Research On Integrated Identification Method Of Optical Intrusion Behavior Based On Feature Fusion

As the optical fiber sensor is with high sensitivity,good insulation,high temperature and corrosion,it is widely used in airports,warehouses and other important places in the perimeter security system.Because the signal of the optical fiber perimeter security system has the characteristics of nonlinearity,nonstationarity,intermittency and vulnerable to the influence of environment and weather factors,it is necessary to eliminate external disturbance and to identify the intrusion signal in time and accurately.In this paper,according to the distribution characteristics of fiber vibration signal,the signal processing and pattern recognition method is used to study the integrated identification method of the intrusion signal in order to improve the accuracy of the intrusion signal recognition of the optical fiber perimeter security system.The main research work of this paper is as follows:1.A method of framing optical fiber vibration signal based on cross-correlation is given.According to the phase space reconstruction method,the minimum frame length of the fiber vibration signal is obtained by the method of cross-correlation function to overcome the non-linear and intermittent characteristics of the signal.2.A method of feature extraction of optical fiber vibration signal in time domain,frequency domain and wavelet domain is adopted.The time-domain,frequency-domain and wavelet-domain characteristics of optical fiber vibration signals are extracted separately.For the feature elements that may be elements which have a bad performance during the classification process in feature space,each fusion feature element is evaluated based on DFPA(Discriminative Function Pruning Analysis),and the feature set with the best classifiability can be obtained.3.An intrusion signal integrated classification method is proposed.Aiming at the non-linearity,non-stationary and intermittent characteristics of the intrusion signal of the optical fiber perimeter security system and the poor generalization ability when using a single classifier,an integrated classifier is built.The simulation results show that the intrusion signal recognition method of optical fiber perimeter security system has the characteristics of high recognition accuracy.