Identification of Damaging Activities for Perimeter Security

Abstract

The advent of fiber sensor has opened up a plenty of opportunities to perimeter security. Damaging activities can cause fiber vibrate whose signals will be collected as data source for detection and identification. In this paper we formulate identifying damaging activities by vibration signals as a classification problem. We design features that characterize vibration signals by combining both the statistic and time-frequency information. In addition, a novel multiclass classification tree of Support Vector Machine (SVM) is introduced to recognize vibration signals. Experimental results show that the proposed feature extraction scheme and classification method yields a high recognition rate of 94.6% for nine different kinds of damaging activities, much better than other results reported ever since.

Intelligent detection and identification in fiber-optical perimeter intrusion monitoring system based on the FBG sensor network

“…Moreover, the FBG sensor response changes linearly with the intrusion behavior impacts, so it is more helpful to identify the event features. However, the event identification is still a pending and challenging problem due to the environmental complexity in practical uses [20][21][22], such as changing climates like wind and rain, and unpredictable wildlife interferences. Even the same person could introduce different effects due to different fence materials or different sensor mount ways, which presents the most difficult problem to extract essentially distinguishable characteristics of the event targets.…”

A novel FBG-based security fence enabling to detect extremely weak intrusion signals from nonequivalent sensor nodes

“…2 and 3, when the intrusion act or the sensor response at the first sensor node is weak and the output is less than the noise level at other nodes, it is impossible to correctly judge whether there is intrusion or not by using any a kind of threshold method. Thus, more and more characteristic parameters should be added together to determine a threat act, such as the event duration time, the signal's frequency components, and so on [20,21]. The normalizing method of different parameters and their proper weighing factors are both critical for a final decision, however, they are always very difficult to obtain.…”

Optical Fiber Sensors in Physical Intrusion Detection Systems: A Review

“…A study by Yan et al [60] focused solely on improved signal processing. They formulate an event classification algorithm for vibrations in a perimeter security system which analyzes both static and dynamic signals.…”