A classify-while-track approach using dynamical tensors

Since a lot of nuclear material (e.g. from hospitals or power plants) is out of the control of the authorities, chances are high that terrorist groups will be able to own such material and mix it with explosives in a ""dirty bomb"" to intensify the scaring effect of an attack. In this paper, a fusion framework based on a tensor with dynamic changing dimensions is presented to solve the association problem of a nuclear source in an open space scenario. Since the decay process of a nuclear source is a random process itself, a likelihood is derived to integrate the distances of all persons to the sensors, the Poisson nature of such a decay, and additive white white noise in the sensing process. As a closed form solution is intractable in general, a Poisson approximation and the saddle point method is proposed. The approach is evaluated in an experimental setup using different scenarios which are motivated from typical situations in a railway station.