CN105512994A - Fault-tolerant perimeter intruder detection method - Google Patents

Abstract

The invention discloses a fault-tolerant perimeter intruder detection method, including three links: detector inner fault tolerance, fault tolerance between detectors, and trust management and fault detector determination. In a first link, the detection output of a detector is obtained; in a second link, the alarm output of the detector is obtained; and in a third link, the credibility of the detector is dynamically adjusted according to results of the first and second links. The method can reduce false alarm and false dismissal in a perimeter intruder detection system caused by ambient noises and detector faults, and find faulted detectors, thereby allowing the system to work normally when a plurality of detectors break down, and substantially improving the fault-tolerant capability and reliability of the perimeter intruder detection system. The method can directly update hardware based on an original perimeter intruder detection system, and possess low enforcement costs. The method can be applied to borderlines, the surrounding of an airport, the surrounding of an electric power facility, coastlines, etc., and reduce dependence on artificial maintenance.

Description

A kind of fault tolerant perimeter intrusion detecting method

Technical field

The invention belongs to technical field of security and protection, particularly relate to a kind of fault tolerant perimeter intrusion detecting method.

Background technology

Perimeter intrusion detecting system, as the important component part of safety-protection system, is the important defence line together assured the safety for life and property of the people.The invasion result detected when personnel enter investigative range, can be transferred to the control center of system, notify staff, and can cause alarm in this locality by perimeter intrusion detecting system in time.Well deterrence and interception are served to intrusion behavior.

While perimeter intrusion detecting system obtains more and more widespread use, its performance also more and more receives publicity.The performance of perimeter intrusion detecting system is primarily of alert rate (false positive mistake) and false dismissed rate (False negative error) are weighed by mistake.Alert rate characterizes detection system when not invading generation and sends the probability of warning by mistake; False dismissed rate characterizes the probability that when invasion occurs, detection system is not reported to the police.This two indices directly affects the fastness in this security of the lives and property defence line, road, and reducing false dismissal, by mistake police is the important goal that perimeter intrusion detecting system is improved.

Summary of the invention

The object of the invention is to for the deficiencies in the prior art, a kind of fault tolerant perimeter intrusion detecting method is provided.

The object of the invention is to be achieved through the following technical solutions: a kind of fault tolerant perimeter intrusion detecting method, the method comprises the following steps:

(1) fault-tolerant in detector: by the result of detection weighted sum in this detector a period of time, to obtain converging the long and, and the detection as detector after this result binaryzation is exported;

(2) fault-tolerant between detector: to infer by the result of detection Nearest Neighbor with Weighted Voting of adjacent detector current time the position that invader is current, and inferred the position in invader's upper moment on this basis.The position in invader by inference upper moment, is weighted ballot to moment result of detection on adjacent detector again.After the result summation of twice Nearest Neighbor with Weighted Voting, through binaryzation, the warning as detector exports.Described weights are relevant with the degree of belief of detector.

(3) trust management and fault detector judge: the warning of the detector that the detection output of the detector obtained according to step 1 and step 2 obtain exports, the degree of belief of dynamic conditioning detector, when the detection output of detector and the warning Output rusults of detector consistent time, increase the degree of belief of detector; When the detection output of detector and the warning Output rusults of detector inconsistent time, reduce the degree of belief of detector; When the degree of belief of detector is lower than degree of belief threshold value, judge this breakdown detector.

Further, described step (1) is specific as follows:

Note detector i is c at the result of detection of moment t _it (), detector weighting on the basis of present detection results gathers the result of detection in L moment in the past, obtains summarized results that is:

$\stackrel{\‾}{c_i\left(t\right)}=\underset{l=0}{\overset{L}{\Σ}}\mathrm{\α}(t-l\×T_0)c_i(t-l\×T_0)$

Wherein t ₀for the work period of detector.

Detection is used to export threshold value Th _dby summarized results binaryzation, obtains a boolean results, and the detection as detector exports O _i(t), that is:

$O_i\left(t\right)=\left\{\begin{array}{c}1,\stackrel{\&OverBar;}{c_i\left(t\right)}\&GreaterEqual;{\mathrm{Th}}_D\\ 0,\stackrel{\&OverBar;}{c_i\left(t\right)}<{\mathrm{Th}}_D\end{array}\right.$

Wherein 1 represents that detector detection exports as reporting to the police, and 0 represents that detector detection exports as mourning in silence.

Further, described step (2) is specific as follows:

If have 1 P can be detected by k detector, and the probability that each detector detects this point is all more than or equal to δ, then this P is detected device network (k, δ) and re-covers lid.In like manner, also claim detector network to meet (k, δ) at P point and re-cover lid.If any point in the Ω of region all meets (k, δ) re-cover lid, then claim region Ω to meet (k, δ) and re-cover lid.

If there is identical (k, δ) heavy overlay area, then these two detectors neighbours' detector each other in two detectors.The set of neighbours' detector of detector i is defined as N _i.

(k, δ) being re-covered lid Region dividing is subregion S set R={SR ₁, SR ₂..., every sub regions is all covered by identical detector set.The detector set covering identical subregion SR with the detection probability being more than or equal to δ is designated as remember that having (k, the δ) overlapped to weigh overlay area with the search coverage of detector i is V _i.For meeting subregion, neighbours' detector set of calculating detector i voting results O _{i, spa}(SR):

Wherein r _j(t-T ₀) for detector j is at (t-T ₀) trust weight in moment;

If invader appears in certain sub regions in t, then multiple detectors that invader may be capped this subregion detect simultaneously, and the Nearest Neighbor with Weighted Voting result of this subregion is the maximal value in neighbouring subregion.Namely invader should at maximum O _{i, spa}(SR) subregion corresponding to, is designated as subregion

If the set of the subregion that invader may be in the upper moment is suppose that the maximal rate of invader is v _max, then calculate by following formula:

Wherein $|SR-{\mathrm{SR}}_{i,spa}^{*}|={\min }_{P_1\&Element;SR.P{2}_{\&Element;}{\mathrm{SR}}_{i,spa}^{*}}|P_1-P_2|.$

For meeting subregion, neighbours' detector set of calculating detector i the voting results O in a upper moment _{i, tem}(SR):

Then in a upper moment, the position that invader may locate should be O of upper moment _{i, tem}(SR) obtain the subregion of maximal value, be designated as

Final summarized results as follows:

$\stackrel{\&OverBar;}{O_i\left(t\right)}=O_i\left(t\right)+O_{i,spa}\left({\mathrm{SR}}_{i,spa}^{*}\right)+O_{i,tem}\left({\mathrm{SR}}_{i,tem}^{*}\right)$

Adopt threshold determination pair carry out binaryzation, note decision threshold is AlarmTh, and the warning of note detector exports as A _i(t), then whether detector is reported to the police and can be determined by following formula:

$A_i\left(t\right)=\left\{\begin{array}{c}1,\stackrel{\&OverBar;}{O_i\left(t\right)}\&GreaterEqual;AlarmTh\\ 0,\stackrel{\&OverBar;}{O_i\left(t\right)}<AlarmTh\end{array}\right.$

Wherein, A _it ()=1 represents that detector i reports to the police, refer to detect invader; A _it ()=0 represents that detector i mourns in silence, refer to not find invader.

When detector i sends alerting signal to control center, if there is A _jt ()=1 meets then the output of this detector is judged to be represent that this detector is correctly reported to the police; Otherwise the output of this detector is judged to be report to the police with representing this detector error.

When detector i sends silent signal to control center, if there is A _jt ()=1 meets then the output of this detector is judged to be mourn in silence with representing this detector error; Otherwise the output of this detector is judged to be represent that this detector is correctly mourned in silence.

Control center accepts the detection information that on circumference, numerous detector returns simultaneously, if fault-tolerant between fault-tolerant and detector in detector, has detector A (t)=1, then system provides a place and reports to the police; If there is multiple detector to meet A (t)=1, then system provides many places warning.

Further, described step (3) is specific as follows:

Keep the misdeed of mourning in silence to be called SilentFault behavior when detector should be sent warning, the behavior sending warning when this being mourned in silence is called AlarmFault behavior.

Note is α (α >0) to the penalty value of SilentFault behavior, when namely each SilentFault behavior occurs, the degree of belief of this detector deducts α.So, when during generation, α/P (A should be deducted in the degree of belief of this detector _i(t)=1), this value is designated as α _s.In like manner, when during generation, the degree of belief of this detector deducts α/P (A _i(t)=0), be denoted as α _a.That is:

α _s＝α/P(A _i(t)＝1)

α _a＝α/P(A _i(t)＝0)

Correctly export behavior for detector, each degree of belief of this detector of giving increases β (β >0).When during generation, the degree of belief of detector increases β/P (A _i(t)=1), be denoted as β _a; When during generation, the degree of belief of this detector increases β/P (A _i(t)=0), be denoted as β _s.That is:

β _a＝β/P(A _i(t)＝1)

β _s＝β/P(A _i(t)＝0)

The average general P (R of invasion is there is by detector search coverage in detector working environment _i(t)=1) estimate P (A _i(t)=1).

Detector i is defined as tr in the degree of belief of moment t _i(t).At each work period T ₀, according to the degree of belief of the output result of determination adjustment detector of detector.

${\mathrm{tr}}_i\left(t\right)=\left\{\begin{array}{c}t{r}_i(t-T_0)-{\mathrm{\&alpha;}}_a,if{I}_i^a\left(t\right)=0\\ t{r}_i(t-T_0)-{\mathrm{\&alpha;}}_s,if{I}_i^s\left(t\right)=0\\ t{r}_i(t-T_0)+{\mathrm{\&beta;}}_a,if{I}_i^a\left(t\right)=1\\ t{r}_i(t-T_0)+{\mathrm{\&beta;}}_s,if{I}_i^s\left(t\right)=1\end{array}\right.$

The detector degree of belief value upper limit is made to be TR ₀, represent that this detector is reliable; Lower limit is 0, represents this breakdown detector.That is:

${\mathrm{tr}}_i\left(t\right)=\left\{\begin{array}{c}T{R}_0,ift{r}_i\left(t\right)>T{R}_0\\ 0,if{\mathrm{tr}}_i\left(t\right)<0\\ t{r}_i\left(t\right),else\end{array}\right.$

If detector i is at the degree of belief tr of moment t _it () is lower than failure determination threshold TR _th, then think that detector i is judged as fault detector at moment t, be designated as m _i(t)=1; If detector i is normal in t, be then designated as m _i(t)=0.That is:

$m_i\left(t\right)=\left\{\begin{array}{c}1,ift{r}_i\left(t\right)<T{R}_{th}\\ 0,others\end{array}\right.$

By detector degree of belief tr _ithe ballot weight of t this detector that () determines fault-tolerant link between detector, is called trust weight r _i(t).Trust weight is provided by following formula:

$r_i\left(t\right)=\left\{\begin{array}{c}\frac{{\mathrm{tr}}_i\left(t\right)}{{\mathrm{TR}}_0},if{m}_i\left(t\right)=0\\ 0,if{m}_i\left(t\right)=1\end{array}\right.$ 。

The invention has the beneficial effects as follows: the inventive method can reduce the mistake police and false dismissal that are caused by neighbourhood noise, breakdown detector in perimeter intrusion detecting system, and can fault detector be found out, system still normally can be run when numerous detector breaks down, greatly improve perimeter intrusion detecting system survivability and reliability.The inventive method can directly based on original perimeter intrusion detecting system hardware upgrading, and implementation cost is low.The inventive method can be applicable to the places such as boundary line, airport periphery, electric power facility periphery, shore line, reduces the dependence to manual maintenance.

Accompanying drawing explanation

Fig. 1 is fault tolerant perimeter intrusion detecting method flow diagram of the present invention;

Fig. 2 is (3, δ) heavy overlay area subregion schematic diagram;

Fig. 3 is invader's two adjacent moment position views.

Embodiment

Below in conjunction with the drawings and specific embodiments, the present invention is described in further detail.

As shown in Figure 1, a kind of fault tolerant perimeter intrusion detecting method provided by the invention, the method comprising the steps of:

(1) fault-tolerant in detector: by the result of detection weighted sum in this detector a period of time, to obtain converging the long and, and the detection as detector after this result binaryzation is exported; Be specially:

Note detector i is c at the result of detection of moment t _it (), detector weighting on the basis of present detection results gathers the result of detection in L moment in the past, obtains summarized results that is:

$\stackrel{\&OverBar;}{c_i\left(t\right)}=\underset{l=0}{\overset{L}{\&Sigma;}}\mathrm{\&alpha;}(t-l\&times;T_0)c_i(t-l\&times;T_0)$

Wherein t ₀for the work period of detector.

Through converging the long and be a serial number, the final detection of detector exports the Boolean that should be and characterize with or without invasion, uses detection to export threshold value Th _dby summarized results binaryzation, obtains a boolean results, and the detection as detector exports O _i(t), that is:

$O_i\left(t\right)=\left\{\begin{array}{c}1,\stackrel{\&OverBar;}{c_i\left(t\right)}\&GreaterEqual;{\mathrm{Th}}_D\\ 0,\stackrel{\&OverBar;}{c_i\left(t\right)}<{\mathrm{Th}}_D\end{array}\right.$

Wherein 1 represents that detector detection exports as reporting to the police, and 0 represents that detector detection exports as mourning in silence.

(2) fault-tolerant between detector: to infer by the result of detection Nearest Neighbor with Weighted Voting of adjacent detector current time the position that invader is current, and inferred the position in invader's upper moment on this basis.The position in invader by inference upper moment, is weighted ballot to moment result of detection on adjacent detector again.After the result summation of twice Nearest Neighbor with Weighted Voting, through binaryzation, the warning as detector exports.Described weights are relevant with the degree of belief of detector.Be specially:

If have 1 P can be detected by k detector, and the probability that each detector detects this point is all more than or equal to δ, then this P is detected device network (k, δ) and re-covers lid.In like manner, also claim detector network to meet (k, δ) at P point and re-cover lid.If any point in the Ω of region all meets (k, δ) re-cover lid, then claim region Ω to meet (k, δ) and re-cover lid.

If there is identical (k, δ) heavy overlay area, then these two detectors neighbours' detector each other in two detectors.The set of neighbours' detector of detector i is defined as N _i.

(k, δ) being re-covered lid Region dividing is subregion S set R={SR ₁, SR ₂..., every sub regions is all covered by identical detector set.The detector set covering identical subregion SR with the detection probability being more than or equal to δ is designated as it is a sub regions of (3, δ) heavy overlay area in each dashed polygon frame in Fig. 2.

Remember that having (k, the δ) overlapped to weigh overlay area with the search coverage of detector i is V _i.For meeting subregion, neighbours' detector set of calculating detector i voting results O _{i, spa}(SR):

Wherein r _j(t-T ₀) for detector j is at (t-T ₀) trust weight in moment;

If invader appears in certain sub regions in t, then multiple detectors that invader may be capped this subregion detect simultaneously, and the Nearest Neighbor with Weighted Voting result of this subregion is the maximal value in neighbouring subregion.Namely invader should at maximum O _{i, spa}(SR) subregion corresponding to, is designated as subregion

If the set of the subregion that invader may be in the upper moment is suppose that the maximal rate of invader is v _max, then calculate by following formula:

Wherein as shown in Figure 3, if invader is in little dotted line frame region in the moment, then go up the position of invader at quarter in a period of time in large dotted line frame region.

Next, calculate the result after the temporal correlation considering detector i, for meeting subregion, neighbours' detector set of calculating detector i the voting results O in a upper moment _{i, tem}(SR):

Then in a upper moment, the position that invader may locate should be O of upper moment _{i, tem}(SR) obtain the subregion of maximal value, be designated as

The result of temporal correlation and spatial coherence is gathered with the result of detection of self, obtains final summarized results

$\stackrel{\&OverBar;}{O_i\left(t\right)}=O_i\left(t\right)+O_{i,spa}\left({\mathrm{SR}}_{i,spa}^{*}\right)+O_{i,tem}\left({\mathrm{SR}}_{i,tem}^{*}\right)$

It is the Boolean whether a sign reports to the police equally that the warning of detector exports, and the same mode of threshold determination that adopts carries out binaryzation.Namely decision threshold is AlarmTh, and the warning of note detector exports as A _i(t), then whether detector is reported to the police and can be determined by following formula:

$A_i\left(t\right)=\left\{\begin{array}{c}1,\stackrel{\&OverBar;}{O_i\left(t\right)}\&GreaterEqual;AlarmTh\\ 0,\stackrel{\&OverBar;}{O_i\left(t\right)}<AlarmTh\end{array}\right.$

Wherein, A _it ()=1 represents that detector i reports to the police, refer to detect invader; A _it ()=0 represents that detector i mourns in silence, refer to not find invader.

A _it () is using correcting errors of exporting as each detector detection of criterion.

When detector i sends alerting signal to control center, if there is A _jt ()=1 meets then the output of this detector is judged to be represent that this detector is correctly reported to the police; Otherwise the output of this detector is judged to be report to the police with representing this detector error.

When detector i sends silent signal to control center, if there is A _jt ()=1 meets then the output of this detector is judged to be mourn in silence with representing this detector error; Otherwise the output of this detector is judged to be represent that this detector is correctly mourned in silence.

Control center accepts the detection information that on circumference, numerous detector returns simultaneously, if fault-tolerant between fault-tolerant and detector in detector, has detector A (t)=1, then system provides a place and reports to the police; If there is multiple detector to meet A (t)=1, then system provides many places warning.Therefore this system may be used for the many places invasion in detection perimeter and can not omit, instead of can only detect place invasion.

(3) trust management and fault detector judge: the warning of the detector that the detection output of the detector obtained according to step 1 and step 2 obtain exports, the degree of belief of dynamic conditioning detector, when the detection output of detector and the warning Output rusults of detector consistent time, increase the degree of belief of detector; When the detection output of detector and the warning Output rusults of detector inconsistent time, reduce the degree of belief of detector; When the degree of belief of detector is lower than degree of belief threshold value, judge this breakdown detector.Be specially:

Keep the misdeed of mourning in silence to be called SilentFault behavior when detector should be sent warning, the behavior sending warning when this being mourned in silence is called AlarmFault behavior.

SilentFault behavior and AlarmFault behavior should catch hell.Note is α (α >0) to the penalty value of SilentFault behavior, when namely each SilentFault behavior occurs, the degree of belief of this detector deducts α.So, when during generation, α/P (A should be deducted in the degree of belief of this detector _i(t)=1), this value is designated as α _s.In like manner, when during generation, the degree of belief of this detector deducts α/P (A _i(t)=0), be denoted as α _a.That is:

α _s＝α/P(A _i(t)＝1)

α _a＝α/P(A _i(t)＝0)

Same, correctly export behavior for detector, each degree of belief of this detector of giving increases β (β >0).When during generation, the degree of belief of detector increases β/P (A _i(t)=1), be denoted as β _a; When during generation, the degree of belief of this detector increases β/P (A _i(t)=0), be denoted as β _s.That is:

β _a＝β/P(A _i(t)＝1)

β _s＝β/P(A _i(t)＝0)

More than process relates to P (A _i(t)=1), but it is not known, and the average general P (R of invasion can be occurred by detector search coverage in detector working environment _i(t)=1) estimate it, namely think P (A _i(t)=1)=P (R _i(t)=1).

Detector i is defined as tr in the degree of belief of moment t _i(t).At each work period T ₀, according to the degree of belief of the output result of determination adjustment detector of detector.

${\mathrm{tr}}_i\left(t\right)=\left\{\begin{array}{c}t{r}_i(t-T_0)-{\mathrm{\&alpha;}}_a,if{I}_i^a\left(t\right)=0\\ t{r}_i(t-T_0)-{\mathrm{\&alpha;}}_s,if{I}_i^s\left(t\right)=0\\ t{r}_i(t-T_0)+{\mathrm{\&beta;}}_a,if{I}_i^a\left(t\right)=1\\ t{r}_i(t-T_0)+{\mathrm{\&beta;}}_s,if{I}_i^s\left(t\right)=1\end{array}\right.$

If do not limited the scope of the degree of belief of detector, then after a period of operation, degree of belief value may become very high to normal detector.Thus, when this detector breaks down, need just to make the degree of belief of this detector reduce for a long time, be unfavorable for this fault detector of Timeliness coverage.Therefore, need detector degree of belief value capping.In like manner, we are also its setting lower limit.The detector degree of belief value upper limit is made to be TR ₀, represent that this detector is reliable; Lower limit is 0, represents this breakdown detector.That is:

${\mathrm{tr}}_i\left(t\right)=\left\{\begin{array}{c}{\mathrm{TR}}_0,if{t}_i\left(t\right)>{\mathrm{TR}}_0\\ 0,if{\mathrm{tr}}_i\left(t\right)<0\\ {\mathrm{tr}}_i\left(t\right),else\end{array}\right.$

No matter there is the fault of which kind of type in detector, and all can cause the decline of its degree of belief because of its SilentFault behavior or AlarmFault behavior, therefore, all faults all can be embodied by degree of belief.If detector i is at the degree of belief tr of moment t _it () is lower than failure determination threshold TR _th, then think that detector i is judged as fault detector at moment t, be designated as m _i(t)=1; If detector i is normal in t, be then designated as m _i(t)=0.That is:

$m_i\left(t\right)=\left\{\begin{array}{c}1,ift{r}_i\left(t\right)<T{R}_{th}\\ 0,others\end{array}\right.$

The detector with higher degree of belief is more reliable, and therefore the ballot weight of each detector fault-tolerant link between detector can be embodied by the degree of belief of this detector.

By detector degree of belief tr _ithe ballot weight of t this detector that () determines fault-tolerant link between detector, is called trust weight r _i(t).Between detector in fault-tolerant link, fault detector does not participate in voting (trust weight is 0).Only have when detector is normal, this detector just can participate in ballot.Trust weight is provided by following formula:

$r_i\left(t\right)=\left\{\begin{array}{c}\frac{{\mathrm{tr}}_i\left(t\right)}{{\mathrm{TR}}_0},if{m}_i\left(t\right)=0\\ 0,if{m}_i\left(t\right)=1\end{array}\right.$ 。

Claims (4)

1. a fault tolerant perimeter intrusion detecting method, is characterized in that, the method comprises the following steps:

(1) fault-tolerant in detector: by the result of detection weighted sum in this detector a period of time, to obtain converging the long and, and the detection as detector after this result binaryzation is exported;

(2) fault-tolerant between detector: to infer by the result of detection Nearest Neighbor with Weighted Voting of adjacent detector current time the position that invader is current, and inferred the position in invader's upper moment on this basis.The position in invader by inference upper moment, is weighted ballot to moment result of detection on adjacent detector again.After the result summation of twice Nearest Neighbor with Weighted Voting, through binaryzation, the warning as detector exports.Described weights are relevant with the degree of belief of detector;

(3) trust management and fault detector judge: the warning of the detector that the detection output of the detector obtained according to step 1 and step 2 obtain exports, the degree of belief of dynamic conditioning detector, when the detection output of detector and the warning Output rusults of detector consistent time, increase the degree of belief of detector; When the detection output of detector and the warning Output rusults of detector inconsistent time, reduce the degree of belief of detector; When the degree of belief of detector is lower than degree of belief threshold value, judge this breakdown detector.

2. a kind of fault tolerant perimeter intrusion detecting method according to claim 1, it is characterized in that, described step (1) is specific as follows:

Note detector i is c at the result of detection of moment t _it (), detector weighting on the basis of present detection results gathers the result of detection in L moment in the past, obtains summarized results that is:

$\stackrel{\&OverBar;}{c_i\left(t\right)}=\underset{l=0}{\overset{L}{\&Sigma;}}\mathrm{\&alpha;}(t-l\&times;T_0)c_i(t-l\&times;T_0)$

Wherein t ₀for the work period of detector.

Detection is used to export threshold value Th _dby summarized results binaryzation, obtains a boolean results, and the detection as detector exports O _i(t), that is:

$O_i\left(t\right)=\left\{\begin{array}{c}1,\stackrel{\&OverBar;}{c_i\left(t\right)}\&GreaterEqual;T{h}_D\\ 0,\stackrel{\&OverBar;}{c_i\left(t\right)}<{\mathrm{Th}}_D\end{array}\right.$

Wherein 1 represents that detector detection exports as reporting to the police, and 0 represents that detector detection exports as mourning in silence.

3. a kind of fault tolerant perimeter intrusion detecting method according to claim 1, it is characterized in that, described step (2) is specific as follows:

If have 1 P can be detected by k detector, and the probability that each detector detects this point is all more than or equal to δ, then this P is detected device network (k, δ) and re-covers lid.In like manner, also claim detector network to meet (k, δ) at P point and re-cover lid.If any point in the Ω of region all meets (k, δ) re-cover lid, then claim region Ω to meet (k, δ) and re-cover lid.

If there is identical (k, δ) heavy overlay area, then these two detectors neighbours' detector each other in two detectors.The set of neighbours' detector of detector i is defined as N _i.

(k, δ) being re-covered lid Region dividing is subregion S set R={SR ₁, SR ₂..., every sub regions is all covered by identical detector set.The detector set covering identical subregion SR with the detection probability being more than or equal to δ is designated as remember that having (k, the δ) overlapped to weigh overlay area with the search coverage of detector i is V _i.For meeting subregion, neighbours' detector set of calculating detector i voting results O _{i, spa}(SR):

Wherein r _j(t-T ₀) for detector j is at (t-T ₀) trust weight in moment;

If invader appears in certain sub regions in t, then multiple detectors that invader may be capped this subregion detect simultaneously, and the Nearest Neighbor with Weighted Voting result of this subregion is the maximal value in neighbouring subregion.Namely invader should at maximum O _{i, spa}(SR) subregion corresponding to, is designated as subregion

If the set of the subregion that invader may be in the upper moment is suppose that the maximal rate of invader is v _max, then calculate by following formula:

Wherein $|SR-{\mathrm{SR}}_{i,spa}^{*}|={\min }_{P_1\&Element;SR,P{2}_{\&Element;}{\mathrm{SR}}_{{}^{i,spa}}^{*}}|P_1-P_2|.$

For meeting subregion, neighbours' detector set of calculating detector i the voting results O in a upper moment _{i, tem}(SR):

Then in a upper moment, the position that invader may locate should be O of upper moment _{i, tem}(SR) obtain the subregion of maximal value, be designated as

Final summarized results as follows:

$\stackrel{\&OverBar;}{O_i\left(t\right)}=O_i\left(t\right)+O_{i,spa}\left({\mathrm{SR}}_{i,spa}^{*}\right)+O_{i,tem}\left({\mathrm{SR}}_{i,tem}^{*}\right)$

Adopt threshold determination pair carry out binaryzation, note decision threshold is AlarmTh, and the warning of note detector exports as A _i(t), then whether detector is reported to the police and can be determined by following formula:

$A_i\left(t\right)=\left\{\begin{array}{c}1,\stackrel{\&OverBar;}{O_i\left(t\right)}\&GreaterEqual;AlarmTh\\ 0,\stackrel{\&OverBar;}{O_i\left(t\right)}<AlarmTh\end{array}\right.$

Wherein, A _it ()=1 represents that detector i reports to the police, refer to detect invader; A _it ()=0 represents that detector i mourns in silence, refer to not find invader.

When detector i sends alerting signal to control center, if there is A _jt ()=1 meets then the output of this detector is judged to be represent that this detector is correctly reported to the police; Otherwise the output of this detector is judged to be report to the police with representing this detector error.

When detector i sends silent signal to control center, if there is A _jt ()=1 meets then the output of this detector is judged to be mourn in silence with representing this detector error; Otherwise the output of this detector is judged to be represent that this detector is correctly mourned in silence.

Control center accepts the detection information that on circumference, numerous detector returns simultaneously, if fault-tolerant between fault-tolerant and detector in detector, has detector A (t)=1, then system provides a place and reports to the police; If there is multiple detector to meet A (t)=1, then system provides many places warning.

4. a kind of fault tolerant perimeter intrusion detecting method according to claim 1, it is characterized in that, described step (3) is specific as follows:

Keep the misdeed of mourning in silence to be called SilentFault behavior when detector should be sent warning, the behavior sending warning when this being mourned in silence is called AlarmFault behavior.

Note is α (α >0) to the penalty value of SilentFault behavior, when namely each SilentFault behavior occurs, the degree of belief of this detector deducts α.So, when during generation, α/P (A should be deducted in the degree of belief of this detector _i(t)=1), this value is designated as α _s.In like manner, when during generation, the degree of belief of this detector deducts α/P (A _i(t)=0), be denoted as α _a.That is:

α _s＝α/P(A _i(t)＝1)

α _a＝α/P(A _i(t)＝0)

Correctly export behavior for detector, each degree of belief of this detector of giving increases β (β >0).When during generation, the degree of belief of detector increases β/P (A _i(t)=1), be denoted as β _a; When during generation, the degree of belief of this detector increases β/P (A _i(t)=0), be denoted as β _s.That is:

β _a＝β/P(A _i(t)＝1)

β _s＝β/P(A _i(t)＝0)

The average general P (R of invasion is there is by detector search coverage in detector working environment _i(t)=1) estimate P (A _i(t)=1).

Detector i is defined as tr in the degree of belief of moment t _i(t).At each work period T ₀, according to the degree of belief of the output result of determination adjustment detector of detector.

${\mathrm{tr}}_i\left(t\right)=\left\{\begin{array}{c}{\mathrm{tr}}_i(t-T_0)-{\mathrm{\&alpha;}}_a,if{I}_i^a\left(t\right)=0\\ {\mathrm{tr}}_i(t-T_0)-{\mathrm{\&alpha;}}_s,if{I}_i^s\left(t\right)=0\\ {\mathrm{tr}}_i(t-T_0)+{\mathrm{\&beta;}}_a,if{I}_i^a\left(t\right)=1\\ {\mathrm{tr}}_i(t-T_0)+{\mathrm{\&beta;}}_s,if{I}_i^s\left(t\right)=1\end{array}\right.$

The detector degree of belief value upper limit is made to be TR ₀, represent that this detector is reliable; Lower limit is 0, represents this breakdown detector.That is:

${\mathrm{tr}}_i\left(t\right)=\left\{\begin{array}{c}T{R}_0,ift{r}_i\left(t\right)>T{R}_0\\ 0,if{\mathrm{tr}}_i\left(t\right)<0\\ t{r}_i\left(t\right),else\end{array}\right.$

If detector i is at the degree of belief tr of moment t _it () is lower than failure determination threshold TR _th, then think that detector i is judged as fault detector at moment t, be designated as m _i(t)=1; If detector i is normal in t, be then designated as m _i(t)=0.That is:

$m_i\left(t\right)=\left\{\begin{array}{c}1,ift{r}_i\left(t\right)<T{R}_{th}\\ 0,others\end{array}\right.$

By detector degree of belief tr _ithe ballot weight of t this detector that () determines fault-tolerant link between detector, is called trust weight r _i(t).Trust weight is provided by following formula:

$r_i\left(t\right)=\left\{\begin{array}{c}\frac{{\mathrm{tr}}_i\left(t\right)}{{\mathrm{TR}}_0},if{m}_i\left(t\right)=0\\ 0,if{m}_i\left(t\right)=1\end{array}\right..$ 。