CN104819382A - Self-adaptive constant false alarm rate vibration source detection method for optical fiber early warning system - Google Patents

Abstract

The invention provides a signal detection method based on self-adaptive constant false alarm rate (GO/SO-CFAR), which is used for realizing vibration source detection in a pipeline optical fiber early warning system and comprises the following steps: continuously receiving N pulse echo data to obtain a frame of echo data, sequentially selecting a distance unit from M distance units in the nth pulse echo data as a detection unit, removing an interference target from a reference unit sequence, calculating a detection threshold value based on the processed reference unit sampling sequence, and finally comparing the detection unit with the detection threshold value to obtain a detection result. The invention utilizes the self-adaptive constant false alarm rate detector to detect and process the optical fiber vibration signal, and the method can ensure constant false alarm rate and also can show stronger detection capability under different backgrounds.

Description

A kind of self-adaption constant false alarm rate vibration source detecting method for predispersed fiber alarm system

Technical field

The present invention relates to a kind of self-adaption constant false alarm rate vibration source detecting method for predispersed fiber alarm system, belong to fiber-optic vibration signal detection technique field.

Background technique

Along with expanding economy, oil and natural gas has become the most important energy of the national economic development.Long distance pipeline, as desirable, the most economic means of transportation of oil and natural gas, is widely used.Oil and gas pipes is the main artery of energy transport, and pipeline, once leakage very easily combustion explosion occurs, not only affects the safety in production of pipeline, also the life and property of giving the country and people masses is brought about great losses.

The detecting and warning system of current application oil and gas pipes mainly contains following several: electronic impulse type fence, microwave wall-alarm, active infrared alarm, leakage cable type perimeter detection alarm system, electret vibration wireline alarm system and optical fiber transducer perimeter alarm system.Compared with electric transducer alarm system, optical fiber transducer has obviously technical advantage in sensing network application: can provide without any need for outdoor Active Device (do not need power supply) safety monitoring reaching 100 kilometers of distances, the terrain environment such as height, complications, turning, bending not by landform limits, and has broken the narrow limitation that infrared rays, microwave wall etc. are only applicable to sighting distance and flat site use.Therefore optical fiber measurement is utilized to vibrate the main method becoming pipeline pre-warning system research.But how rationally effective analysis is carried out to fiber laser arrays signal, set up which type of event model just more effective, become the large focus of one in research and difficult point.

Also there is obvious deficiency in current fiber-optic vibration signal transacting, its mentality of designing is generally the method for direct-detection, the signal detected directly is delivered to display device, the changes in amplitude of clutter and noise is shown simultaneously, the detectability of target signal is determined the supervision of display device by operator.But the monitoring result error that this processing method obtains is large, it is inaccurate to report to the police, and makes long distance complex vibration detect and faces a severe challenge, be badly in need of carrying out the research of long-distance optical fiber vibration detecting method on this basis.

The subject matter that existing research exists does not set up suitable model, particularly do not set up suitable input model, thus the long distance early warning system efficiency making to have put into production is lower even lies idle, the signal transacting link in predispersed fiber alarm system has become the main bottleneck of system and industry development.Therefore, need to propose a kind of suitable model to realize the detection of vibration event, to improve the stability of detection probability and false-alarm probability.

Summary of the invention

The present invention is based on the predispersed fiber alarm system vibration source detection algorithm of self-adaption constant false alarm rate (GO/SO-CFAR) to solve the problem to fiber-optic vibration input.

Based on the predispersed fiber alarm system vibration source detecting method of self-adaption constant false alarm rate (GO/SO-CFAR), it comprises: gather fiber-optic vibration signal; Set up self-adaption constant false alarm rate (GO/SO-CFAR) detector model; Carry out GO/SO-CFAR detection to friction signal, actual false alarm rate and the relation of setting false alarm rate, ensure that detector normally works; Homogeneous background signal is detected, obtains the detection perform under homogeneous background; Multiple target background signal is detected, obtains the detection perform under multiple target background; Adjust signal to noise ratio in both environments respectively, then carry out GO/SO-CFAR detection, draw detection perform curve.

Wherein, detection threshold meets following relation:

S＝TZ

Wherein, S is threshold value; Z is power level, and its value is divided into two kinds of situations to consider, if result of determination k value is greater than the half of reference unit, namely during k>R/2, otherwise t is the normalizing factor, and its value is according to false-alarm probability P _fatry to achieve,

$P_{\mathrm{fa}}=\left\{\begin{array}{c}\left(\begin{array}{c}R\\ k\end{array}\right)\underset{j=1}{\overset{k}{\mathrm{\Π}}}{[T+\frac{R-j+1}{k-j+1}]}^{-1},k>R/2\\ \left(\begin{array}{c}R\\ k\end{array}\right){(1+T)}^{-(N-k-1)}\·\underset{j=0}{\overset{k}{\mathrm{\Σ}}}\left(\begin{array}{c}k\\ j\end{array}\right){(-1)}^j{(1+\frac{j}{N-k}+T)}^{-1},k\≤R/2\end{array}\right.$

Self-adaption constant false alarm rate (GO/SO-CFAR) detector set up by such scheme, can be ensured that the false alarm rate detected is constant, and have stronger detectability under homogeneous background and multiple target background.

According to an aspect of the present invention, provide a kind of signal detecting method based on self-adaption constant false alarm rate, it is characterized in that comprising:

Set up self adaption false alarm rate detector;

Friction laboratory data detected through self-adaption constant false alarm rate detector, ensure that the false alarm rate of actual false alarm rate and setting is extremely close, namely absolute error ensures in the 1e-4 order of magnitude;

Self-adaption constant false alarm rate is used to detect the signal under homogeneous background and under multiple target background respectively, try to achieve its detection probability respectively, and change the signal to noise ratio of data, multi-group data is detected, obtain the detection probability under different signal to noise ratio, show its detection perform.

Accompanying drawing explanation

The enforcement of Fig. 1 the inventive method and proof procedure;

Fig. 2 self-adaption constant false alarm rate (GO/SO-CFAR) detector skeleton diagram;

Fig. 3 self-adaption constant false alarm rate (GO/SO-CFAR) deletion algorithm flow chart;

Self-adaption constant false alarm rate (GO/SO-CFAR) testing result under Fig. 4 homogeneous background;

Self-adaption constant false alarm rate (GO/SO-CFAR) testing result under Fig. 5 multiple target background;

The detection perform curve of Fig. 6 self-adaption constant false alarm rate (GO/SO-CFAR) under homogeneous background;

The detection perform curve of Fig. 7 self-adaption constant false alarm rate (GO/SO-CFAR) under multiple target background;

Specific embodiments

Below in conjunction with the accompanying drawing in the embodiment of the present invention, the technological scheme in the embodiment of the present invention is clearly and completely described.Obviously, described embodiment is only the present invention's part embodiment, instead of whole embodiment.Based on the embodiment in the present invention, for those of ordinary skill in the art, under the prerequisite not paying creative work, other embodiments can also be obtained according to these accompanying drawings, all belong to the scope of protection of the invention.

Shown in Figure 1, Fig. 1 is the flow chart of the predispersed fiber alarm system vibration source detection algorithm based on self-adaption constant false alarm rate (GO/SO-CFAR) of the present invention.As shown in Figure 1, the predispersed fiber alarm system vibration source detection algorithm based on self-adaption constant false alarm rate (GO/SO-CFAR) that the present embodiment discloses comprises:

S101: fiber-optic vibration signal is gathered;

S102: set up self-adaption constant false alarm rate (GO/SO-CFAR) detector model;

S103: self-adaption constant false alarm rate (GO/SO-CFAR) is carried out to friction signal and detects, actual false alarm rate and the relation of setting false alarm rate, ensure that detector normally works;

S104: detect homogeneous background signal, obtains the detection perform under homogeneous background;

S105: detect multiple target background signal, obtains the detection perform under multiple target Beijing;

S106: adjust signal to noise ratio in both environments respectively, then carry out self-adaption constant false alarm rate (GO/SO-CFAR) detection, draw detection perform curve.

In S101, the signal of fiber-optic vibration system is gathered, receive N number of pulse echo data continuously and obtain a frame echo data, and each pulse echo data are processed and sampled.Wherein, comprise M in each pulse echo data according to echo time tactic range unit from left to right, different columns represents echo position.Particularly, each the pulse echo data received are carried out processing and sampling, form the signal arranged in the matrix form.

In S102, set up self-adaption constant false alarm rate (GO/SO-CFAR) detector model, self-adaption constant false alarm rate (GO/SO-CFAR) detector model as shown in Figure 2, part deletion is carried out to the data forming matrix, obtain threshold value again, finally detection unit and threshold value are compared, draw testing result.Intermediate portions is wherein delete procedure, and as shown in Figure 3, its method disclosed comprises:

S301: to reference unit sampling sequence x ₍₁₎≤ x ₍₂₎≤ ... ≤ x _(R);

S302: then to the individual lower orderly sampling summation of k

$Z_k=\underset{i=1}{\overset{k}{\mathrm{\Σ}}}{x}_{\left(i\right)}$

Wherein, x is the value of reference unit; Z _kit is front k item reference unit clutter power;

S303: obtain the normalizing factor T of threshold value according to given wrong probability of erasure _k, the wrong probability of erasure P of delete procedure kth step _fCbe defined as

P _FC＝Pr{D _k>0|H _N} (1)

Wherein, P _fCfor wrong probability of erasure, H _nfor weak clutter is sampled, D _kfor test statistics,

D _k＝x _(k+1)-T _kZ _k(2)

Wherein, x _(k+1)be (k+1) item sampled value, T _kfor the normalizing factor of kth item threshold value;

Mistake probability of erasure P _fCcan be write as integrated form,

$P_{\mathrm{FC}}=-\frac{1}{2\mathrm{\πi}}\∫w^{-1}{\mathrm{\Φ}}_{D_k|H_N}\left(w\right)\mathrm{dw}---\left(3\right)$

Wherein, for at H _nsuppose lower D _kmoment generating function, definition is

${\mathrm{\Φ}}_{D_k|H_c}\left(w\right)=E\left[\exp (-D_{k}w)\right]---\left(4\right)$

(2) formula is substituted in (4) formula,

${\mathrm{\Φ}}_{D_k|H_c}\left(w\right)=E\left\{\exp \right[-(x_{(k+1)}-T_k{Z}_k)w\left]\right\}---\left(5\right)$

Ordered Statistic D ₍₁₎, D ₍₂₎... D _(k+1)joint moment generating function be defined as

${\mathrm{\Φ}}_{D_{\left(1\right)},...,D_{(k+1)}|H_t}(w_1,w_2,...,w_{k+1})=E\left[\exp \right\{-\underset{j=1}{\overset{k+1}{\mathrm{\Σ}}}\left(w_j{D}_{\left(j\right)}\right)\left\}\right]---\left(6\right)$

Relatively (5), (6) two formulas, can draw

w ₁＝w ₂＝…＝w _k＝-T _kw

And

w _k+1＝w

Therefore

${\mathrm{\Φ}}_{D_k|H_N}\left(w\right)=\frac{R!}{(R-k+1)!}\underset{j=1}{\overset{k+1}{\mathrm{\Π}}}\frac{1}{(R-j+1)+w[1-(k-j+1)T_k]}---\left(7\right)$

Wherein, R is reference unit number, and wushu (7) substitutes into formula (3),

$P_{\mathrm{FC}}=\left(\begin{array}{c}R\\ k\end{array}\right)\frac{1}{{[1+T_k(R-k)]}^k}---\left(8\right)$

According to formula (8), the normalizing factor of threshold value can be obtained when given wrong probability of erasure;

S304: try to achieve adaptive threshold T _kz _k, kth+1 reference unit value and adaptive threshold are compared;

S305: if x _(k+1)be less than T _kz _k, then certainly 1 is increased to k, repeats above-mentioned steps, otherwise directly stop circulation, namely x is described _(k+1)..., x _(R)belong to the sampling of strong clutter district;

S306: then k value and R/2 value are compared;

S307: if k>R/2, then judge that detection unit is in weak clutter district, by x ₍₁₎..., x _(k)form clutter power Z, namely the moment generating function definition of its test statistics D is

${\mathrm{\Φ}}_{D_k|H_N}\left(w\right)=\left(\begin{array}{c}R\\ k\end{array}\right)\underset{j=1}{\overset{k}{\mathrm{\Π}}}{[w+\frac{R-j+1}{k-j+1}]}^{-1}---\left(9\right)$

The definition of false-alarm probability is

P _fa＝Φ(T) (10)

Wherein, P _fafalse-alarm probability, then simultaneous formula (9), formula (10),

$P_{\mathrm{fa}}=\left(\begin{array}{c}R\\ k\end{array}\right)\underset{j=1}{\overset{k}{\mathrm{\Π}}}{[T+\frac{R-j+1}{k-j+1}]}^{-1}---\left(11\right)$

S308: if k<R/2, then judge that detection unit is in strong clutter district, by x _(k+1)..., x _(R)form clutter power Z namely the moment generating function definition of its test statistics D is

${\mathrm{\&Phi;}}_{D_k|H_N}\left(w\right)=\left(\begin{array}{c}R\\ k\end{array}\right){(1+w)}^{-(N-k-1)}\&CenterDot;\underset{j=0}{\overset{k}{\mathrm{\&Sigma;}}}\left(\begin{array}{c}k\\ j\end{array}\right){(-1)}^j{(1+\frac{j}{N-k}+w)}^{-1}---\left(12\right)$

Equally by formula (12) and false-alarm definition (10) simultaneous

$P_{\mathrm{fa}}=\left(\begin{array}{c}R\\ k\end{array}\right){(1+T)}^{-(N-k-1)}\&CenterDot;\underset{j=0}{\overset{k}{\mathrm{\&Sigma;}}}\left(\begin{array}{c}k\\ j\end{array}\right){(-1)}^j{(1+\frac{j}{N-k}+T)}^{-1}---\left(13\right)$

In sum, normalizing factor T is according to false-alarm probability P _fatry to achieve, as shown in the formula

$P_{\mathrm{fa}}=\left\{\begin{array}{c}\left(\begin{array}{c}R\\ k\end{array}\right)\underset{j=1}{\overset{k}{\mathrm{\&Pi;}}}{[T+\frac{R-j+1}{k-j+1}]}^{-1},k>R/2\\ \left(\begin{array}{c}R\\ k\end{array}\right){(1+T)}^{-(N-k-1)}\&CenterDot;\underset{j=0}{\overset{k}{\mathrm{\&Sigma;}}}\left(\begin{array}{c}k\\ j\end{array}\right){(-1)}^j{(1+\frac{j}{N-k}+T)}^{-1},k\&le;R/2\end{array}\right.$

Finally calculate threshold value

S＝TZ

Wherein, S is threshold value.

In S103, to fiber-optic vibration system acquisition to friction data carry out GO/SO-CFAR detection, optimum configurations and detect after false-alarm probability as shown in table 1.

Table 1

Can find out actual false-alarm probability and setting false-alarm probability between error in 1e-4 level, illustrate that self-adaption constant false alarm rate (GO/SO-CFAR) detector normally works.

In S104, the repetition data under same group of homogeneous background being carried out to self-adaption constant false alarm rate (GO/SO-CFAR) detects for 10000 times, and optimum configurations is as shown in table 2,

Table 2

Get 16 width testing result figure respectively, as shown in Figure 4, then pass through the record of testing result data, as shown in table 3, the detection perform of self-adaption constant false alarm rate (GO/SO-CFAR) detector under homogeneous background is described.

Table 3

In S105, the repetition data under one group of multiple target background being carried out respectively to self-adaption constant false alarm rate (GO/SO-CFAR) detects for 10000 times, and optimum configurations is as shown in table 4

Table 4

Get 16 width testing result figure respectively, as shown in Figure 5, then pass through the record of testing result data, as shown in table 5, the detection perform of self-adaption constant false alarm rate (GO/SO-CFAR) detector under multiple target background more can be described.

Table 5

Detection mode	Average detected probability
		Self-adaption constant false alarm rate (GO/SO-CFAR)	0.9950

In S106, the data of different signal to noise ratio are detected respectively, can obtain when signal to noise ratio is different, the detection probability value of self-adaption constant false alarm rate (GO/SO-CFAR) detector, i.e. detection perform curve, under homogeneous background and under multiple target background, detection perform respectively as shown in Figure 6,7.Can find out, the detectability of self-adaption constant false alarm rate (GO/SO-CFAR) is stronger.

The present invention has the following advantages compared with existing detecting method

(1), the method can reach fiber-optic vibration input;

(2), constant false alarm rate detecting method can ensure that false-alarm probability is constant, stable performance;

(3), self-adaption constant false alarm rate (GO/SO-CFAR) all shows higher detection probability under homogeneous background and under multiple target background, and namely this self-adaption constant false alarm rate (GO/SO-CFAR) method can ensure the detectability under different background.

Claims (4)

1., based on the signal detecting method of self-adaption constant false alarm rate, it is characterized in that comprising:

Set up self adaption false alarm rate detector;

Friction laboratory data detected through self-adaption constant false alarm rate detector, ensure that the false alarm rate of actual false alarm rate and setting is extremely close, namely absolute error ensures in the 1e-4 order of magnitude;

Self-adaption constant false alarm rate is used to detect the signal under homogeneous background and under multiple target background respectively, try to achieve its detection probability respectively, and change the signal to noise ratio of data, multi-group data is detected, obtain the detection probability under different signal to noise ratio, show its detection perform.

2. signal detecting method according to claim 1, is characterized in that comprising further:

Set up self-adaption constant false alarm rate detector model, the data comprised forming matrix are carried out part deletion, then are obtained threshold value, finally compare detection unit and threshold value, draw testing result,

Wherein said part is deleted and is comprised:

(1), to the i-th row sampling sequence x of reference unit ₍₁₎≤ x ₍₂₎≤ ... ≤ x _(R), wherein x is reference unit value, and R is total reference unit number;

(2), x is used ₍₁₎represent clutter sampling, compare x ₍₂₎and T ₁x ₍₁₎, wherein T ₁meet given wrong probability of erasure P _fCthe normalizing factor of threshold value, if x ₍₂₎<T ₁x ₍₁₎, judge x ₍₁₎and x ₍₂₎be same profile samples, then carry out next step; If x ₍₂₎>T ₁x ₍₁₎, judge x ₍₂₎..., x _(R)the echo signal in strong clutter region, by that analogy, by i-th row all dividing elements Chu Qiang clutter district and weak clutter district two-part;

(3), suppose that dividing position is k, if k>R/2, wherein R is reference unit number, then judge that detection unit is in weak clutter district, by x ₍₁₎..., x _(k)form clutter power Z; If k<R/2, then judge that detection unit is in strong clutter district, by x _(k+1)..., x _(R)form clutter power Z;

(4), according to S=TZ obtain threshold value, wherein S is threshold value, and T is the normalizing factor, and when detection unit is greater than threshold value S, this detection unit should be judged to noise, and namely in optical fiber early warning, this signal should be reported to the police.

3. signal detecting method according to claim 2, is characterized in that comprising further:

The normalizing factor T of threshold value is obtained according to given wrong probability of erasure _k, the wherein normalizing factor T of threshold value _kcalculate according to following formula

$P_{\mathrm{FC}}=\left(\begin{array}{c}R\\ k\end{array}\right)\frac{1}{{[1+T_k(R-k)]}^k}---\left(1\right)$

Wherein, R is reference unit number, P _fCthe wrong probability of erasure of delete procedure kth step

P _FC＝Pr{D _k>0|H _N} (2)

Wherein, P _fCfor wrong probability of erasure, H _nfor weak clutter is sampled, D _kfor test statistics,

D _k＝x _(k+1)-T _kZ _k(3)

Wherein, x _(k+1)be (k+1) item sampled value, T _kfor the normalizing factor of kth item threshold value, Z _ksample in order for front k item and.

4. signal detecting method according to claim 2, is characterized in that normalizing factor T is according to false-alarm probability P _fatry to achieve, as shown in the formula

$P_{\mathrm{fa}}=\left\{\begin{array}{cc}\left(\begin{array}{c}R\\ k\end{array}\right)\underset{j=1}{\overset{k}{\mathrm{\&Pi;}}}{[T+\frac{R-j+1}{k-j+1}]}^2,k>R/2& \\ \left(\begin{array}{c}R\\ k\end{array}\right){(1+T)}^{-(N-k-1)}\&CenterDot;\underset{j=0}{\overset{k}{\mathrm{\&Sigma;}}}\left(\begin{array}{c}k\\ j\end{array}\right){(-1)}^j{(1+\frac{j}{N-k}+T)}^{-1},k\&le;R/2& \end{array}\right.---\left(4\right)$