Non-negative Matrix Factorization of Gamma-Ray Spectra for Background Modeling, Detection, and Source Identification

Bilton, K. J.; Joshi, T. H. Y.; Bandstra, Mark S.; Curtis, Joseph C.; Quiter, Brian J.; Cooper, R. J.; Vetter, K.

doi:10.1109/tns.2019.2907267

Cited by 21 publications

(24 citation statements)

References 34 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In addition, by finding background components with physical origins, NMF could be leveraged in new algorithms for anomaly detection and background estimation. NMF has already been shown to be competitive with PCA-based methods for spectral anomaly detection, which may be due to its accurate treatment of Poisson statistics and ability to be consistent with physics [27]. The different temporal variability of the components could be exploited by Kalman filters or low-pass filters to find anomalous behavior.…”

Section: Discussionmentioning

confidence: 99%

“…NMF has been presented here retrospectively like NASVD, but both NMF and NASVD components can be fit to future measurements under the assumption that the components derived from the training set are also effective at reducing the dimensionality of future data. Using a trained NMF model to decompose spectra has been shown to be an effective tool for spectral anomaly detection in a mobile detection system [27].…”

Section: Discussionmentioning

confidence: 99%

“…Recently Non-negative Matrix Factorization (NMF) [25], [26] was introduced as a technique for performing dimensionality reduction of gamma-ray background data [27] and for gamma-ray source identification [28]. NMF has already been used for spectral deconvolution in other fields of spectroscopy (e.g., [29], [30], [31]).…”

Section: E Non-negative Matrix Factorizationmentioning

confidence: 99%

“…Since both methods exactly reconstruct each spectrum's gross counts, it is also important to compare how well they fit the shape of the spectrum. For this purpose we introduce the Spectral Anomaly Detection (SAD) metric, the L2 norm between the ith measured spectrum and the ith reconstructed spectrum [41], [27]:…”

Section: E Comparison To Nasvdmentioning

confidence: 99%

See 3 more Smart Citations

Modeling Aerial Gamma-Ray Backgrounds Using Non-negative Matrix Factorization

Bandstra

Joshi

Bilton

et al. 2020

IEEE Trans. Nucl. Sci.

Self Cite

View full text Add to dashboard Cite

Airborne gamma-ray surveys are useful for many applications, ranging from geology and mining to public health and nuclear security. In all these contexts, the ability to decompose a measured spectrum into a linear combination of background source terms can provide useful insights into the data and lead to improvements over techniques that use spectral energy windows. Multiple methods for the linear decomposition of spectra exist but are subject to various drawbacks, such as allowing negative photon fluxes or requiring detailed Monte Carlo modeling. We propose using Non-negative Matrix Factorization (NMF) as a data-driven approach to spectral decomposition. Using aerial surveys that include flights over water, we demonstrate that the mathematical approach of NMF finds physically relevant structure in aerial gamma-ray background, namely that measured spectra can be expressed as the sum of nearby terrestrial emission, distant terrestrial emission, and radon and cosmic emission. These NMF background components are compared to the background components obtained using Noise-Adjusted Singular Value Decomposition (NASVD), which contain negative photon fluxes and thus do not represent emission spectra in as straightforward a way. Finally, we comment on potential areas of research that are enabled by NMF decompositions, such as new approaches to spectral anomaly detection and data fusion.

show abstract

Section: Discussionmentioning

confidence: 99%

Section: Discussionmentioning

confidence: 99%

Section: E Non-negative Matrix Factorizationmentioning

confidence: 99%

Section: E Comparison To Nasvdmentioning

confidence: 99%

See 2 more Smart Citations

Modeling Aerial Gamma-Ray Backgrounds Using Non-negative Matrix Factorization

Bandstra

Joshi

Bilton

et al. 2020

IEEE Trans. Nucl. Sci.

Self Cite

View full text Add to dashboard Cite

show abstract

“…Previous studies (e.g., refs. [4][5][6]) have demonstrated benefits to using full-spectrum techniques, namely, increased detection sensitivity over methods that only consider regions of the gamma-ray spectrum. Specifically, methods that incorporate information from the entire gamma-ray spectrum are capable of building more accurate models of gamma-ray sources and background.…”

Section: Introductionmentioning

confidence: 99%

Neural Network Approaches for Mobile Spectroscopic Gamma-Ray Source Detection

Bilton

Joshi

Bandstra

et al. 2021

JNE

Self Cite

View full text Add to dashboard Cite

Artificial neural networks (ANNs) for performing spectroscopic gamma-ray source identification have been previously introduced, primarily for applications in controlled laboratory settings. To understand the utility of these methods in scenarios and environments more relevant to nuclear safety and security, this work examines the use of ANNs for mobile detection, which involves highly variable gamma-ray background, low signal-to-noise ratio measurements, and low false alarm rates. Simulated data from a 2” × 4” × 16” NaI(Tl) detector are used in this work for demonstrating these concepts, and the minimum detectable activity (MDA) is used as a performance metric in assessing model performance.In addition to examining simultaneous detection and identification, binary spectral anomaly detection using autoencoders is introduced in this work, and benchmarked using detection methods based on Non-negative Matrix Factorization (NMF) and Principal Component Analysis (PCA). On average, the autoencoder provides a 12% and 23% improvement over NMF- and PCA-based detection methods, respectively. Additionally, source identification using ANNs is extended to leverage temporal dynamics by means of recurrent neural networks, and these time-dependent models outperform their time-independent counterparts by 17% for the analysis examined here. The paper concludes with a discussion on tradeoffs between the ANN-based approaches and the benchmark methods examined here.

show abstract