LAUSR

Recognition of different kinds of human intrusions from the environmental disturbance with high efficiency is still a challenging task in perimeter security monitoring with fiber optic vibration sensor, since the vibration signals induced by these events are highly similar to each other, and it can not be directly discriminated by the signals. In this paper, an intelligent event recognition scheme is proposed to improve its performance in complicated environmental applications. In this event recognition scheme, a variational mode decomposition based kurtosis feature combined with a zero crossing rate feature is used as the input feature vectors. A support vector machine is used to classify the input feature vectors into the corresponding categories. A series of field tests show that the proposed scheme can accurately and rapidly classify wind disturbance and three typical patterns of human intrusions such as waggling the fence, climbing the fence and knocking the fence. The average identification rate of 100.0% and 96.9% are achieved for the wind disturbance and human intrusion events, respectively. The recognition processing time can be controlled less than 0.4 s. Thus, the intelligent event recognition scheme can fully satisfy the online monitoring requirements for practical applications.