arXiv subjects

Peter Lalor

Publications and source records attributed to Peter Lalor.

7 recordsLinked to original sources

arXiv2026

Computer vision-based neural networks for radioisotope identification in urban environments

Algorithm development for radioisotope identification in mobile urban search scenarios face significant challenges from non-uniform backgrounds, momentary source encounters, and severe class imbalance between rare threat signatures and background measurements. We present a machine learning-based approach to this problem that converts list-mode gamma-ray data into two-dimensional waterfall spectrograms and applies computer vision architectures to the resulting images. Rather than treating waterfalls as conventional images, we employ a representation where consecutive time spectra can form input channels, similar to RGB channels in color images. This representation encodes both spectral and temporal information, enabling neural networks to more effectively learn patterns that distinguish source signatures from background fluctuations. We evaluate three architectures, a multilayer perceptron (MLP), convolutional neural network (CNN), and vision transformer (ViT), on the Radiological Anomaly Detection and Identification (RADAI) benchmark dataset. At a false positive rate of less than one false alarm per hour, our CNN outperforms the previous-best non-negative matrix factorization (NMF) method across all global metrics, achieving true detection, classification, and identification rates of 0.4334, 0.3965, and 0.2950 respectively, compared to 0.4151, 0.3611, and 0.2625 for NMF. At lower false positive rate constraints, the neural network approaches show comparable but ultimately lower performance than NMF, indicating opportunities for further research.

physics.ins-det↗

arXiv2026

Unsupervised domain adaptation for radioisotope identification in gamma spectroscopy

Training machine learning models for radioisotope identification using gamma spectroscopy remains an elusive challenge for many practical applications, largely stemming from the difficulty of acquiring and labeling large, diverse experimental datasets. Simulations can mitigate this challenge, but the accuracy of models trained on simulated data can deteriorate substantially when deployed to an out-of-distribution operational environment. In this study, we demonstrate that unsupervised domain adaptation (UDA) can improve the ability of a model trained on synthetic data to generalize to a new testing domain, provided unlabeled data from the target domain is available. Conventional supervised techniques are unable to utilize this data because the absence of isotope labels precludes defining a supervised classification loss. We compare a range of different UDA techniques, finding that feature alignment strategies, particularly via maximum mean discrepancy (MMD) minimization or domain-adversarial training, yield the most consistent improvement to testing scores. For instance, using a custom transformer-based neural network, we achieve a testing accuracy of $0.904 \pm 0.022$ on an experimental LaBr$_3$ test set after performing unsupervised feature alignment via MMD minimization, compared to $0.754 \pm 0.014$ before alignment. Overall, our results highlight the potential of using UDA to adapt a radioisotope classifier trained on synthetic data for real-world deployment.

cs.LG↗

arXiv2024

Atomic number estimation of dual energy cargo radiographs: initial experimental results using a semiempirical transparency model

To combat the risk of nuclear smuggling, radiography systems are deployed at ports to scan cargo containers for concealed illicit materials. Dual energy radiography systems enable a rough elemental analysis of cargo containers due to the Z-dependence of photon attenuation, allowing for improved material detection. This work presents our initial experimental findings using a novel approach to predict the atomic number of dual energy images of a loaded cargo container. We consider measurements taken by a Rapiscan Sentry Portal scanner, which is a dual energy betatron-based system used to inspect cargo containers and large vehicles. We demonstrate the ability to accurately fit our semiempirical transparency model to a set of calibration measurements. We then use the calibrated model to reconstruct the atomic number of an unknown material by minimizing the chi-squared error between the measured pixel values and the model predictions. We apply this methodology to two experimental scans of a loaded cargo container. First, we incorporate an image segmentation routine to group clusters of pixels into larger, roughly homogeneous objects. By considering groups of pixels, the subsequent atomic number reconstruction step produces a lower noise result. We demonstrate the ability to accurately reconstruct the atomic number of blocks of steel and high density polyethylene. Furthermore, we are able to identify the presence of two high-Z lead test objects, even when embedded within lower-Z organic shielding. These results demonstrate the significant potential of this methodology to yield improved performance characteristics over existing methods when applied to commercial dual energy systems.

physics.ins-det↗

arXiv2024

Sim-to-real supervised domain adaptation for radioisotope identification

Machine learning has the potential to improve the speed and reliability of radioisotope identification using gamma spectroscopy. However, meticulously labeling an experimental dataset for training is often prohibitively expensive, while training models purely on synthetic data is risky due to the domain gap between simulated and experimental measurements. In this research, we demonstrate that supervised domain adaptation can substantially improve the performance of radioisotope identification models by transferring knowledge between synthetic and experimental data domains. We consider two domain adaptation scenarios: (1) a simulation-to-simulation adaptation, where we perform multi-label proportion estimation using simulated high-purity germanium detectors, and (2) a simulation-to-experimental adaptation, where we perform multi-class, single-label classification using measured spectra from handheld lanthanum bromide (LaBr) and sodium iodide (NaI) detectors. We begin by pretraining a spectral classifier on synthetic data using a custom transformer-based neural network. After subsequent fine-tuning on just 64 labeled experimental spectra, we achieve a test accuracy of 96% in the sim-to-real scenario with a LaBr detector, far surpassing a synthetic-only baseline model (75%) and a model trained from scratch (80%) on the same 64 spectra. Furthermore, we demonstrate that domain-adapted models learn more human-interpretable features than experiment-only baseline models. Overall, our results highlight the potential for supervised domain adaptation techniques to bridge the sim-to-real gap in radioisotope identification, enabling the development of accurate and explainable classifiers even in real-world scenarios where access to experimental data is limited.

cs.LG↗

arXiv2024

Atomic number estimation of dual energy cargo radiographs using a semiempirical transparency model

Dual energy cargo inspection systems are sensitive to both the area density and the atomic number of an imaged container due to the $Z$ dependence of photon attenuation. The ability to identify cargo contents by their atomic number enables improved detection capabilities of illicit materials. This work introduces a novel method for atomic number reconstruction by minimizing the chi-squared error between measured transparency values and a semiempirical transparency model. This method is tested using two Geant4 Monte Carlo simulated radiographic phantoms, demonstrating the ability to obtain accurate material predictions on noisy input images, even in the presence of shielding. Furthermore, we provide a simple procedure for porting this method to a commercial system, requiring an approximate model of the scanner's beam spectra and detector response, along with only three calibration measurements.

eess.IV↗

arXiv2024

Fundamental limitations of dual energy X-ray scanners for cargo content atomic number discrimination

To combat the risk of nuclear smuggling, radiography systems are deployed at ports to scan cargo containers for concealed illicit materials. Dual energy radiography systems enable a rough elemental analysis of cargo containers due to the $Z$-dependence of photon attenuation, allowing for improved material detection. This work studies the capabilities for atomic number discrimination using dual energy MeV systems by considering dual energy $\{6, 4\}$ MeV, $\{10, 6\}$ MeV, and $\{10, 4\}$ MeV bremsstrahlung beams. Results of this analysis show that two different pure materials can sometimes produce identical transparency measurements, leading to a fundamental ambiguity when differentiating between materials of different atomic numbers. Previous literature has observed this property, but the extent of the limitation is poorly understood and the cause of the degeneracy is generally inadequately explained. This non-uniqueness property stems from competition between photoelectric absorption and pair production and is present even in systems with perfect resolution and zero statistical noise. These findings are validated through Monte Carlo transparency simulations. Results of this study show that currently deployed commercial radiographic systems are fundamentally incapable of distinguishing between high-$Z$ nuclear materials and miscellaneous mid-$Z$ cargo contents.

nucl-ex↗

arXiv2023

A Semiempirical Transparency Model for Dual Energy Cargo Radiography Applications

Cargo containers passing through ports are scanned by non-intrusive inspection systems to search for concealed illicit materials. By using two photon beams with different energy spectra, dual energy inspection systems are sensitive to both the area density and the atomic number of cargo contents. Most literature on the subject assumes a simple exponential attenuation model for photon intensity in which only free streaming photons are detected. However, this approximation neglects second order effects such as scattering, leading to a biased model and thus incorrect material predictions. This work studies the accuracy of the free streaming model by comparing it to simulation outputs, finding that the model shows poor atomic number reconstruction accuracy at high-$Z$ and suffers significantly if the source energy spectra and detector response function are not known exactly. To address these challenges, this work introduces a semiempirical transparency model which modifies the free streaming model by rescaling different components of the mass attenuation coefficient, allowing the model to capture secondary effects ignored by the free streaming model. The semiempirical model displays improvement agreement with simulated results at high-$Z$ and shows excellent extrapolation to materials and thicknesses which were not included during the calibration step. Furthermore, this work demonstrates that the semiempirical model yields accurate atomic number predictions even when the source spectra and detector response are not known exactly. Using the semiempirical model, manufacturers can perform a simple calibration to enable more precise $Z$ reconstruction capabilities, which has the potential to significantly improve the performance of existing radiographic systems.

physics.ins-det↗