CN106023499A - Fiber security signal dual identification method and system - Google Patents

Abstract

The invention discloses a fiber security signal dual identification method, comprising the steps of: S100, obtaining the phase difference of optical signals in a fiber monitoring area; S200, obtaining a time domain analysis signal package; S300, dividing the time domain analysis signal package according to a preset time span to obtain an event signal; S400, obtaining the time domain signal of the event signal; S500, obtaining the frequency domain signal of the event signal; S600, determining the type of a signal according to the position of a frequency domain signal continuous zero point; S700, constructing a characteristic formwork; and S800, determining whether a signal which can not be determined is an abnormal signal according to the characteristic formwork. The method reduces a rate of false alarm.

Description

A kind of dual recognition methods of optical fiber security signal and system

Technical field

The invention belongs to safety-security area, particularly to a kind of dual recognition methods of optical fiber security signal and system.

Background technology

Perimeter security system all has very important application in national defence and civil area, and it is mainly used in boundary line, army The circumference intrusion detection of the important areas such as thing base, warehouse, barracks, government facility, airport, nuclear power station and prison.Current Circumference security and guard technology mainly has leaky cable, microwave correlation, infrared emission and optical fiber sensing technology etc..Optical fiber perimeter security protection system System is that a kind of accident to threatening area safety is monitored and the modern defense system of alarm, is based on distribution type fiber-optic Sensing technology is applied to the new system of circumference monitoring protection.Owing to optical fiber and Fibre Optical Sensor have, volume is little, lightweight, anti-to be done Ability of disturbing is strong, highly sensitive, functional reliability is high, with low cost and can be used as signal without advantages such as outfield power The feature of transmission channel allows it show one's talent in other circumference security and guard technologies.In practical engineering application, sensor fibre is most Number is exposed in external environment, and the sensor fibre of this design uniqueness is very sensitive to motion, pressure and vibration.It can be along this Fence, enclosure wall are laid and are detected climbing and tap, it is possible to be laid under soil lawn detection and trample.But the high sensitivity of optical fiber must So bring the substantial amounts of warning of system, and be not enough to distinguish a large amount of event early warning based on time-domain signal energy quantitative analysis system, from And cause higher rate of false alarm.The Chinese patent CN201610046738.0 of applicant's earlier application discloses a kind of optical fiber security protection Signal processing method and system, but need the time domain obtaining event signal according to time-domain signal and frequency-region signal special in that patent Seek peace frequency domain character, reconstructed frequency domain feature and temporal signatures is a packet, carries out all data in packet afterwards Processing, such system is accomplished by processing substantial amounts of data, reduces system running speed.

Summary of the invention

For the defect of prior art, the invention provides a kind of dual recognition methods of optical fiber security signal and system.

A kind of dual recognition methods of optical fiber security signal, comprises the following steps: S100: obtains optical fiber and deploys to ensure effective monitoring and control of illegal activities light letter in region Number phase difference；S200: the amplitude difference that the sample rate that the phase difference of optical signal is temporally gone up is changed on the signal of telecommunication, Obtain time-domain analysis signal packet；S300: divide described time-domain analysis signal packet with default time span, obtain event signal； S400: extract the AC portion of described event signal, obtains the time-domain signal of event signal；S500: described time-domain signal is entered Row fast fourier transform obtains the frequency-region signal of event signal；S600: connect the location determination of zero point according to described frequency-region signal The kind of signal, described signal kinds includes: abnormal signal, non-abnormal signal, and can not judge signal；S700: to described not Can judge that signal carries out cluster analysis, construction feature template；S800: signal can not be judged according to the judgement of described feature templates Whether it is abnormal signal.

A kind of optical fiber security signal dual identification system, including with lower module: phase difference acquisition module, is used for obtaining light Fibre is deployed to ensure effective monitoring and control of illegal activities the phase difference of optical signal in region；Time-domain analysis signal packet acquisition module, for pressing the phase difference of optical signal The amplitude difference that temporal sample rate is changed on the signal of telecommunication, it is thus achieved that time-domain analysis signal packet；Divide module, for preset Time span divide described time-domain analysis signal packet, obtain event signal；First acquisition module, extracts described event signal AC portion, obtains the time-domain signal of event signal；Second acquisition module, for carrying out fast Flourier to described time-domain signal Conversion obtains the frequency-region signal of described event signal；First judge module, for connecting the position of zero point according to described frequency-region signal Judging the kind of signal, described signal kinds includes: abnormal signal, non-abnormal signal, and can not judge signal；Feature templates structure To described, modeling block, for not judging that signal carries out cluster analysis, construction feature template；Second judge module, for basis Described feature templates can not judge whether signal is abnormal signal described in judging.

The invention has the beneficial effects as follows: the present invention on the basis of fully analyzing behavior of men feature, propose a kind of based on Time domain denoising, the signal identification new method of frequency domain filtering.In the case of ensureing not fail to report, collect most event signal, By even zero judgement, only signal to be determined is sent to subsequent step and processes judgement, so greatly reduce and need number to be processed According to amount, improve processing speed；Then all kinds of event-templates by building with cluster analysis carry out similarity analysis.The present invention By the most double identification, in the case of ensureing recognition time and alarm rate, reduce rate of false alarm, for optical fiber perimeter safety-protection system Important support is provided.

Accompanying drawing explanation

Fig. 1 is the flow chart of the dual recognition methods of optical fiber security signal of the present invention；

Fig. 2 is the flow chart of step S500；

Fig. 3 is the schematic diagram of step S600；

Fig. 4 is the flow chart of step S700；

Fig. 5 is the structural representation of optical fiber security signal of the present invention dual identification system.

Detailed description of the invention

Understandable, below in conjunction with the accompanying drawings to the present invention for enabling the above-mentioned purpose of the present invention, feature and advantage to become apparent from Detailed description of the invention be described in detail, make above and other purpose, feature and the advantage of the present invention to become apparent from.Entirely The part that reference instruction identical in portion's accompanying drawing is identical.Accompanying drawing the most drawn to scale, it is preferred that emphasis is illustrate this Bright purport.

Embodiment 1

First the dual recognition methods of optical fiber security signal to the present invention is introduced, and refers to Fig. 1, the security protection letter of the present invention Number processing method is on the basis of fully analyzing behavior of men feature, it is proposed that a kind of based on time domain denoising, the letter of frequency domain filtering Number recognition methods, collects most event signal, extracts representative frequency domain by time domain denoising, frequency domain filtering compression Feature, then all kinds of event-templates by building with cluster analysis carry out similarity analysis, thus judge whether to invade Event.

S100: obtain optical fiber and deploy to ensure effective monitoring and control of illegal activities the phase difference of optical signal in region.Described light can be obtained by M-Z type interferometer The phase difference of signal, uses M-Z type optical fiber perimeter safety-protection system Real-time Feedback to deploy to ensure effective monitoring and control of illegal activities the optical signal in region.If optical signal There are unusual fluctuations in phase place, then it is assumed that event occurs.

S200: the amplitude difference that the sample rate that the phase difference of optical signal is temporally gone up is changed on the signal of telecommunication, it is thus achieved that Time-domain analysis signal packet.When event occurs, the sample rate that light sensing unit (M-Z type) is temporally gone up is by optical signal phase place Difference is transformed into the difference of amplitude on the signal of telecommunication, it is thus achieved that time-domain analysis signal packet.

S300: divide described time-domain analysis signal packet with default time span, obtain event signal.Time domain pretreatment: When fiber phase produces instantaneous unusual fluctuation when, typically last for the time less than 1 second by analysis.Corresponding, the instantaneous row of people For being persistently close with the temporal effect of natural phenomena on the impact of fiber phase.But, divide from the angle of body mechanics Analysis, the transient behavior of people is in certain scope of application.Such as 60～80 steps of normally walking for a minute for each person.By 100 meters Distance carries out analysis of running, and the paces of people are about 60 steps, and the time of running is about 12 seconds.In 1 second, the punch number of times of people is 5 Secondary left and right.For the shortest stress time, the behavior of people has trend, is limited to the mechanics of body parts muscle and skeleton Characteristic, everything all cannot be beyond the speed of limb action and the limits of capacity of frequency.So, people on the impact of optical fiber effect from For strength speed, depend on direction and the movement locus of action；For frequency, depend on position and the mode of action.People Behavior be typically series of actions, analyze on the whole, the interference of natural environment is also likely to be discontinuously, interruption a succession of Interference.The interference of animal also belongs to the speed in a segment limit and frequency because meeting the mechanical characteristic of muscle and skeleton.From mechanics Analyze in angle, will persistently keep stress, be necessary for doing work the most effectively.And persistently do work be typically machine vibration, abnormal Natural environment.

So analyzing from time domain, can be 10s for realizing the time series analysis upper limit of optical fiber perimeter security protection.Simultaneously from work The domain analysis of journey practice, an early warning system needs to distinguish the event of triggering within the shortest time.Take into full account people in engineering Behavioural trait, substantially divides single time-domain analysis signal packet into 0.1～1s.Event by experiment test different time span Packet, present invention discover that the when that packet time span being about 0.25 second in intrusion behavior, actual zone indexing is the highest.Institute Set up with the present invention with 0.25 second time domain system as event data span analysis time.

S400: extract the AC portion of described event signal, obtains the time-domain signal of event signal.The event letter collected Some direct current signals can be there are in number, have only to extract signal communication Partial Feature in event analysis, it is therefore desirable to by calculating Direct current signal in signal is filtered by method.Every frame signal is defined as S_i(n), definition signal average is:

$\stackrel{\‾}{S_i}=\[\underset{n=0}{\overset{N}{\Σ}}{s}_i\left(n\right)\]/n$

The DC signal component part is made to beThen AC portion isMeanwhile, Invasion signal is substantially within 100KHz.By down-sampled to event signal.Down-sampled compressed after packet.

S500: described time-domain signal is carried out fast fourier transform and obtains the frequency-region signal of described event signal.Due to Having some difference on the energy of the time domain waveform of artificial invasion signal and environment noise and zero-crossing rate, therefore the present invention is according to signal Short-time zero-crossing rate and short-time energy size extract the temporal signatures of event signal.As in figure 2 it is shown, it comprises the following steps:

S501: calculate short-time energy and short-time zero-crossing rate.The time-domain signal that definition receives is S_iN (), every frame signal is short Shi Nengliang is

$E_i=\underset{n=0}{\overset{N}{\Σ}}\left|S_i\left(n\right)\right|$

If the short-time zero-crossing rate of every frame is

$Z_i=\underset{n=1}{\overset{N}{\Σ}}|\mathrm{sgn}\[S_i\left(n\right)\]-\mathrm{sgn}\[S_i(n-1)\]|$

The short-time energy of computing environment noise and short-time zero-crossing rate, it is assumed that front 10 frame signals are environment noise, first obtain The mean square deviation of every frame noise, using the average of these 10 mean square deviations as the direct current biasing of signal short-time zero-crossing rate.Obtain front 10 frames Short-time energy and average Z of short-time zero-crossing rate_mean、E_meanWith standard deviation Z_std、E_std, it is possible to obtain its initial value Z₀= Z_mean+2*Z_stdAnd E₀=E_mean+2*E_std, two coefficient E are set_coefAnd Z_coefAs threshold value, change two coefficient values and be used for adjusting The sensitivity of joint system.

S502: often cross Preset Time, repeat step S501, only calculates the frame less than threshold value, revises threshold value；The time preset Can be 1 hour, 1 day etc..

S503: extract time-domain signal x (n) of event signal.If now there being frame signal S_iN () is judged as invasion letter Number, take out former frame S of this frame signal_i-1N () is bisected into 5 subframes, calculate in short-term of these subframes the most respectively Amount.Take out short-time energy in subframe and, more than several subframes of threshold value, invade the starting point of signal as this time.Same method is found out The terminal of invasion signal.Extract signal x (n).

S504: frequency domain character processes: after event signal extracts feature in time domain, and event signal passes through FFT again, Again by frequency domain data normalization, as frequency domain character.Time-domain signal x (n) is done Fourier transformation and draws frequency domain characteristic:

$X\left(k\right)={\Σ}_{n=0}^{N-1}x\left(n\right)W_N^{kn},k=0,1,\mathrm{2...},N-1;W_N=e^{-j\frac{2\mathrm{\π}}{N}}$

Then frequency-region signal is normalized:

$X_{norm}\left(k\right)=\frac{X\left(k\right)-X{\left(k\right)}_{min}}{X{\left(k\right)}_{max}-X{\left(k\right)}_{min}}$

S600: connect the kind of the location determination signal of zero point according to frequency domain character, described signal kinds includes: abnormal letter Number, non-abnormal signal, and signal can not be judged.

It is found by the applicant that can produce different optical signals for optical fiber different types of invasion in region of deploying to ensure effective monitoring and control of illegal activities, animal class enters Invade caused optical signal with velocity attenuation faster, and artificially to invade caused optical signal and can decline with slower speed Subtract, and some signal attenuation speed is between invasion and non-intrusive, need to do judgement further.Based on this, the present invention is by even The rate of decay of reflected optical signal is carried out in the region that continuous zero-signal occurs, and then judges kind signal occur accordingly, i.e. connects zero Judge.Signal kinds includes: abnormal signal, non-abnormal signal, and can not judge signal.

S601: obtain normalized optical signal frequency domain character X_normAfter (k), by frequency domain character X_normK () is by adjustable less than one Parameter V_minIt is filtered, makes X_norm(k)=0；if X_norm(k)<V_min, k=0,1 ..., N.X_normK the value of () reflects invasion The power of signal, X when not having intrusion behavior or a non-artificial invasion_normK () value is 0 or a smaller value, when occurring artificially to invade X_normK () becomes big.Will be less than adjustable parameter V_minX_normK the value of () all processes by 0, the most just obtain some continuous print zero Signal.

S602: adjustable parameter v is set_{judg 1},v_{judg 2}, whole frequency domain is divided into 3 parts

f_{judg 1}: (0, v_{judg 1}],

f_{judg 2}: (v_{judg 1}, v_{judg 2}],

f_{judg 3}: (v_{judg 2},+∞)

S603: definite value V is set_zero, when for the first time the signal meeting following formula occurring in a frequency domain, record f_area+ V_zeroRegion.

X_norm(k)=0；If k=f_area,…,f_area+V_zero

S604: if f_area+V_zeroAt f_{judg 1}Then it is judged to non-intrusive data (non-abnormal signal)；If at f_{judg 3}Then It is judged to invade data (abnormal signal)；If at f_{judg 2}(can not judge signal) then enters the algorithm of cluster analysis.

Refer to Fig. 3, tri-optical signal frequency domain character X of A, B, C in Fig. 3_norm(k) curve represent respectively three kinds different Invasion situation, wherein A curve is non-artificial invasion (non-abnormal signal), and B is undecidable situation (can not judge signal), and C is Artificial invasion (abnormal signal), according to company zero determining method of the present invention, for A class situation, X_normK () is at k=f_areaPlace is less than Adjustable parameter V_min, now start to record the number of times occurring zero continuously, recording definite value V_zeroAfter secondary zero occurs, X_norm(f_area+ V_zero)=0, now f_area+V_zeroIt is positioned at f_{judg 1}∶(0,v_{judg 1}] this region, then can determine that A belongs to non-artificial invasion.With A Decision method identical, can be determined that B belongs to undecidable situation by even zero judgement, and C belongs to artificial invasion.

By this judgement, signal has been divided into non-intrusive signal, signal to be determined and invasion signal by the present invention, for light For fine safety-protection system, most of signals belong to non-intrusive signal or the invasion signal being easier to judge, this kind of signal is not Need to do further judgement again to process, therefore only signal to be determined can be sent to subsequent step and process judgement, the biggest Decrease greatly and need data volume to be processed, improve processing speed.

S700: can not judge that signal carries out cluster analysis, construction feature template to described.Before system formally uses, poly- The feature closing a large amount of event carries out cluster analysis, constructs various types of another characteristic template by cluster analysis.Again by each class Other template is divided into 2 classes: a class is artificial；One class is non-artificial.The present invention is by the time-frequency characteristics data of each invasion signal Regard one group of vector as.Application K means clustering method finds out one or several masterplate vector of every class invasion signal, refers to Fig. 4.

S701: determine classification number k according to practical situation, from the object x that n data object choice 2 distance is maximum_i1, x_i2 As accumulation；

S702: select the 3rd accumulation x_i3, meet following relation:

min{d(x_i3, x_ir)=max{min{d (x_j, x_ir)}}

Wherein: r=1,2；j≠i₁,i₂；

S703: repeat step S702 until selecting kth accumulation X_ik, obtain the set of the initial accumulation of k:

$L^{\left(0\right)}=\{x_1^{\left(0\right)},x_2^{\left(0\right)},\mathrm{...},x_k^{\left(0\right)}\}$

Principle of classification is designated as:

$G_i^{\left(0\right)}=\{x:d(x,x_1^{\left(0\right)})\&le;d(x,x_j^{\left(0\right)})\}$

Then, sample is divided into disjoint k class, obtains a preliminary classification:

$G^{\left(0\right)}=\{G_1^{\left(0\right)},G_2^{\left(0\right)},...,G_k^{\left(0\right)}\}$

S704: recalculating accumulation according to preliminary classification, rule is as follows:

$x_i^{\left(1\right)}=\frac{1}{n_i}\underset{x_i\&Element;G_i^{\left(0\right)}}{\&Sigma;}{x}_i,i=1,2,\mathrm{...},k$

After S705: repetition step S704m time, current class G^(m)With front subseries G^(m-1)Time equal, i.e. G^(m)=G^(m-1) Time, then calculate and terminate；

S706: willPreserve as one group of masterplate.

S800: can not judge whether signal is artificially to invade signal according to the judgement of described feature templates.To count in real time Contrast similarity according to feature masterplate, take the class template that similarity is maximum.Judge templet is belonging to artificially also right and wrong the most again Artificially.In this, as output, it is artificially generated alarm signal, non-artificial generation cue.Owing to cosine similarity is the biggest, then table Bright invasion signal is the highest in the similarity of masterplate.Calculate invasion signal x and masterplateBetween cosine similarity:

${\mathrm{xcor}}_i=\frac{x\&CenterDot;x_i^{\left(m\right)}}{\sqrt{x^2}\&CenterDot;\sqrt{x_i^{{\left(m\right)}^2}}},i=1,2,\mathrm{...},k$

Obtain one group of cosine similarity array xcor.Therefore maximum xcor in xcor is obtained_iWith classification i, and this is entered Invade signal as the i-th class invasion signal.

Embodiment 2

Accordingly, as it is shown in figure 5, present invention also offers a kind of optical fiber security signal dual identification system, a kind of optical fiber Security signal processing system, including with lower module: phase difference acquisition module, deploys to ensure effective monitoring and control of illegal activities optical signal in region for obtaining optical fiber Phase difference；Time-domain analysis signal packet acquisition module, for being changed into the sample rate that the phase difference of optical signal is temporally gone up Amplitude difference on the signal of telecommunication, it is thus achieved that time-domain analysis signal packet；Divide module, during for dividing described with default time span Domain analysis signal packet, obtains event signal；First acquisition module, extracts the AC portion of described event signal, obtains event letter Number time-domain signal；Second acquisition module, obtains described event letter for described time-domain signal carries out fast fourier transform Number frequency-region signal；First judge module, for connecting the kind of the location determination signal of zero point, institute according to described frequency-region signal State signal kinds to include: abnormal signal, non-abnormal signal, and signal can not be judged；Feature templates builds module, for described Can not judge that signal carries out cluster analysis, construction feature template；Second judge module, for judging institute according to described feature templates State and can not judge whether signal is abnormal signal.

Elaborate a lot of detail in the above description so that fully understanding the present invention.But above description is only Presently preferred embodiments of the present invention, the present invention can implement to be much different from alternate manner described here, therefore originally Invention is not limited by disclosed above being embodied as.The most any those skilled in the art are without departing from the technology of the present invention In the case of aspects, technical solution of the present invention is made many possible by the method and the technology contents that all may utilize the disclosure above Variation and modification, or it is revised as the Equivalent embodiments of equivalent variations.Every content without departing from technical solution of the present invention, according to this The technical spirit of invention, to any simple modification made for any of the above embodiments, equivalent variations and modification, all still falls within skill of the present invention In the range of the protection of art scheme.

Claims (10)

1. the dual recognition methods of optical fiber security signal, it is characterised in that comprise the following steps:

S100: obtain optical fiber and deploy to ensure effective monitoring and control of illegal activities the phase difference of optical signal in region；

S200: the amplitude difference that the sample rate that the phase difference of optical signal is temporally gone up is changed on the signal of telecommunication, it is thus achieved that time domain Analyze signal packet；

S300: divide described time-domain analysis signal packet with default time span, obtain event signal；

S400: extract the AC portion of described event signal, obtains the time-domain signal of event signal；

S500: described time-domain signal is carried out fast fourier transform and obtains the frequency-region signal of event signal；

S600: connect the kind of the location determination signal of zero point according to described frequency-region signal, described signal kinds includes: abnormal letter Number, non-abnormal signal, and tentatively can not judge signal；

S700: gather the frequency-region signal of intrusion event and environment by cluster analysis, construction feature template；

S800: can not judge whether signal is abnormal signal according to the judgement of described feature templates.

The dual recognition methods of optical fiber security signal the most according to claim 1, it is characterised in that by M-Z type interferometer Obtain the phase difference of described optical signal.

The dual recognition methods of optical fiber security signal the most according to claim 1, it is characterised in that described default time span Degree is 0.1-1S.

The dual recognition methods of optical fiber security signal the most according to claim 1, it is characterised in that described default time span Degree is 0.25S.

The dual recognition methods of optical fiber security signal the most according to claim 1, it is characterised in that described step S400 is concrete Including: every frame event signal is defined as S_iN (), definition event signal average is

In formula, n represents n-th each sampled point, makes the DC signal component part be

Then AC portion is

The dual recognition methods of optical fiber security signal the most according to claim 1, it is characterised in that described step S500 is concrete Including:

S501: the every frame event signal received is S_i(n), the short-time energy E of every frame event signal_iFor

If the short-time zero-crossing rate Z of every frame_iFor

In formula, sgn is sign function；

The short-time energy of computing environment noise and short-time zero-crossing rate, it is assumed that front 10 frame event signals are environment noise, first obtain The mean square deviation of every frame noise, using the average of these 10 mean square deviations as the direct current biasing of signal short-time zero-crossing rate, obtains front 10 frames Short-time energy and average Z of short-time zero-crossing rate_mean、E_meanWith standard deviation Z_std、E_std, obtain its initial value Z₀=Z_mean+2* Z_stdAnd E₀=E_mean+2*E_std, two coefficient E are set_coefAnd Z_coefAs threshold value, two threshold values are used for regulating system sensitivity；

S502: often cross Preset Time, repeat step S501, only calculates the frame less than threshold value, revises threshold value；

S503: if now there being frame signal S_iN () is judged as invading signal, take out former frame S of this frame signal_i-1N () divides equally It is 5 subframes, calculates the short-time energy of these subframes the most respectively, take out short-time energy in subframe and be more than the several of threshold value Individual subframe, as the starting point of this invasion signal, in like manner finds out the terminal of invasion signal, extracts the time-domain signal of event signal x(n)；

S504: time-domain signal x (n) is done Fourier transformation and draws frequency-region signal:

Then frequency-region signal is normalized:

The dual recognition methods of optical fiber security signal the most according to claim 1, it is characterised in that described step S600 is concrete Including:

S601: obtain normalized optical signal frequency domain character X_normAfter (k), by frequency domain character X_normK () is by less than an adjustable parameter V_minIt is filtered, makes X_norm(k)=0；if X_norm(k)<V_min, k=0,1 ..., N；

S602: adjustable parameter v is set_judg1,v_judg2, whole frequency domain is divided into 3 parts,

f_judg1∶(0,v_judg1],f_judg2∶(v_judg1,v_judg2],f_judg3∶(v_judg2,+∞)；

S603: definite value V is set_zero, when occurring meeting in a frequency domain for the first time

X_norm(k)=0；If k=f_area,…,f_area+V_zero

Signal time, record f_area+V_zeroRegion；

S604: if f_area+V_zeroAt f_judg1Then it is judged to non-abnormal signal；If at f_judg3Then it is judged to abnormal signal；As Fruit is at f_judg2Then enter step S700.

The dual recognition methods of optical fiber security signal the most according to claim 1, it is characterised in that described step S700 is concrete Including:

S701: determine classification number k, from the object x that n data object choice 2 distance is maximum_i1, x_i2As accumulation；

S702: select the 3rd accumulation x_i3, meet following relation:

min{d(x_i3, x_ir)=max{min{d (x_j, x_ir)}}

Wherein: r=1,2；j≠i₁, i₂；

S703: repeat step S702 until selecting kth accumulation X_ik, obtain the set of the initial accumulation of k:

Principle of classification is designated as:

Then, sample is divided into disjoint k class, obtains a preliminary classification:

S704: recalculating accumulation according to preliminary classification, rule is as follows:

After S705: repetition step S704m time, obtain G^(m)=G^(m-1), then calculate and terminate；

S706: willPreserve as one group of masterplate.

The dual recognition methods of optical fiber security signal the most according to claim 1, it is characterised in that described step S800 bag Include:

Calculate invasion signal x and masterplateBetween cosine similarity:

Obtain one group of cosine similarity array xcor, obtain maximum xcor in xcor_iWith classification i, and using this invasion signal as I-th class invasion signal.

10. an optical fiber security signal dual identification system, it is characterised in that include with lower module:

Phase difference acquisition module, deploys to ensure effective monitoring and control of illegal activities the phase difference of optical signal in region for obtaining optical fiber；

Time-domain analysis signal packet acquisition module, for being changed into the signal of telecommunication by the sample rate that the phase difference of optical signal is temporally gone up On amplitude difference, it is thus achieved that time-domain analysis signal packet；

Divide module, for dividing described time-domain analysis signal packet with default time span, obtain event signal；

First acquisition module, extracts the AC portion of described event signal, obtains the time-domain signal of event signal；

Second acquisition module, obtains the frequency domain letter of described event signal for described time-domain signal carries out fast fourier transform Number；

First judge module, for connecting the kind of the location determination signal of zero point, described signal kinds according to described frequency-region signal Including: abnormal signal, non-abnormal signal, and signal can not be judged；

Feature templates builds module, for gathering the frequency-region signal of intrusion event and environment by cluster analysis, construction feature mould Plate；

Second judge module, can not judge whether signal is abnormal signal described in judging according to described feature templates.