Research On Pattern Recognition For Distributed Optical Fiber Vibration Sensing Security System Based On Coherent OTDR

With the development of social economy and the rising level of science and technology,the demand for intelligent security system is getting higher and higher.The security system based on distributed optical fiber vibration sensing plays an increasingly important role in modern security system due to its unique advantages.The security system uses optical fiber as the transmission and perception of environmental information,it can be aware of the surrounding vibration events while the signal can also be transmitted to the system terminal for processing.In the fence and other perimeter scenes,sensor fiber can be laid in the target area,you can monitor the climbing,stampede and other intrusion events,it take the role of the perimeter security;If the fiber is laid along the oil or gas transmission line or the national border,the above long-distance area can be safely monitored.However,the actual surrounding environment is complex and varied,the signals collected by the system show different data characteristics,process the sensing signal effectively,and then identify and classify the intrusion events is the key technology to realize the intelligent of the security system.The actual effect of its pattern recognition module will directly affect the monitoring performance of the security system.In this paper,the research on vibration signal analysis and pattern recognition intelligent realization of the distributed optical fiber security system are taken.The main achievements are described as following:1.The structure and working principle of the fiber-optic vibration sensing system based on phase-sensitive optical time domain reflectometry are analyzed.The fiber-optic vibration sensing technology based on coherent detection is applied to the actual security system,and mathematical model analysis and some preprocessing are also applied to sensing signals to provide support for subsequent pattern recognition algorithm design.2.Depth research on the theory of pattern recognition and its application in the intelligent implementation of optical fiber security system,and the algorithm requirements analysis and pattern recognition algorithm framework design are also taken for perimeter environment and longdistance security application field respectively.3.In this paper,the simulation of perimeter security in laboratory environment is carried out.Through spectrum analysis,three-dimensional eigenvector is constructed,combined with multi-class SVM,the pattern recognition algorithm is designed and implemented.A 98.6% of the intrusion detection recognition rate and a 91.2% of the event classification recognition rate are achieved.The simulation of long distance security in off-site environment is also carried out.This paper proposes some event detection algorithms which combine the time-space-frequency multidimensional features and the time-sequence signal processing method,and has successfully monitored the event of walking and driving.The experimental results show that the perimeter environment pattern recognition algorithm proposed in this paper can classify intrusion events intelligently with high accuracy.The proposed event detection algorithm can realize intrusion event dynamic monitoring in off-site environment.