A robust hypothesis test for the sensitive detection of constant speed radiation moving sources

Abstract

International audienceRadiation Portal Monitors are deployed in linear networks to detect radiological material in motion. As a complement to single and multichannel detection algorithms, inefficient under too low signal-to-noise ratios, temporal correlation algorithms have been introduced. Test hypothesis methods based on empirically estimated mean and variance of the signals delivered by the different channels have shown significant gain in terms of a tradeoff between detection sensitivity and false alarm probability. This paper discloses the concept of a new hypothesis test for temporal correlation detection methods, taking advantage of the Poisson nature of the registered counting signals, and establishes a benchmark between this test and its empirical counterpart. The simulation study validates that in the four relevant configurations of a pedestrian source carrier under respectively high and low count rate radioactive backgrounds, and a vehicle source carrier under the same respectively high and low count rate radioactive backgrounds, the newly introduced hypothesis test ensures a significantly improved compromise between sensitivity and false alarm. It also guarantees that the optimal coverage factor for this compromise remains stable regardless of signal-to-noise ratio variations between 2 and 0.8, therefore allowing the final user to parametrize the test with the sole prior knowledge of background amplitude

Recent Developments in Count Rate Processing Associated with Radiation Monitoring Systems

“…When the network is linear and the source carrier has a constant velocity, this phase corresponds to the periodic echo of the signal increase on the first channel as seen on the other channels. The difference in nature between both methods introduces a quantitative information gain, and thus a potential improvement in detection capability [27][28][29].…”

“…If ∀j ∈ 1; n ½ ½ &∀φ ∈ 1; It has been proven in [27][28][29], the largely significant added-value in term of detection capability permits by the implementation of the correlation based detection. The true detection rate is increased while maintaining very low false alarm rate.…”

“…However, an algorithmic refinement is achieved with the introduction of a Normal law a priori on count rate distributions [29]. Thus, a modified variance σ 2 (R ϕ ) is obtained by the estimation of individual variances σ 2 (ρ j, i ) provides by the nonlinear filter (cf.…”