Robust Localization of an Arbitrary Distribution of Radioactive Sources for Aerial Inspection

TL;DR

This paper introduces a robust, scalable algorithm for localizing complex distributions of radioactive sources in large and high-dimensional spaces, significantly improving accuracy over previous methods.

Contribution

The paper presents a novel multi-layer sequential Monte Carlo algorithm combined with clustering for accurate localization of arbitrary radiation source distributions.

Findings

Achieves F1-scores greater than 0.95 in source localization.

Scales effectively to large regions and higher-dimensional spaces.

Handles complex source distributions with high accuracy.

Abstract

Radiation source detection has seen various applications in the past decade, ranging from the detection of dirty bombs in public places to scanning critical nuclear facilities for leakage or flaws, and in the autonomous inspection of nuclear sites. Despite the success in detecting single point sources or a small number of spatially separated point sources, most of the existing algorithms fail to localize sources in complex scenarios with a large number of point sources or non-trivial distributions & bulk sources. Even in simpler environments, most existing algorithms are not scalable to larger regions and/or higher dimensional spaces. For effective autonomous inspection, we not only need to estimate the positions of the sources, but also the number, distribution, and intensities of each of them. In this paper, we present a novel algorithm for the robust localization of an arbitrary distribution of radiation sources using multi-layer sequential Monte Carlo methods coupled with suitable clustering algorithms. We achieve near-perfect accuracy, in terms of F1-scores (> 0.95), while allowing the algorithm to scale, both to large regions in space and to higher dimensional spaces (5 tested).