Aggregating seismic, acoustic and vibration sensor outputs for enhancing threat detection performance and estimating threat-level

In this paper, a sensor fusion technique with enhanced performance in assets´ protection is introduced. The presented fusion technique models activity dynamics in the protected area by combining acoustic, seismic and vibration sensors outputs. The proposed algorithm learns underlying normal activities in the protected area; and detects abnormal activities - possible threat using sensor outputs. The activity learning in the smart fence evolves through time, and it is independent of prior assumption of threat models. The simulation result developed for cargo train protection shows more than 98% performance in possible threat detection, which performs at least 3% better than naive detection technique. Activity dynamics in large scale areas - airports and military basis - can be modeled using the proposed fusion technique, in which the computational complexity for threat detection is not significant. The capability of the methodology to adjust its free parameters through time makes the threat detection process robust to existing environmental and activity dynamics changes.