CN109827652A - One kind being directed to Fibre Optical Sensor vibration signal recognition and system - Google Patents
One kind being directed to Fibre Optical Sensor vibration signal recognition and system Download PDFInfo
- Publication number
- CN109827652A CN109827652A CN201811416611.9A CN201811416611A CN109827652A CN 109827652 A CN109827652 A CN 109827652A CN 201811416611 A CN201811416611 A CN 201811416611A CN 109827652 A CN109827652 A CN 109827652A
- Authority
- CN
- China
- Prior art keywords
- signal
- vibration
- image
- channel
- decomposition
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Landscapes
- Measurement Of Mechanical Vibrations Or Ultrasonic Waves (AREA)
Abstract
The invention discloses one kind to be directed to Fibre Optical Sensor vibration signal recognition, is modulated first to the optical signal in fiber optic cables；Acquire the vibration signal in fiber optic cables；Vibration signal is monitored, abnormal end point signal is tentatively judged whether there is；Abnormal end point signal if it exists, then intercept abnormal signal from abnormal end point signal, into next step；Otherwise this step is repeated；The abnormal signal of interception is demodulated to obtain the one-dimensional vibration signal of analysis to be identified；Transducing signal sense of hearing conspicuousness two dimensional image is converted by one-dimensional vibration signal；Judge whether abnormal signal is disturbing signal based on transducing signal sense of hearing conspicuousness two dimensional image；If abnormal signal is disturbing signal, enter in next step；Otherwise it returns and the vibration signal of acquisition is monitored；Disturbing signal scene is positioned.The present invention realize Fibre Optical Sensor vibration signal it is quick, accurately identify, solve the low technical problem of Fibre Optical Sensor disturbing signal detecting discrimination.
Description
Technical field
The invention belongs to signal detection technique fields, and in particular to one kind for Fibre Optical Sensor vibration signal recognition and
System.
Background technique
Illegal invasion and destructive behavior set the bases such as military base, communication cable, oil and gas pipeline and circumference security protection
Standby, facility safety causes to seriously threaten.For the various illegal invasions and destruction for jeopardizing cable, pipe safety and security system
Sexual behaviour (hereinafter referred to as disturbing signal), effectively identification has important researching value and realistic meaning accurately and in time.
Distributing optical fiber sensing Disturbance Detection technology, because its anti-electromagnetic interference capability is strong, high sensitivity, dynamic range are big,
Convenient for quickly capture with the technological merits such as high precision monitor dynamic-change information, have become the mainstream of intrusion behavior safety monitoring
Technology.However, the objective factors such as, optical path noise complicated and changeable by natural environment and route shot noise are influenced, and extraneous
Enter to invade the randomness and diversity that destructive behavior is presented, Fibre Optical Sensor disturbing signal detecting method is caused to improve identification
There is substantive difficult point in terms of reducing false alarm rate in rate.Therefore need that a kind of simple, fast and accurately method and system detect
Fibre optical sensor signal, realization fast and accurately identify disturbing signal.
In nature, in interacting for a long time with the adaptation of environment, transducing signal Auditory Perception behavior has very well animal
Self-repairing flexible ability, it is fixed that perception is that good state is fitted.Auditory system distinguishes that object space face shows brilliant performance in audition,
Even if in a noisy environment, can also adaptively improve the entrance loss under various complex environments, priori knowledge and mesh are utilized
Mark adjustment of features cognitive behavior achievees the purpose that accurately to detect and identify voice signal.This just with light distributed under complex environment
Fine vibrating sensing disturbing signal real-time detection task has deep similitude.
Existing distributing optical fiber sensing invasion signal detecting method is aided with mostly from light channel structure optimization design is improved
The method of intelligence computation is made up.But obtain the complete information of disturbing signal it is more difficult, along with disturbing signal not really
Qualitative and randomness is also difficult to construct complete knowledge data base even if increasing priori knowledge, so that the inspection of available data driving
Survey method still can't resolve such ill ill-posed problem.
Summary of the invention
To solve the above problems, the present invention proposes that one kind for Fibre Optical Sensor vibration signal recognition and system, is realized
Fibre Optical Sensor vibration signal it is quick, accurately identify, solve the low technical problem of Fibre Optical Sensor disturbing signal detecting discrimination.
The present invention adopts the following technical scheme that one kind is directed to Fibre Optical Sensor vibration signal recognition, the specific steps are as follows:
S1: the optical signal in fiber optic cables is modulated；
S2: the vibration signal in acquisition fiber optic cables；
S3: being monitored the vibration signal of acquisition, tentatively judges whether there is abnormal end point signal；Abnormal end if it exists
Point signal, then intercept abnormal signal from abnormal end point signal, enter step S4；Abnormal end point signal if it does not exist, then repeat
Execute step S3；
S4: the abnormal signal of interception is demodulated to obtain the one-dimensional vibration signal of analysis to be identified；
S5: transducing signal sense of hearing conspicuousness two dimensional image is converted by one-dimensional vibration signal；
S6: judge whether abnormal signal is disturbing signal based on transducing signal sense of hearing conspicuousness two dimensional image；If abnormal letter
Number be disturbing signal, then enter step S7；If judging, abnormal signal is not disturbing signal, return step S3；
S7: disturbing signal scene is positioned.
Preferably, the method that abnormal end point signal is judged whether there is in the step S3 is short-time energy diagnostic method.
Preferably, the step of one-dimensional vibration signal is converted transducing signal sense of hearing conspicuousness two dimensional image by the step S5
It is as follows:
51) Fast Fourier Transform (FFT) is utilized to the one-dimensional vibration signal after demodulation, obtains signal time-frequency image；
52) vision significance processing is carried out to signal time-frequency image, obtains transducing signal sense of hearing conspicuousness two dimensional image.
Preferably, the method that the step 52) obtains transducing signal sense of hearing conspicuousness two dimensional image is Itti algorithm, specifically
Steps are as follows:
521) multiple dimensioned multiple features channel decomposition is carried out to signal time-frequency image and obtains multiple dimensioned multiple features channel image；
Multiple features channel decomposition includes: color space channel decomposition, Strength Space channel decomposition and gradient space channel point
Solution, specifically:
Color space channel decomposition is carried out to the signal time-frequency image, i.e., by the multichannel cromogram of signal time-frequency image
As being decomposed into multiple single channel gray level images；
Strength Space channel decomposition is carried out to the signal time-frequency image, that is, calculates the signal of each pixel of signal time-frequency image
Intensity；
Gradient space channel decomposition is carried out to the signal time-frequency image, that is, calculates the gradient of each pixel of signal time-frequency image
Direction；
Multi-resolution decomposition is specially color space channel decomposition, Strength Space channel decomposition and the gradient space channel
The calculating of decomposition carries out under conditions of multiple dimensioned, i.e., carries out change of scale to signal time-frequency image, obtain n scale hypostome
Decomposition image under colour space channel, Strength Space channel and gradient space channel；
522) by each feature channel decomposition figure under the supreme one layer of scale of feature channel decomposition image scaling each under different scale
The size of picture calculates separately decomposition image and upper one layer of ruler under color space channel, Strength Space channel, gradient space channel
Hamming distance under degree between each feature channel decomposition image, it is empty respectively as color space channel, Strength Space channel, gradient
Between characteristic pattern on scale is corresponded on channel；
523) characteristic pattern in different characteristic channel is normalized, by feature under each scale in different characteristic channel
Scheme point-to-point direct addition to merge respectively as the characteristic pattern under respective feature channel；
524) by color space channel, Strength Space channel, gradient space channel the point-to-point direct addition of characteristic pattern after
Its average value is sought, transducing signal sense of hearing conspicuousness two dimensional image is obtained.
Preferably, judge whether abnormal signal is to disturb based on transducing signal sense of hearing conspicuousness two dimensional image in the step S6
The method of dynamic signal is to carry out classification knowledge to transducing signal sense of hearing conspicuousness two dimensional image using preparatory trained convolutional network
Not.
One kind be directed to Fibre Optical Sensor vibration signal identifying system, comprising: signal modulation module, optical fiber vibration sensing device and
Monitoring system, wherein
The signal modulation module is for being modulated the optical signal in fiber optic cables；
The optical fiber vibration sensing device is used to acquire the vibration signal in fiber optic cables, and vibration signal is uploaded to prison
In control system；
The monitoring system includes: fiber-optic vibration detection module, optical fibre interrogation module, fiber-optic vibration conversion module, optical fiber
Vibration identification module and fiber-optic vibration locating module；
The vibration signal that the fiber-optic vibration detection module is used to acquire optical fiber vibration sensing device is monitored, tentatively
Abnormal end point signal is judged whether there is, abnormal end point signal, then intercept abnormal signal from abnormal end point signal if it exists, defeated
Enter optical fibre interrogation module；Abnormal end point signal if it does not exist, then continue to monitor；
The optical fibre interrogation module is for demodulating the abnormal signal of interception to obtain the one-dimensional vibration of analysis to be identified
Signal, input optical fibre vibrate conversion module；
The fiber-optic vibration conversion module is used to convert transducing signal sense of hearing conspicuousness X-Y scheme for one-dimensional vibration signal
Picture, input optical fibre Vibration identification module；
The fiber-optic vibration identification module is used to judge that abnormal signal is based on transducing signal sense of hearing conspicuousness two dimensional image
No is disturbing signal；If abnormal signal is disturbing signal, abnormal signal input optical fibre is vibrated into locating module；If abnormal signal
It is not disturbing signal, then returns to optical fiber vibration detection module and be monitored；
The fiber-optic vibration locating module is for positioning disturbing signal scene.
Preferably, the method that the fiber-optic vibration detection module judges whether there is abnormal end point signal is sentenced for short-time energy
Other method.
Preferably, the fiber-optic vibration conversion module includes that vibration signal time-frequency image acquiring unit and visual saliency map obtain
Take unit, in which:
Vibration signal time-frequency image acquiring unit is used to utilize Fast Fourier Transform (FFT) to the one-dimensional vibration signal after demodulation,
Obtain signal time-frequency image；
Visual saliency map acquiring unit is used to carry out vision significance processing to signal time-frequency image, obtains transducing signal and listens
Feel conspicuousness two dimensional image.
Preferably, the method that the visual saliency map acquiring unit obtains transducing signal sense of hearing conspicuousness two dimensional image is
Itti algorithm, the specific steps are as follows:
821) multiple dimensioned multiple features channel decomposition is carried out to signal time-frequency image and obtains multiple dimensioned multiple features channel image；
Multiple features channel decomposition includes: color space channel decomposition, Strength Space channel decomposition and gradient space channel point
Solution, specifically:
Color space channel decomposition is carried out to the signal time-frequency image, i.e., by the multichannel cromogram of signal time-frequency image
As being decomposed into multiple single channel gray level images；
Strength Space channel decomposition is carried out to the signal time-frequency image, that is, calculates the signal of each pixel of signal time-frequency image
Intensity；
Gradient space channel decomposition is carried out to the signal time-frequency image, that is, calculates the gradient of each pixel of signal time-frequency image
Direction；
Multi-resolution decomposition is specially color space channel decomposition, Strength Space channel decomposition and the gradient space channel
The calculating of decomposition carries out under conditions of multiple dimensioned, i.e., carries out change of scale to signal time-frequency image, obtain n scale hypostome
Decomposition image under colour space channel, Strength Space channel and gradient space channel；
822) by each feature channel decomposition figure under the supreme one layer of scale of feature channel decomposition image scaling each under different scale
The size of picture calculates separately decomposition image and upper one layer of ruler under color space channel, Strength Space channel, gradient space channel
Hamming distance under degree between each feature channel decomposition image, it is empty respectively as color space channel, Strength Space channel, gradient
Between characteristic pattern on scale is corresponded on channel；
823) characteristic pattern in different characteristic channel is normalized, by feature under each scale in different characteristic channel
Scheme point-to-point direct addition to merge respectively as the characteristic pattern under respective feature channel；
824) by color space channel, Strength Space channel, gradient space channel the point-to-point direct addition of characteristic pattern after
Its average value is sought, transducing signal sense of hearing conspicuousness two dimensional image is obtained.
Preferably, based on the abnormal letter of transducing signal sense of hearing conspicuousness two dimensional image judgement in the fiber-optic vibration identification module
It number whether be the method for disturbing signal is using preparatory trained convolutional network to transducing signal sense of hearing conspicuousness two dimensional image
Carry out Classification and Identification.
Invent achieved the utility model has the advantages that the present invention is a kind of for Fibre Optical Sensor vibration signal recognition and system,
Realize Fibre Optical Sensor vibration signal it is quick, accurately identify, solve the low technology of Fibre Optical Sensor disturbing signal detecting discrimination and ask
Topic.Non-stationary, discontinuity, short-term stationarity etc. and environmental sound signal phase when the present invention has long using fiber-optic vibration signal
As characteristic, by detecting abnormal disturbances signal segment, and it is significant by one-dimensional model of vibration to convert the two-dimentional transducing signal sense of hearing
Image finally continues to classify, realizes fiber-optic vibration signal and accurately identify with convolutional neural networks to it；The present invention passes through
The transducing signal sense of hearing specific image of fiber-optic vibration signal determines the type of fiber-optic vibration signal, can be used for various based on optical fiber vibration
Dynamic security system has stronger accuracy rate and adaptability.
Detailed description of the invention
Fig. 1 is a kind of flow diagram for Fibre Optical Sensor vibration signal recognition of the embodiment of the present invention；
Fig. 2 is a kind of one-dimensional vibration signal flow path switch schematic diagram of the embodiment of the present invention.
Specific embodiment
Below according to attached drawing and technical solution of the present invention is further elaborated in conjunction with the embodiments.
Fig. 1 is a kind of flow diagram for Fibre Optical Sensor vibration signal recognition of the embodiment of the present invention, a kind of
For Fibre Optical Sensor vibration signal recognition, comprising the following steps:
S1: the optical signal in fiber optic cables is modulated；The parameter of optical signal modulation is according to live situation and optical fiber
The type of cable is determined.
S2: the vibration signal in acquisition fiber optic cables；Vibration signal includes because what illegal invasion or malicious sabotage generated disturbs
Dynamic signal and the vibration signal because of generations such as normal ambulations.
S3: being monitored the vibration signal of acquisition, tentatively judges whether there is abnormal end point signal；Abnormal end if it exists
Point signal, then intercept abnormal signal, i.e., doubtful disturbing signal enters step S4 from obtained abnormal end point signal；If not depositing
In abnormal end point signal, then step S3 is repeated；The disturbance letter issued for finding doubtful illegal invasion and destructive behavior
Number segment.
The method for judging whether there is abnormal end point signal is short-time energy diagnostic method；
S4: the abnormal signal of interception is demodulated to obtain the one-dimensional vibration signal of analysis to be identified；The parameter root of demodulation
It is determined according to the modulation parameter of application.
S5: transducing signal sense of hearing conspicuousness two dimensional image is converted by one-dimensional vibration signal, as shown in Figure 2；
The step of converting transducing signal sense of hearing conspicuousness two dimensional image for one-dimensional vibration signal is as follows:
51) Fast Fourier Transform (FFT) is utilized to the one-dimensional vibration signal after demodulation, obtains signal time-frequency image；The signal
The format of time-frequency image, comprising: the common image storage format such as RGB, YUV, YCrCb, Lab.
52) vision significance processing is carried out to signal time-frequency image, obtains transducing signal sense of hearing conspicuousness two dimensional image (two
Tie up time-frequency image)；
521) multiple dimensioned multiple features channel decomposition is carried out to signal time-frequency image and obtains multiple dimensioned multiple features channel image；
Multiple features channel decomposition includes: color space channel decomposition, Strength Space channel decomposition and gradient space channel point
Solution, specifically:
Color space channel decomposition is carried out to the signal time-frequency image, i.e., by the multichannel cromogram of signal time-frequency image
As being decomposed into multiple single channel gray level images；The result that the color space decomposes determined by signal time-frequency image format, with
For rgb format, three decomposition result, that is, R (red), G (green), B (blue) single channel gray level images.
Strength Space channel decomposition is carried out to the signal time-frequency image, that is, calculates the signal of each pixel of signal time-frequency image
Intensity；The calculation method of the signal strength includes: Hamming distance calculating method, Euclidean distance calculating method etc..With signal time-frequency figure
As storing in an rgb format, for Hamming distance calculating method, the signal strength I of each pixel is
Wherein R, G, B respectively represent gray value red, under green and blue channel.
Gradient space channel decomposition is carried out to the signal time-frequency image, that is, calculates the gradient of each pixel of signal time-frequency image
Direction.The calculation method of the gradient direction includes: hog operator, Jacobian etc..
Multi-resolution decomposition is specially color space channel decomposition, Strength Space channel decomposition and the gradient space channel
The calculating of decomposition carries out under conditions of multiple dimensioned, i.e., carries out change of scale to signal time-frequency image, obtain n scale hypostome
Decomposition image under colour space channel, Strength Space channel and gradient space channel；
The change of scale size includes: 0.25,0.5,2 and 4.
522) by each feature channel decomposition figure under the supreme one layer of scale of feature channel decomposition image scaling each under different scale
The size of picture, using bilinear interpolation by supreme one layer of feature channel decomposition image scaling each under different scale in the present embodiment
The size of each feature channel decomposition image under scale, calculates separately color space channel, Strength Space channel, gradient space channel
Under the Hamming distance decomposed under image and upper one layer of scale between each feature channel decomposition image, it is logical respectively as color space
Road, Strength Space channel correspond to characteristic pattern on scale on gradient space channel；
523) characteristic pattern in different characteristic channel is normalized, bilinear interpolation pair is used in the present embodiment
The characteristic pattern in different characteristic channel is normalized, by the point-to-point direct phase of characteristic pattern under each scale in different characteristic channel
Adduction and respectively as the characteristic pattern under respective feature channel；
524) by color space channel, Strength Space channel, gradient space channel the point-to-point direct addition of characteristic pattern after
Its average value is sought, transducing signal sense of hearing conspicuousness two dimensional image is obtained.
S6: judge whether abnormal signal is disturbing signal based on transducing signal sense of hearing conspicuousness two dimensional image；If abnormal letter
Number be disturbing signal, then enter step S7；If judging, abnormal signal is not disturbing signal, return step S3；
Judge that abnormal signal whether be the method for disturbing signal is utilization based on transducing signal sense of hearing conspicuousness two dimensional image
Preparatory trained convolutional network carries out Classification and Identification to transducing signal sense of hearing conspicuousness two dimensional image.
The structure of convolutional neural networks and the data of pre-training are total to by site environment and possible abnormal disturbances signal type
With decision.The convolutional neural networks method of operation includes but is not limited to: CPU, GPU, FPGA, special circuit etc..
It is described for differentiate the convolutional neural networks structure of transducing signal sense of hearing conspicuousness two dimensional image classification by convolutional layer,
Pond layer, activation primitive layer, normalization layer and the end to end composition of full articulamentum.Wherein:
The convolutional neural networks include convolutional layer, activation primitive layer, pond layer, normalization layer and full articulamentum.
The convolutional layer is used to extract transducing signal sense of hearing conspicuousness two dimensional image can self study, adaptive image spy
Sign, the convolutional layer convolution kernel size are 3*3.
The activation primitive layer quickly approaches initial data distribution situation, the activation letter using all kinds of nonlinear functions
Number includes: tanh, sigmod, ReLU, pReLU etc..The present embodiment uses ReLU as activation primitive.
The pond layer carries out quick sampling to activation primitive processing result, and the method for sampling includes maximum value Chi Huahe mean value
The methods of pond, it acts as: operating rate can be promoted in the case where guaranteeing that discrimination is not decreased obviously.
The normalization layer is realized using Batch Normalization algorithm or Group Normalization to institute
The intermediate result for stating convolutional layer output is standardized, and reduces the gradient disperse problem in network.
The full articulamentum is used to that the final feature that extraction obtains to be learnt and be differentiated, the full articulamentum includes
The number of plies and every layer of nodal point number codetermined by the dimension of feature and classification number to be sorted.
S7: disturbing signal scene is positioned.
One kind be directed to Fibre Optical Sensor vibration signal identifying system, comprising: signal modulation module, optical fiber vibration sensing device and
Monitoring system, wherein
The signal modulation module obtains vibration signal for being modulated to the optical signal in fiber optic cables；The light
Fiber-optic cable is layed in around area to be monitored.
The parameter of optical signal modulation is determined according to the situation at scene and the type of fiber optic cables.
The optical fiber vibration sensing device is used to acquire the vibration signal in fiber optic cables, and vibration signal is uploaded to prison
In control system；The vibration signal may be the disturbing signal generated by illegal invasion or malicious sabotage, be also possible to because normally walking
The dynamic vibration signal for waiting generations.The optical fiber vibration sensing device is installed on fiber optic cables at the access of monitoring center.
The monitoring system includes: fiber-optic vibration detection module, optical fibre interrogation module, fiber-optic vibration conversion module, optical fiber
Vibration identification module and fiber-optic vibration locating module；
The fiber-optic vibration detection module for carrying out the vibration signal that optical fiber vibration sensing device acquires incessantly
Monitoring, tentatively judges whether there is abnormal end point signal, and abnormal end point signal, then intercept different from abnormal end point signal if it exists
Regular signal, i.e., doubtful disturbing signal, input optical fibre demodulation module；Abnormal end point signal if it does not exist, then continue to monitor；Judgement is
The no method that there is abnormal end point signal is short-time energy diagnostic method；Fiber-optic vibration detection module is mounted in monitoring center, is used
In the disturbing signal segment for finding that doubtful illegal invasion and destructive behavior issue.
The optical fibre interrogation module is for demodulating the abnormal signal of interception to obtain the one-dimensional vibration of analysis to be identified
Signal, input optical fibre vibrate conversion module；The parameter of demodulation is determined according to the modulation parameter of application.
The fiber-optic vibration conversion module is used to convert transducing signal sense of hearing conspicuousness X-Y scheme for one-dimensional vibration signal
Picture, input optical fibre Vibration identification module；Method for transformation is codetermined by field condition and possible abnormal disturbances signal type.
Fiber-optic vibration conversion module includes vibration signal time-frequency image acquiring unit and visual saliency map acquiring unit,
In:
Vibration signal time-frequency image acquiring unit is used to utilize Fast Fourier Transform (FFT) to the one-dimensional vibration signal after demodulation,
Obtain signal time-frequency image, the format of the signal time-frequency image, comprising: the common image storage such as RGB, YUV, YCrCb, Lab
Format.
Visual saliency map acquiring unit is used to carry out vision significance processing to signal time-frequency image, obtains transducing signal and listens
Feel conspicuousness two dimensional image.
The method that the visual saliency map acquiring unit obtains transducing signal sense of hearing conspicuousness two dimensional image is Itti algorithm,
Specific step is as follows:
821) multiple dimensioned multiple features channel decomposition is carried out to signal time-frequency image and obtains multiple dimensioned multiple features channel image；
Multiple features channel decomposition includes: color space channel decomposition, Strength Space channel decomposition and gradient space channel point
Solution, specifically:
Color space channel decomposition is carried out to the signal time-frequency image, i.e., by the multichannel cromogram of signal time-frequency image
As being decomposed into multiple single channel gray level images；The result that the color space decomposes determined by signal time-frequency image format, with
For rgb format, three decomposition result, that is, R (red), G (green), B (blue) single channel gray level images.
Strength Space channel decomposition is carried out to the signal time-frequency image, that is, calculates the signal of each pixel of signal time-frequency image
Intensity；The calculation method of the signal strength includes: Hamming distance calculating method, Euclidean distance calculating method etc..With signal time-frequency figure
As storing in an rgb format, for Hamming distance calculating method, the signal strength I of each pixel is
Wherein R, G, B respectively represent gray value red, under green and blue channel.
Gradient space channel decomposition is carried out to the signal time-frequency image, that is, calculates the gradient of each pixel of signal time-frequency image
Direction.The calculation method of the gradient direction includes: hog operator, Jacobian etc..
Multi-resolution decomposition is specially color space channel decomposition, Strength Space channel decomposition and the gradient space channel
The calculating of decomposition carries out under conditions of multiple dimensioned, i.e., carries out change of scale to signal time-frequency image, obtain n scale hypostome
Decomposition image under colour space channel, Strength Space channel and gradient space channel；
The change of scale size includes: 0.25,0.5,2 and 4.
822) by each feature channel decomposition figure under the supreme one layer of scale of feature channel decomposition image scaling each under different scale
The size of picture, using bilinear interpolation by supreme one layer of feature channel decomposition image scaling each under different scale in the present embodiment
The size of each feature channel decomposition image under scale, calculates separately color space channel, Strength Space channel, gradient space channel
Under the Hamming distance decomposed under image and upper one layer of scale between each feature channel decomposition image, it is logical respectively as color space
Road, Strength Space channel correspond to characteristic pattern on scale on gradient space channel；
823) characteristic pattern in different characteristic channel is normalized, bilinear interpolation pair is used in the present embodiment
The characteristic pattern in different characteristic channel is normalized, by the point-to-point direct phase of characteristic pattern under each scale in different characteristic channel
Adduction and respectively as the characteristic pattern under respective feature channel；
824) by color space channel, Strength Space channel, gradient space channel the point-to-point direct addition of characteristic pattern after
Its average value is sought, transducing signal sense of hearing conspicuousness two dimensional image is obtained.
The fiber-optic vibration identification module is used to judge that abnormal signal is based on transducing signal sense of hearing conspicuousness two dimensional image
No is disturbing signal；If abnormal signal is disturbing signal, abnormal signal input optical fibre is vibrated into locating module；If abnormal signal
It is not disturbing signal, then returns to optical fiber vibration detection module and be monitored；
Whether abnormal signal is judged based on transducing signal sense of hearing conspicuousness two dimensional image in the fiber-optic vibration identification module
Method for disturbing signal is to be divided using preparatory trained convolutional network transducing signal sense of hearing conspicuousness two dimensional image
Class identification.The structure of convolutional neural networks and the data of pre-training are common by site environment and possible abnormal disturbances signal type
It determines.The convolutional neural networks method of operation includes but is not limited to: CPU, GPU, FPGA, special circuit etc..
For differentiate the convolutional neural networks structure of transducing signal sense of hearing conspicuousness two dimensional image classification include: convolutional layer,
Pond layer, activation primitive layer, normalization layer and full articulamentum, in which:
The convolutional layer is used to extract transducing signal sense of hearing conspicuousness two dimensional image can self study, adaptive image spy
Sign, the convolutional layer convolution kernel size are 3*3.
The activation primitive layer quickly approaches initial data distribution situation, the activation letter using all kinds of nonlinear functions
Number includes: tanh, sigmod, ReLU, pReLU etc..The present embodiment uses ReLU as activation primitive.
The pond layer carries out quick sampling to activation primitive processing result, and the method for sampling includes maximum value Chi Huahe mean value
The methods of pond, it acts as: operating rate can be promoted in the case where guaranteeing that discrimination is not decreased obviously.
The normalization layer can utilize Batch Normalization algorithm or Group Normalization realization pair
The intermediate result of the convolutional layer output is standardized, and reduces the gradient disperse problem in network.
The full articulamentum is used to that the final feature that extraction obtains to be learnt and be differentiated, the full articulamentum includes
The number of plies and every layer of nodal point number codetermined by the dimension of feature and classification number to be sorted.
The fiber-optic vibration locating module is for positioning disturbing signal scene.Fiber-optic vibration locating module peace
At monitoring center fiber-optic signal access, abnormality informing signal type and generation position.
The foregoing is merely the preferred embodiment of the present invention, to illustrate technical solution of the present invention, rather than limit it
System；It is noted that, modifying the technical solutions described in the foregoing embodiments, or part of technical characteristic is carried out
Equivalent replacement, the spirit and scope for technical solution of various embodiments of the present invention that it does not separate the essence of the corresponding technical solution.
Claims (10)
1. one kind is directed to Fibre Optical Sensor vibration signal recognition, which comprises the following steps:
S1: the optical signal in fiber optic cables is modulated；
S2: the vibration signal in acquisition fiber optic cables；
S3: being monitored the vibration signal of acquisition, tentatively judges whether there is abnormal end point signal；Abnormal endpoint letter if it exists
Number, then abnormal signal is intercepted from abnormal end point signal, enters step S4；Abnormal end point signal if it does not exist, then repeat
Step S3；
S4: the abnormal signal of interception is demodulated to obtain the one-dimensional vibration signal of analysis to be identified；
S5: transducing signal sense of hearing conspicuousness two dimensional image is converted by one-dimensional vibration signal；
S6: judge whether abnormal signal is disturbing signal based on transducing signal sense of hearing conspicuousness two dimensional image；If abnormal signal is
Disturbing signal then enters step S7；If judging, abnormal signal is not disturbing signal, return step S3；
S7: disturbing signal scene is positioned.
2. according to claim 1 a kind of for Fibre Optical Sensor vibration signal recognition, which is characterized in that the step
The method that abnormal end point signal is judged whether there is in S3 is short-time energy diagnostic method.
3. according to claim 1 a kind of for Fibre Optical Sensor vibration signal recognition, which is characterized in that the step
The step of one-dimensional vibration signal is converted transducing signal sense of hearing conspicuousness two dimensional image by S5 is as follows:
51) Fast Fourier Transform (FFT) is utilized to the one-dimensional vibration signal after demodulation, obtains signal time-frequency image；
52) vision significance processing is carried out to signal time-frequency image, obtains transducing signal sense of hearing conspicuousness two dimensional image.
4. according to claim 3 a kind of for Fibre Optical Sensor vibration signal recognition, which is characterized in that the step
52) method for obtaining transducing signal sense of hearing conspicuousness two dimensional image is Itti algorithm, the specific steps are as follows:
521) multiple dimensioned multiple features channel decomposition is carried out to signal time-frequency image and obtains multiple dimensioned multiple features channel image；
Multiple features channel decomposition includes: color space channel decomposition, Strength Space channel decomposition and gradient space channel decomposition, tool
Body are as follows:
Color space channel decomposition is carried out to the signal time-frequency image, i.e., by the multichannel color image of signal time-frequency image point
Solution is multiple single channel gray level images；
Strength Space channel decomposition is carried out to the signal time-frequency image, i.e. the signal of calculating each pixel of signal time-frequency image is strong
Degree；
Gradient space channel decomposition is carried out to the signal time-frequency image, that is, calculates the gradient side of each pixel of signal time-frequency image
To；
Multi-resolution decomposition is specially color space channel decomposition, Strength Space channel decomposition and the gradient space channel decomposition
Calculating carried out under conditions of multiple dimensioned, i.e., change of scale is carried out to signal time-frequency image, obtains under n scale color sky
Between decomposition image under channel, Strength Space channel and gradient space channel；
522) by each feature channel decomposition image under the supreme one layer of scale of feature channel decomposition image scaling each under different scale
Size calculates separately under the decomposition image and upper one layer of scale under color space channel, Strength Space channel, gradient space channel
Hamming distance between each feature channel decomposition image, it is logical respectively as color space channel, Strength Space channel, gradient space
The characteristic pattern on scale is corresponded on road；
523) characteristic pattern in different characteristic channel is normalized, by characteristic pattern point under each scale in different characteristic channel
Point is directly added and is merged respectively as the characteristic pattern under respective feature channel；
524) by color space channel, Strength Space channel, gradient space channel the point-to-point direct addition of characteristic pattern after seek
Its average value obtains transducing signal sense of hearing conspicuousness two dimensional image.
5. according to claim 1 a kind of for Fibre Optical Sensor vibration signal recognition, which is characterized in that the step
Judge that abnormal signal whether be the method for disturbing signal is using preparatory based on transducing signal sense of hearing conspicuousness two dimensional image in S6
Trained convolutional network carries out Classification and Identification to transducing signal sense of hearing conspicuousness two dimensional image.
6. one kind is directed to Fibre Optical Sensor vibration signal identifying system characterized by comprising signal modulation module, fiber-optic vibration
Sensing device and monitoring system, wherein
The signal modulation module is for being modulated the optical signal in fiber optic cables；
The optical fiber vibration sensing device is used to acquire the vibration signal in fiber optic cables, and vibration signal is uploaded to monitoring system
In system；
The monitoring system includes: fiber-optic vibration detection module, optical fibre interrogation module, fiber-optic vibration conversion module, fiber-optic vibration
Identification module and fiber-optic vibration locating module；
The vibration signal that the fiber-optic vibration detection module is used to acquire optical fiber vibration sensing device is monitored, preliminary to judge
With the presence or absence of abnormal end point signal, abnormal end point signal, then intercept abnormal signal, input light from abnormal end point signal if it exists
Fine demodulation module；Abnormal end point signal if it does not exist, then continue to monitor；
The optical fibre interrogation module is used to be demodulated to obtain the one-dimensional vibration signal of analysis to be identified to the abnormal signal of interception,
Input optical fibre vibrates conversion module；
The fiber-optic vibration conversion module is used to convert transducing signal sense of hearing conspicuousness two dimensional image for one-dimensional vibration signal, defeated
Enter fiber-optic vibration identification module；
The fiber-optic vibration identification module be used for based on transducing signal sense of hearing conspicuousness two dimensional image judge abnormal signal whether be
Disturbing signal；If abnormal signal is disturbing signal, abnormal signal input optical fibre is vibrated into locating module；If abnormal signal is not
Disturbing signal then returns to optical fiber vibration detection module and is monitored；
The fiber-optic vibration locating module is for positioning disturbing signal scene.
7. according to claim 6 a kind of for Fibre Optical Sensor vibration signal identifying system, which is characterized in that the optical fiber
The method that vibration detection module judges whether there is abnormal end point signal is short-time energy diagnostic method.
8. according to claim 6 a kind of for Fibre Optical Sensor vibration signal identifying system, which is characterized in that the optical fiber
Vibrating conversion module includes vibration signal time-frequency image acquiring unit and visual saliency map acquiring unit, in which:
Vibration signal time-frequency image acquiring unit is used to utilize Fast Fourier Transform (FFT) to the one-dimensional vibration signal after demodulation, obtains
Signal time-frequency image；
Visual saliency map acquiring unit is used to carry out vision significance processing to signal time-frequency image, and it is aobvious to obtain the transducing signal sense of hearing
Work property two dimensional image.
9. according to claim 8 a kind of for Fibre Optical Sensor vibration signal identifying system, which is characterized in that the vision
The method that notable figure acquiring unit obtains transducing signal sense of hearing conspicuousness two dimensional image is Itti algorithm, the specific steps are as follows:
821) multiple dimensioned multiple features channel decomposition is carried out to signal time-frequency image and obtains multiple dimensioned multiple features channel image；
Multiple features channel decomposition includes: color space channel decomposition, Strength Space channel decomposition and gradient space channel decomposition, tool
Body are as follows:
Color space channel decomposition is carried out to the signal time-frequency image, i.e., by the multichannel color image of signal time-frequency image point
Solution is multiple single channel gray level images；
Strength Space channel decomposition is carried out to the signal time-frequency image, i.e. the signal of calculating each pixel of signal time-frequency image is strong
Degree；
Gradient space channel decomposition is carried out to the signal time-frequency image, that is, calculates the gradient side of each pixel of signal time-frequency image
To；
Multi-resolution decomposition is specially color space channel decomposition, Strength Space channel decomposition and the gradient space channel decomposition
Calculating carried out under conditions of multiple dimensioned, i.e., change of scale is carried out to signal time-frequency image, obtains under n scale color sky
Between decomposition image under channel, Strength Space channel and gradient space channel；
822) by each feature channel decomposition image under the supreme one layer of scale of feature channel decomposition image scaling each under different scale
Size calculates separately under the decomposition image and upper one layer of scale under color space channel, Strength Space channel, gradient space channel
Hamming distance between each feature channel decomposition image, it is logical respectively as color space channel, Strength Space channel, gradient space
The characteristic pattern on scale is corresponded on road；
823) characteristic pattern in different characteristic channel is normalized, by characteristic pattern point under each scale in different characteristic channel
Point is directly added and is merged respectively as the characteristic pattern under respective feature channel；
824) by color space channel, Strength Space channel, gradient space channel the point-to-point direct addition of characteristic pattern after seek
Its average value obtains transducing signal sense of hearing conspicuousness two dimensional image.
10. according to claim 6 a kind of for Fibre Optical Sensor vibration signal identifying system, which is characterized in that the light
In fine Vibration identification module based on transducing signal sense of hearing conspicuousness two dimensional image judge abnormal signal whether be disturbing signal side
Method is to carry out Classification and Identification to transducing signal sense of hearing conspicuousness two dimensional image using preparatory trained convolutional network.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811416611.9A CN109827652A (en) | 2018-11-26 | 2018-11-26 | One kind being directed to Fibre Optical Sensor vibration signal recognition and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811416611.9A CN109827652A (en) | 2018-11-26 | 2018-11-26 | One kind being directed to Fibre Optical Sensor vibration signal recognition and system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109827652A true CN109827652A (en) | 2019-05-31 |
Family
ID=66859841
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811416611.9A Pending CN109827652A (en) | 2018-11-26 | 2018-11-26 | One kind being directed to Fibre Optical Sensor vibration signal recognition and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109827652A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111157099A (en) * | 2020-01-02 | 2020-05-15 | 河海大学常州校区 | Distributed optical fiber sensor vibration signal classification method and identification classification system |
| CN112364836A (en) * | 2020-12-07 | 2021-02-12 | 无锡科晟光子科技有限公司 | Vibration optical fiber signal classification method based on full convolution neural network |
| CN112464777A (en) * | 2020-11-20 | 2021-03-09 | 电子科技大学 | Intelligent estimation method for vertical distance of optical fiber vibration source |
| CN112488179A (en) * | 2020-11-26 | 2021-03-12 | 中国舰船研究设计中心 | Rotary machine fault diagnosis method based on GRU |
| CN112861883A (en) * | 2021-03-18 | 2021-05-28 | 上海壁仞智能科技有限公司 | Image saliency region detection method and device |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105095624A (en) * | 2014-05-15 | 2015-11-25 | 中国电子科技集团公司第三十四研究所 | Method for identifying optical fibre sensing vibration signal |
| CN105574866A (en) * | 2015-12-15 | 2016-05-11 | 努比亚技术有限公司 | Image processing method and apparatus |
| CN106600869A (en) * | 2016-12-31 | 2017-04-26 | 上海华魏光纤传感技术有限公司 | Fence intrusion identification method for fiber fence security protection system |
| US20180283935A1 (en) * | 2017-03-28 | 2018-10-04 | Oki Electric Industry Co., Ltd. | Vibration sensing optical fiber sensor and vibration sensing method |
-
2018
- 2018-11-26 CN CN201811416611.9A patent/CN109827652A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105095624A (en) * | 2014-05-15 | 2015-11-25 | 中国电子科技集团公司第三十四研究所 | Method for identifying optical fibre sensing vibration signal |
| CN105574866A (en) * | 2015-12-15 | 2016-05-11 | 努比亚技术有限公司 | Image processing method and apparatus |
| CN106600869A (en) * | 2016-12-31 | 2017-04-26 | 上海华魏光纤传感技术有限公司 | Fence intrusion identification method for fiber fence security protection system |
| US20180283935A1 (en) * | 2017-03-28 | 2018-10-04 | Oki Electric Industry Co., Ltd. | Vibration sensing optical fiber sensor and vibration sensing method |
Non-Patent Citations (2)
| Title |
|---|
| XU XIAOYAN 等: "Intrusion Recognition for Optic Fiber Vibration Sensor Based on the Selective Attention Mechanism", 《PROCEEDINGS OF SPIE》 * |
| 孟琭: "《计算机视觉原理与应用》", 30 November 2012 * |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111157099A (en) * | 2020-01-02 | 2020-05-15 | 河海大学常州校区 | Distributed optical fiber sensor vibration signal classification method and identification classification system |
| CN111157099B (en) * | 2020-01-02 | 2022-07-15 | 河海大学常州校区 | Distributed optical fiber sensor vibration signal classification method and identification classification system |
| CN112464777A (en) * | 2020-11-20 | 2021-03-09 | 电子科技大学 | Intelligent estimation method for vertical distance of optical fiber vibration source |
| CN112488179A (en) * | 2020-11-26 | 2021-03-12 | 中国舰船研究设计中心 | Rotary machine fault diagnosis method based on GRU |
| CN112364836A (en) * | 2020-12-07 | 2021-02-12 | 无锡科晟光子科技有限公司 | Vibration optical fiber signal classification method based on full convolution neural network |
| CN112861883A (en) * | 2021-03-18 | 2021-05-28 | 上海壁仞智能科技有限公司 | Image saliency region detection method and device |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109827652A (en) | One kind being directed to Fibre Optical Sensor vibration signal recognition and system | |
| Bao et al. | Threat of adversarial attacks on DL-based IoT device identification | |
| CN112380952A (en) | Power equipment infrared image real-time detection and identification method based on artificial intelligence | |
| CN102496157B (en) | Image detection method based on Gaussian multi-scale transform and color complexity | |
| CN113537070B (en) | Detection method, detection device, electronic equipment and storage medium | |
| CN114818839B (en) | Deep learning-based optical fiber sensing underwater acoustic signal identification method and device | |
| CN112364768A (en) | Distributed optical fiber intrusion identification method based on airspace features and machine learning | |
| CN111062319B (en) | Driver call detection method based on active infrared image | |
| WO2019088333A1 (en) | Method for recognizing human body activity on basis of depth map information and apparatus therefor | |
| CN103927518B (en) | A kind of face feature extraction method for human face analysis system | |
| CN115964670A (en) | Frequency spectrum anomaly detection method | |
| US20230024104A1 (en) | Identification of false transformer humming using machine learning | |
| CN114354744B (en) | Power cable external damage prevention monitoring method, system and device | |
| CN115860144A (en) | Machine learning system for anti-electricity-stealing site | |
| CN109241819A (en) | Based on quickly multiple dimensioned and joint template matching multiple target pedestrian detection method | |
| CN115060184A (en) | Optical fiber perimeter intrusion detection method and system based on recursive graph | |
| CN111951505B (en) | Fence vibration intrusion positioning and mode identification method based on distributed optical fiber system | |
| CN114925722A (en) | Perimeter security intrusion signal detection method based on generalized S transformation and transfer learning | |
| CN116189026A (en) | Pedestrian re-recognition method and device and storage medium | |
| Mao et al. | A method for detecting image information leakage risk from electromagnetic emission of computer monitors | |
| Cong et al. | Salient man-made object detection based on saliency potential energy for unmanned aerial vehicles remote sensing image | |
| CN110334671A (en) | A kind of violence infringement detection system and detection method based on Expression Recognition | |
| CN108171168A (en) | Weak target changes intelligent image detection method and device | |
| CN114243925B (en) | Intelligent fusion terminal-based distribution substation allergy sensing method and system | |
| Huang et al. | Invasion detection on transmission lines using saliency computation |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190531 |
|
| RJ01 | Rejection of invention patent application after publication |