Investigation On Pattern Recognition Based On SVM Algorithm For ?-OTDR Fiber-optic Distributed Disturbance Sensing System

Distributed optical fiber-optic disturbance sensor system,which has many distinguished advantages such as simple structure,high sensitivity,long detection range and resistance to electromagnetic interference,is widely applied in the fields of perimeter security,pipelines monitoring and communication cable monitoring.Phase-sensitive optical time domain reflectometer(?-OTDR)optical fiber distributed sensing system has become a hot research area due to its important advantages such as high stability,high resolution,accurate positioning and multipoint disturbance recognition.In this thesis,we focus on an in-depth theoretical and experimental research of pattern recognition of disturbance event in ?-OTDR optical fiber distributed disturbance sensor system.The main research work is described as follows:(1)Through analyzing and comparing the scheme of optical time domain reflectometer(OTDR),(?-OTDR,polarization optical time domain reflectometer(P-OTDR)and brillouin optical time domain reflectometer(B-OTDR),a ?-OTDR system was setup in the laboratory and in-situ sensor system,respectively and the relevant experimental research was conducted as well.(2)A method of recognizing disturbances and disturbance pattern based on Support Vector Machines(SVM)is presented.Through extracting the single mean,variance,standard deviation and signal power characteristics of time-domain and frequency-domain and using the left binary tree structure,a classification based on SVM algorithm is established,whose function is to distinguish disturbance and study disturbed pattern.According to the characteristics of the sensor signal,three-stage classifier is used.Classifier I is used to distinguish if disturbance signals exists.The following classifier II,classifier III and IV are respectively applied to identify the types of disturbance signals.The main identified disturbance patterns include stepping disturbance,water disturbance,and percussion disturbance.(3)GPS navigation App based on Baidu Maps is developed.GPS navigation has three functions,i.e.Determining the position of disturbance,receiving disturbing information and navigating.When disturbance occurs,the three parts work together to guarder line guard to arrive at the scene soon.(4)?-OTDR fiber distributed disturbance sensing system is established and the SVM classification algorithm proposed is evaluated through experiments.100 groups of experiment to distinguish the existence of disturbance are performed at 500m,6500m and 21500m respectively.At 500m;three conditions of walking,watering and percussion need 100 groups of tests respectively.The test data have been thus acquired:The correct rate of recognizing disturbance is above 96%,the correct rate of recognizing three modes is above 94%,and the false and missing rate is less than 6%.