Research On Intrusion Event Recognition Algorithm Based On Michelson Interferometer Fiber Perimeter Security System

In recent years,the in-depth study of distributed optical fiber sensing technology has made continuous progress.The fiber perimeter security system based on Michelson distributed vibration fiber sensing technology has become more and more perfect,and has a wide range of development space in many perimeter security fields.However,at present,such security systems cannot fully meet the actual needs in terms of identification of intrusion event types,recognition accuracy,and event signal feature vector extraction efficiency.In view of the above fundamental problems in pattern recognition,this paper based on distributed vibration sensing technology,briefly introduces the current status of fiber perimeter security system.Based on the analysis and comparison of the advantages and disadvantages of different security systems,the following research work is mainly carried out and completed:1.Introduce the sensing principle and intrusion judgment principle of Michelson-type fiber perimeter security system,analyze the significance of intrusion event pattern recognition to security system,and analyze the software and hardware components and implementation of Michelson perimeter security system.Explained in detail.1.Introduce the sensing principle and intrusion judgment principle of Michelsontype fiber perimeter security system,analyze the significance of intrusion event pattern recognition to security system,and analyze the software and hardware components and implementation of Michelson perimeter security system.Explained in detail.2.Using the zero-crossing rate as a feature to analyze the time-frequency of the intrusion event signal,compare the zero-crossing rate distribution of different types of intrusion events,analyze and deal with the difference between the external noise signal and the zero-crossing rate of the intrusion disturbance event signal,and eliminate the difference.The influence of external noise and interference factors on system event identification.3.Considering the requirement of pattern recognition for feature vector,segmentation zero-crossing rate is extracted for the signal of intrusion event,and each type of event extracts 6-segment zero-crossing rate as a set of feature vectors,and multiple groups of each event The feature vector is trained as a training/test sample for the event recognition model.A method of using the zero-crossing rate as the feature vector and combining the support vector machine to efficiently identify many different types of events is proposed.Event classification was performed on 600 sets of data of shearing,shaking,tapping,climbing fence events and non-invasive events.The recognition rate of each of the five events was above 90%,the average recognition rate was over 97%,and the recognition time was 0.046 s.The recognition efficiency is high.4.In the personal computer,the feasibility of event model training and event recognition is verified.The method of using FPGA to realize efficient recognition of intrusion events is proposed.The support vector machine pattern recognition algorithm is written in Verilog HDL language,and the software of FPGA-based security system is applied.Designs such as signal acquisition and processing are detailed.The system was tested in the field.100 sets of simulated intrusion experiments were carried out on shearing,shaking,tapping and climbing.The false positive rate of the experimental results was 2.25%,and the accuracy of event identification was 97.75%,which can meet the actual engineering.Application requirements.