Research On Intrusion Detection And Recognition Technology Based On Distributed Optical Fiber Acoustic Sensing

Nowadays,optical fiber Sensing technology is developing rapidly.DAS,(Distributed Acoustic Sensing)has been widely used in the field of security and defense due to its advantages of long distance,high sensitivity and anti-interference.But in field application,because of the complex application environment and the changeable intrusion means,there are high false alarm rate and missing alarm rate.That leads to huge property damage and also endanger the people’s life safety and important site security.It’s fascinating for the researchers to design an optical fiber distributed security system which can detect the intrusion information accurately and quickly.In our research,we find that the vibration ranges of multiple vibration signals mixes with each other when multiple intrusion events occur in close distance and at the same time.This kind of mixed event is very common in field application,but there are few researches on mixed event recognition.In the field application,in order to ensure the safety of the security area,it is necessary to carry out real-time vibration detection and identification of the vibration signal of the whole optical cable.Therefore,the number of samples that need to be identified is huge.While the external intrusion signals only occur in some locations of the cable,the rest signals of the cable are mostly environmental and background noise.When the identification algorithm with complex structure is used directly to identify the vibration signal of the whole optical cable,it is easy to cause data jam and more false alarms,which leads to result in the poor performance of the system.In order to solve these two problems,this thesis conducts effective research from DAS technology,pattern recognition method and vibration signal detection method to propose the secondary vibration detection scheme that can not only filter out a large amount of environmental noise but also can identify mixed events which are mixed by two kinds of single events.The main contents of this thesis are as follows:(1)In this thesis,under the guidance of acoustic modulation mechanism of light in optical fiber and acoustic phase sensing principle,a distributed optical fiber acoustic sensing system is built.Nine kinds of typical vibration events(four of that are mixed events of two kinds of events)are collected,and the data set of DAS vibration signals is constructed.(2)In order to solve the more complicated problem of mixed intrusion signals,a feature extraction and classification algorithm based on group roll neural network(100GNET)is proposed.It can effectively extract the characteristic information of DAS signal in the spatial direction and classify intrusion events accurately.Experimental results show that its performance on DAS data set is superior to classical VggNet,InceptionNet,and ResNet.In the validation set,the accuracy rate of the 100G-Net can reach 97.5%.(3)In order to solve the problem of false alarms caused by many environments and complex background noise,secondary vibration detection that based on the distributed and long-distance detection characteristics of DAS technology is proposed.The first part is the SVM vibration detection part,which is only used to detect the occurrence of unknown signals,that can filter out a lot of background noise,reduce false alarm and release the burden of recognition algorithm.Only when the danger signal is identified,the more accurate second part of classification could carry out.The second part classification part adopts 100G-NET to conduct pattern recognition for the danger events to accurately recognize the types of intrusion signals.