Research On Analysis Technology Of Detection Signal Of Optical Fiber Perimeter Security System Based On Machine Learning

With the improvement of people's awareness of safety,the perimeter security system has gradually become a research hotspot.Compared with the traditional perimeter security sensor index,the optical fiber perimeter security system has the advantages of small size,light weight,and corrosion resistance.The optical fiber perimeter security system uses optical fiber vibration sensors as sensing elements,which can sensitively detect external vibration,touch,squeeze and other actions.After detect the vibration signal,it is necessary to classify the signal through the optical fiber vibration signal classification system.If it is judged as an intrusion,the alarm system will be activated.How to detect the vibration signal and how to classify the vibration signal are two key research techniques in this system.Therefore,this paper has carried out analysis and technical research on the detection and analysis of fiber optic vibration signal.The main research contents are as follows:(1)The working principle of optical fiber vibration sensing is introduced,and the working principles of Sagnac,Mach-Zander,Michelson and their advantages and disadvantages are analyzed in detail.(2)The end point detection of optical fiber vibration sensing signal is analyzed in detail,and feature extraction is performed from several aspects of time domain,frequency domain and wavelet domain.A composite characteristic dynamic threshold detection method based on short-time over-two-level rate,short-term energy,wavelet coefficient energy variance,and frequency band variance is proposed,which effectively improves the system's vibration recognition accuracy and reduces the system's false alarm rate.(3)Introduces the current feature-based engineering method commonly used in fiber optic vibration signal classification.This method is based on the time-domain and frequency-domain features of the signal and MFCC and other features,and uses machine learning classification algorithms for classification,but the classification accuracy of the signal is insufficient.(4)An optical fiber vibration signal recognition system based on multi-scale one-dimensional convolutional neural network is proposed and designed.Compared with two-dimensional convolutional neural network,it has the advantages of short model training time and fast model inference speed,compared with one-dimensional convolution The recognition accuracy of the product neural network model is higher.