• Acta Optica Sinica
  • Vol. 39, Issue 6, 0628002 (2019)
Kuan Peng1、2, Cheng Feng1, Senmao Wang1, Fan Ai1, Hao Li1, Deming Liu1, and Qizhen Sun1、*
Author Affiliations
  • 1 School of Optical and Electronic Information, Huazhong University of Science and Technology, Wuhan, Hubei 430074, China
  • 2 Wuhan Fisilink Microelectronics Technology Company Limited, Wuhan, Hubei 430074, China
  • show less
    DOI: 10.3788/AOS201939.0628002 Cite this Article Set citation alerts
    Kuan Peng, Cheng Feng, Senmao Wang, Fan Ai, Hao Li, Deming Liu, Qizhen Sun. Event Discrimination Method for Distributed Optical Fiber Intrusion Sensing System Based on Integrated Time/Frequency Domain Feature Extraction[J]. Acta Optica Sinica, 2019, 39(6): 0628002 Copy Citation Text show less
    Principle diagram of WLI-DOFVS system
    Fig. 1. Principle diagram of WLI-DOFVS system
    Flow chart of data processing
    Fig. 2. Flow chart of data processing
    Schematic of Mallat multiresolution decomposition
    Fig. 3. Schematic of Mallat multiresolution decomposition
    Signals before and after denoising by wavelet threshold. (a) Original signal; (b) denoised signal
    Fig. 4. Signals before and after denoising by wavelet threshold. (a) Original signal; (b) denoised signal
    Signal segmentation results of several typical intrusion events. (a) Footsteps of passerby; (b) bicycle; (c) knocking on fence; (d) cutting of optical cable
    Fig. 5. Signal segmentation results of several typical intrusion events. (a) Footsteps of passerby; (b) bicycle; (c) knocking on fence; (d) cutting of optical cable
    Photograph of sensing fiber laying on fence and ground
    Fig. 6. Photograph of sensing fiber laying on fence and ground
    Typical features in time domain. (a) Average fragment interval; (b)fragment length; (c) PAR
    Fig. 7. Typical features in time domain. (a) Average fragment interval; (b)fragment length; (c) PAR
    Energy distributions in frequency domain. (a) Footsteps of passerby; (b) bicycle rolling; (c) knocking on the fence; (d) cutting of optical cable
    Fig. 8. Energy distributions in frequency domain. (a) Footsteps of passerby; (b) bicycle rolling; (c) knocking on the fence; (d) cutting of optical cable
    ParameterValue
    L0 /km50
    Ld /km1
    W /nm≈40
    P /dBm9.4
    Table 1. Main parameters of WLI-DOFVS system
    Intrusion eventTime domain feature +SVMFrequency domain feature +SVMTime/frequency domain feature +RBF NNTime/frequency domain feature +SVM
    Footsteps of passerby10052.547.597.5
    Bicycle rolling75.097.5100100
    Knocking on fence90.097.587.5100
    Cutting of optical cable10072.587.595.0
    Average recognition rate91.2580.0080.6398.13
    Table 2. Recognition rate of 4 different methods for the 1st training set%
    Intrusion eventTime domain feature +SVMFrequency domain feature +SVMTime/frequency domain feature +RBF NNTime/frequency domain feature +SVM
    Footsteps of passerby10050.082.598.75
    Bicycle rolling90.095.097.595.0
    Knocking on fence85.095.087.5100
    Cutting of optical cable10060.072.5100
    Average recognition rate93.7575.0085.0098.50
    Table 3. Recognition rate of 4 different methods for the 2nd training set%
    Intrusion eventTime domain feature +SVMFrequency domain feature +SVMTime/frequency domain feature +RBF NNTime/frequency domain feature +SVM
    Footsteps of passerby10052.552.5100
    Bicycle rolling92.597.592.597.5
    Knocking on fence87.597.597.5100
    Cutting of optical cable97.570.087.5100
    Average recognition rate94.3879.3882.5099.38
    Table 4. Recognition rate of 4 different methods for the 3rd training set%
    Intrusion eventTime domain feature +SVMFrequency domain feature +SVMTime/frequency domain feature +RBF NNTime/frequency domain feature +SVM
    Footsteps of passerby10040.040.0100
    Bicycle rolling85.090.095.097.5
    Knocking on fence85.097.595.0100
    Cutting of optical cable10065.055.0100
    Average recognition rate92.5073.1371.2599.38
    Table 5. Recognition rate of 4 different methods for the 4th training set%
    Intrusion eventTime domain feature +SVMFrequency domain feature +SVMTime/frequency domain feature +RBF NNTime/frequency domain feature +SVM
    Footsteps of passerby10095.030.097.5
    Bicycle rolling92.597.595.0100
    Knocking on fence75.090.082.592.5
    Cutting of optical cable97.52.542.595.0
    Average recognition rate91.2571.5062.5096.25
    Table 6. Recognition rate of 4 different methods for the testing set%
    Recognition methodMean /%Variance /10-4
    Time domain feature +SVM92.632.04
    Frequency domain feature +SVM75.8014.18
    Time/frequency domain feature +RBF NN76.3887.20
    Time/frequency domain feature +SVM98.331.65
    Table 7. Mean and variance of recognition rate of 4 different methods for the 5 groups of data
    Kuan Peng, Cheng Feng, Senmao Wang, Fan Ai, Hao Li, Deming Liu, Qizhen Sun. Event Discrimination Method for Distributed Optical Fiber Intrusion Sensing System Based on Integrated Time/Frequency Domain Feature Extraction[J]. Acta Optica Sinica, 2019, 39(6): 0628002
    Download Citation