Research On Perimeter Alarm System Based On Distributed Optical Fiber Sensor

At present,fence,patrol,infrared and other methods are used by China's airport to prevent intrusion.These traditional technical methods have the disadvantages of high costs,short detection distance and high false alarm rate,and can not afford the demand of safe operation of the airport.As a kind of high performance perimeter alarm system,distributed optical fiber sensing system has the advantages of high sensitivity,high anti-interference,long distance sensing and so on.This is a very ideal airport perimeter alarm system.Based on this,we have developed a perimeter alarm system based on ?-OTDR and Mach-Zehnder.?-OTDR system is adopted to locate the disturbance,and Mach-Zehnder system is used to recognize the disturbance.The main work of this paper is as follows:Firstly,the sensing principle of distributed ?-OTDR sensor is studied in this paper.Based on the detailed deduction of the theoretical formula of ?-OTDR,we design and build a set of ?-OTDR experimental platform.Finally,we analyze the selection of the devices in the system.Secondly,in order to solve the problem that the ?-OTDR backscattering signal is weak,we use the method of cumulative average and wavelet denoising.This method is verified by real test data.In view of the traditional positioning algorithms used to exist false alarm,we use area positioning algorithm.The experimental results show that the positioning accuracy is above 80%.Thirdly,combined with the hardware system,we have designed software system.The software system adopts the producer consumer model to get a separation between acquisition and data processing,and solve the problem of data waiting.In addition,the cumulative average and wavelet denoising algorithm and area location algorithm are implemented.Finally,Mach-Zehnder is used to recognize the disturbance.We study on disturbance identification algorithm.According to the non-stationary characteristics of signal,a method of signal feature extraction and recognition based on CEEMD singular value entropy and multiple kernel SVM is proposed.Recognition rate of this method is proved to be 90%.