Physical Sciences Inc. VG14-048 Utilization of advanced clutter suppression algorithms for improved standoff detection and identification of radionuclide threats Bogdan R. Cosofret, Kirill Shokhirev and Phil Mulhall, Andover MA cosofret@psicorp.com David Payne and Bernard Harris Raytheon Integrated Defense Systems, Tewksbury MA 2014 SPIE DSS 6-8 May 2014 ACKNOWLEDGEMENT: This work has been supported by the US Department of Homeland Security, Domestic Nuclear Detection Office, under competitively awarded contract/iaa HSHQDC - 10 - C - 00171 and HSHQDC-11-C-00117. This support does not constitute an express or implied endorsement on the part of the Government. 20 New England Business Center Andover, MA 01810
Agenda VG14-048 -1 Motivation Overview of PCS Algorithm and Integration with Detector Systems Field Measurements and Results Standoff Radiation Detection System Radiation Portal Handheld isotope identifier Conclusions
Counts Motivation and Key Challenges General Objective: Increase capability of detection systems in low SNR regimes encountered in portal and urban detection applications 30 25 20 6 Background VG14-048 -2 Background + 1 mci 137 Cs at 40 m 1 mci 137 Cs at 40 m 4 Technical Approach: Develop advanced algorithms that employ accurate probabilistic modeling of radiological backgrounds for noise and clutter suppression 15 10 5 2 0 400 500 600 700 800 900 Key Challenges: Low SNR regimes encountered at short integration times Poisson noise and clutter mask weak signals, while significant spectral variation is present in complex urban environments Handheld isotope identifiers are characterized by both low SNR and low count regimes due use of small scintillator crystals 0 0 500 1000 1500 2000 2500 3000 E, kev Typical count levels observed when processing 1 sec integration time gamma spectra acquired by a 2 x4 x16 NaI detector. Weak threat signals are masked by noise and background variability. New advanced algorithms aimed at signal processing in low SNR regimes
VG14-048 -3 Poisson/Clutter Split Model (PCS): Conceptual Approach
GLRT Framework GLRT Methodology for Threat Detection Background Estimation Data Set: Background-only Spectra VG14-048 -4 Likelihood of H 0 Statistical Model (no threat) Likelihood ratio CFAR threshold Detection and ID Algorithm Likelihood of H 1 (threat present) Test Spectra: Background + (Threat Signal) PCS algorithm is based on the GLRT framework, where the background is estimated within the Poisson/Clutter model
PCS Model: Separation of Poisson and Clutter Noise The variability among background radiological spectra can be attributed to two mechanisms: Background clutter, i.e. the changes of the energy-dependent count rate due to variations in isotopic composition depending on particular environments, weather conditions, etc. The random process of radioactive decay, described by Poisson statistics VG14-048 -5 The key innovations behind the Poisson Clutter Split (PCS) algorithm are: The use of a novel probabilistic representation of radiological backgrounds Accurate modeling of gamma counts based on Poisson statistics The use of the Generalized Likelihood Ratio Test (GLRT) to simultaneously perform detection and identification of sources. PCS algorithm s non-linear probabilistic model provides a better characterization of the radiological environment than traditional linear methods
PCS Model: Separation of Poisson and Clutter Noise Observed counts x, obey Poisson statistics corresponding to the local background rate, b, and the integration time, Δt. ~ P( b t) PCS calculates the mean rate as a function of energy and the dominant modes of spectral variations, z k, observed across sampled environments: The underlying rate, b, can be accurately parameterized with a limited number of coefficients which determine the spectral variability of the rate: Clutter is reflected in the varying parameters, w z k capture the spectral features of the environment x b b b({ z 1,.., z }, w) w ( w 1,.., w K ) 1 x D vector, D is the number of channels In the presence of a radioactive source, the background rate, b, is elevated by an energy-dependent contribution from the source: b s b( w) s ( w, ) k p b pb ( w) f ( w) x Test f(w) is the probability distribution of the clutter parameters ~ P( t) VG14-048 -6
PCS-based Background Estimation and Threat Detection and Identification Estimate the background (train) in single or multiple environments VG14-048 -7 N background spectra, { x i }, i = 1,..,N, Estimate the modes of spectral variability and find parameter combinations, w i for each spectrum Fit a probabilistic model to the distribution of w s z k, k 1,.., K distribution of w s Detection and ID: analyze new spectrum x Given new spectrum x, maximize likelihood under two hypotheses: H 0 : x is generated from a rate consistent with the background model H 1 : x is generated from the rate consistent with the estimated background spectrally perturbed by a threat isotope Alarm and ID if likelihood ratio exceeds threshold
PCS Integration for Real-Time Processing ASP GUI SORDS GUI VG14-048 -8 PCS is implemented in C ++ as a native executable code for real-time processing. It is compiled for both Windows OS and Linux. The code makes use of multicore architectures and parallel processing using the Intel-based MKL 8.0.1 library PCS processes a given spectrum in ~ 20 msec against a library of 16 isotopes on an Intel Core i7 2.80GHz, 4GB RAM (Win7) Current efforts are directed at the PCS implementation on a dual core ARM Cortex A-9 architecture.
Examples of Performance Enhancements Using PCS Integrated with Existing Detector Systems Advanced Spectroscopic Portal (ASP) Using 1/3 of available detectors, PCS demonstrated P d,id > 90%, CFAR = 1 in 1000 occupancies against cargo moving at 30 mph and carrying weak and shielded sources (< 8 µci) Capability represents a factor of 6 improvement in system throughput compared to what s currently possible, while providing operation at lower cost VG14-048 -9 Handheld Isotope Identifier (e.g. 1.5 x1.5 LaBr 3 ) Demonstrated increased detection sensitivity while offering isotope identification in real time (1 Hz update rate) Demonstrated significant reduction in false warning rates during operation in cluttered containerized environments Standoff Radiation Detection System (SORDS) Demonstrated reduction in operational false alarms and a factor of 2-3 increase in detection sensitivity over capability offered by standard processing algorithms Current efforts focused on extracting better PCS performance through fusion of information generated by orthogonal sensing technologies
VG14-048 -10 Standoff Radiation Detection System With Embedded PCS Capability
Estimation of SORDS Background Model and Isotope-Specific CFAR Threshold Determination VG14-048 -11 Raytheon s Trimodal Imager (TMI) was used for background acquisition over a long duration sampling of various environments Data collection during extensive on-the-move sampling of DC Philadelphia Boston Over 20 hours of data acquired in urban, rural, and highway type environments Background estimation and CFAR threshold determination: A total of 35,000 two second integration time spectra were generated using the NaI back and front arrays Random subset of the background set was used to develop the PCS background model (6,000 spectra) Remaining spectra (29,000 spectra ~ 16 hours) were analyzed via PCS to determine responses as a function of time and to determine isotope-specific CFAR thresholds 28 isotopes in the library CFAR threshold: 1 in 8 hours
PCS signal PCS signal Distance to source Distance to source 100 80 60 40 20 80 60 40 20 0-20 PCS Detection/ID Results: Offline Processing Savannah River, 180 and 90 μci 131 I sources Response for I-131. Source: 180 Ci 131 I 100 200 300 400 500 600 700 180 Ci 90 Ci 100 200 300 400 500 600 700 Combined test time, s CFAR Threshold = 1 in 8 hours 100 80 60 40 20 30 20 10 0-10 Detection No alarm VG14-048 -12 Response for I-131. Source: 90 Ci 131 I 50 100 150 200 250 300 50 100 150 200 250 300 Combined test time, s CFAR Threshold = 1 in 8 hours 180 μci source detected in 10/10 passes (source detected from ~ 40 m away) 90 μci source detected in 5/5 passes (source detected from ~ 30 m away) 28 isotopes extensively analyzed to date for false alarm and cross-talk Demonstrated results consistent with SORDS operational requirements with low probability of isotope mis-identification
PCS Integration with TMI VG14-048 -13 Combined front + back array (57 out of 65 detectors) to generate 2 sec spectra PCS C-based code processes an incoming TMI spectrum against 28 isotopes in ~ 100 msec We implemented the reporting of a Confidence Level (CL) corresponding to each PCS alarm All PCS responses are displayed on the primary system GUI
TMI-PCS Performance Evaluation: Outdoor Environment Static P d,id VG14-048 -14 Detection sensitivity as a function of range was evaluated using Cs-137 (~8 µci) and Ba-133 (7 µci) sources Example shown: 7 µci Ba-133 at 5, 8 and 10 m standoff range CFAR operating conditions: 1 in 8 hr Results based on static acquisition: P d,id = 100% at 5 m, 90% at 8 m, and 45% at 10 m No false alarms observed in sampling RTN Tewksbury environment No misidentifications reported against the other 27 isotopes Similar results obtained against 8 µci Cs-137
TMI-PCS Performance Evaluation: Rain Conditions Static P d,id VG14-048 -15 Detection statistics against 8 µci Cs-137 in the rain are similar to the performance demonstrated in dry conditions PCS response against 226 Ra was shown to closely track the intensity of the rain. The elevated 226 Ra responses in rainy conditions are consistent with previously reported washout of atmospheric Radon ( 222 Rn) Results indicate the ability of PCS to track atmospheric effects and retain good performance under elevated background levels
SORDS made several passes at 15 mph with closest approach distance of 8 m (front side). Back side standoff range ~ 12 m. TMI-PCS Performance Evaluation: On-the-Move Acquisition (7 8 µci sources) VG14-048 -16 Results: Detected and correctly identified both sources in 8/8 passes from 8 m closest approach distance No misidentifications against all other 27 isotopes One false alarm (Cd-109) reported in 5 hours of sampling cities of Lowell & Lawrence Demonstrated performance consistent with previous results 7 μci, moving at 10-15 mph at 8 m is equivalent to 98 μci, moving at 38-56 mph at 30 m
VG14-048 -17 Spectroscopic Portal System With Embedded PCS Capability* [1] Cosofret, B.R., Shokhirev, K.N., Mulhall, P.A., Payne, D., Harris, B., Arsenault, E., and Moro, R., Utilization of * advanced clutter suppression algorithms for improved spectroscopic portal capability against radionuclide threats, HST 2013, IEEE International Conference of Technologies for Homeland Security, Waltham, MA, 12 14 November 2013.
ASP-PCS Objectives Improve overall performance of current ASP systems using advanced algorithms for noise and clutter suppression VG14-048 -18 Demonstrate the achievement of ASP Key Performance Parameters under improved and cost effective operational capability: Utilization of only 4 out of 12 NaI detectors currently integrated with RTN s ASP units Vehicle speeds through the portal in excess of 20 mph (> 6x improvement over current throughput) ASP Key Performance Parameters targeted: False alarm rate of 1 in 1000 occupancies P d,id > 90% for weak activity sources (< 10 µci)
Truck with Shielded Cs-137 (attenuated 8 µci) + Salt Vehicle Speeds: 20 and 30 mph VG14-048 -19 20 mph 30 mph PCS results with 4/12 detectors: P d,id = 93% at CFAR of 1 in 1000 occ. against shielded Cs-137 at 20-30 mph (detected in 14 out of 15 runs) No false alarms or mis-identifications were reported Note: Standard ASP software with all 12/12 ASP detectors yielded P d,id = 0%
VG14-048 -20 Handheld Isotope Identifier With Embedded PCS Capability
PCS Integration with RadSeeker VG14-048 -21 High rate isotope ID is critical to end users to be able to quickly differentiate real threats from alarms introduced by nuisance isotopes or elevated NORM levels RIID devices work primarily as gross-counting devices until an elevated count level is encountered followed by acquisition of 30 sec integration time spectra to output an accurate isotope identification PCS was demonstrated to operate at lower false alarm rates with improved detection sensitivity, while providing simultaneous isotope identification at 1 Hz update rate
PCS-RadSeeker Capability Tested In Cargo Environment VG14-048 -22 Two acquisition modes: Search: 3 ft/sec dynamic measurements Interrogation: 2-4 min static collection at pre-defined locations
PCS Responses Against Cargo Background and Constant False Warning Rate Assessment Conducted ~ 4 hours of continuous sampling (0.5 sec IT spectra) of containerized environment VG14-048 -23 Observed significant variability in gross count as a function of location Single false alarm observed during the duration of background collection (Ba-133) consistent with isotope specific 1 in 4 hr CFWR thresholds
PCS-RadSeeker Detection/ID Sensitivity PCS-RadSeeker sensitivity was assessed by observing check sources as a function of range VG14-048 -24 Detection/ID statistics were generated based on processing of 1-10 second integration time spectra: 1 sec IT: P d,id ~ 90% against the 8.8 µci 133 Ba was demonstrated from 1 m standoff 10 sec IT: P d,id ~ 82% against a 7.8 µci 57 Co source from 2 meters standoff The elevated and cluttered radiological background encountered in the containerized environment reduces the detection sensitivity of the system when compared to its operation in an indoor area
On-the-Move, Scenario-based Tests 1 Hz Isotope ID Reporting Rate VG14-048 -25 PCS detection and ID output rate: ~ 1 Hz PCS correctly detected and identified all sources (4-8 µci) in all passes at ~ 3 ft/sec No mis-identifications or false alarms reported against any of the other isotopes in the library
VG14-048 -26 Conclusions
Conclusions VG14-048 -27 A novel statistical approach for radiological background estimation was developed that improves the performance of medium resolution detectors in low SNR regimes The PCS algorithm is flexible and can be integrated with a range of detector systems The algorithm has demonstrated high probability of source detection and discrimination with low integration time spectra, while significantly reducing operational false alarms We are currently integrating PCS with small, low power processors such as those found on smartphone devices to allow for a flexible, high performance platform that can interface directly existing instruments