The Research And Implementation Of Detection And Recognition Algorithm For Optical Fiber Intrusion Signals

Optical fiber sensing signals have many advantages such as anti-electromagnetic interference,passive sensors and good electrical insulation.It also own high sensitivity and adaptability to a wide range of complex environmental monitoring.Therefore,optical fiber pre-warning system is widely used to monitor and protect national borders,airports,military bases,and power generation,plants and oil pipelines and other fields.The optical fiber pre-warning system is designed primarily for harmful intrusions and uses the detected vibration signals to identify the type of intrusion.The detection and identification algorithm of the optical fiber pre-warning system needs to be implemented on an embedded processor composed of a digital signal processor(DSP).The memory resources of the signal processing board are limited,so it is necessary to design a reasonable software architecture which is ensure that the algorithm can be implemented effectively and succinctly.In this paper,the detection and recognition algorithm of optical fiber intrusion signal is studied.The algorithm is divided into two parts:location detection of vibration signals and identification of intrusion types.A set of intrusion signal detection and recognition scheme implemented on DSP is proposed based on the embedded platform of the fiber early warning system.For the location detection of vibration signals,in the aspect of algorithm design,this paper adapts the more efficient Cell Averaging Constant False Alarm Rate(CA-CFAR)algorithm in the constant false alarm algorithm to detect the vibration signal.However,the performance of this method is often limited to the distribution characteristics of the actual signals.In order to achieve uniform clutter background,the background noise is standardized.Based on this method,the false alarm probability and detection probability of the signal are changed from 6%and 80.5%to 2.75%and 91%.In other words,the false alarm probability is reduced by 3.25%and the detection probability is increased by 10.5%.For intrusion identification,this paper proposes a method based on time domain duty cycle feature and domain features that achieved by fast Fourier transform(FFT)to identify fiber intrusion signals.This method is simple and easy to implement,and can effectively distinguish the three types of intrusion,electric drilling,walking and picking,and the average recognition rate reached 93.5%.Among them,the recognition rate of the electric drilling signal is 86.6%,the recognition rate of the walking signal is 96.7%,and the recognition rate of the picking signal is 97.3%.In terms of algorithm implementation,this paper designs an architecture based on parallel processing of two DSPs,and optimizes the processing flow of sliding window summation to further improve the execution efficiency of the algorithm.The detection result is used as the input of identification,which reduced the amount of data in the subsequent links.The interaction function between the DSP and the host computer is designed,and visually displaying the final recognition result on the host computer.We performed on-site observation data acquisition of external intrusions and conducted actual tests on different types of signals to verify the algorithm.Experimental results show that the algorithm can effectively detect and identify intrusion events and achieve real-time performance requirements.