Abstract:This paper presents the development and application of an AI-based method for particle track identification using scintillating fibres read out with imaging sensors. We propose a variational autoencoder (VAE) to efficiently filter and identify frames containing signal from the substantial data generated by SPAD array sensors. Our VAE model, trained on purely background frames, demonstrated a high capability to distinguish frames containing particle tracks from background noise. The performance of the VAE-based anomaly detection was validated with experimental data, demonstrating the method's ability to efficiently identify relevant events with rapid processing time, suggesting a solid prospect for deployment as a fast inference tool on hardware for real-time anomaly detection. This work highlights the potential of combining advanced sensor technology with machine learning techniques to enhance particle detection and tracking.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is how to efficiently identify particle trajectories when using an imaging sensor readout in a scintillating fiber detector. Specifically, the paper proposes a method based on the variational auto - encoder (VAE) for quickly screening and identifying frames containing signals from a large amount of background noise data. This is especially important in particle physics experiments, because these experiments usually generate a large amount of data, and distinguishing signal events from background noise is a challenge. ### Main problems: 1. **Large amount of data**: The scintillating fiber detector will generate a large amount of data during the readout process, and an efficient processing method is required to reduce the data processing time. 2. **Distinguishing between signal and background noise**: In the case of high background noise, how to accurately identify the signal frames containing particle trajectories. 3. **Real - time processing ability**: Develop a method that can implement real - time anomaly detection on hardware to improve the efficiency and response speed of the experiment. ### Solutions: - **Variational auto - encoder (VAE)**: The paper proposes an anomaly detection method based on VAE. By learning the typical patterns of background noise, VAE can effectively identify frames containing signals. - **Training method**: The VAE model is trained only with pure background frames, so it can distinguish between background noise and signal events. - **Performance verification**: The effectiveness of the VAE method is verified through experimental data, showing its potential in quickly processing and accurately identifying signal frames. ### Key technical points: - **VAE architecture**: It includes an encoder with six convolutional layers and a decoder with six transposed convolutional layers, as well as a fully - connected layer for generating a 32 - dimensional latent space. - **Loss function**: Combines the binary cross - entropy (BCE) loss and the Kullback - Leibler divergence (KLD) loss to optimize the performance of the model. - **Dimensionality reduction techniques**: Use methods such as UMAP and t - SNE to project the high - dimensional latent space onto a low - dimensional space for easy visualization and analysis. ### Conclusion: The paper shows the application potential of the VAE - based anomaly detection method in scintillating fiber detectors, especially in efficiently identifying signal frames and achieving real - time processing. Future work will explore the integration of this method on hardware to further improve its real - time processing ability.

AI-based particle track identification in scintillating fibres read out with imaging sensors

Particle Identification in Particle Tracking Velocimetry Using Two-Stage Neural Networks

Demonstration of particle tracking with scintillating fibres read out by a SPAD array sensor and application as a neutrino active target

Fast Particle-based Anomaly Detection Algorithm with Variational Autoencoder

Event-Based Processing of Single Photon Avalanche Diode Sensors

Artificial intelligence for improved fitting of trajectories of elementary particles in inhomogeneous dense materials immersed in a magnetic field

Smartpixels: Towards on-sensor inference of charged particle track parameters and uncertainties

Scintillation event imaging with a single photon avalanche diode camera

Development of scintillating fiber detector technology for high rate particle tracking

Autoencoders for Real-Time SUEP Detection

Convolutional neural networks automate detection for tracking of submicron scale particles in 2D and 3D

High-speed particle detection and tracking in microfluidic devices using event-based sensing

A first application of machine and deep learning for background rejection in the ALPS II TES detector

Resilient VAE: Unsupervised Anomaly Detection at the SLAC Linac Coherent Light Source

Deep Learning-based Alpha Particles Spectroscopy with Solid-state Nuclear Track Detector CR-39

Zero-bias new particle searches using autoencoders in UPCs and diffractive events

Deep Learning-Based Spatiotemporal Multi-Event Reconstruction for Delay Line Detectors

A 1 mm Scintillating Fibre Tracker Readout by a Multi-anode Photomultiplier

Tracking hidden objects with a single-photon camera

Event-based Moving Object Detection and Tracking