Research On Signal To Noise Ratio Enhancement And Pattern Recognition Method Of ?-OTDR Distributed Optical Fiber Sensing System

The distributed optical fiber sensing system based on phase sensitive optical time domain reflectometer(?-OTDR)has the advantages of simple structure,stable performance,long sensing distance and strong anti-interference ability,and has a wide range of applications.However,in practical application,the signal to noise ratio(SNR)of sensing system will decrease due to the influence of noise.The low SNR of the sensing signal will make it very difficult for the system to locate the disturbance point,and the system false alarm rate will also increase.At the same time,with the complexity and diversification of the application scenarios of ?-OTDR system,the simple disturbance location functions do not meet the needs of users.The user also wants to know the type of disturbance signal while obtaining the disturbance location information.Therefore,the SNR enhancement and disturbance signal pattern recognition of sensor systems have become key problems to be solved in the application of systems.This paper focuses on SNR enhancement technology and disturbance signal pattern recognition technology for ?-OTDR system.The research content is as follows.1.In order to enhance the SNR of the disturbance location curve,the time-distance 2D image is synthesized by using the sensing signal of the ?-OTDR system.Then,the base of the image is removed and the guided filter is used to filter the image.The experimental results show that the spatial resolution of the disturbed location points does not decrease,and the SNR of the disturbance location curve can increase 11.05 dB in the best case.2.In order to identify the type of the disturbance signal,four kinds of disturbance signals such as environmental noise,knocking,shaking,and shearing are collected during the experiment.Then,two different pattern recognition methods are designed.One of the pattern recognition methods uses wavelet energy spectrum as the signal feature,and uses BP neural network to realize the pattern recognition of the disturbance signal.Another pattern recognition method uses the short-time Fourier transform time-frequency image as a signal feature,and uses a convolutional neural network to realize the pattern recognition of the disturbance signal.At the same time,genetic algorithm is used to optimize the initial parameters of BP neural network.Finally,the recognition accuracy of the two pattern recognition methods is compared.The experimental results show that the accuracy of pattern recognition of BP neural network increases from 82% to 90% after optimization the parameters by genetic algorithm.The short-time Fourier transform feature image contains the information of signal frequency change with time,so the pattern recognition accuracy of convolutional neural network is better than BP neural network.The recognition accuracy rate of convolutional neural network is about 93.5%.3.The software of the ?-OTDR system is compiled by C++ which can realize the function of signal acquisition and disturbance location.Further,two types of pattern recognition systems based on BP neural network and convolutional neural network are realized by MATLAB GUI and C++ mixed programming.Then,in order to realize the communication between the ?-OTDR software and the pattern recognition system,the communication protocol between the systems is designed.Finally,?-OTDR software and two different pattern recognition systems are tested based on ?-OTDR experimental platform.