To improve the convergence and precision of intrusion localization in optical-fiber sensing perimeter protection applications, we present an algorithm based on an unscented particle filter (UPF). The algorithm employs particle swarm optimization (PSO)
IntroductionOptical-fiber sensor-based intrusion detection technologies are widely used in perimeter security protection systems. Recently, the optical fiber sensing technologies available for intrusion detection include the interferometer-based optical fiber sensors and the optical time domain reflectometry (OTDR)-based optical fiber sensors, and each of which has characters [1]- [8]. Among the technologies, the interferometer-based optical-fiber sensors are preferred in intrusion detection for their high sensitivity to vibrational signals and low cost. As is known, it is important to localize the intruder when an intrusion signal is detected in a perimeter protection system. Generally, the underground intrusion signals to be detected are acoustic (or vibrational) signals generated by the intruder. When an intrusion occurs, the time of arrival (TOA) of the intrusion signal is used to locate the position of the intruder approximately [9].As the interferometer-based systems use consecutive laser pulses, the interval between the laser being sent out and the intrusion signal arriving at the receiver cannot be determined. Thus, the TOA of the intrusion signal cannot be accurately measured, which affects the precision of intrusion localization. To get the precise TOAs of the intrusion signals, many signal processing algorithms were employed [11]. However, the approaches suffer from the measurement errors for the fast speed of the laser propagating in the optical fiber, the errors of the time limit the precision of the intrusion localization to tens of meters [12]. Many researchers have worked on this problem and various signal processing algorithms have been employed to obtain a precise TOA in optical-fiber sensing localization [13], [14]. To improve the precision of intrusion localization, we have previously used the geometrical positions of the distributed sensors and the differences in relative TOAs to estimate the position of the intruder [15]-[16]. State estimation-based methods demonstrate high precision. However, as the measurement equation is nonlinear, the convergence speed is poor and the precision is subject to measurement errors.In this paper, a particle filter is used to handle the problem of nonlinearity in optical-fiber sensing intrusion localization. To avoid the degeneracy problem and sample impoverishment after resampling, we employ the unscented particle filter (UPF) and use particle swarm optimization (PSO) to maintain diversity in the particles. A series of simulations demonstrate that the proposed algorithm improves precision and convergence when there is no prior intruder location.