Abstract
Optical fiber has been widely employed for intrusion detection in both civilian and military applications. Currently, most intrusion detection systems collect data properties from optical fiber before detecting intrusion via feature analysis. We present two intrusion detection methods based on system identification in this paper to solve the problem of optical fiber intrusion characteristics that are weak and difficult to extract under windy conditions.The first algorithm estimates the noise covariance of the system at each point. The Frobenius norm of the difference between the noise covariance of different points is used to detect the intrusion. The second algorithm applies the state parameters of non-intrusion point to reconstruct the output of each point by using Kalman filter, and the cross reconstruction error is calculated to recognize intrusion. The two algorithms are tested by data collected from field experiments. The results of the tests demonstrate that the proposed algorithms are both effective. When compared, the first algorithm has greater robustness, whereas the second takes less time to compute.