Detection of particle sources with directional detector arrays

We consider the problem of detecting far-field particle sources, such as nuclear, radioactive, optical, or cosmic. This problem arises in applications such as security, surveillance, visual systems, and astronomy. We propose a mean-difference test (MDT) with cubic and spherical detector arrays, assuming Poisson distributed measurement models. Through performance analysis, such as computing the probability of detection (P _d) for a given probability of false alarm (P _fa), we show that the MDT has a number of advantages over the generalized likelihood-ratio test (GLRT): computational efficiency, higher probability of detection, asymptotic constant false-alarm rate (CFAR), and applicability to low signal-tonoise ratio (SNR). For each array, we also present an estimator to find the source direction. We conduct Monte-Carlo numerical examples that confirm our analysis.