A Detection System for Dirty Bombs in Open Environments

Abstract

In this paper, a system is presented, which addresses the first question of source assignment in an open
environment of unordered person flows using sophisticated data fusion algorithms for a set of gamma ray
detectors and 3D time-of-flight cameras for tracking. The basis of such a system are precise, reliable, and real
time updated tracks of all persons within the region of interest. In our system, a set of 3D time-of-flight cameras
is used to extract position information of persons in a room.
In parallel, a set of gamma ray sensors is used to measure the current intensity spatially distributed in the room.
Nuclear decay is a random process, where the statistics can be described with a Poisson distribution with high
variance. Thus, it directly follows that the gamma detection sensors have a poor spatial resolution for allocation
of a source. Therefore, the inference of which person is a carrier of radio active material must be based on
multiple sensors in order to reduce ambiguities. In the system proposed in this paper, this assignment problem is
solved based on Bayesian estimation. At each time step, a prior probability for all possible assignments within
the range of a sensor is updated by a likelihood which is modeled as a convolution of a Poisson process and
Gaussian distributed measurement noise.