A Hybrid Detector Network for Nuclear and Radioactive Threat Detection

Abstract:

The research is along two directions and a number of results have been published or submitted. Physical side: Algorithms have been developed to detect isotopes based on the ideas of group positive Lasso and the total least squares. Two papers have been either published or to appear in Radiation Measurements. Cyber-side: The research focused on detection of unknown sources.

• Detection by an binary sensor network has been developed assuming the unknown trajectory of the source is piecewise linear. The result is reported in this poster and submitted for publication.
• Triangulation is a common tool for localizing the source. Our work demonstrated that triangulation is very sensitive to noise and a robust way is proposed. The result has been published in IEEE Signal Processing Letter.
• Localization and tracking of a source assuming no model of the signal strength is available. Convergence results have been established. The result is submitted for publication.
• Localization of the unknown number of sources were studied and an algorithm has been developed based on the Bayesian approach together with the Markov Chain Monte Carlo numerical algorithm. The results are submitted for publication.

Reported in this poster is an binary sensor network for source localization. Our idea is based on a simple observation: In a number of real applications, a very complex representation of a source movement may be unnecessarily. With a reasonable degree of accuracy, some approximations can and should be considered, e.g., piecewise linear approximation both in direction and speed. Further, the accuracy can be improved by increasing the number of line segments. This idea is reminiscent of approximating an arbitrary curve by piecewise linear functions. It is shown in the paper that if an object moves in a piecewise linear fashion with an unknown speed in each line segment and passes through the sensing ranges of three or more sensors, the object's movement including the path and speed can be determined uniquely with probability one in some probability sense. Further, the results hold even the density of the binary sensors goes to zero.