CN109523729A - Based on the optical fiber perimeter security protection intrusion event recognition methods modeled entirely and identifier - Google Patents

Based on the optical fiber perimeter security protection intrusion event recognition methods modeled entirely and identifier Download PDF

Info

Publication number
CN109523729A
CN109523729A CN201811286536.9A CN201811286536A CN109523729A CN 109523729 A CN109523729 A CN 109523729A CN 201811286536 A CN201811286536 A CN 201811286536A CN 109523729 A CN109523729 A CN 109523729A
Authority
CN
China
Prior art keywords
output
probability
optical fiber
classifier
security protection
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN201811286536.9A
Other languages
Chinese (zh)
Inventor
黄翔东
杨琳
刘铁根
刘琨
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Tianjin University
Original Assignee
Tianjin University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Tianjin University filed Critical Tianjin University
Priority to CN201811286536.9A priority Critical patent/CN109523729A/en
Publication of CN109523729A publication Critical patent/CN109523729A/en
Pending legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G08—SIGNALLING
    • G08B—SIGNALLING OR CALLING SYSTEMS; ORDER TELEGRAPHS; ALARM SYSTEMS
    • G08B13/00—Burglar, theft or intruder alarms
    • G08B13/02—Mechanical actuation
    • G08B13/12—Mechanical actuation by the breaking or disturbance of stretched cords or wires
    • G08B13/122—Mechanical actuation by the breaking or disturbance of stretched cords or wires for a perimeter fence
    • G08B13/124—Mechanical actuation by the breaking or disturbance of stretched cords or wires for a perimeter fence with the breaking or disturbance being optically detected, e.g. optical fibers in the perimeter fence

Landscapes

  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Alarm Systems (AREA)

Abstract

The invention discloses a kind of based on the optical fiber perimeter security protection intrusion event recognition methods modeled entirely and identifier, method includes: that application endpoints detect determining event endpoint, output signal v (n) carries out AR modeling to signal v (n), and calculates zero-crossing rate；AR coefficient and ZCR are combined, constitutive characteristic vector；M SVM is combined together by classifier by AdaBoost algorithm, by the trained classifier of feature vector feed-in, parallel output M group decision value；Using sigmoid pattern fitting method, probability value is converted by decision value }, finally, being based on AdaBoost algorithm weights, each classifier output probability is added, obtains final probability output, and judge final output classification, realization classifies to six class things.Identifier includes: analog-to-digital conversion device and DSP device, the different invasion movement of 6 kinds of identification.

Description

Based on the optical fiber perimeter security protection intrusion event recognition methods modeled entirely and identifier
Technical field
A kind of entered the present invention relates to digital signal processing technique field more particularly to based on the optical fiber perimeter security protection modeled entirely Invade event recognition method and identifier, and in particular in DMZI (double Mach-Zehnder interferometers) optical fiber perimeter security system It carries out accurate to intrusion event and efficiently classifies.
Background technique
In recent years, distribution type fiber-optic technology is widely used in circumference safety-security area[1][2].As a kind of typical distribution Formula optical fiber sensing system, DMZI distributed optical fiber sensing system[3][4]Have the advantages that highly sensitive and reaction speed is fast, by It is widely used in submarine cable security protection[5], pipeline leakage testing[6]Etc. all kinds of safety-security areas[7]In.
Determine in general, the processing that a DMZI distribution type fiber-optic security system acts invasion generally includes invasion Position[8][9][10], end-point detection[11][12]Classify with intrusion event[13][14]Three parts.Currently, intrusion classification is still in immature Stage.The intrusion event classification method mature for one, should have the characteristics that following four:
1) feature extraction complexity is as low as possible；2) identifying system can not only export court verdict, moreover it is possible to export all kinds of Other probability of happening.3) classification accuracy is as high as possible；4) event as much as possible is identified.
However, existing method[15][16][17][18]Above four features cannot be taken into account to the classification of intrusion event.Document [16] it proposes a kind of based on wavelet decomposition and support vector machines[19](support vector machine, SVM) is combined Method, but the time-consuming of this method and operand can increase with Decomposition order into exponential increase, and identify mistake Rate is relatively high, it is not easy to accomplish flexible configuration, not be able to satisfy practical application request；Document [17] proposes a kind of based on empirical mode (Empirical Mode Decomposition, EMD) and RBF (radial basis function, RBF) are decomposed through network The method combined, the kurtosis value that this method passes through calculating intrinsic mode function (Intrinsic Mode Function, IMF) Invasion signal is described as feature vector, and then can identify 4 class intrusion events, but this method calculating process needs experience more Secondary iteration, so that whole system calculation amount increases, efficiency is reduced.
Summary of the invention
The present invention provides a kind of based on the optical fiber perimeter security protection intrusion event recognition methods modeled entirely and identifier, this hair It bright the characteristics of invasion signal is described, can accurately reflect invasion signal in terms of frequency domain, time domain two respectively, can be with Higher accuracy rate realization classifies to six class things, described below:
It is a kind of based on the optical fiber perimeter security protection intrusion event recognition methods modeled entirely, the described method comprises the following steps: answering Determine that event endpoint, output signal v (n) carry out AR modeling to signal v (n), and calculate zero-crossing rate with end-point detection；By AR system Several and ZCR is combined, constitutive characteristic vector；
M SVM is combined together by classifier by AdaBoost algorithm, by the trained classification of feature vector feed-in Device, parallel output M group decision value；
Using sigmoid pattern fitting method, probability value is converted by decision value, finally, weighing based on AdaBoost algorithm Each classifier output probability is added, obtains final probability output, and judge final output classification by weight, realizes to six classes Things is classified.
Wherein, the AR coefficient is used to determine the spectral shape of power spectral density.The method only needs 3 AR coefficients, energy The power spectrum of one invasion signal is described.
Further, which comprises give a feature vector x, decision value is switched to the probability of happening of event, is Invasive biology and probability Estimation provide basis.
Wherein, described to be based on AdaBoost algorithm weights, each classifier output probability is added, it is defeated to obtain final probability P out1,,…,pQ, and judge final output classification specifically:
Input vector resampling, the adjustment according to identification error and weight normalization, the classifier of M appropriate precision pass through M weighted value ω1,...,ωMWeighting can gather the classifier as a higher precision,
Final probability output value p1,...,pQAre as follows:
Therefore, final type deterministic conversion are as follows:
Further, the method can be used for output probability of happening.
A kind of identifier based on the optical fiber perimeter security protection intrusion event recognition methods modeled entirely, the identifier include:
Signal to be filtered is passed through into end-point detection, the starting point for judging that outgoing event occurs from the off will below one section The signal of time is sent into analog-to-digital conversion device and samples to obtain sample sequence x (n), enters DSP device in the form that parallel data inputs, Feature vector x is calculated by AR modeling and zero-crossing rate, feed-in AdaBoostSVM identifies 6 kinds of different invasion movements.
The beneficial effect of the technical scheme provided by the present invention is that:
1, the present invention proposes the feature vector combined based on AR coefficient and zero-crossing rate, respectively in terms of frequency domain, time domain two The characteristics of being described to invasion signal, capable of accurately reflecting invasion signal, guarantees the accuracy rate of identification；
2, the present invention is based on the support vector machines of sigmoid models fitting, can not only preferably embody proposed feature Advantage improves accuracy of identification, and can export the probability of happening of all kinds of events；
3, AdaBoost method of the present invention focuses in the event being difficult to differentiate between, and can further increase The discrimination of system, method proposed by the invention can identify further types of event, and discrimination can satisfy practical need It asks.
4, frequency domain, temporal signatures of the present invention in combination with intrusion event, construct the multi-feature vector simplified, entirely Face accurately describes the characteristics of every class intrusion event, realizes and comprehensively describes to all kinds of intrusion events；Meanwhile the present invention is used Feature vector extremely simplify, only 6 class events can accurately be described with 4 elements；
5, present invention implementation pattern identification classification in DMZI optical fiber sensing system, can identify 6 class events, pass through experiment 6 class events can be accurately identified by demonstrating pattern recognition classifier device proposed by the present invention, and average recognition accuracy reaches 87.14%, accuracy of identification is able to satisfy actual requirement.
Detailed description of the invention
Fig. 1 is the schematic diagram of the DMZI distributed optical fiber sensing system of the prior art；
Fig. 2 is the flow chart provided by the invention based on the optical fiber perimeter security protection intrusion event recognition methods modeled entirely；
Fig. 3 is the flow chart of feature extraction provided by the invention and pattern-recognition；
Fig. 4 is the schematic diagram of the data model of modern spectrum analysis provided by the invention；
Fig. 5 is the schematic diagram of Sigmoid Function Fitting provided by the invention；
Fig. 6 is the schematic diagram of 6 kinds of invasions provided by the invention movement；
Fig. 7 is the schematic diagram of invasion signal waveform provided by the invention:
(a) it shakes；(b) it shears；(c) it climbs；(d) it taps；(e) it hits；(f) it kicks.
Fig. 8 is the schematic diagram of averaged feature vector provided by the invention:
(a) it shakes；(b) it shears；(c) it climbs；(d) it taps；(e) it hits；(f) it kicks.
Fig. 9 is that hardware of the invention implements figure；
Figure 10 is DSP internal processes flow graph.
Table 1 is Average zero-crossing rate；
Table 2 is the comparison of 4 class intrusion event experimental precisions；
Table 3 is the comparison of 6 class event experimental precisions；
Table 4 is the estimation of 6 class event occurrence rates.
Specific embodiment
To make the object, technical solutions and advantages of the present invention clearer, embodiment of the present invention is made below further Ground detailed description.
Embodiment 1
The embodiment of the present invention provide it is a kind of based on the optical fiber perimeter security protection intrusion event recognition methods modeled entirely, referring to fig. 2, The recognition methods the following steps are included:
As shown in Fig. 2, the event intrusion system that the embodiment of the present invention is proposed includes 3 stages: 1) passing through end-point detection Algorithm determines the time of origin point of event；2) suitable feature vector x is extracted；3) suitable classifier, identification invasion thing are used Part classification c, while exporting all kinds of probability of happening p1,…,pQ。
Stage1, application endpoints, which detect, determines event endpoint, output signal v (n).And the 2nd stage and the 3rd stage is specific Process is as shown in Figure 3.
Stage2: AR modeling is carried out to signal v (n), and calculates zero-crossing rate.By AR coefficient a1,…,apIt is combined with ZCR, Constitutive characteristic vector x=[a1,…,ap,ZCR]。
Stage3: M SVM is combined together by classifier by AdaBoost algorithm, and feature vector feed-in is trained Classifier, parallel output M group decision value { f1,1,…,f1,Q},…,{fM,1,…,fM,Q}。
Then, using sigmoid pattern fitting method, probability value is converted by decision value:
{p1,1,…,p1,Q},…,{pM,1,…,pM,Q}。
Finally, being based on AdaBoost algorithm weights, each classifier output probability is added, final probability output is obtained p1,,…,pQ, and final output classification is judged according to following formula
In conclusion the embodiment of the present invention using double Mach-Zehnder distributed optical fiber sensing systems as background, proposes one Kind is based on the intrusion event recognition methods modeled entirely.The recognition methods passes through AR first[20][21](autoregressive) it models Technology models signal, and extracts AR coefficient and zero-crossing rate combination as feature vector；Secondly, passing through sigmoid[22]Mould The method that type fitting and SVM are combined enables this method to export probability of happening；Finally, in conjunction with AdaBoost[23]Technical combinations Multiple SVM are to improve the accuracy rate of identification.
Embodiment 2
The scheme in embodiment 1 is further introduced below with reference to Fig. 1-Fig. 5, specific calculation formula, example, It is described below:
1, double Mach-Zehnder distributed optical fiber sensing systems
The structural principle of distributed optical fiber sensing system based on DMZI principle is as shown in Figure 1:
Such as Fig. 1, P point is disturbance point, and sensing optic cable length is L.The light that laser issues after isolator by passing through by coupling Two-beam line is equally divided into after device C1, this two-beam is injected in the double Mach-Zehnder interferometers being made of C2, C3, later Two-beam in sensing loop respectively to propagate clockwise and counterclockwise, and on the coupler of opposite end (C3 or C2) It interferes and is output on detector PD1 and PD2.Optical signal is converted to electric signal by detector, by right after stopping direct current The high-speed collection card (Data Acquisition, DAQ) answered collects.Difference according to actual needs, capture card DAQ are set For different sample rates.Particularly, the DAQ1 in Fig. 1 classifies for end-point detection and intrusion event, and DAQ2 is for invading thing Part positioning.Finally, pass through (Industrial Personal Computer, IPC) execution related algorithm in industrial computer It realizes required function (such as end-point detection, furnace-incoming coal and pattern classification).
2, based on the feature extracting method of modeling
1) latent structure: the construction of feature vector is based primarily upon the following:
1, feature vector should include the integrated information in relation to disturbing signal, to include not only time-domain information, also to include Frequency domain information；
2, the length of feature vector should be short as far as possible, so as to reduce the complexity of subsequent mode identification procedure Degree；
3, feature vector should be able to inherently reflect the whole feature of signal, and this point can pass through data modeling It realizes.
2) based on the frequency spectrum description of AR modeling: it is generally known that, spectrum analysis is divided into two kinds: classical spectrum analysis and modern spectrum point Analysis.Classical spectrum analysis is based on Fourier transformation, actually a kind of method of signal decomposition, therefore feature difficult to realize is brief Description.On the contrary, as shown in figure 4, modern spectrum analysis is a kind of method based on data modeling, by an observation data v (n) It is considered as what white noise w (n) was obtained by linear system H (z).
According to statistic line loss rate theory, auto-correlation function r is observedvv(n) and white noise auto-correlation function rwwIt (n) can be by Following formula indicates:
rvv(n)=rww(n)*h(n)*h (2)
Wherein, h (n) indicates the shock response of linear system, and h (- n) indicates that the h (n) after overturning, " * " indicate convolution.It is right (2) formula carries out Fourier transformation, can obtain:
Pvv(ω)=Pww(ω)·|H(jω)|2 (3)
Wherein, Pvv(ω), Pww(ω) respectively indicates the power spectral density of observation signal and noise, and H (j ω) indicates linear The frequency response of system, it may be assumed that
Wherein, H (z) is system function.
In view of white noise is flat in frequency band, therefore:
Wherein,It is noise variance, (3) formula changes in turn are as follows:
Under normal conditions, H (z) can be expressed from the next:
Wherein, as B (z)=1, H (z) is referred to as AR (autoregressive) model；If A (z)=1, H (z) quilt Referred to as MA (moving avearge) model；When A (z) ≠ 1 and B (z) ≠ 1, H (z) is referred to as arma modeling.In three of the above In type, AR is most popular one kind, because it can only match wide range of random process by seldom coefficient.
Obviously, when selecting AR model, power spectral density Pvv(ω) becomes:
It may determine that by above formula, Pvv(ω) is by noise varianceIt is determined with AR coefficient, that is to say, that AR coefficient can be with As signal characteristic.In addition,Determine PvvThe amplitude of (ω), AR coefficient determine PvvThe spectral shape of (ω), and DMZI system is only adjusted The phase of optical fiber disturbance signal processed rather than amplitude, therefore, between various types of signalGap is little, not by it as feature.
It is well known that AR coefficient can be calculated by p rank Yule-Walker linear equation:
Wherein, Cv(i), i=0 ..., p is covariance value, can be by being calculated.
The covariance matrix of above formula is Hermitian and Toeplitz, thus, coefficient a1,...,apIt can pass through Levinson-Durbin algorithm calculates.The experimental results showed that the AR model of low order just matches the invasion not of the same race letter of DMZI enough Number.Experiment shows that, as p=3, effect is best, that is to say, that as long as 3 AR coefficients, it will be able to one invasion of effective description The power spectrum of signal.
3) based on the time-domain description of zero-crossing rate: as a kind of common time-domain response criterion, zero-crossing rate (zero- Crossing rate, ZCR) it is able to reflect out the degree of signal intensity, and defined by following formula:
Wherein, L is sample length, and " sign " indicates following operation:
ZCR is added into feature vector x, so that its information for being included further is enriched, to further increase mould The precision of formula classification.It is emphasized that because a1,a2,a3With ZCR can from whole, the substantially condensed information of height, therefore, Extracted feature vector can be used in distinguishing the intrusion event of larger class.
3. the classifier design based on modeling
1) SVM decision value switchs to probability output:
In order to identify a greater variety of intrusion events, mentioned method uses Multi- class SVM classifier.As shown in figure 3, when given One feature vector x is dedicated to switching to decision value into the probability of happening of event, to provide base for invasive biology and probability Estimation Plinth.
Assuming that Q decision function f will be generated in the training stage for a SVM classifierq(x), q=1 ..., Q,
The final types of decision-making can be determined by following formula:
Further, for training sample x label y (y ∈ { 1 ..., Q }), label y is further converted to length For Q vector v=[v (1) ..., v (q) ..., v (Q)]T, and meetWherein
That is, decision value fq(x) bigger, then the probability of c=y is bigger, correspondingly, decision value fqIt is (x) smaller, So probability of c=y is with regard to smaller.As shown in figure 5, being based on training sample, the statistic curve that can obtain discrimination (uses '+' table Show), and corresponding matched curve is indicated by the solid line.
As seen from Figure 5, it is determined as the probability of happening and SVM decision function f of y classy(x) relation curve between, can To be obtained by sigmoid Function Fitting.
Wherein, function representation are as follows:
That is, this relationship can pass through determination (Ay,By) this parameter is to determination.Therefore, single SVM will Determine Q parameter pair.Because of the application of M class SVM, a shared QM parameter will be to will be determined.In addition, as shown in figure 3, testing Stage, Q decision value f of m-th of SVM classifierm,1,...,fm,QProbability value p will be converted directly intom,1,...,pm,Q。
2) AdaBoost: in order to further increase identification probability, the embodiment of the present invention introduces AdaBoost method to handle M The output valve of a SVM.Particularly, for single SVM, it is only necessary to moderately high accuracy rate.It is a series of by utilizing Boosting method (input vector resampling, the adjustment according to identification error and weight normalization etc.), this M appropriate precision Classifier pass through M weighted value ω1,...,ωMWeighting can gather the classifier as a higher precision.Therefore,
Final probability output value p1,...,pQAre as follows:
Therefore,
Final type deterministic can also convert into:
In conclusion the embodiment of the present invention is respectively described invasion signal in terms of frequency domain, time domain two, it can be accurate Reflection invade signal the characteristics of, guarantee identification accuracy rate.
Embodiment 3
Experimental facilities used by testing is as shown in Figure 1.Laser source is the distributed feedback laser of 1550nm, and intensity is 3.5Mw.The sample rate of DAQ1 is 10kHz, and the intra-record slack byte of each movement is 0.3s.Sensing optic cable is the single-mode optics of 2.25 kms Fibre is fastened around on fence up and down, as shown in Figure 6.
The embodiment of the present invention altogether tests six kinds of movements (shake, shear, climb, tap, hit and kick).It is each Class experiment repeat 120 times, wherein 50 times as training, 70 times as test.
1) feature extraction:
Fig. 7 provides 6 class intrusion event signal waveforms.Fig. 8 gives the averaged feature vector of 6 class events.It can be with from Fig. 8 Find out 6 class events in frequency domain character [a1,a2,a3]TThere is apparent difference.In addition, the addition of temporal signatures ZCR is but also each Difference becomes more apparent upon between class.
2) 4 class events are distinguished:
This trifle will to this method and EMD method for 4 class events (shake, shearing, climbing and tap) identification into Row compares.For EMD method, the feature vector length used is 6, and mentioned method length is 4.Table 2 gives two kinds of sides The accuracy of identification of method compares.From Table 2, it can be seen that the average recognition rate of the method based on EMD be 85.75% (99.7%, 85.1%, 87.3%, 70.9% average value), lower than mentioned method average recognition rate 92.785% (100%, 82.07%, 100%, 89.07% average value).
3) 6 class event category:
In order to further compare the ability that two methods identify more types of events, this method is added to the new thing of two classes Part (hits and kicks).Table 3 gives two methods for the accuracy of identification of 6 class events.
1, when event type increases, the Average Accuracy of the method based on EMD is reduced to 63.33% from 85.75%.This It is because EMD method is a kind of method based on signal decomposition, when facing polymorphic type event, it can not all kinds of things of accurate description The feature of part.
1 Average zero-crossing rate of table
The comparison of 24 class intrusion event experimental precision of table
The comparison of 36 class event experimental precision of table
2, in contrast, this method institute is impacted smaller.
Average recognition rate for 6 class events is 87.14%, only fewer than 4 classes 5.645%.This is because feature mainly according to It is obtained by data modeling, can there is brief comprehensive reflection signal characteristic.
3, mentioned method can not only provide identification types, additionally it is possible to export probability of happening.Its significance lies in that these occur Probability can provide reference in comprehensive possibility assessment, especially for the sample of mistake point.
Particularly, table 4 gives the example of 4 mistakes point, and provides all kinds of probability of happening.By taking first row as an example, although Maximum probability is p2=57.03%, corresponding classification is shearing, and the second high probability is p4=40.32%, still remind percussion It should be regarded as a reasonable recognition result.
The estimation of 46 class event occurrence rate of table
Embodiment 4
The embodiment of the invention provides a kind of based on the optical fiber perimeter security protection intrusion event identifier modeled entirely, referring to Fig. 9 And Figure 10, the identifier be it is corresponding with the recognition methods in Examples 1 and 2, the structure of the identifier is specific as follows:
Referring to Fig. 9, signal x (t) to be filtered is first passed around into end-point detection, the starting point that outgoing event occurs is judged, from Point starts, and the signal of a period of time below is sent into A/D (analog-to-digital conversion device) sampling and obtains sample sequence x (n), with parallel data The form of input enters DSP device, and feature vector x, feed-in AdaBoostSVM identification is calculated by AR modeling and zero-crossing rate 6 kinds of different invasion movements).
The internal processes process of DSP device is as shown in Figure 10.
Figure 10 process is divided into the following steps:
(1) it needs to require (signal passband bandwidth such as to be filtered) according to concrete application first, constructs filter.
(2) end-point detection is carried out to signal using filter.
(3) then, CPU main controller reads sampled data from the port I/O, into internal RAM.
(4) AR model is established, extracts signal AR coefficient, and calculate signal zero-crossing rate, construction feature vector.
(5) vector feed-in AdaBoostSVM is identified, exports classification and probability.
It may be noted that realized due to using DSP, so that entire parameter estimation operation becomes more flexible, it can be according to signal The concrete condition for the various components for being included is arranged by the inner parameter that flexible in programming changes algorithm.
The embodiment of the present invention to the model of each device in addition to doing specified otherwise, the model of other devices with no restrictions, As long as the device of above-mentioned function can be completed.
Bibliography
[1]J.C.Juarez and H.F.Taylor,“Distributed fiber optic intrusion sensor system,”IEEE J.Lightw.Technol,vol.23,no.6,pp.2081–2087,2005.
[2]L.Chen,T.Zhu,X.Bao,and Y.Lu,“Distributed vibration sensor based on coherent detection of phase-OTDR”IEEE J.Lightw.Technol,vol.28,no.22,pp.3243– 3249,2010.
[3]Q.Sun,D.Liu,H.Liu,Y.He,and J.Yuan,“Distributed disturbance sensor based on a novel Mach-Zehnder interfero meter with a fiber-loop,”Proc.SPIE, vol.6344,pp.63 440K–1–63 440K–7,Jun.2006.
[4]Q.Sun,D.Liu,H.Liu,and P.Shum,“Distributed fiber-optic sensor with a ring Mach-Zehnder interferometer,”Proc.SPIE,vol.6781,pp.67 814D–1–67 814D– 8,Nov.2007.
[5]S.Xie,M.Zhang,S.Lai,and Y.Liao,“Positioning method for dual Mach- Zehnder interferometric submarine cable security system,”Proc.SPIE,vol.7677, pp.76 770A–1–76 770A–4,Apr.2010.
[6]Y.Zhou,S.Jin,and Z.Qu,“Study on the distributed optical fiber sensing technology for pipeline leakage protection,”Proc.SPIE,vol.6344,pp.634 435–1–634 435–6,Jun.2006.
[7]L.Jiang and R.Yang,“Identification technique for the intrusion of airport enclosure based on double Mach-Zehnder interferometer,”Journal of Computers,vol.7,no.6,pp.1453–1459,2012.
[8]K.Liu,M.Tian,J.Jiang,J.An,T.Xu,C.Ma,L.Pan,T.Wang,.Li,and W.Zheng, “An improved positioning algorithm in a longrange asymmetric perimeter security system,”IEEE J.Lightw.Technol.,vol.34,no.22,pp.5278–5283,2016.
[9]C.Ma,T.Liu,K.Liu,J.Jiang,Z.Ding,L.Pan,and M.Tian,“Longrange distributed fiber vibration sensor using an asymmetric dual Mach-Zehnder interferometers,”IEEE J.Lightw.Technol.,vol.34,no.9,pp.2235–2239,2016.
[10]Q.Chen,T.Liu,K.Liu,J.Jiang,Z.Shen,Z.Ding,H.Hu,X.Huang,L.Pan,and C.Ma,“An improved positioning algorithm with high precision for dual Mach- Zehnder interferometry disturbance sensing system,”IEEE J.Lightw.Technol., vol.33,no.10,pp.1954–1960,2015.
[11]X.Huang,J.Yu,K.Liu,T.Liu,and Q.Chen,“Configurable filter-based endpoint detection in DMZI vibration system,”IEEE Photon.Technol.Lett., vol.26,no.19,pp.1956–1959,Oct.2014.
[12]X.Huang,Y.Wang,K.Liu,T.Liu,C.Ma,and M.Tian,“High-efficiency endpoint detection in optical fiber perimeter security,”IEEE J.Lightw.Technol,vol.34,no.21,pp.5049–5055,2016.
[13]X.Huang,Y.Wang,K.Liu,T.Liu,C.Ma,and Q.Chen,“Event discrimination of fiber disturbance based on filter bank in DMZI sensing system,”IEEE Photon.J.,vol.8,no.3,pp.1–14,2016.
[14]X.Huang,H.Zhang,K.Liu,T.Liu,Y.Wang,and C.Ma,“Hybrid feature extraction based intrusion discrimination in optical fiber perimeter security system,”IEEE Photon.J.,vol.9,no.1,pp.1–12,2017.
[15]S.S.Mahmoud,Y.Visagathilagar,and J.Katsifolis,“Real-time distributed fiber optic sensor for security systems:Performance,event classification and nuisance mitigation,”Photon.Sensors,vol.2,no.3,pp.225–236, 2012.
[16]L.Liu,W.Sun,Y.Zhou,Y.Li,J.Zheng,and B.Ren,“Security event classification method for fiber-optic perimeter security system based on optimized incremental support vector machine,”in Pattern Recognition.Springer,2014,pp.595–603.
[17]K.Liu,M.Tian,T.Liu,J.Jiang,Z.Ding,Q.Chen,C.Ma,C.He,H.Hu,and X.Zhang,“A high-efficiency multiple events discrimination method in optical fiber perimeter security system,”IEEE J.Lightw.Technol.,vol.33,no.23,pp.4885– 4890,2015.
[18]H.Qu,T.Zheng,L.Pang,and X.Li,“A new detection and recognition method for optical fiber pre-warning system,”Optik–International Journal for Light and Electron Optics,vol.137,2017.
[19]C.C.Chang and C.J.Lin,LIBSVM:A library for support vector machines.ACM,2011.
[20]Y.Tang,J.Tang,A.Gong,and W.Wang,“Classifying EEG signals based HMM-AR,”in The International Conference on Bioinformatics and Biomedical Engineering,2008,pp.2111–2114.
[21]Z.Y.He and L.W.Jin,“Activity recognition from acceleration data using AR model representation and SVM,”in International Conference on Machine Learning and Cybernetics,2008,pp.2245–2250.
[22]J.C.Platt,“Probabilistic outputs for support vector machines and comparisons to regularized likelihood methods,”Advances in Large Margin Classifiers,vol.10,no.4,pp.61–74,2000.
[23]X.Li,L.Wang,and E.Sung,“Adaboost with SVM-based component classifiers,”Eng.Appl.Artif.Intell.,vol.21,no.5,pp.785–795,2008.
It will be appreciated by those skilled in the art that attached drawing is the schematic diagram of a preferred embodiment, the embodiments of the present invention Serial number is for illustration only, does not represent the advantages or disadvantages of the embodiments.
The foregoing is merely presently preferred embodiments of the present invention, is not intended to limit the invention, it is all in spirit of the invention and Within principle, any modification, equivalent replacement, improvement and so on be should all be included in the protection scope of the present invention.

Claims (7)

1. a kind of based on the optical fiber perimeter security protection intrusion event recognition methods modeled entirely, which is characterized in that the method includes with Lower step:
Application endpoints, which detect, determines event endpoint, and output signal v (n) carries out AR modeling to signal v (n), and calculates zero-crossing rate； AR coefficient and ZCR are combined, constitutive characteristic vector；
M SVM is combined together by classifier by AdaBoost algorithm, by the trained classifier of feature vector feed-in, and Row output M group decision value；
Using sigmoid pattern fitting method, probability value is converted by decision value, finally, AdaBoost algorithm weights are based on, it will Each classifier output probability is added, and obtains final probability output, and judge final output classification, realize to six class things into Row classification.
2. according to claim 1 a kind of based on the optical fiber perimeter security protection intrusion event recognition methods modeled entirely, feature It is, the AR coefficient is used to determine the spectral shape of power spectral density.
3. according to claim 1 a kind of based on the optical fiber perimeter security protection intrusion event recognition methods modeled entirely, feature It is, the method only needs 3 AR coefficients, can describe the power spectrum of an invasion signal.
4. according to claim 1 a kind of based on the optical fiber perimeter security protection intrusion event recognition methods modeled entirely, feature It is, which comprises give a feature vector x, decision value is switched to the probability of happening of event, for invasive biology and generally Rate estimation provides basis.
5. according to claim 1 a kind of based on the optical fiber perimeter security protection intrusion event recognition methods modeled entirely, feature It is, it is described to be based on AdaBoost algorithm weights, each classifier output probability is added, final probability output p is obtained1,,…, pQ, and judge final output classification specifically:
Input vector resampling, the adjustment according to identification error and weight normalization, the classifier of M appropriate precision pass through M Weighted value ω1,...,ωMWeighting can gather the classifier as a higher precision,
Final probability output value p1,...,pQAre as follows:
Therefore, final type deterministic conversion are as follows:
6. according to claim 1 a kind of based on the optical fiber perimeter security protection intrusion event recognition methods modeled entirely, feature It is, the method can be used for output probability of happening.
7. a kind of for a kind of based on the optical fiber perimeter security protection modeled entirely invasion described in any claim in claim 1-6 The identifier of event recognition method, which is characterized in that the identifier includes:
Signal to be filtered is passed through into end-point detection, judges the starting point that outgoing event occurs, it from the off, will below for a period of time Signal be sent into analog-to-digital conversion device sample to obtain sample sequence x (n), with parallel data input form enter DSP device, pass through Feature vector x is calculated in AR modeling and zero-crossing rate, and feed-in AdaBoostSVM identifies 6 kinds of different invasion movements.
CN201811286536.9A 2018-10-31 2018-10-31 Based on the optical fiber perimeter security protection intrusion event recognition methods modeled entirely and identifier Pending CN109523729A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201811286536.9A CN109523729A (en) 2018-10-31 2018-10-31 Based on the optical fiber perimeter security protection intrusion event recognition methods modeled entirely and identifier

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201811286536.9A CN109523729A (en) 2018-10-31 2018-10-31 Based on the optical fiber perimeter security protection intrusion event recognition methods modeled entirely and identifier

Publications (1)

Publication Number Publication Date
CN109523729A true CN109523729A (en) 2019-03-26

Family

ID=65773539

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201811286536.9A Pending CN109523729A (en) 2018-10-31 2018-10-31 Based on the optical fiber perimeter security protection intrusion event recognition methods modeled entirely and identifier

Country Status (1)

Country Link
CN (1) CN109523729A (en)

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110988804A (en) * 2019-11-11 2020-04-10 浙江大学 Radar radiation source individual identification system based on radar pulse sequence
CN111398201A (en) * 2020-06-08 2020-07-10 翼捷安全设备（昆山）有限公司 Optical gas detector
CN111833561A (en) * 2019-11-20 2020-10-27 杭州四方博瑞科技股份有限公司 Method and system for judging abnormal conditions of perimeter of enclosing wall
CN115060184A (en) * 2022-05-18 2022-09-16 武汉迪信达科技有限公司 Optical fiber perimeter intrusion detection method and system based on recursive graph

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20120301146A1 (en) * 2011-05-27 2012-11-29 Nec Laboratories America, Inc. Equivalent-Link Backward Propagation Method for Nonlinearity Compensation in Fiber Transmission Systems
CN105928604A (en) * 2016-05-27 2016-09-07 深圳艾瑞斯通技术有限公司 Signal acquisition and processing method of optical fiber sensor and device
CN106384463A (en) * 2016-11-24 2017-02-08 天津大学 Method for identifying opening fiber surrounding security invasion events based on mixed characteristic extraction
CN107180521A (en) * 2017-04-19 2017-09-19 天津大学 Optical fiber perimeter security protection intrusion event recognition methods and device based on comprehensive characteristics
EP3321901A1 (en) * 2015-09-02 2018-05-16 Nuctech Company Limited Distributed optical fiber perimeter security system, and sound restoration system and method
CN108133559A (en) * 2016-11-30 2018-06-08 光子瑞利科技（北京）有限公司 Application of the optical fiber end-point detection in circumference early warning system

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20120301146A1 (en) * 2011-05-27 2012-11-29 Nec Laboratories America, Inc. Equivalent-Link Backward Propagation Method for Nonlinearity Compensation in Fiber Transmission Systems
EP3321901A1 (en) * 2015-09-02 2018-05-16 Nuctech Company Limited Distributed optical fiber perimeter security system, and sound restoration system and method
CN105928604A (en) * 2016-05-27 2016-09-07 深圳艾瑞斯通技术有限公司 Signal acquisition and processing method of optical fiber sensor and device
CN106384463A (en) * 2016-11-24 2017-02-08 天津大学 Method for identifying opening fiber surrounding security invasion events based on mixed characteristic extraction
CN108133559A (en) * 2016-11-30 2018-06-08 光子瑞利科技（北京）有限公司 Application of the optical fiber end-point detection in circumference early warning system
CN107180521A (en) * 2017-04-19 2017-09-19 天津大学 Optical fiber perimeter security protection intrusion event recognition methods and device based on comprehensive characteristics

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
XIANG-DONG HUANG, HAO-JIE ZHANG, KUN LIU, TIE-GEN LIU: "Fully modelling based intrusion discrimination in optical fiber perimeter security system", 《OPTICAL FIBER TECHNOLOGY》 *

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110988804A (en) * 2019-11-11 2020-04-10 浙江大学 Radar radiation source individual identification system based on radar pulse sequence
CN110988804B (en) * 2019-11-11 2022-01-25 浙江大学 Radar radiation source individual identification system based on radar pulse sequence
CN111833561A (en) * 2019-11-20 2020-10-27 杭州四方博瑞科技股份有限公司 Method and system for judging abnormal conditions of perimeter of enclosing wall
CN111833561B (en) * 2019-11-20 2021-11-09 杭州四方博瑞科技股份有限公司 Method and system for judging abnormal conditions of perimeter of enclosing wall
CN111398201A (en) * 2020-06-08 2020-07-10 翼捷安全设备（昆山）有限公司 Optical gas detector
CN115060184A (en) * 2022-05-18 2022-09-16 武汉迪信达科技有限公司 Optical fiber perimeter intrusion detection method and system based on recursive graph

Similar Documents

Publication Publication Date Title
CN109523729A (en) Based on the optical fiber perimeter security protection intrusion event recognition methods modeled entirely and identifier
Xu et al. Pattern recognition based on time-frequency analysis and convolutional neural networks for vibrational events in φ-OTDR
Biswas et al. Application of machine learning algorithms to the study of noise artifacts<? format?> in gravitational-wave data
CN111442827B (en) Optical fiber passive online monitoring system for transformer winding vibration
Do et al. Convolutional-neural-network-based partial discharge diagnosis for power transformer using UHF sensor
CN109974835A (en) A kind of vibration detection identification and space-time localization method and system based on fiber-optic signal feature
CN108986363A (en) Optical fiber security protection intrusion event recognition methods and device based on ARMA modeling
CN108832478A (en) A kind of efficient laser control system and control method
Ristea et al. Emotion recognition system from speech and visual information based on convolutional neural networks
Gu et al. An improved sensor fault diagnosis scheme based on TA-LSSVM and ECOC-SVM
Meena et al. Gender recognition using in-built inertial sensors of smartphone
CN115510909A (en) Unsupervised algorithm for DBSCAN to perform abnormal sound features
Huang et al. Fully modelling based intrusion discrimination in optical fiber perimeter security system
Xie et al. Internal defect inspection in magnetic tile by using acoustic resonance technology
Dawood et al. Power quality disturbance classification based on efficient adaptive Arrhenius artificial bee colony feature selection
CN115577249A (en) Transformer acoustic signal identification method, system and medium with multi-view feature fusion
Zhang et al. Artificial Intelligence‐Based Joint Movement Estimation Method for Football Players in Sports Training
Bublin Machine learning for distributed acoustic sensors, classic versus image and deep neural networks approach
Balachandran et al. Classification of Resident Space Objects by shape and spin motion using neural networks and photometric light curves
Bi et al. A hybrid deep learning method for network attack prediction
Wu et al. Real-time activity identification in a smart FBG-based fiber-optic perimeter intrusion detection system
Li et al. Semi-supervised learning for fault identification in electricity distribution networks
Hosgurmath et al. Grey wolf optimizer with linear collaborative discriminant regression classification based face recognition
Huang et al. Fall detection model based on AlphaPose combined with LSTM and Lightgbm
Liu et al. Research on anomaly intrusion detection based on rough set attribute reduction

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
RJ01 Rejection of invention patent application after publication
RJ01 Rejection of invention patent application after publication

Application publication date: 20190326