Title: Detecting, identifying, and localizing radiological material in urban environments using scan statistics

Abstract: A method is proposed, based on scan statistics, to detect, identify, and localize illicit radiological material using mobile sensors in an urban environment. Our method handles varying levels of background radiation that change according to an (unknown) environment. Our method can accurately determine if a source is present along a street segment as well as identify which of six possible sources generated the radiation. Our method can also localize the source, when detected, to within a few seconds. We have presented our results across a range of decision thresholds allowing stakeholders to evaluate the performance at different false alarm rates. Due to the simplicity of our approach, our models can be trained in a few minutes with very little training data and holds the potential to score a run in real-time. Our method was one of the top performing submissions in the 'Detecting Radiological Threats in Urban Areas' competition.