The key technology and main difficulty for optical fiber perimeter system is the extraction and recognition of intrusion signals, vibration signals normally consist of noises, intrusion and disturb signals. Firstly, a new detection method combining constant false alarm rate (CFAR) method and Level Crossing (LC) method was proposed to distinguish the intrusion and no-intrusion signal before recognition. The former can produce adaptive thresholds to eliminate noise and disturb signals according to the background homogeneity, the later can ensure the integrity of the intrusion signal and further reduce disturb signal. Second, multi-feature parameters including traditional timedomain features, wavelet packet energy Shannon entropy and wavelet packet energy, energy proportion, kurtosis, skewness are accurately extracted from the intrusion signal. Finally, use support vector machine (SVM) identify multi-feature vectors of different types of vibration signals. The proposed method was experimented on Sagnac optical fiber pre-warning system. The result show that the method can extract vibration signal effectively form sensing signals, improve the system recognition rate.

Date Published: 18 December 2019
PDF: 7 pages
Proc. SPIE 11340, AOPC 2019: Optical Fiber Sensors and Communication, 113400L (18 December 2019); doi: 10.1117/12.2542482

Published in SPIE Proceedings Vol. 11340:
AOPC 2019: Optical Fiber Sensors and Communication
Jie Zhang; Songnian Fu; Jun Yang, Editor(s)