Detection of radiation signatures in urban environment

Detection of radiation sources in an urban environment requires efficient strategies to accurately identify and track the potential threat to the urban population. In this work, we propose strategies to identify the location and intensity of a radiation source located in a simulated urban neighborhood based on synthetic measurements. The radioactive decay and detection are Poisson random processes so we employ maximum likelihood estimators based on this distribution. Due to the domain geometry and the proposed response model, the negative logarithm of the likelihood has multiple local minimum points and discontinuities. We employ three hybrid algorithms consisting of mixed optimization techniques. For the global optimization technique, we propose stochastic and heuristic approaches including Simulated Annealing, Particle Swarm and Genetic Algorithm. These methods rely only on objective function evaluations which makes them computationally demanding. Equipped with early stopping criteria, the global optimization methods are able to generate pseudo-optima points. These are utilized subsequently as the initial value by the local, non-smooth, deterministic Implicit Filtering optimization method to finish the search. These novel approaches have the potential to address the likelihood non-smoothness issue with multiple local minima and their performances are compared for computational efficiency.