CN101860866A - Intrusion detection and positioning method of non-uniform sensitivity nodes of anti-intrusion system sensing network - Google Patents

Abstract

The invention discloses an intrusion detection and positioning method of non-uniform sensitivity nodes of an anti-intrusion system sensing network. The intrusion detection and positioning method is characterized in that the existence of an intrusion signal is judged by using the related characteristic of a signal that the self-correlation time of a signal with intrusion is longer than that of a signal without intrusion, calculating a self-correlation function of the signal and comparing the signal correlation coefficient value at the non-zero delay tap. The method is suitable for detecting the non-uniform sensitivity sensing nodes without performing consistent software or hardware calibration, has extremely high detection and positioning accuracy and low error alarm rate, and is applicable to the anti-intrusion application of a large or long-distance circumferential rail made of many kinds of mixed materials. Moreover, by adopting the method, the multi-point weak intrusion can be accurately detected.

Description

The intrusion detection of non-uniform sensitivity nodes of anti-intrusion system sensing network and localization method

Technical field

The present invention relates to security and guard technology and sensing network signal processing method field, be specifically related to non-uniform sensitivity nodes of anti-intrusion system sensing network intrusion detection and localization method.

Background technology

In anti-intrusion system sensing network, no matter quasi-distributed array sensing node adopts traditional electric sensor, still adopt have anti-strong electromagnetic, passive Fibre Optical Sensor, because the difference of the difference of each sensor node of network self hardware condition and mounting condition, installation environment, fence material, cause each node transducer sensitivity inconsistent, or identical invasion signal is different in the output response at each node place, influence is the accurate detection and the location of invasion signal truly, therefore need carry out the consistency calibration of hardware and software to each sensing node.Disclosed China applied for a patent on October 22nd, 2008: 200810059962.9 be exactly for address this problem proposition at a kind of consistency calibration method that shakes electric sensor.Yet along with the increase of anti-intrusion detection circumference or the expansion of sensing network scale, the sensor node number can increase greatly, consistency calibration method not only bother and cost higher, and the demarcation back is along with the growth of monitoring time, each node transducer self-condition changes, or the difference of the installation environment consistency variation gradually still that causes each node in the sensing network, system needs carry out consistency calibration at set intervals again, otherwise can influence the accuracy that anti-intrusion system detects equally.Especially for optical pickocff, the transducer sensitivity height, carry out relatively difficulty of consistency calibration, cost is higher, and the uneven sensing node noise energy of sensitivity difference is bigger, and change in time, if the noise energy at certain node place and faint invasion signal energy quite or greater than the invasion signal energy, can cause omission or flase drop situation.Invade situation because inhomogeneous being difficult to more of sensitivity judged and discern for multiple spot.

Summary of the invention

Problem to be solved by this invention is: intrusion detection and localization method that how a kind of non-uniform sensitivity nodes of anti-intrusion system sensing network is provided, this method can overcome existing defective in the prior art, need not to carry out consistency software or hardware demarcates, the accuracy rate height that detects, the alert rate of mistake is low.

Technical problem proposed by the invention is to solve like this: intrusion detection and localization method that a kind of non-uniform sensitivity nodes of anti-intrusion system sensing network is provided, it is characterized in that, utilize the correlation properties of signal self, the signal autocorrelation time span that invasion is promptly arranged is greater than the signal autocorrelation time span of not having invasion, auto-correlation function by signal calculated, relatively the signal correction coefficient value at non-zero time delay tap place is distinguished having or not of invasion signal.The signal that does not have invasion mainly is made of various noises such as system, environment, no matter its amplitude intensity and energy have much, a little less than the signal correlation, the downward trend of signal auto-correlation function curve is very fast, except time delay is that the signal autocorrelation coefficient value at 0 place is 1, the auto-correlation coefficient value at other time delay tap places all is close to 0; And the signal of invasion is arranged, the power of signal no matter, the correlation of signal self is stronger, and the downward trend of signal auto-correlation function curve is slow, the auto-correlation coefficient value at the time delay tap place non-0 is much larger than 0, and approaching 1 near 0 time delay tap place auto-correlation coefficient.

If the transducing signal array that receives is X=x _i(n), i=1,2 ..., N; N=1,2 ..., M}, N are the node number of sensing network, and M is the analysis length of each node signal, and the normalized autocorrelation functions of each node signal is C _i={ E _i(x (n), x (n-l)), l=0,1 ..., M}, l are time delay, unit is a sampled point, | C _i|≤1; If the I node is not for there being the node of invasion, signal time auto-correlation length is not l when having invasion ₁, the J node is the node that invasion is arranged, signal time auto-correlation length was l when invasion was arranged ₂, the two, length relation always had 0≤l correlation time ₁＜l ₂

Select a definite time delay l ₁＜L＜l ₂, normalizated correlation coefficient E must be arranged ₁(x (n), x (n-L))＜E _J(x (n), x (n-L)), and have

Or E _J(x (n), x (n-L))＞＞0.Set a normalized autocorrelation coefficient threshold value η _c, 0.5＜η _c＜1, can judge whether the invasion signal exists according to following formula:

$R_i=\left\{\begin{array}{cc}1,& \left|E_i(x\left(n\right),x(n-L))\right|\&GreaterEqual;{\mathrm{\&eta;}}_{\mathrm{<figure-callout\; id="0"\; label="c"\; filenames="HSA00000132368400011.png"\; state="\{\{state\}\}">c</figure-callout>}}\\ 0,& \left|E_i(x\left(n\right),x(n-L))\right|<{\mathrm{\&eta;}}_c\end{array}\right.---\left(1\right)$

When

The time, judge not invasion of circumference; When

The time, judge that circumference has invasion, signal that invasion is arranged is judged it is to be caused or environmental change causes by artificial invasion according to its frequecy characteristic or the sensing node number that is influenced and scope again; To defining artificial invasion after the environmental factor eliminating, corresponding R _i≠ 0 node subscript i is the sensing node that invasion action is influenced, the location of invading according to the installation site of this node.

If Be the single-point invasion; If

To R _iContinuous several node i of=1 are carried out cluster, several nodes that can cluster judge it is that invasion action by a point causes, the number of the class of finally determining after the cluster is actual invasion and counts, and the node ordinal number in each class is asked on average round again, carries out the location of multiple spot invasion.

Useful technique effect of the present invention is, avoid the trouble of multinode consistency calibration in the sensing network, save installation, maintenance and the whole system operating cost of anti-intrusion system greatly, and this method is from distinguishing true invasion and nothing invasion signal in essence, system's intrusion detection highly sensitive is fit to the faint invasion input of hard fence.After defining the existence of invasion signal, artificially invade the difference with environmental change such as wind influence again, this method detects and the accuracy rate height of location invasion, and the alert rate of mistake is low, the accurate differentiation of suitable multiple spot invasion, and be applicable to the anti-invasion application of circumference fence that various material is mixed.

Description of drawings

Fig. 1 is the signal auto-correlation function comparison diagram of one embodiment of the invention.

Fig. 2 is an embodiment of invasion detecting device of the present invention, based on the fiber fence anti-intrusion system of quasi-distributed FBG transducer.

Embodiment

Below in conjunction with accompanying drawing the present invention is further described:

The invention provides a kind of intrusion detection and localization method of non-uniform sensitivity nodes of anti-intrusion system sensing network.The present invention utilizes self correlation properties of invasion signal to discern the existence of having or not of invasion or faint invasion, makes each node of anti-intrusion system sensing network need not to carry out consistency software or hardware is demarcated.Described signal self correlation properties are: length correlation time of signal self.The signal that does not have invasion mainly is made of various noises such as system, environment, no matter its amplitude intensity and energy have much, a little less than the signal correlation, the autocorrelative time is shorter, shown in Fig. 1 solid line, except time delay is that the auto-correlation coefficient at 0 place is 1, the auto-correlation coefficient at other time delay tap places all is close to 0, and signal auto-correlation function suddenly descends; And the signal of invasion is arranged, the power of signal no matter, the correlation of signal self is stronger, the signal auto-correlation function curve is slow downward trend, shown in Fig. 1 dotted line, the signal autocorrelation coefficient value much larger than 0, and is that 1 or 2 sample place auto-correlation coefficient values still approach 1 in time delay tap at non-0 time delay tap place.According to this difference, can compare the signal autocorrelation coefficient value at non-zero time delay tap place by the auto-correlation function of signal calculated, from distinguishing signal that invasion is arranged or the signal that does not have invasion in essence.This method is simple and practical, has stronger noise inhibiting ability, and is therefore less demanding to the preliminary treatment such as front end denoising of anti-intrusion system.

Be example with fiber fence anti-intrusion system below, in conjunction with the accompanying drawings specific implementation method of the present invention be further described based on quasi-distributed FBG transducer:

At first introduce the invasion detecting device that the specific embodiment of the invention one is adopted, as shown in Figure 2, by three the part form: be hung on the fence or be embedded in circumferentially under the armouring optical cable (be in series with N FBG Fibre Optical Sensor, can detect N point), be used for the vibration or the strain signal of invading on the perception circumference; The signal (FBG) demodulator is used to provide light source, and light signal is carried out demodulation, opto-electronic conversion and A/D analog-to-digital conversion; Warning system or processing host are used for the sensing network node signal that transmits is handled in real time, judge that invasion has or not and carries out sound and light alarm, provide and show intrusion detection and positioning result thereof.What next introduce is exactly specific implementation method of the present invention:

The light transducing signal that carries invasion information is transferred to (FBG) demodulator by optical cable, through demodulation, opto-electronic conversion and A/D analog-to-digital conversion, be transferred to processing host by Ethernet or serial ports form, main frame is handled the signal of all nodes of sensing network of receiving in real time, judges having or not and positioning according to the signal energy size of invasion.

To Signal Processing is key point of the present invention.Because each node sensitivity of sensing network is inhomogeneous, directly carry out energy and relatively judge having or not of invasion signal, cause false dismissal and mistake alert easily, and the preset threshold size is relevant with conditions such as fence material, weather, environment, be not easy to determine, therefore need a kind of new intrusion detection method.As embodiment two, the detection method of invasion is:

If the transducing signal array that receives is X={x _i(n), i=1,2 ..., N; N=1,2 ..., M}, N are the node number of sensing network, and M is the analysis length of each node signal, and the normalized autocorrelation functions of each node signal is C _i={ E _i(x (n), x (n-l)), l=0,1 ..., M}, l are time delay, unit is a sampled point, | C _i|≤1; If the I node is not for there being the node of invasion, signal time auto-correlation length is not l when having invasion ₁, the J node is the node that invasion is arranged, signal time auto-correlation length was l when invasion was arranged ₂, the two, length relation always had 0≤l correlation time ₁＜l ₂

Select a definite time delay l ₁＜L＜l ₂, normalizated correlation coefficient E must be arranged _I(x (n), x (n-L))＜E _J(x (n), x (n-L)), and have

Or E _J(x (n), x (n-L))＞＞0.Set a normalized autocorrelation coefficient threshold value η _c, 0.5＜η _c＜1, can judge whether the invasion signal exists according to following formula:

$R_i=\left\{\begin{array}{cc}1,& \left|E_i(x\left(n\right),x(n-L))\right|\&GreaterEqual;{\mathrm{\&eta;}}_{\mathrm{<figure-callout\; id="0"\; label="c"\; filenames="HSA00000132368400011.png"\; state="\{\{state\}\}">c</figure-callout>}}\\ 0,& \left|E_i(x\left(n\right),x(n-L))\right|<{\mathrm{\&eta;}}_c\end{array}\right.---\left(1\right)$

When

The time, judge not invasion of circumference; When The time, judge that circumference has invasion, signal that invasion is arranged is judged it is to be caused or environmental change causes by artificial invasion according to its frequecy characteristic or the sensing node number that is influenced and scope again; To defining artificial invasion after the environmental factor eliminating, corresponding R _i≠ 0 node subscript i is the sensing node that invasion action is influenced, the location of invading according to the installation site of this node.

If

Be the single-point invasion; If

To R _iContinuous several node i of=1 are carried out cluster, several nodes that can cluster judge it is that invasion action by a point causes, the number of the class of finally determining after the cluster is actual invasion and counts, and the node ordinal number in each class is asked on average round again, carries out the location of multiple spot invasion.

The specific implementation method of the anti-intrusion system of enumerating in the embodiment of the invention that is based on quasi-distributed FBG Fibre Optical Sensor, the signal processing method of this invention can be applied to based on other optics, electric class fully or mix in the anti-intrusion system of quasi-distributed sensing network.

Claims (3)

1. the intrusion detection of a non-uniform sensitivity nodes of anti-intrusion system sensing network and localization method, it is characterized in that, utilize the correlation properties of signal self, the signal autocorrelation time span that invasion is promptly arranged is greater than the signal autocorrelation time span of not having invasion, auto-correlation function by signal calculated, the signal correction coefficient value that compares non-zero time delay tap place, distinguish having or not of invasion signal:

The signal that does not have invasion: mainly constitute by system and ambient noise, no matter its amplitude intensity and energy have much, a little less than signal self correlation, the auto-correlation time span of signal is short, the downward trend of signal auto-correlation function curve is very fast, except time delay is that the signal autocorrelation coefficient value at 0 place is 1, the auto-correlation coefficient value at other time delay tap places all is close to 0;

The signal that invasion is arranged: the power of no matter invading signal, the correlation of this signal is strong, the auto-correlation time of signal is long, the downward trend of signal auto-correlation function curve is slow, much larger than 0, and still approach 1 in the auto-correlation coefficient value at non-0 time delay tap place at auto-correlation coefficient near 0 time delay tap place.

2. the intrusion detection of non-uniform sensitivity nodes of anti-intrusion system sensing network according to claim 1 and localization method is characterized in that:

If the transducing signal array that receives is X={x _i(n), i=1,2 ..., N; N=1,2 ..., M}, N are the node number of sensing network, and M is the analysis length of each node signal, and the normalized autocorrelation functions of each node signal is C _i={ E _i(x (n), x (n-l)), l=0,1 ..., M}, l are time delay, unit is a sampled point, | C _i|≤1; If the I node is not for there being the node of invasion, signal time auto-correlation length is not l when having invasion ₁, the J node is the node that invasion is arranged, signal time auto-correlation length was l when invasion was arranged ₂, the two, length relation always had 0≤l correlation time ₁＜l ₂

Select a definite time delay l ₁＜L＜l ₂, normalizated correlation coefficient E must be arranged _I(x (n), x (n-L))＜E _J(x (n), x (n-L)), and have

,

Or E _JA normalized autocorrelation coefficient threshold value η is set in (x (n), x (n-L))＞＞0 _c, 0.5＜η _c＜1, judge according to following formula whether the invasion signal exists:

$R_i=\left\{\begin{array}{cc}1,& \left|E_i(x\left(n\right),x(n-L))\right|\&GreaterEqual;{\mathrm{\&eta;}}_c\\ 0,& \left|E_i(x\left(n\right),x(n-L))\right|<{\mathrm{\&eta;}}_c\end{array}\right.---\left(1\right)$

When

The time, judge not invasion of circumference;

When

The time, judge that circumference has invasion, signal that invasion is arranged is judged it is to be caused or environmental change causes by artificial invasion according to its frequecy characteristic or the sensing node number that is influenced and scope again;

To defining artificial invasion after the environmental factor eliminating, corresponding R _i≠ 0 node subscript i is the sensing node that invasion action is influenced, the location of invading according to the installation site of this node.

3. the intrusion detection of non-uniform sensitivity nodes of anti-intrusion system sensing network according to claim 2 and localization method is characterized in that:

If

Be the single-point invasion;

If

To R _iContinuous several node i of=1 are carried out cluster, several nodes that can cluster judge it is that invasion action by a point causes, the number of the class of finally determining after the cluster is actual invasion and counts, and the node ordinal number in each class is asked on average round again, carries out the location of multiple spot invasion.

Images