Target event positioning method and system for simultaneous adjacent sensing units

Abstract

The invention discloses a target event positioning method and system for simultaneous adjacent sensing units. Relates to the technical field of target positioning. According to the scheme, a plurality of adjacent sensing units form an abnormal signal cluster for feature extraction, a plurality of classifiers are fused to make a decision on each sensing unit in the cluster and obtain confidence distribution of a corresponding event type, confidence distribution feature information of the event type is extracted, and the event type corresponding to the abnormal signal cluster is obtained by combining a BP network model. Determining the position information of the sensing unit where the target event is actually located; high-precision mode recognition and positioning of external disturbance events are realized; according to the scheme, the event confidence spatial distribution of each sensing unit in the cluster is calculated based on the event confidence spatial distance weighting formula, and the problems that the acoustic vibration signal propagation process is complex and the actual disturbance source is difficult to position due to the unknown buried condition of the optical cable are effectively solved.

Description

Technical Field

[0001] The present invention relates to the technical field of target positioning, and in particular to a target event positioning method and system combining adjacent sensing units. Background Art

[0002] Distributed acoustic vibration sensing technology based on optical fiber plays an important role in the safety monitoring of national and urban infrastructure, such as optical cables, electric cables, long-distance oil and gas pipelines, bridges and tunnels, and regional perimeters. When an external disturbance event acts near a certain sensor unit of the optical fiber, the acoustic vibration signal generated by the disturbance event will propagate along the medium to several sensor units near this sensor unit, causing the corresponding sensor units in this area to be affected by the external acoustic vibration.

[0003] The traditional pattern recognition method based on the data of a single sensor unit will identify the data of all sensor units in this area one by one, which will cause the recognition system to output multiple disturbance event judgment results. In addition, due to the influence of the unknown underground conditions of the optical cable, the propagation process of the acoustic vibration signal is complex. The sensor units in this area may output different event type results, but in fact there is only one disturbance source in the outside world, which greatly restricts the application of distributed optical fiber acoustic vibration sensing technology in safety monitoring. On the other hand, facing the burial conditions of complex underground structures in non-uniform media, traditional positioning methods based on vibration waveform delay differences and waveform energy transfer attenuation characteristics are difficult to apply directly. Therefore, the precise positioning of the disturbance source in this area has also become a challenging problem.

[0004] To this end, a pattern recognition and target positioning technology that combines adjacent sensor units is needed, which can not only effectively eliminate the influence of false alarms in system identification caused by the propagation of external sound wave vibrations in adjacent sensor units, but also accurately locate the actual position of the disturbance source in this area, further improving the intelligent recognition performance of external vibration sources based on distributed optical fiber acoustic wave vibration sensing systems. Summary of the invention

[0005] The technical problem to be solved by the present invention is that the traditional pattern recognition method based on single sensor unit data will output the judgment results of multiple disturbance events due to one disturbance source, which affects the precise positioning effect of the disturbance source; the purpose of the present invention is to provide a target event positioning method and system that jointly connects adjacent sensor units, and multiple adjacent sensor units are combined into an abnormal signal group for feature extraction, and multiple classifiers are integrated to make decisions for each sensor unit in the group and obtain the confidence distribution of the corresponding event type, extract the confidence distribution feature information of the event type, and combine with the BP network model to finally obtain the event type corresponding to the abnormal signal group, and determine the actual location information of the sensor unit where the target event is located; to achieve high-precision pattern recognition and positioning of external disturbance events.

[0006] The present invention is achieved through the following technical solutions:

[0007] The present invention provides a method for locating a target event by combining adjacent sensing units, comprising the steps of:

[0008] S1, obtaining the acoustic vibration signal detected by each sensor unit in the optical cable;

[0009] S2, performing abnormality discrimination on the acoustic vibration signal, and combining the abnormal acoustic vibration signals of multiple adjacent sensing units into an abnormal cluster signal;

[0010] S3, extracting characteristic information parameters of each abnormal acoustic wave vibration signal in the abnormal signal group in turn, classifying the characteristic information parameters of each abnormal acoustic wave vibration signal into events based on multiple classifiers, and fusing the decision results of multiple classifiers to obtain the confidence of each event;

[0011] S4, mining the spatial distribution information of confidence of various events, and inputting it into the BP network model to identify the event pattern corresponding to the abnormal signal group;

[0012] S5, based on the event confidence spatial distribution information of each sensor unit in the abnormal signal group, the position information of the sensor unit where the target event is located is calculated using the event confidence spatial distance weighted formula.

[0013] Working principle of this scheme: When an external disturbance event acts near a certain sensor unit of the optical fiber, the acoustic vibration signal generated by the disturbance event will propagate along the medium to several sensor units near this sensor unit, causing the corresponding sensor units in this area to be affected by the external acoustic vibration. The traditional pattern recognition algorithm based on the data of a single sensor unit will identify the data of all sensor units in this area one by one, which will cause the recognition system to output the judgment results of multiple disturbance events. In addition, due to the influence of the unknown underground conditions of the optical cable, the propagation process of the acoustic vibration signal is complicated. The sensor units in this area may output different event type results, but in fact there is only one disturbance source in the outside world, which greatly restricts the application of distributed optical fiber acoustic vibration sensing technology in security monitoring. On the other hand, facing the burial conditions of complex underground structures in non-uniform media, traditional positioning methods based on vibration waveform delay differences and waveform energy transfer attenuation characteristics are difficult to apply directly.

[0014] This scheme provides a method for locating target events by combining adjacent sensor units. Multiple adjacent sensor units are grouped into an abnormal signal group for feature extraction. Multiple classifiers are integrated to make decisions for each sensor unit in the group and obtain the confidence distribution of the corresponding event type. The confidence distribution feature information of the event type is extracted, and combined with the BP network model, the event type corresponding to the abnormal signal group is finally obtained, and the actual location information of the sensor unit where the target event is located is determined; high-precision pattern recognition and positioning of external disturbance events are achieved. It can be applied to the fields of national and urban infrastructure, communication optical, cable, long-distance oil and gas pipelines, and security monitoring of regional perimeters, and has the characteristics of wide monitoring area, high recognition accuracy, high real-time performance, and high stability.

[0015] The further optimized solution is that the method for distinguishing abnormal acoustic vibration signals includes:

[0016] Select the acoustic vibration signal of each sensing unit corresponding to a duration of T;

[0017] The selected acoustic vibration signal is divided into N T =T*fs / l time domain segments, where the segment length is l and the sampling rate is fs;

[0018] Calculate the time domain energy value S corresponding to each time domain segment respectively i Sum signal kurtosis value P i ;

[0019]

[0020]

[0021] where x ij is the jth data point of the i-th time domain segment, S i and P i are the time domain energy value and signal kurtosis value corresponding to the i-th time domain segment, R i is the background noise value corresponding to the i-th time domain segment.

[0022] Calculate N T The average of the signal kurtosis values ​​of the time domain segments is taken as the kurtosis value P of the current acoustic vibration signal; the statistical time domain energy value S i The number of segments that exceed the time domain energy threshold M T ;

[0023] If M T ≥Fragment threshold M th , or P ≥ signal kurtosis threshold P th When the current sound wave vibration signal is abnormal, it is determined that the current sound wave vibration signal is an abnormal sound wave vibration signal; otherwise, it is determined that the current sound wave vibration signal is a normal sound wave vibration signal.

[0024] The further optimization scheme is that the construction method of abnormal group signal includes:

[0025] Set the interval parameter L_chan for the sensor unit cluster;

[0026] Compare the intervals between the two adjacent abnormal acoustic vibration signals and the sensing units to see if they are ≤L_chan. If so, the sensing units corresponding to the two current abnormal acoustic vibration signals belong to the same signal group; otherwise, the sensing unit corresponding to the abnormal acoustic vibration signal is used as the starting sensing unit of another signal group.

[0027] The abnormal sound wave vibration signals of the same signal group constitute an abnormal signal group.

[0028] A further optimization scheme is that the characteristic information parameters include time domain and frequency domain characteristics;

[0029] The time domain and frequency domain features include: signal energy mean of time domain segments, signal time domain peaks, and energy distribution characteristics of different frequency bands;

[0030] The signal energy mean of the time domain segment includes the current abnormal acoustic vibration signal N T The average value of the time domain energy values ​​corresponding to the time domain segments.

[0031] A further optimization scheme is that the method for extracting the signal time domain peak includes:

[0032] First, the peak-finding threshold parameter th is set, and the first peak-finding operation is performed on the current abnormal acoustic vibration signal, and the peak average value of the peak-finding result is calculated, recorded as TH, and the peak-finding threshold parameter th is replaced by TH;

[0033] Then perform a second peak search operation on the current abnormal acoustic vibration signal to obtain the number of time domain peaks Np obtained by the second peak search and the peak value P corresponding to each time domain peak N ;

[0034] For all peak values ​​P N Sum and average to obtain the time domain peak-to-mean value P of the current abnormal acoustic vibration signal; calculate the peak-to-noise ratio P of the current abnormal acoustic vibration signal SNR , P SNR = time domain peak-to-average value P / noise floor value of time domain signal;

[0035] The method for extracting energy distribution of different frequency bands includes:

[0036] The spectrum information of the current abnormal sound wave vibration signal is obtained based on fast Fourier transform, and the spectrum of the current abnormal sound wave vibration signal is divided into N f The energy sum in each frequency band is calculated respectively, and the energy sum in each frequency band is used as the energy distribution in different frequency bands of the current abnormal sound wave vibration signal.

[0037] A further optimization scheme is that S3 also includes the following sub-steps:

[0038] S31, composing a multi-dimensional feature matrix with feature information parameters extracted from the current abnormal acoustic vibration signal;

[0039] S32, perform two rounds of feature screening on the multidimensional feature matrix to eliminate redundant features:

[0040] The first round of screening: calculate the variance value E of the multidimensional feature matrix under each dimension of the entire data set in turn, set the threshold d1, if the variance value E under the current feature is greater than d1, then retain the current feature, otherwise, remove the current feature;

[0041] The second round of screening: map the feature information obtained in the first round of screening to the feature space, find the center position of each type of event features in each dimension, and calculate the ratio B, B = L1 / L2, L1 is the distance between the center positions of different types of event features; L2 is the distance between different features in the same type of event and the center position of the event feature of that type; set the threshold d2, if B is greater than d2, retain the current feature, otherwise, remove and retain the current feature.

[0042] The further optimization scheme is that the methods for obtaining the confidence of various events include:

[0043] Configure n different classifier models and fuse them in parallel. The training of each classifier model is independent of each other. Each classifier model will match the confidence distribution results of various events corresponding to the abnormal acoustic vibration signal as the output of the model (z i1 ,…,z iM ), where M is the number of target events, z ij The confidence level corresponding to the i-th classifier model identifying the matched abnormal acoustic vibration signal as the j-th type of event;

[0044] Set a weight coefficient for each classifier model to weight the output results of the classifier model:

[0045]

[0046] Among them, C i is the weight coefficient of the i-th classifier, Z1,…,Z M It is the decision result of multiple classifier fusion, that is, the confidence distribution corresponding to M types of events.

[0047] A further optimization scheme is that S4 includes the following sub-steps:

[0048] S41, extracting event confidence spatial distribution information corresponding to M types of events in the current abnormal signal group, wherein the event confidence spatial distribution information includes: confidence variance of each type of event, confidence mean of each type of event, and distribution of confidence values ​​of different ranges of each type of event;

[0049] S42, inputting the extracted event confidence spatial distribution information into the BP network model to identify the event pattern type corresponding to the current abnormal signal group.

[0050] A further optimization scheme is that the method for calculating the position information of the sensor unit where the target event is located includes:

[0051] The event confidence spatial distance distribution E1, E2, …, E of the m sensor units in the abnormal signal group is calculated based on the following formula: m :

[0052]

[0053] Among them l i,j It indicates the spatial distance between the i-th sensor unit and the j-th sensor unit when the j-th sensor unit is assumed to be the location of the external disturbance event; the external disturbance event pattern type corresponding to the abnormal signal group is g; Z gi is the confidence of the g-th event pattern corresponding to the i-th sensor unit in the abnormal signal group;

[0054] Find E1, E2, …, E m The sensor unit corresponding to the maximum value is output as the position information of the sensor unit where the target event is located.

[0055] The present solution also provides a target event positioning system of joint adjacent sensing units, which is used to implement the above-mentioned target event positioning method of joint adjacent sensing units, including:

[0056] An acquisition module is used to obtain the acoustic vibration signal detected by each sensor unit in the optical cable;

[0057] A construction module is used to identify abnormalities in acoustic vibration signals and construct abnormal cluster signals from abnormal acoustic vibration signals of multiple adjacent sensor units;

[0058] The confidence calculation module is used to extract the characteristic information parameters of each abnormal acoustic wave vibration signal in the abnormal signal group in turn, classify the characteristic information parameters of each abnormal acoustic wave vibration signal based on multiple classifiers, and integrate the decision results of multiple classifiers to obtain the confidence of each event;

[0059] The recognition module is used to mine the spatial distribution information of confidence of various events and input the BP network model to identify the event pattern corresponding to the abnormal signal group;

[0060] The calculation module is used to calculate the position information of the sensor unit where the target event is located based on the event confidence spatial distribution information of each sensor unit in the abnormal signal group and using the event confidence spatial distance weighted formula.

[0061] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0062] The present invention provides a method and system for locating a target event by combining adjacent sensor units. A plurality of adjacent sensor units are grouped into an abnormal signal group for feature extraction, a plurality of classifiers are integrated to make a decision on each sensor unit in the group and obtain the confidence distribution of the corresponding event type, the confidence distribution feature information of the event type is extracted, and combined with a BP network model, finally the event type corresponding to the abnormal signal group is obtained, and the position information of the sensor unit where the target event is actually located is determined; and high-precision pattern recognition and positioning of external disturbance events are achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] In order to more clearly illustrate the technical solutions of the exemplary embodiments of the present invention, the following briefly introduces the drawings required for use in the embodiments. It should be understood that the following drawings only illustrate certain embodiments of the present invention and should not be regarded as limiting the scope. For those of ordinary skill in the art, other relevant drawings can be obtained based on these drawings without creative work. In the drawings:

[0064] Figure 1 A flowchart of a method for locating a target event by combining adjacent sensing units;

[0065] Figure 2 A specific flow chart of a method for locating a target event by combining adjacent sensing units;

[0066] Figure 3 A diagram of the process of forming an abnormal signal group for adjacent sensing units;

[0067] Figure 4 A process diagram for obtaining the confidence distribution of various events of the sensor unit for multiple classifier fusion decisions;

[0068] Figure 5 It is a diagram of the pattern recognition process based on the confidence distribution characteristics of various event patterns;

[0069] Figure 6 Schematic diagram of feature screening process;

[0070] Figure 7 Schematic diagram of the target event positioning process based on event confidence spatial distribution information. DETAILED DESCRIPTION

[0071] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with embodiments and drawings. The exemplary embodiments of the present invention and their description are only used to explain the present invention and are not intended to limit the present invention.

[0072] Example 1

[0073] This embodiment provides a method for locating a target event by combining adjacent sensor units. Figure 1 and Figure 2 As shown, the steps include:

[0074] S1, obtaining the acoustic vibration signal detected by each sensor unit in the optical cable;

[0075] S2, performing abnormality discrimination on the acoustic vibration signal, and combining the abnormal acoustic vibration signals of multiple adjacent sensing units into an abnormal cluster signal;

[0076] The methods for distinguishing abnormal acoustic vibration signals include:

[0077] Select the acoustic vibration signal of each sensing unit corresponding to a duration of T;

[0078] The selected acoustic vibration signal is divided into N T =T*fs / l time domain segments, where the segment length is l and the sampling rate is fs;

[0079] Calculate the time domain energy value S corresponding to each time domain segment respectively i Sum signal kurtosis value P i ;

[0080] Calculate N T The average of the signal kurtosis values ​​of the time domain segments is taken as the kurtosis value P of the current acoustic vibration signal; the statistical time domain energy value S i The number of segments that exceed the time domain energy threshold M T ;

[0081] If M T ≥Fragment threshold M th , or P ≥ signal kurtosis threshold P th When the current sound wave vibration signal is abnormal, it is determined that the current sound wave vibration signal is an abnormal sound wave vibration signal; otherwise, it is determined that the current sound wave vibration signal is a normal sound wave vibration signal.

[0082] like Figure 3 As shown, the method for constructing the abnormal group signal includes:

[0083] Set the interval parameter L_chan for the sensor unit cluster;

[0084] Compare the intervals between the two adjacent abnormal acoustic vibration signals and the sensing units to see if they are ≤L_chan. If so, the sensing units corresponding to the two current abnormal acoustic vibration signals belong to the same signal group; otherwise, the sensing unit corresponding to the abnormal acoustic vibration signal is used as the starting sensing unit of another signal group.

[0085] The abnormal sound wave vibration signals of the same signal group constitute an abnormal signal group.

[0086] S3, such as Figure 4 and Figure 5 As shown, characteristic information parameters of each abnormal acoustic wave vibration signal in the abnormal signal group are extracted in turn, and event classification is performed on the characteristic information parameters of each abnormal acoustic wave vibration signal based on multiple classifiers, and the confidence of each type of event is obtained by integrating the decision results of multiple classifiers;

[0087] The characteristic information parameters include time domain and frequency domain characteristics;

[0088] The time domain and frequency domain features include: signal energy mean of time domain segments, signal time domain peaks, and energy distribution characteristics of different frequency bands;

[0089] The signal energy mean of the time domain segment includes the current abnormal acoustic vibration signal N T The average value of the time domain energy values ​​corresponding to the time domain segments.

[0090] The method for extracting the signal time domain peak comprises:

[0091] First, the peak-finding threshold parameter th is set, and the first peak-finding operation is performed on the current abnormal acoustic vibration signal, and the peak average value of the peak-finding result is calculated, recorded as TH, and the peak-finding threshold parameter th is replaced by TH;

[0092] Then perform a second peak search operation on the current abnormal acoustic vibration signal to obtain the number of time domain peaks Np obtained by the second peak search and the peak value P corresponding to each time domain peak N ;

[0093] For all peak values ​​P N Sum and average to obtain the time domain peak-to-mean value P of the current abnormal acoustic vibration signal; calculate the peak-to-noise ratio P of the current abnormal acoustic vibration signal SNR , P SNR = time domain peak-to-average value P / noise floor value of time domain signal;

[0094] The method for extracting energy distribution of different frequency bands includes:

[0095] The spectrum information of the current abnormal sound wave vibration signal is obtained based on fast Fourier transform, and the spectrum of the current abnormal sound wave vibration signal is divided into N fThe energy sum in each frequency band is calculated respectively, and the energy sum in each frequency band is used as the energy distribution in different frequency bands of the current abnormal sound wave vibration signal.

[0096] S3 also includes the following sub-steps:

[0097] S31, composing a multi-dimensional feature matrix with feature information parameters extracted from the current abnormal acoustic vibration signal;

[0098] S32, perform two rounds of feature screening on the multidimensional feature matrix to eliminate redundant features:

[0099] The first round of screening: calculate the variance value E of the multidimensional feature matrix under each dimension of the entire data set in turn, set the threshold d1, if the variance value E under the current feature is greater than d1, then retain the current feature, otherwise, remove the current feature;

[0100] The second round of screening: map the feature information obtained in the first round of screening to the feature space, find the center position of each type of event features in each dimension, and calculate the ratio B, B = L1 / L2, L1 is the distance between the center positions of different types of event features; L2 is the distance between different features in the same type of event and the center position of the event feature of that type; set the threshold d2, if B is greater than d2, retain the current feature, otherwise, remove and retain the current feature.

[0101] like Figure 6 As shown, the center position of class A events (classA) is A1, the center position of class B events (classB) is B1, and the center position of class C events (classC) is C1; L1 is the distance between A1, B1 and C1, and L2 is the distance from each feature in classA / classB / classC to their respective center positions.

[0102] Methods for obtaining the confidence of various events include:

[0103] Configure n different classifier models and fuse them in parallel. The training of each classifier model is independent of each other. Each classifier model will match the confidence distribution results of various events corresponding to the abnormal acoustic vibration signal as the output of the model (z i1 ,…,z iM ), where M is the number of target events, z ij The confidence level corresponding to the i-th classifier model identifying the matched abnormal acoustic vibration signal as the j-th type of event;

[0104] Set a weight coefficient for each classifier model to weight the output results of the classifier model:

[0105]

[0106] Among them, C i is the weight coefficient of the i-th classifier, Z1,…,Z M It is the decision result of multiple classifier fusion, that is, the confidence distribution corresponding to M types of events.

[0107] S4, mining the spatial distribution information of confidence of various events, and inputting the distribution feature information into the BP network model to identify the event pattern corresponding to the abnormal signal group;

[0108] S4, such as Figure 7 As shown, it includes the following sub-steps:

[0109] S41, extracting event confidence spatial distribution information corresponding to M types of events in the current abnormal signal group, wherein the event confidence spatial distribution information includes: confidence variance of each type of event, confidence mean of each type of event, and distribution of confidence values ​​of different ranges of each type of event;

[0110] S42, inputting the extracted event confidence spatial distribution information into the BP network model to identify the event pattern type corresponding to the current abnormal signal group.

[0111] S5, based on the event confidence spatial distribution information of each sensor unit in the abnormal signal group, the position information of the sensor unit where the target event is located is calculated using the event confidence spatial distance weighted formula.

[0112] The method for calculating the position information of the sensor unit where the target event is located includes:

[0113] The event confidence spatial distance distribution E1, E2, …, E of the m sensor units in the abnormal signal group is calculated based on the following formula: m :

[0114]

[0115] Among them l i,j It indicates the spatial distance between the i-th sensor unit and the j-th sensor unit when the j-th sensor unit is assumed to be the location of the external disturbance event; the external disturbance event pattern type corresponding to the abnormal signal group is g; Z gi is the confidence of the g-th event pattern corresponding to the i-th sensor unit in the abnormal signal group;

[0116] Find E1, E2, …, E m The sensor unit corresponding to the maximum value is output as the position information of the sensor unit where the target event is located.

[0117] Example 2

[0118] This embodiment provides a target event positioning system of joint adjacent sensing units, which is used to implement the target event positioning method of joint adjacent sensing units of the above embodiment, including:

[0119] An acquisition module is used to obtain the acoustic vibration signal detected by each sensor unit in the optical cable;

[0120] A construction module is used to identify abnormalities in acoustic vibration signals and construct abnormal cluster signals from abnormal acoustic vibration signals of multiple adjacent sensor units;

[0121] The confidence calculation module is used to extract the characteristic information parameters of each abnormal acoustic wave vibration signal in the abnormal signal group in turn, classify the characteristic information parameters of each abnormal acoustic wave vibration signal based on multiple classifiers, and integrate the decision results of multiple classifiers to obtain the confidence of each event;

[0122] The recognition module is used to mine the spatial distribution information of confidence of various events and input the BP network model to identify the event pattern corresponding to the abnormal signal group;

[0123] The calculation module is used to calculate the position information of the sensor unit where the target event is located based on the event confidence spatial distribution information of each sensor unit in the abnormal signal group and using the event confidence spatial distance weighted formula.

[0124] The present invention uses two rounds of peak-finding operations to find the time-domain peak features of the signal. First, the peak-finding threshold parameter th is set, and the first peak-finding operation is performed on the sensor unit signal data. Then, the average value of these peak values ​​is calculated, recorded as TH, and TH replaces the previous peak-finding threshold parameter th, and then the sensor unit signal data is subjected to a second peak-finding operation. Through two peak-finding operations, more accurate signal time-domain peak features can be extracted. In addition, two rounds of feature screening are performed on the multidimensional feature matrix of the signal to eliminate redundant features, further improving the accuracy of abnormal signal pattern recognition.

[0125] The present invention performs group operation on adjacent abnormal signal sensing units, and groups abnormal signals of adjacent sensing units caused by the same external disturbance source into the same group of abnormal signals for pattern recognition, which can effectively eliminate the influence of system identification false alarms caused by the propagation of external sound wave vibrations on adjacent sensing units.

[0126] The present invention provides an identification method based on the confidence distribution of adjacent abnormal signal sensor unit events. The confidence of each type of event of each sensor unit in the abnormal signal group is obtained by using multiple classifier fusion decisions, and then the event pattern corresponding to the abnormal signal group is output in combination with the BP network model according to the event confidence distribution characteristics of each sensor unit in the group. By fusing multiple classifier decisions to improve the confidence accuracy of various events of the sensor unit, more accurate event confidence distribution characteristics are obtained, and the event recognition accuracy of the system is further improved.

[0127] The present invention provides a target positioning method based on event confidence spatial distance weighted calculation. The event confidence spatial distance weighted formula is used to calculate the event confidence spatial distribution of each sensor unit in the group to determine the actual location information of the sensor unit where the target event is located. This method can effectively solve the problem that the propagation process of the acoustic vibration signal is complicated and the actual disturbance source positioning is difficult due to the unknown buried conditions of the optical cable.

[0128] Example 3

[0129] This embodiment provides a computer-readable medium having a computer program stored thereon. The computer program is executed by a processor to implement a method for locating a target event by combining adjacent sensing units as described in Embodiment 1.

[0130] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for locating a target event by combining adjacent sensing units, characterized in that: Includes steps: S1, obtaining the acoustic vibration signal detected by each sensor unit in the optical cable; S2, performing abnormality discrimination on the acoustic vibration signal, and combining the abnormal acoustic vibration signals of multiple adjacent sensing units into an abnormal cluster signal; S3, extracting characteristic information parameters of each abnormal acoustic wave vibration signal in the abnormal signal group in turn, classifying the characteristic information parameters of each abnormal acoustic wave vibration signal into events based on multiple classifiers, and fusing the decision results of multiple classifiers to obtain the confidence of each event; S4, mining the spatial distribution information of confidence of various events, and inputting it into the BP network model to identify the event pattern corresponding to the abnormal signal group; S5, based on the event confidence spatial distribution information of each sensor unit in the abnormal signal group, the position information of the sensor unit where the target event is located is calculated using the event confidence spatial distance weighted formula.

2. The method for locating a target event by combining adjacent sensing units according to claim 1, characterized in that: The methods for distinguishing abnormal acoustic vibration signals include: Select the acoustic vibration signal of each sensing unit corresponding to a duration of T; The selected acoustic vibration signal is divided into N T =T*fs / l time domain segments, where the segment length is l and the sampling rate is fs; Calculate the time domain energy value S corresponding to each time domain segment respectively i Sum signal kurtosis value P i ; Calculate N T The average of the signal kurtosis values ​​of the time domain segments is taken as the kurtosis value P of the current acoustic vibration signal; the statistical time domain energy value S i The number of segments that exceed the time domain energy threshold M T ; If M T ≥Fragment threshold M th , or P ≥ signal kurtosis threshold P th When the current sound wave vibration signal is abnormal, it is determined that the current sound wave vibration signal is an abnormal sound wave vibration signal; otherwise, it is determined that the current sound wave vibration signal is a normal sound wave vibration signal.

3. The method for locating a target event by combining adjacent sensing units according to claim 2, characterized in that: The construction method of the abnormal group signal includes: Set the interval parameter L_chan for the sensor unit cluster; Compare the intervals between the two adjacent abnormal acoustic vibration signals and the sensing units to see if they are ≤L_chan. If so, the sensing units corresponding to the two current abnormal acoustic vibration signals belong to the same signal group; otherwise, the sensing unit corresponding to the abnormal acoustic vibration signal is used as the starting sensing unit of another signal group. The abnormal sound wave vibration signals of the same signal group constitute an abnormal signal group.

4. The method for locating a target event by combining adjacent sensing units according to claim 2, characterized in that: The characteristic information parameters include time domain and frequency domain characteristics; The time domain and frequency domain features include: signal energy mean of time domain segments, signal time domain peaks, and energy distribution characteristics of different frequency bands; The signal energy mean of the time domain segment includes the current abnormal acoustic vibration signal N T The average value of the time domain energy values ​​corresponding to the time domain segments.

5. The method for locating a target event by combining adjacent sensing units according to claim 4, characterized in that: The method for extracting the signal time domain peak comprises: First, the peak-finding threshold parameter th is set, and the first peak-finding operation is performed on the current abnormal acoustic vibration signal, and the peak average value of the peak-finding result is calculated, recorded as TH, and the peak-finding threshold parameter th is replaced by TH; Then perform a second peak search operation on the current abnormal acoustic vibration signal to obtain the number of time domain peaks Np obtained by the second peak search and the peak value P corresponding to each time domain peak N ; For all peak values ​​P N Sum and average to obtain the time domain peak-to-mean value P of the current abnormal acoustic vibration signal; calculate the peak-to-noise ratio P of the current abnormal acoustic vibration signal SNR , P SNR = time domain peak-to-average value P / noise floor value of time domain signal; The method for extracting energy distribution of different frequency bands includes: The spectrum information of the current abnormal sound wave vibration signal is obtained based on fast Fourier transform, and the spectrum of the current abnormal sound wave vibration signal is divided into N f The energy sum in each frequency band is calculated respectively, and the energy sum in each frequency band is used as the energy distribution in different frequency bands of the current abnormal sound wave vibration signal.

6. The method for locating a target event by combining adjacent sensing units according to claim 1, characterized in that: S3 also includes the following sub-steps: S31, composing a multi-dimensional feature matrix with feature information parameters extracted from the current abnormal acoustic vibration signal; S32, perform two rounds of feature screening on the multidimensional feature matrix to eliminate redundant features: First round Screening: Calculate the variance value E of the entire data set under each dimension of the multidimensional feature matrix in turn, set the threshold d1, if the variance value E under the current feature is greater than d1, retain the current feature, otherwise, remove the current feature; The second round of screening: map the feature information obtained in the first round of screening to the feature space, find the center position of each type of event features in each dimension, and calculate the ratio B, B = L1 / L2, L1 is the distance between the center positions of different types of event features; L2 is the distance between different features in the same type of event and the center position of the event feature of that type; set the threshold d2, if B is greater than d2, retain the current feature, otherwise, remove and retain the current feature.

7. The method for locating a target event by combining adjacent sensing units according to claim 1, characterized in that: Methods for obtaining the confidence of various events include: Configure n different classifier models and fuse them in parallel. The training of each classifier model is independent of each other. Each classifier model will match the confidence distribution results of various events corresponding to the abnormal acoustic vibration signal as the output of the model (z i1 ,…,z iM ), where M is the number of target events, z ij The confidence level corresponding to the i-th classifier model identifying the matched abnormal acoustic vibration signal as the j-th type of event; Set a weight coefficient for each classifier model to weight the output results of the classifier model: Among them, C i is the weight coefficient of the i-th classifier, Z1,…,Z M It is the decision result of multiple classifier fusion, that is, the confidence distribution corresponding to M types of events.

8. The method for locating a target event by combining adjacent sensing units according to claim 1, characterized in that: S4 includes the following sub-steps: S41, extracting event confidence spatial distribution information corresponding to M types of events in the current abnormal signal group, wherein the event confidence spatial distribution information includes: confidence variance of each type of event, confidence mean of each type of event, and distribution of confidence values ​​of different ranges of each type of event; S42, inputting the extracted event confidence spatial distribution information into the BP network model to identify the event pattern type corresponding to the current abnormal signal group.

9. The method for locating a target event by combining adjacent sensing units according to claim 1, characterized in that: The method for calculating the position information of the sensor unit where the target event is located includes: The event confidence spatial distance distribution E1, E2, …, E of the m sensor units in the abnormal signal group is calculated based on the following formula: m : Among them l i,j It indicates the spatial distance between the i-th sensor unit and the j-th sensor unit when the j-th sensor unit is assumed to be the location of the external disturbance event; the external disturbance event pattern type corresponding to the abnormal signal group is g; Z gi is the confidence of the g-th event pattern corresponding to the i-th sensor unit in the abnormal signal group; Find E1, E2, …, E m The sensor unit corresponding to the maximum value is output as the position information of the sensor unit where the target event is located.

10. A target event location system combining adjacent sensing units, characterized in that: A method for locating a target event by using joint adjacent sensing units according to any one of claims 1 to 9, comprising: An acquisition module is used to obtain the acoustic vibration signal detected by each sensor unit in the optical cable; A construction module is used to identify abnormalities in acoustic vibration signals and construct abnormal cluster signals from abnormal acoustic vibration signals of multiple adjacent sensor units; The confidence calculation module is used to extract the characteristic information parameters of each abnormal acoustic wave vibration signal in the abnormal signal group in turn, classify the characteristic information parameters of each abnormal acoustic wave vibration signal based on multiple classifiers, and integrate the decision results of multiple classifiers to obtain the confidence of each event; The recognition module is used to mine the spatial distribution information of confidence of various events and input the BP network model to identify the event pattern corresponding to the abnormal signal group; The calculation module is used to calculate the position information of the sensor unit where the target event is located based on the event confidence spatial distribution information of each sensor unit in the abnormal signal group and using the event confidence spatial distance weighted formula.

Images