Research On Optical Fiber Perimeter Security System Based On Linear Sagnac Structure

Focusing on the field of perimeter security of optical fiber,this paper conducts the following research on the problems of large volume and high cost of dedicated acquisition card in linear Sagnac structure,easy noise interference of vibration signal and low recognition rate in signal multiple classification:Firstly,the basic principle and characteristics of Sagnac optical fiber sensor are introduced,and the advantages of Sagnac interferometric structure are analyzed.The scheme of reducing sampling rate and using embedded instead of special acquisition card is adopted.Specifically,STM32+AD7606 is used as signal acquisition module to build a linear Sagnac optical fiber sensing system based on embedded.The system cost is reduced,the structure of the system signal acquisition part is smaller,and it is easier to install multiple signal acquisition modules in the defense area type security system.Secondly,the phase demodulation technology of optical fiber is studied,and the demodulation algorithm of 3x3 coupler is mainly analyzed.In order to facilitate data transmission from embedded system to computer,the demodulation algorithm is transplanted to embedded system,which is realized by c code.In the process of signal acquisition,aiming at the problem of low signal-to-noise ratio,this paper proposes a joint de-noising algorithm of PE-VMD-WT by combining permutation entropy(PE),variational mode decomposition(VMD)and wavelet threshold denoising(WT).Through permutation entropy self-adaptation,the decomposition layers k of important parameters of VMD are determined.The VMD was decomposed to obtain k intrinsic modal components(IMF),and the IMF components were further processed by wavelet threshold denoising to obtain a clean IMF.Finally,the signal was reconstructed,and the effectiveness of the method was verified by MATLAB simulation and the acquisition of real signals by the established experimental system.Thirdly,to solve the problem of low recognition rate of traditional feature extraction methods of Sagnac fiber vibration signals,a new feature extraction method,multi-domain hybrid feature extraction based on singular value and weighted permutation entropy,is proposed.The basic principle and solving process of singular value and weighted permutation entropy are analyzed in detail,and the multi-domain feature vectors of singular value in time domain,singular value in frequency domain and VMD weighted permutation entropy are extracted.The combined effects of random forest and support vector machine(SVM)and different feature extraction algorithms are compared and analyzed.The average recognition rate of random forest plus feature extraction method in this paper is up to 97%.Finally,the experimental system is built,the upper computer monitoring software is written,and the complete experimental test of the system is completed,which proves the feasibility and effectiveness of the system scheme.