiments, the instructions further indicate that the presence of the source has been semi-deterministically detected when the metric meets a source presence threshold. In some embodiments, the instructions for generating a detection metric divide the product of the projection vector and the measured value by the square root of the expected variance.

[0101] In some embodiments, a method is provided for generating a metric related to a radiation source in sequentially collected radiation measurements, which is executed by one or more computer systems. The method accesses the source signature of the source, the estimated background, and the background basis vectors. The method generates a projection vector as represented by the following equation, T = S t W(I - B(B t WB) -1 B t W) where T represents the projection vector, S represents the source signature, B represents the estimated background, and W represents the background basis vectors. The method accesses the measurements. The method generates a metric as represented by the following equation,

Equation

[0102] Hybrid matching filter In some embodiments, a method is provided for generating a metric related to a radiation source in measured radiation values, which is executed by one or more computer systems. The method accesses the source signature of the source and the estimated background. The method generates a background vector that removes the source signature. The method generates a projection vector based on a weight matrix derived from the estimated background, the source signature, and the estimated background. The method accesses the measured values. The method generates a metric based on the measured values, the projection vector, and an expected variance an average background. In some embodiments, the source signature is a histogram representing an energy range divided into energy bins, and each energy bin has a value representing the number of photons emitted by the source over a certain time interval. In some embodiments, the weight matrix is generated from the estimated background, and the variance of each energy bin is based on the expected background. In some embodiments, the source signature represents the shielding of the source. In some embodiments, the method generates an aggregated measurement of different numbers of measured values, and for each aggregated measurement, generates a metric based on the aggregated measurement and the projection vector. In some embodiments, the method indicates that the presence of the source has been detected when the metric meets a source presence threshold. In some embodiments, the method indicates that the presence of the source has been semi - deterministically detected when the metric meets a source presence threshold.

[0103] In some embodiments, one or more computer systems are provided for generating a metric related to a radiation source in measured radiation values. The one or more computer systems include one or more computer-readable storage media storing computer-executable instructions and one or more processors executing the computer-executable instructions stored on the one or more computer-readable storage media. The instructions, when executed, access a source signature of the source and an estimated background. The instructions generate a background vector that removes the source signature. The instructions generate a projection vector based on a weight matrix derived from the estimated background, the source signature, and the estimated background. The instructions access the measured values. The instructions generate a metric based on the measured values, the projection vector, and an expected variance mean background. In some embodiments, the source signature hs is a histogram representing an energy range divided into energy bins, where each energy bin has a value representing the number of photons emitted by the source over a certain time interval. In some embodiments, the weight matrix is generated from the estimated background, and the variance of each energy bin is based on the expected background. In some embodiments, the source signature represents shielding of the source. In some embodiments, the instructions further generate an aggregated measurement of a different number of measured values and, for each aggregated measurement, generate a metric based on the aggregated measurement and the projection vector. In some embodiments, the instructions further indicate that the presence of the source has been detected when the metric meets a source presence threshold. In some embodiments, the instructions further indicate that the presence of the source has been semi-deterministically detected when the metric meets a source presence threshold.

[0104] In some embodiments, a method is provided for generating a metric indicating the presence of a radiation source in sequentially collected measured radiation values, performed by one or more computer systems. The method accesses a source signature of the source, an estimated background, and a background basis vector. The method generates a projection vector as represented by the following equation,

Equation

[0105] Bidimensional (「BD」) matching filter In some embodiments, a method is provided for generating a metric related to a radiation source in measured radiation values, which is executed by one or more computer systems. The method accesses the measured values, the source signature of the source, and the estimated background. The method generates a background vector that removes the source signature. The method generates a projection vector based on a weight matrix derived from the estimated background, the source signature, and the estimated background. The method accesses the measured values. The method generates a metric based on the measured values, the projection vector, and the expected variance mean background. In some embodiments, the source signature is a histogram representing an energy range divided into energy bins, and each energy bin has a value representing the number of photons emitted by the source over a certain time interval. In some embodiments, the weight matrix is generated from the estimated background, and the variance of each energy bin is based on the expected background. In some embodiments, the source signature represents the shielding of the source. In some embodiments, the method further generates an aggregated measurement of a different number of measured values, and for each aggregated measurement, generates a metric based on the aggregated measurement and the projection vector. In some embodiments, the method indicates that the presence of the source has been detected when the metric meets a source presence threshold. In some embodiments, the method indicates that the presence of the source has been semi - deterministically detected when the metric meets a source presence threshold.

[0106] In some embodiments, one or more computer systems are provided for generating a metric related to a radiation source in measured values of radiation. The one or more computer systems include one or more computer-readable storage media storing computer-executable instructions for controlling the one or more computer systems, and one or more processors executing the computer-executable instructions stored in the one or more computer-readable storage media. The instructions access the measured values, the source signature of the source, and the estimated background. The instructions generate a background vector that removes the source signature. The instructions generate a projection vector based on a weight matrix derived from the estimated background, the source signature, and the estimated background. The instructions access the measured values. The instructions generate a metric based on the measured values, the projection vector, and the expected variance average background. In some embodiments, the source signature is a histogram representing an energy range divided into energy bins, and each energy bin has a value representing the number of photons emitted by the source over a certain time interval. In some embodiments, the weight matrix is generated from the estimated background, and the variance of each energy bin is based on the expected background. In some embodiments, the source signature represents the shielding of the source. In some embodiments, the instructions further generate an aggregated measurement of a different number of measured values, and for each aggregated measurement, generate a metric based on the aggregated measurement and the projection vector. In some embodiments, the instructions further indicate that the presence of the source has been detected when the metric meets a source presence threshold. In some embodiments, the instructions indicate that the presence of the source has been semi-deterministically detected when the metric meets a source presence threshold.

[0107] In some embodiments, a method is provided for generating a metric indicating the presence of a radiation source in measured values of radiation collected in sequence, performed by one or more computer systems. The method accesses the source signature of the source, the estimated background, and the background basis vector. The method solves the following equation,

Equation

Equation

Equation

[0108] LRT matching filter In some embodiments, a method is provided for generating a metric related to a radiation source in radiation measurement values, which is executed by one or more computer systems. The method accesses the source signature of the source and the estimated background. The method generates a first hypothesis based on the estimated background and the estimated background rate. The method generates a second hypothesis based on the estimated background, the estimated background rate, and the source signature. The method generates a first likelihood of the first hypothesis given a Poisson distribution. The method generates a second likelihood of the second hypothesis given a Poisson distribution. The method generates a projection vector based on the first likelihood and the second likelihood. The method accesses the measurement values. The method generates a metric based on the measurement values, the projection vector, and the estimated background rate. In some embodiments, the source signature is a histogram representing an energy range divided into energy bins. The energy bins have values representing the number of photons emitted by the source over a certain time interval. In some embodiments, the source signature represents the shielding of the source. In some embodiments, the method further generates an aggregated measurement of different numbers of measurement values, and for each aggregated measurement value, generates a metric based on the aggregated measurement value and the projection vector. In some embodiments, when the metric meets a source presence threshold, it indicates that the presence of the source has been detected. The method indicates that when the metric meets the source presence threshold, the presence of the source has been semi - deterministically detected.

[0109] Provide one or more computer systems for generating a metric related to a radiation source in a measured value of radiation. The one or more computer systems include one or more computer-readable storage media storing computer-executable instructions for controlling the one or more computer systems, and one or more processors executing the computer-executable instructions stored in the one or more computer-readable storage media. The instructions access a source signature of the source and an estimated background. The instructions generate a first hypothesis based on the estimated background and an estimated background rate. The instructions generate a second hypothesis based on the estimated background, the estimated background rate, and the source signature. The instructions generate a first likelihood of the first hypothesis given a Poisson distribution. The instructions generate a second likelihood of the second hypothesis given a Poisson distribution. The instructions generate a projection vector based on the first likelihood and the second likelihood. The instructions access the measured value. The instructions generate a metric based on the measured value, the projection vector, and the estimated background rate. In some embodiments, the source signature is a histogram representing an energy range divided into energy bins, and each energy bin has a value representing the number of photons emitted by the source over a fixed time interval. In some embodiments, the source signature represents shielding of the source. In some embodiments, the instructions generate an aggregated measurement of a different number of measured values and, for each aggregated measurement, generate a metric based on the aggregated measurement and the projection vector. In some embodiments, the instructions indicate that the presence of the source has been detected when the metric meets a source presence threshold. In some embodiments, the instructions indicate that the presence of the source has been semi-deterministically detected when the metric meets a source presence threshold.

[0110] In some embodiments, provide a method for generating a metric indicating the presence of a radiation source in sequentially collected measured values of radiation, performed by one or more computer systems. The method constructs a hypothesis represented by the following equation: H0=B s B r Δt

Number

Number

Number

Number

[0111] Rank average In some embodiments, provided is a method of adjusting a measurement count of a measurement value to account for a transient change in the measurement value, executed by one or more computer systems. The method accesses the measurement value. For each measurement value, the method generates a measurement count of that measurement value. The method generates an upper average measurement count that is an average of measurement counts higher than a threshold count. The method calculates an expected offset of a probability distribution based on the upper average measurement count and measurement counts that exceed the threshold. The method subtracts an estimated offset from the measurement count of each measurement value. In some embodiments, the measurement value is a measurement value of radiation and the transient change is a result of a change in background radiation. In some embodiments, the method further detects when the measurement value includes a radiation source. In some embodiments, the detection is performed using a gross count detection algorithm. In some embodiments, the upper average measurement count is based on a probability distribution.

[0112] In some embodiments, one or more computer systems that adjust the measurement count of a measurement value to account for temporary changes in the measurement value. The one or more computer systems include one or more computer-readable storage media that store computer-executable instructions for controlling the one or more computer systems, and one or more processors that execute the computer-executable instructions stored in the one or more computer-readable storage media. The instructions generate a measurement count for each of a plurality of measurement values. The instructions generate a top average measurement count that is the average of the measurement counts that exceed a threshold count. The instructions calculate an expected offset of a probability distribution based on the top average measurement count and the measurement counts that exceed the threshold. The instructions subtract an estimated offset from the measurement count of each measurement value. In some embodiments, the measurement value is a measurement value of radiation and the temporary change is the result of a change in background radiation. In some embodiments, the instructions further detect when the measurement value includes a radiation source. In some embodiments, the detection is performed using a gross count detection algorithm. In some embodiments, the top average measurement count is based on a probability distribution.

[0113] Progressive Projection In some embodiments, a method is provided for estimating a background histogram of measurement counts of measurements collected to detect the presence of a source histogram of a source in measurement values, executed by one or more computer systems. The method accesses a spanning vector of a histogram spanning a linear combination of previous background histograms derived from previous measurement histograms. The method calculates a current background histogram of a current measurement histogram sequence. The method determines whether the current background histogram has a strong correlation or strong separability with the spanning vector. If the method determines that the current background histogram has a strong correlation or strong separability with the spanning vector, the method modifies the spanning vector based on the current background histogram. After modifying this spanning vector, the method adjusts the spanning vector to increase diversity. In some embodiments, the adjustment attempts to maximize an eigenvalues association with a correlation matrix formed by the spanning vector. In some embodiments, the method applies a detection algorithm that detects the presence of a source in target measurement values considering the adjusted spanning vector. In some embodiments, the source is a radiation source. In some embodiments, the previous measurement histogram is a predetermined histogram. In some embodiments, the previous measurement histogram includes a histogram collected during the same collection process prior to the collection of the sequence of current measurement histograms. In some embodiments, previous measurement histograms that are not collected during the same collection process.

[0114] In some embodiments, one or more computer systems are provided for estimating a background histogram of measurement counts of measurement values collected to detect the presence of a line source histogram of a line source in the measurement values. The one or more computer systems include one or more computer-readable storage media storing computer-executable instructions for controlling the one or more computer systems, and one or more processors executing the computer-executable instructions stored on the one or more computer-readable storage media. The instructions access a spanning vector of a histogram that spans a linear combination of previous background histograms derived from previous measurement histograms. The instructions calculate a current background histogram of a current measurement histogram sequence. The instructions determine whether the current background histogram has a strong correlation or strong separability with the spanning vector. If the instructions determine that the current background histogram has a strong correlation or strong separability with the spanning vector, the instructions modify the spanning vector based on the current background histogram. After modifying the spanning vector, the instructions adjust the spanning vector to increase diversity. In some embodiments, the instructions attempt to maximize the eigenvalue correlation with a correlation matrix formed by the spanning vector. In some embodiments, the instructions further apply a detection algorithm that detects the presence of a line source in a target measurement value that takes into account the adjusted spanning vector. In some embodiments, the line source is a radiation source. In some embodiments, the previous measurement histogram is a predetermined histogram. In some embodiments, the previous measurement histogram includes a histogram collected during the same collection process and before the collection of the current measurement histogram sequence. In some embodiments, the previous measurement histogram is not collected during the same collection process.

[0115] The subject matter has been described in terms of language specific to structural features and / or acts, but it should be understood that the subject matter defined in the appended claims is not necessarily limited to the specific features or acts described above. Rather, the specific features and acts described above are disclosed as examples of forms for carrying out the claims. Accordingly, the present invention is not limited by the appended claims.

Explanation of Signs

[0116] Estimate 900 Background RA 901 i = 0 902 ++ i > N 903 x[] = sort(x[i,t]) 904 Calculate the average of the top x counts 905 Off = Expected offset 906 Adjust x[i,t] by the offset

Claims

1. one or more computer systems for identifying a radiation source, receive radiation readings, generating a detection metric for one or more sources that indicates whether the source is present in the measurements; evaluating the detection metric to detect whether a source is present within the measurements; computer-executable instructions for a radiation source detector; When the presence of the source is detected in the measurements, generating, for one or more sources, a discrimination metric that indicates whether a source is present in the measurements; generating a null hypothesis metric indicating whether or not a source is present in the measurements; evaluating the one or more discrimination metrics and the null hypothesis metric to identify a source, if present, within the measurements; computer-executable instructions for a radiation source identifier; one or more computer-readable storage media storing the one or more processors that execute the computer-executable instructions stored on the one or more computer-readable storage media; 1. One or more computer systems comprising:

2. the radiation source detector generates the detection metric of the source based on a detected source signature of the source, and the radiation source detector generates the identification metric of the source based on a plurality of identification source signatures of the source.

10. One or more computer systems according to claim 1.

3. the source identification signatures representing the source with different shielding; 3. One or more computer systems according to claim 2.

4. the radiation source identifier identifies a radiation source class of the identified source; 10. One or more computer systems according to claim 1.

5. the measurements are calibrated to account for drift in the detector that collected the measurements; 10. One or more computer systems according to claim 1.

6. the radiation source identifier generates an aggregate metric for each source; 10. One or more computer systems according to claim 1.

7. the measurements within the window are aggregated, and radiation is detected and identified based on the aggregate measurements.

10. One or more computer systems according to claim 1.

8. 1. A method for detecting the presence of a radiation source from radiation measurements, implemented by a computer system, comprising: For one or more sources: For one or more detection algorithms: generating a metric indicative of the similarity between the current measurement and a source signature of the source using the detection algorithm that takes into account an estimate of background radiation generated based on previous measurements of the current measurement; generating an aggregate metric for the source from one or more metrics for the source; analyzing the aggregate metrics of one or more sources to determine whether a source is present within the current measurement; indicating that the presence of a source has been detected upon determining that the source is likely present within the current measurements; A method comprising:

9. generating an estimate of background radiation based on previous measurements excluding the previous measurement in which the presence of the source was detected; The method of claim 8.

10. the estimate of background radiation is a weighted average based on measured and prior estimates of background radiation; 10. The method of claim 9.

11. The source signature relates to the identity of the source and the shielding of said source. The method of claim 8.

12. For each detection algorithm, a metric is generated for each source signature associated with the source; The method of claim 11.

13. The measurements are represented by an energy range histogram containing the number of photons in each energy range. The method of claim 8.

14. generating a metric for each of a plurality of window measurements, each window including the current measurement and a number of adjacent previous measurements; The method of claim 8.

15. the detection algorithm is associated with coefficients, and further includes dynamically adjusting the coefficients based on an estimate of background radiation. The method of claim 8.

16. the analyzing the aggregation metrics includes applying a classifier trained using training data including feature vectors having characteristics including a detection algorithm, a window size, and an aggregation metric, and a label for each feature vector indicating whether a source is present; The method of claim 8.

17. the feature vector is generated from collected radiation measurements; 17. The method of claim 16.

18. the feature vector is generated from simulated radiation measurements; 17. The method of claim 16.

19. calibrating the current measurements to account for drift associated with the detector used to collect the measurements.

20. The method of claim 18.

20. 1. A method for identifying a radiation source from radiation measurements, implemented by a computer system, comprising: receiving a current measurement in which the presence of a source is detected based on the detected source signature; for each of a plurality of discrimination source signatures more comprehensive than the detected source signature, generating a metric indicative of the similarity between the current measurement and the source signature of the source using a discrimination algorithm that takes into account an estimate of background radiation generated in detecting the presence of the source; generating, for each source, a source metric based on one or more metrics generated using the source signature of the source; generating a null hypothesis metric indicating the similarity between the current measurement and the background radiation estimate, targeting the source and the background radiation estimate, and the source metric and the null hypothesis metric as target metrics; generating, for each target, a target probability based on the target metric of the target, which probability indicates the presence of the target in the current measurement; A method comprising:

21. generating, for each of a plurality of radiation source classes, a radiation source class probability that the current measurement represents a source of the radiation source class based on the target probability; 21. The method of claim 20.

22. generating the target probabilities takes into account the prior probability of each target; 21. The method of claim 20.

23. The discrimination algorithm is based on the orthonormal subspace projection matched filter algorithm, 21. The method of claim 20.

24. 1. A method for generating a metric relating to a radiation source in a radiation measurement, the method being performed by a computer system, the method comprising: accessing a source signature, an estimated background, and background basis vectors for the source; generating a projection vector based on the source signature, the estimated background, and the background basis vectors; accessing the measurements; generating a metric based on the source signature, the estimated background, and the background basis vectors; A method comprising:

25. the source signature is a histogram representing an energy range divided into energy bins, each energy bin having a value representing the number of photons emitted by the source over a time interval; 25. The method of claim 24.

26. generating the projection vector includes generating a weight matrix from the estimated background, wherein the variance of each energy bin is based on the expected background; 26. The method of claim 25.

27. generating the projection vector further comprises generating a source signature weighted projection and a background weighted projection; and setting the projection vector to the difference between the source signature weighted projection and the background weighted projection.

26. The method of claim 25.

28. the source signature represents the shielding of said source; 25. The method of claim 24.

29. generating aggregate measurements of different numbers of measurements; and for each aggregate measurement, generating a metric based on the aggregate measurement, the projection vector, and the expected variance of the aggregate measurement.

25. The method of claim 24.

30. further comprising indicating that the presence of the source has been detected when the metric meets a source presence threshold.

25. The method of claim 24.

31. further comprising the step of indicating that the presence of the source has been semi-conclusively detected when the metric meets a source presence threshold.

25. The method of claim 24.

32. generating the detection metric includes dividing the product of the projection vector and the measurement by the square root of the expected variance; 25. The method of claim 24.

33. 1. One or more computer systems for generating metrics relating to a radiation source in a radiation measurement, comprising: one or more computer-readable storage media storing computer-executable instructions; One or more processors that execute computer-executable instructions stored in the one or more computer-readable storage media, comprising, the computer-executable instructions cause the one or more computer systems to, access the source signature, estimated background, and background basis vectors of the source, generate a projection vector based on the source signature, estimated background, and the background basis vectors, access measurement values, generate the metric based on the measurement values, the projection vector, and the expected variance of the measurement values, such that it is controlled, characterized by one or more computer systems. **Claim 34** The source signature is a histogram representing an energy range divided into energy bins, and each energy bin has a value representing the number of photons emitted by the source over a certain time interval. The one or more computer systems according to claim 33. **Claim 35** The instructions for generating the projection vector generate a weight matrix from the estimated background, and the variance of each energy bin is based on the expected background. The one or more computer systems according to claim 34. **Claim 36** The instructions for generating the projection vector further generate a source-signature weighted projection and a background weighted projection, and set the projection vector to the difference between the source-signature weighted projection and the background weighted projection. The one or more computer systems according to claim 35. **Claim 37** The source signature represents the shielding of the source. The one or more computer systems according to claim 33. **Claim 38** The instructions further generate aggregated measurement values of different numbers of measurement values, and for each aggregated measurement value, generate a metric based on the aggregated measurement value, the projection vector, and the expected variance of the aggregated measurement value. The one or more computer systems according to claim 33. **Claim 39** The instructions further indicate that the presence of the source has been detected when the metric meets a source presence threshold. The one or more computer systems according to claim 10. **Claim 40** The instructions further indicate that the presence of the source has been semi-deterministically detected when the metric meets a source presence threshold. The one or more computer systems according to claim 33.

41. The instructions for generating the detection metric divide the product of the projection vector and the measurement value by the square root of the expected variance. One or more computer systems according to claim 33.

42. A method of generating a metric related to a radiation source in sequentially collected radiation measurement values, executed by a computer system, the method comprising: accessing a source signature, an estimated background, and a background basis vector of the source; generating the projection vector as represented by the following equation, where the projection vector is T, the source signature is S, the estimated background is B, and the background basis vector is W: T = S t W(I - B(B t WB) -1 B t W) accessing the measurement values; Set the metric as DM, the measured value as X, and the expected variance of the measured value as |X| 1 generating the metric as represented by the following equation; 【Number 1】 A method characterized by including the above steps.

43. The method according to claim 42, further comprising detecting the presence of the source based on the metric. The method according to claim 42.

44. The method according to claim 42, further comprising identifying the source based on the metric. The method according to claim 42.

45. The source signature is a histogram representing an energy range divided into energy bins, and each energy bin has a value representing the number of photons emitted by the source over a certain time interval. The method according to claim 42.

46. The source signature represents the shielding of the source. The method according to claim 42.

47. The method according to claim 42, further comprising generating an aggregated measurement value of different numbers of measurement values, and for each aggregated measurement value, generating a metric based on the aggregated measurement value, the projection vector, and the expected variance of the aggregated measurement value. The method according to claim 42.

Images