Research And Implementation Of Optical Fiber Intrusion Feature Extraction Algorithm Based On ?OTDR

The distributed optical fiber pre-warning system based on ?OTDR is a new type of security system.It has the advantages of strong anti-interference and high sensitivity.It can monitor in real-time the long-distance and concurrent intrusion behavior.However,the intrusion signal detected in the practical application of optical fiber pre-warning system is various in types and complicated in features,and thus requires effective preprocessing and feature extraction.This dissertation focuses on this technical difficulty and carries out the corresponding algorithm and engineering research.The main content is as follows:This paper firstly summarizes the current situation of optical fiber pre-warning system and the existing methods of intrusion signal feature extraction,summing up their implementation bottlenecks,which leads to the research idea of the thesis.Then,the paper introduces the composition and working principle of the optical fiber pre-warning system.In the aspect of fiber intrusion signal pre-processing,the time-domain correlation analysis of the intrusion signal is carried out in accordance with the birefringence of light.It is found that there is a significant difference between the false alarm and the intrusion signal in the time domain.And then,this paper proposes a preprocessing method based on time-domain correlation coefficient.In order to improve the efficiency of the correlation coefficient,the optimization of the implementation is carried out for the two modules of wavelet de-noising and correlation calculation.The experimental data verify that the preprocessing method can effectively remove the false alarm signal.In terms of feature extraction of the fiber intrusion signal,a variety of signal extraction methods and their engineering realization are studied.In the time domain,the duty cycle and the pitch period are extracted,and the frequency center feature is extracted in the frequency domain.The energy fraction of the wavelet coefficients is extracted in the time-frequency domain.The eigenvector constructed by the multidimensional feature has a strong distinction between the mechanical and the artificial signal.In this paper,the author proposes a method to extract the pitch period from the pre-screening based on endpoint detection,which efficiently improves the accuracy of feature extraction.The realization of feature extraction algorithm is the major research content of this paper.According to the characteristics of feature extraction of intrusion signal,this paper is based on DSP platform and design feature extraction and implementation architecture,and uses modularization to divide the function,including DSP and host computer data interaction logic,data synchronization between two DSP and DSP data storage.The actual data is used to verify the fiber intrusion signal preprocessing,feature extraction method,and its implementation proposed in this paper.