(53d) Machine Learning Detection of Radioactive Sources Using Maximum Detection Distance for Environmental Remediation

The role of radioactivity detection is of paramount importance in upholding public safety and effectively managing environmental risks. This study centers on the development of a methodology, harnessing machine learning techniques, to discern radioactivity in relation to background radiation levels. The data utilized is derived from field survey points collected through the deployment of a 4-wheel drive all-terrain vehicle (ATV) equipped with a Global Positioning System (GPS) and thallium-activated sodium iodide, NaI(Tl), gamma scintillation detector, within a 260-acre region of interest. This research's implications extend to critical applications, notably in the realm of emergency preparedness and mission planning, especially pertinent in scenarios involving the remediation of contaminated sites. The core innovation lies in the adaptation of a Maximum Detection Distances (MDD) calculation algorithm, harmonized with the localization function of a mobile gamma-ray detection system. When exploring specific routes, such as former roadways and earthen berms, the ability to detect a radioactive source is constrained by the distance from the path – therefore, knowing the MDD is significant to declare an area “clean”.

Distinct from the prevalent reliance on simulation-based data in many existing studies, this research adheres to the utilization of empirical data gleaned from field operations. This approach fosters a close alignment between the research and practical real-world applications. Search teams involved in emergency remedial clean-up can optimize survey routes to ensure that environmental monitoring areas are adequately covered where concerns of potential contamination may be present. The algorithm yields agreement with field measurements, considering the features such as gamma readings and coordinates, as well as showing scalability for very large regions, when predicting the presence of radioactive sources. Ultimately, the outcome of this study holds the potential to bridge the chasm between theoretical advancements and their practical implementation, rending radiological monitoring technology more accessible and cost-effective for an expanded array of applications.