CN110660184A - Adaboost-based railway perimeter early warning method of fiber laser radar - Google Patents
Adaboost-based railway perimeter early warning method of fiber laser radar Download PDFInfo
- Publication number
- CN110660184A CN110660184A CN201910883514.9A CN201910883514A CN110660184A CN 110660184 A CN110660184 A CN 110660184A CN 201910883514 A CN201910883514 A CN 201910883514A CN 110660184 A CN110660184 A CN 110660184A
- Authority
- CN
- China
- Prior art keywords
- railway
- optical fiber
- adaboost
- signal
- light
- 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
Images
Classifications
-
- G—PHYSICS
- G08—SIGNALLING
- G08B—SIGNALLING OR CALLING SYSTEMS; ORDER TELEGRAPHS; ALARM SYSTEMS
- G08B13/00—Burglar, theft or intruder alarms
- G08B13/02—Mechanical actuation
- G08B13/12—Mechanical actuation by the breaking or disturbance of stretched cords or wires
- G08B13/122—Mechanical actuation by the breaking or disturbance of stretched cords or wires for a perimeter fence
- G08B13/124—Mechanical actuation by the breaking or disturbance of stretched cords or wires for a perimeter fence with the breaking or disturbance being optically detected, e.g. optical fibers in the perimeter fence
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01V—GEOPHYSICS; GRAVITATIONAL MEASUREMENTS; DETECTING MASSES OR OBJECTS; TAGS
- G01V8/00—Prospecting or detecting by optical means
- G01V8/10—Detecting, e.g. by using light barriers
- G01V8/12—Detecting, e.g. by using light barriers using one transmitter and one receiver
- G01V8/16—Detecting, e.g. by using light barriers using one transmitter and one receiver using optical fibres
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Life Sciences & Earth Sciences (AREA)
- Geophysics (AREA)
- Optical Radar Systems And Details Thereof (AREA)
Abstract
The invention belongs to the technical field of railway safety monitoring, and particularly relates to a railway perimeter early warning method based on an Adaboost fiber laser radar, which comprises the following steps: s1, injecting pulse light into the sensing optical fiber; s2, receiving the backscattered light in the sensing optical fiber by using a coherent demodulation module; s3, light intensity signals received by the coherent demodulation module are collected through a collection card and are further processed by an upper computer, high-sensitivity sensing of the surrounding environment along the railway track is achieved through a mode of combining an optical fiber coherent Rayleigh principle and heterodyne detection, the precision is high, the structure is simple, the remote monitoring cost is low, the operation is convenient, the communication optical cable has double functions of sensing and transmission, third-party construction or detection of man-made and wild animal invasion events near the track is achieved through the optical fiber Rayleigh scattering and phase demodulation principle, and the positioning precision of construction events can reach 100 meters.
Description
Technical Field
The invention belongs to the technical field of railway safety monitoring, and particularly relates to a railway perimeter early warning method based on an Adaboost fiber laser radar.
Background
Along with the leap-type development of railways in China, the passenger transport high-speed degree and the freight transport heavy-load degree are continuously improved, and a new challenge is provided for a railway driving safety guarantee system. The railway perimeter protection monitoring system is one of important subsystems of a railway disaster prevention safety monitoring system, is mainly used for monitoring the peripheral environment of a high-speed railway, important sites, railway bridge crossings and other places for processing foreign matter invasion problems, aims to endow a vision system with the capability of observing and analyzing scene contents, realizes automation and intellectualization of monitoring, and has shown huge development potential in railway engineering application. When suspicious personnel enter the monitoring range, the suspicious personnel can be automatically identified, namely, the suspicious personnel are snapshotted and images at that time are transmitted to a management center, and an alarm signal is output from the management center, but the detection effect is poor.
Some intrusion events, such as illegal damage or crossing of a railway guardrail, placement of foreign matters on a track, theft of cables along the railway and other dangerous or malicious behaviors, can cause serious accident potential for railway safety operation. Of course, environmental factors can also bring great hidden dangers to railway safety, such as debris flow, landslide, collapse, earthquake and the like can damage railway perimeter protection, and third-party factors such as manual construction, excavator operation and the like can bring potential safety hazards to optical cables along the railway though the third-party factors are not malicious behaviors.
For the above hidden dangers that may damage the optical cable along the railway or the railway, an effective monitoring means is needed to dynamically detect the safety state of the optical cable, so as to ensure the safety of the railway communication system.
Disclosure of Invention
The invention provides a railway perimeter early warning method based on an Adaboost fiber laser radar, which aims to solve the problems in the background technology.
In order to achieve the purpose, the invention provides the following technical scheme: a railway perimeter early warning method based on Adaboost fiber laser radar comprises the following steps:
s1, injecting pulse light into the sensing optical fiber;
s2, receiving the backscattered light in the sensing optical fiber by using a coherent demodulation module;
s3, collecting the light intensity signals received by the coherent demodulation module through a collection card, and handing the light intensity signals to an upper computer for further processing;
in S3, the local oscillation light interferes with the signal light which is back scattered by the optical fiber in the reverse direction, the difference operation is carried out on the Rayleigh signal curves at the front and the back moments through photoelectric conversion and amplification processing of a detector, and the position of the interference light intensity signal on the difference curve is changed;
aiming at the signal after photoelectric conversion by the detector, according to the time domain amplitude and duration of the signal, the frequency domain distribution and the frequency spectrum duration of the signal and the time variation characteristic of the signal frequency, an Adaboost algorithm is adopted, and on the basis of a plurality of weak classifiers, the frequency spectrum and the time-frequency comprehensive information characteristics of the railway threat event are extracted and identified.
Preferably, in S1, the pulsed light is modulated and converted by an acousto-optic modulator (AOM) by a laser, and the pulsed light is power-amplified by an erbium-doped fiber amplifier (EDFA), and injected into the sensing fiber through a circulator.
Preferably, the laser is a narrow linewidth laser.
Preferably, in S2, the coherent demodulation module receives the pulsed light, and generates backward rayleigh scattered light due to the non-uniform distribution of the medium in the optical fiber during the forward propagation along the optical fiber.
Preferably, due to the non-uniform distribution of the medium in the optical fiber, backward rayleigh scattered light is generated and propagates along the sensing fiber in the reverse direction to be received by the coherent demodulation module through the circulator.
Preferably, the position of the interference light intensity signal on the difference curve is displayed on a display screen through a waterfall graph through software.
Preferably, a plurality of weak classifiers are initially trained, and if the accuracy of a weak classifier is high, the weight of the weak classifier is higher, and otherwise, the weight of the weak classifier is lower.
Preferably, the initial training is divided into several rounds, and the classification error rate (weighted error function) of each weak classifier on the training data set of the current round is calculated: the weight error function focuses on the weight distribution of the data set in the current round, and does not focus on the parameters inside the weak classifier; greater punishment is given to the error of the high probability distribution (the data with important attention) of the round; model coefficients calculated from the classification errors of the weak classifiers of the current round on the data set: representing the importance degree of the weak classifier obtained in the current round; the smaller the classification error rate in the current round, the more the basic classifier plays a role in the final classifier; updating the weight distribution of the training data set of the next round; in this training round, the weights of the misclassified samples by the basic classifier are expanded, while the weights of the correctly classified samples are reduced in the next round.
Compared with the prior art, the invention has the beneficial effects that:
1. the invention realizes high-sensitivity perception of the surrounding environment along the railway track by combining the optical fiber coherent Rayleigh principle and heterodyne detection, and has the advantages of high precision, simple structure, low remote monitoring cost and convenient operation.
2. The communication optical cable has double functions of sensing and transmission, realizes the third-party construction near the track or the detection of man-made and wild animal invasion events by the aid of optical fiber Rayleigh scattering and phase demodulation principles, and has the positioning accuracy of 100 meters for construction events.
3. The technical method adopted by the invention can realize the continuous monitoring of the speed of the train in a waterfall diagram mode.
4. The sensor is uncharged and can be applied to a strong electromagnetic radiation environment of a railway environment.
5. Adaboost has high classification precision as a classifier, and is beneficial to improving the accurate detection of the railway security threat event.
6. Under the framework of Adaboost, various regression classification models can be used for constructing the weak learner, and the weak learner is very flexible.
7. When the binary classifier is used as a simple binary classifier, the structure is simple, and over-fitting is not easy to occur compared with algorithms such as a BP neural network and an SVM support vector machine, so that the identification accuracy is improved.
Drawings
The accompanying drawings, which are included to provide a further understanding of the invention and are incorporated in and constitute a part of this specification, illustrate embodiments of the invention and together with the description serve to explain the principles of the invention and not to limit the invention. In the drawings:
FIG. 1 is a schematic structural view of the present invention;
FIG. 2 is a time domain waterfall plot (s/km) of passenger cars passing through in embodiment 1 of the present invention;
FIG. 3 is a time domain waterfall plot (s/m) of passenger cars passing through in embodiment 1 of the present invention;
fig. 4 is a time domain signal and a short-time fourier analysis spectrogram when a section of truck passes through in embodiment 2 of the present invention;
fig. 5 is a waterfall diagram of excavator signal detection in embodiment 3 of the present invention;
FIG. 6 is a schematic diagram of a model for identifying a railway perimeter intrusion signal using the ADAboost algorithm according to the present invention;
FIG. 7 is a flowchart illustrating the Adaboost algorithm for processing intrusion events on the perimeter of a railway.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Example 1
Referring to fig. 1-3, the present invention provides the following technical solutions: a railway perimeter early warning method based on Adaboost fiber laser radar comprises the following steps:
s1, injecting pulse light into the sensing optical fiber;
s2, receiving the backscattered light in the sensing optical fiber by using a coherent demodulation module;
and S3, collecting the light intensity signals received by the coherent demodulation module through a collection card, and handing the light intensity signals to an upper computer for further processing.
Specifically, in S1, the pulsed light is modulated and converted by an acousto-optic modulator (AOM) by a laser, and the pulsed light is power-amplified by an erbium-doped fiber amplifier (EDFA), and injected into the sensing fiber by a circulator; the laser adopts a narrow linewidth laser; in S2, during the forward propagation of the pulse light along the optical fiber, the coherent demodulation module generates backward rayleigh scattered light due to the non-uniform distribution of the medium in the optical fiber; due to uneven medium distribution in the optical fiber, backward Rayleigh scattered light can be generated and reversely propagated along the sensing optical fiber, and the backward Rayleigh scattered light is received by the coherent demodulation module through the circulator; in S3, the local oscillation light interferes with the signal light which is back scattered by the optical fiber in the reverse direction, the difference operation is carried out on the Rayleigh signal curves at the front and the back moments through photoelectric conversion and amplification processing of a detector, and the position of the interference light intensity signal on the difference curve is changed; the position of the interference light intensity signal on the difference curve is displayed on a display screen through a waterfall graph through software.
In this embodiment: continuous light waves output by the continuous narrow-linewidth laser are modulated and converted into pulse light by an acousto-optic modulator (AOM), then are subjected to power amplification by an erbium-doped fiber amplifier (EDFA), and are injected into the sensing fiber by a circulator. During the forward propagation of the pulsed light along the optical fiber, backward rayleigh scattered light is generated due to the uneven distribution of the medium in the optical fiber. The scattered light is reversely propagated along the sensing optical fiber, is received by the coherent demodulation module through the circulator, and is collected by the collection card to be further processed by the upper computer.
The fiber laser radar uses the ultra-narrow linewidth laser to realize the interference effect between the backward Rayleigh scattering light in the pulse width range, and when receiving external intrusion interference along a certain position on an optical fiber circuit, the refractive index of the optical fiber at the corresponding position changes, so that the optical phase at the position changes. The change of interference effect phase can cause the change of backward Rayleigh scattering light intensity, and the interference of local oscillation light and signal light which is scattered back through the optical fiber is processed by photoelectric conversion and amplification of a detector. And performing difference operation on the Rayleigh signal curves at the front and rear moments, wherein the position of the interference light intensity signal on the difference curve, which changes, corresponds to the position of the disturbance.
In this embodiment, a feature library of railway perimeter signals is established by collecting a large number of signals of threat events and noise events. Wherein the threat event mainly comprises: constructing a third party around the railway, illegally entering a railway perimeter area, and crossing over a railway fence; the noise events mainly include: noise of highways and other railways along the perimeter of the railways, small animal invasion along the railways, and the like.
For railway security threat events and noise events, it can be described by the following signal characteristics: 1. the time domain amplitude and duration of the signal, the frequency domain distribution and the frequency spectrum duration of the signal, and the time-varying characteristic of the signal frequency;
adopting Adaboost algorithm, based on a plurality of weak classifiers, extracting and identifying the frequency spectrum and time-frequency comprehensive information characteristics of the railway threat event, as shown in FIG. 6;
the AdaBoost algorithm is used for the detection of railway perimeter threat events. The railway perimeter threat event mainly refers to the illegal climbing caused by the construction of a third party along the railway; the main noise interference sources are: river crossing along the railway, vehicle vibration on parallel roads, agricultural machinery cultivation and the like. The specific implementation mode comprises the following steps of learning a series of weak classifiers or basic classifiers from a railway database acquired by a laser radar system, and linearly combining the weak classifiers into a strong classifier.
According to the AdaBoost algorithm, if the accuracy of a base classifier is high, its weight is a little higher, otherwise it is lower. Firstly, carrying out initialization training on the acquired intrusion vibration signals, wherein the weight distribution of data (N represents the number of samples):
D＝{w11,w1i...w1N},i＝1,2...N
1. the assumption that these vibration signal data have a uniform weight distribution, i.e. each training sample acts the same in the learning of the basic classifier, ensures that the first step is able to learn the basic classifier G1(x) on the original train signal;
2. assuming that the train round for the railway intrusion signal is M (until some predetermined sufficiently small error rate is reached or a pre-specified maximum number of iterations is reached), M is processed as follows:
and a, learning by using a training data set with weight distribution Dm (a data set corresponding to the weight distribution of the current round) to obtain a basic classifier of the current round.
b, calculating the classification error rate (weight error function) of each weak classifier on the training data set of the current round: the weight error function focuses on the weight distribution of the current round of data sets and not on the parameters inside the weak classifiers. More penalties are given to errors in the high probability distribution (data of major interest) of the current round.
c, calculating the model coefficient according to the classification error of the weak classifier of the current round on the data set: representing the importance of the weak classifiers obtained in this round. The smaller the classification error rate in the current round, the more the basic classifier will have a role in the final classifier.
And d, updating the weight distribution of the training data set of the next round. In this training round, the weights of the misclassified samples by the basic classifier are expanded, while the weights of the correctly classified samples are reduced in the next round. Compared with the two phases, the weight of the misclassified intrusion signal sample is amplified, so that the misclassified sample plays a greater role in the next learning round. The distribution of the weight values of the training data is continuously changed without changing the given training data, so that the training data continuously optimizes weak classification in the learning of the basic classifier, and the aim of a strong classification model is finally fulfilled.
Please refer to fig. 7;
in each round of training, the weight distribution of the training samples is constantly changed, and simultaneously 1, the weight distribution plays a direct proportion role in the importance degree of the weak classifier in the round in the final linear classifier combination; 2. and the weight adjustment of the sample in the next round is inversely proportional.
The train signal is about 6.5 km away from the head station as shown in fig. 2, the running track of the train can be seen through amplification in fig. 3, and the speed of the train can be calculated through the slope of the track.
Example 2
Referring to fig. 1 and 4, the present invention provides the following technical solutions: a railway perimeter early warning method based on Adaboost fiber laser radar comprises the following steps:
s1, injecting pulse light into the sensing optical fiber;
s2, receiving the backscattered light in the sensing optical fiber by using a coherent demodulation module;
and S3, collecting the light intensity signals received by the coherent demodulation module through a collection card, and handing the light intensity signals to an upper computer for further processing.
Specifically, in S1, the pulsed light is modulated and converted by an acousto-optic modulator (AOM) by a laser, and the pulsed light is power-amplified by an erbium-doped fiber amplifier (EDFA), and injected into the sensing fiber by a circulator; the laser adopts a narrow linewidth laser; in S2, during the forward propagation of the pulse light along the optical fiber, the coherent demodulation module generates backward rayleigh scattered light due to the non-uniform distribution of the medium in the optical fiber; due to uneven medium distribution in the optical fiber, backward Rayleigh scattered light can be generated and reversely propagated along the sensing optical fiber, and the backward Rayleigh scattered light is received by the coherent demodulation module through the circulator; in S3, the local oscillation light interferes with the signal light which is back scattered by the optical fiber in the reverse direction, the difference operation is carried out on the Rayleigh signal curves at the front and the back moments through photoelectric conversion and amplification processing of a detector, and the position of the interference light intensity signal on the difference curve is changed; the position of the interference light intensity signal on the difference curve is displayed on a display screen through a waterfall graph through software.
In this embodiment: continuous light waves output by the continuous narrow-linewidth laser are modulated and converted into pulse light by an acousto-optic modulator (AOM), then are subjected to power amplification by an erbium-doped fiber amplifier (EDFA), and are injected into the sensing fiber by a circulator. During the forward propagation of the pulsed light along the optical fiber, backward rayleigh scattered light is generated due to the uneven distribution of the medium in the optical fiber. The scattered light is reversely propagated along the sensing optical fiber, is received by the coherent demodulation module through the circulator, and is collected by the collection card to be further processed by the upper computer.
The fiber laser radar uses the ultra-narrow linewidth laser to realize the interference effect between the backward Rayleigh scattering light in the pulse width range, and when receiving external intrusion interference along a certain position on an optical fiber circuit, the refractive index of the optical fiber at the corresponding position changes, so that the optical phase at the position changes. The change of interference effect phase can cause the change of backward Rayleigh scattering light intensity, and the interference of local oscillation light and signal light which is scattered back through the optical fiber is processed by photoelectric conversion and amplification of a detector. And performing difference operation on the Rayleigh signal curves at the front and rear moments, wherein the position of the interference light intensity signal on the difference curve, which changes, corresponds to the position of the disturbance.
In this embodiment, a feature library of railway perimeter signals is established by collecting a large number of signals of threat events and noise events. Wherein the threat event mainly comprises: constructing a third party around the railway, illegally entering a railway perimeter area, and crossing over a railway fence; the noise events mainly include: noise of highways and other railways along the perimeter of the railways, small animal invasion along the railways, and the like.
For railway security threat events and noise events, it can be described by the following signal characteristics: 1. the time domain amplitude and duration of the signal, the frequency domain distribution and the frequency spectrum duration of the signal, and the time-varying characteristic of the signal frequency;
adopting Adaboost algorithm, based on a plurality of weak classifiers, extracting and identifying the frequency spectrum and time-frequency comprehensive information characteristics of the railway threat event, as shown in FIG. 6;
the AdaBoost algorithm is used for the detection of railway perimeter threat events. The railway perimeter threat event mainly refers to the illegal climbing caused by the construction of a third party along the railway; the main noise interference sources are: river crossing along the railway, vehicle vibration on parallel roads, agricultural machinery cultivation and the like. The specific implementation mode comprises the following steps of learning a series of weak classifiers or basic classifiers from a railway database acquired by a laser radar system, and linearly combining the weak classifiers into a strong classifier.
According to the AdaBoost algorithm, if the accuracy of a base classifier is high, its weight is a little higher, otherwise it is lower. Firstly, carrying out initialization training on the acquired intrusion vibration signals, wherein the weight distribution of data (N represents the number of samples):
D＝{w11,w1i...w1N},i＝1,2...N
1. the assumption that these vibration signal data have a uniform weight distribution, i.e. each training sample acts the same in the learning of the basic classifier, ensures that the first step is able to learn the basic classifier G1(x) on the original train signal;
2. assuming that the train round for the railway intrusion signal is M (until some predetermined sufficiently small error rate is reached or a pre-specified maximum number of iterations is reached), M is processed as follows:
and a, learning by using a training data set with weight distribution Dm (a data set corresponding to the weight distribution of the current round) to obtain a basic classifier of the current round.
b, calculating the classification error rate (weight error function) of each weak classifier on the training data set of the current round: the weight error function focuses on the weight distribution of the current round of data sets and not on the parameters inside the weak classifiers. More penalties are given to errors in the high probability distribution (data of major interest) of the current round.
c, calculating the model coefficient according to the classification error of the weak classifier of the current round on the data set: representing the importance of the weak classifiers obtained in this round. The smaller the classification error rate in the current round, the more the basic classifier will have a role in the final classifier.
And d, updating the weight distribution of the training data set of the next round. In this training round, the weights of the misclassified samples by the basic classifier are expanded, while the weights of the correctly classified samples are reduced in the next round. Compared with the two phases, the weight of the misclassified intrusion signal sample is amplified, so that the misclassified sample plays a greater role in the next learning round. The distribution of the weight values of the training data is continuously changed without changing the given training data, so that the training data continuously optimizes weak classification in the learning of the basic classifier, and the aim of a strong classification model is finally fulfilled.
Please refer to fig. 7;
in each round of training, the weight distribution of the training samples is constantly changed, and simultaneously 1, the weight distribution plays a direct proportion role in the importance degree of the weak classifier in the round in the final linear classifier combination; 2. and the weight adjustment of the sample in the next round is inversely proportional.
Fig. 4 shows the vibration signal of the optical fiber when the freight train passes through, the left side is the original signal, the right side is the time-domain differential signal, the signal is in periodic segments, the period length is about 1s, the train speed is about 15m/s and 54km/h, the corresponding length is 15 meters, and the length is equivalent to the length of the ordinary freight train carriage, so the period should be caused by the connection vibration between the carriages.
Example 3
Referring to fig. 1 and 5, the present invention provides the following technical solutions: a railway perimeter early warning method based on Adaboost fiber laser radar comprises the following steps:
s1, injecting pulse light into the sensing optical fiber;
s2, receiving the backscattered light in the sensing optical fiber by using a coherent demodulation module;
and S3, collecting the light intensity signals received by the coherent demodulation module through a collection card, and handing the light intensity signals to an upper computer for further processing.
Specifically, in S1, the pulsed light is modulated and converted by an acousto-optic modulator (AOM) by a laser, and the pulsed light is power-amplified by an erbium-doped fiber amplifier (EDFA), and injected into the sensing fiber by a circulator; the laser adopts a narrow linewidth laser; in S2, during the forward propagation of the pulse light along the optical fiber, the coherent demodulation module generates backward rayleigh scattered light due to the non-uniform distribution of the medium in the optical fiber; due to uneven medium distribution in the optical fiber, backward Rayleigh scattered light can be generated and reversely propagated along the sensing optical fiber, and the backward Rayleigh scattered light is received by the coherent demodulation module through the circulator; in S3, the local oscillation light interferes with the signal light which is back scattered by the optical fiber in the reverse direction, the difference operation is carried out on the Rayleigh signal curves at the front and the back moments through photoelectric conversion and amplification processing of a detector, and the position of the interference light intensity signal on the difference curve is changed; the position of the interference light intensity signal on the difference curve is displayed on a display screen through a waterfall graph through software.
In this embodiment, a feature library of railway perimeter signals is established by collecting a large number of signals of threat events and noise events. Wherein the threat event mainly comprises: constructing a third party around the railway, illegally entering a railway perimeter area, and crossing over a railway fence; the noise events mainly include: noise of highways and other railways along the perimeter of the railways, small animal invasion along the railways, and the like.
For railway security threat events and noise events, it can be described by the following signal characteristics: 1. the time domain amplitude and duration of the signal, the frequency domain distribution and the frequency spectrum duration of the signal, and the time-varying characteristic of the signal frequency;
adopting Adaboost algorithm, based on a plurality of weak classifiers, extracting and identifying the frequency spectrum and time-frequency comprehensive information characteristics of the railway threat event, as shown in FIG. 6;
the AdaBoost algorithm is used for the detection of railway perimeter threat events. The railway perimeter threat event mainly refers to the illegal climbing caused by the construction of a third party along the railway; the main noise interference sources are: river crossing along the railway, vehicle vibration on parallel roads, agricultural machinery cultivation and the like. The specific implementation mode comprises the following steps of learning a series of weak classifiers or basic classifiers from a railway database acquired by a laser radar system, and linearly combining the weak classifiers into a strong classifier.
According to the AdaBoost algorithm, if the accuracy of a base classifier is high, its weight is a little higher, otherwise it is lower. Firstly, carrying out initialization training on the acquired intrusion vibration signals, wherein the weight distribution of data (N represents the number of samples):
D＝{w11,w1i...w1N},i＝1,2...N
1. the assumption that these vibration signal data have a uniform weight distribution, i.e. each training sample acts the same in the learning of the basic classifier, ensures that the first step is able to learn the basic classifier G1(x) on the original train signal;
2. assuming that the train round for the railway intrusion signal is M (until some predetermined sufficiently small error rate is reached or a pre-specified maximum number of iterations is reached), M is processed as follows:
and a, learning by using a training data set with weight distribution Dm (a data set corresponding to the weight distribution of the current round) to obtain a basic classifier of the current round.
b, calculating the classification error rate (weight error function) of each weak classifier on the training data set of the current round: the weight error function focuses on the weight distribution of the current round of data sets and not on the parameters inside the weak classifiers. More penalties are given to errors in the high probability distribution (data of major interest) of the current round.
c, calculating the model coefficient according to the classification error of the weak classifier of the current round on the data set: representing the importance of the weak classifiers obtained in this round. The smaller the classification error rate in the current round, the more the basic classifier will have a role in the final classifier.
And d, updating the weight distribution of the training data set of the next round. In this training round, the weights of the misclassified samples by the basic classifier are expanded, while the weights of the correctly classified samples are reduced in the next round. Compared with the two phases, the weight of the misclassified intrusion signal sample is amplified, so that the misclassified sample plays a greater role in the next learning round. The distribution of the weight values of the training data is continuously changed without changing the given training data, so that the training data continuously optimizes weak classification in the learning of the basic classifier, and the aim of a strong classification model is finally fulfilled.
Please refer to fig. 7;
in each round of training, the weight distribution of the training samples is constantly changed, and simultaneously 1, the weight distribution plays a direct proportion role in the importance degree of the weak classifier in the round in the final linear classifier combination; 2. and the weight adjustment of the sample in the next round is inversely proportional.
Fig. 5 is a continuous waterfall graph of 1500-meter length optical cable line-along-line ground vibration signals, which are collected by the fiber laser radar early warning system, changing with time, and it can be seen that at a distance of 250 meters from the system starting end, continuous strong vibration signals exist, and for excavator operation events, intrusion behaviors are determined according to the combined amplitude and frequency distribution change conditions of the signals in time domain and frequency domain.
Finally, it should be noted that: although the present invention has been described in detail with reference to the foregoing embodiments, it will be apparent to those skilled in the art that changes may be made in the embodiments and/or equivalents thereof without departing from the spirit and scope of the invention. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (7)
1. A railway perimeter early warning method based on Adaboost fiber laser radar is characterized in that: the method comprises the following steps:
s1, injecting pulse light into the sensing optical fiber;
s2, receiving the backscattered light in the sensing optical fiber by using a coherent demodulation module;
s3, collecting the light intensity signals received by the coherent demodulation module through a collection card, and handing the light intensity signals to an upper computer for further processing;
in S3, the local oscillation light interferes with the signal light which is back scattered by the optical fiber in the reverse direction, the difference operation is carried out on the Rayleigh signal curves at the front and the back moments through photoelectric conversion and amplification processing of a detector, and the position of the interference light intensity signal on the difference curve is changed;
aiming at the signal after photoelectric conversion by the detector, according to the time domain amplitude and duration of the signal, the frequency domain distribution and the frequency spectrum duration of the signal and the time variation characteristic of the signal frequency, an Adaboost algorithm is adopted, and on the basis of a plurality of weak classifiers, the frequency spectrum and the time-frequency comprehensive information characteristics of the railway threat event are extracted and identified.
2. The railway perimeter early warning method based on the Adaboost fiber laser radar as claimed in claim 1, wherein: in S1, the pulsed light is modulated and converted by an acousto-optic modulator (AOM) by a laser, and the pulsed light is power-amplified by an erbium-doped fiber amplifier (EDFA), and injected into the sensing fiber by a circulator.
3. The railway perimeter early warning method based on the Adaboost fiber laser radar as claimed in claim 1, wherein: in S2, the coherent demodulation module receives the pulsed light, and generates backward rayleigh scattered light due to the non-uniform distribution of the medium in the optical fiber during the forward propagation along the optical fiber.
4. The railway perimeter early warning method based on the Adaboost fiber laser radar as claimed in claim 1, wherein: the position of the interference light intensity signal on the difference curve is displayed on a display screen through a waterfall graph through software.
5. The railway perimeter early warning method based on the Adaboost fiber laser radar as claimed in claim 1, wherein: and (4) initially training a plurality of weak classifiers, wherein if the accuracy of one weak classifier is high, the weight of the weak classifier is higher, and otherwise, the weight is lower.
6. The railway perimeter early warning method based on the Adaboost fiber laser radar as claimed in claim 5, wherein: the initial training is divided into a plurality of rounds, and the classification error rate of each weak classifier on the training data set of the round is calculated: the weight error function focuses on the weight distribution of the current round of data sets and not on the parameters inside the weak classifiers.
7. The railway perimeter early warning method based on the Adaboost fiber laser radar as claimed in claim 6, wherein: giving larger punishment to the error of the high probability distribution of the round; model coefficients calculated from the classification errors of the weak classifiers of the current round on the data set: representing the importance degree of the weak classifier obtained in the current round; the smaller the classification error rate in the current round, the more the basic classifier plays a role in the final classifier; updating the weight distribution of the training data set of the next round; in this training round, the weights of the misclassified samples by the basic classifier are expanded, while the weights of the correctly classified samples are reduced in the next round.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910883514.9A CN110660184A (en) | 2019-09-18 | 2019-09-18 | Adaboost-based railway perimeter early warning method of fiber laser radar |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910883514.9A CN110660184A (en) | 2019-09-18 | 2019-09-18 | Adaboost-based railway perimeter early warning method of fiber laser radar |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110660184A true CN110660184A (en) | 2020-01-07 |
Family
ID=69038208
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910883514.9A Pending CN110660184A (en) | 2019-09-18 | 2019-09-18 | Adaboost-based railway perimeter early warning method of fiber laser radar |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110660184A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112883802A (en) * | 2021-01-20 | 2021-06-01 | 刘信 | Method for identifying destructive event of pipeline optical fiber vibration safety early warning system |
| CN116204784A (en) * | 2022-12-30 | 2023-06-02 | 成都天仁民防科技有限公司 | DAS-based subway tunnel external hazard operation intrusion recognition method |
| CN118243626A (en) * | 2024-05-21 | 2024-06-25 | 西安科技大学 | Photothermal spectrum gas detection device and method based on micro-nano optical fiber array |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102360519A (en) * | 2011-07-29 | 2012-02-22 | 电子科技大学 | Intrusion detecting and positioning method for distributed optical fiber fence |
| CN103824406A (en) * | 2014-02-08 | 2014-05-28 | 上海波汇通信科技有限公司 | Optical fiber sensing-based tunnel cable invasive detection method and system |
| CN107369282A (en) * | 2017-08-23 | 2017-11-21 | 武汉理工光科股份有限公司 | A kind of buried COTDR circumferences intrusion alarm system and signal processing method |
| CN108286991A (en) * | 2017-01-10 | 2018-07-17 | 光子瑞利科技（北京）有限公司 | Application based on the optical fiber safety-protection system of rayleigh scattering principle in campus |
-
2019
- 2019-09-18 CN CN201910883514.9A patent/CN110660184A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102360519A (en) * | 2011-07-29 | 2012-02-22 | 电子科技大学 | Intrusion detecting and positioning method for distributed optical fiber fence |
| CN103824406A (en) * | 2014-02-08 | 2014-05-28 | 上海波汇通信科技有限公司 | Optical fiber sensing-based tunnel cable invasive detection method and system |
| CN108286991A (en) * | 2017-01-10 | 2018-07-17 | 光子瑞利科技（北京）有限公司 | Application based on the optical fiber safety-protection system of rayleigh scattering principle in campus |
| CN107369282A (en) * | 2017-08-23 | 2017-11-21 | 武汉理工光科股份有限公司 | A kind of buried COTDR circumferences intrusion alarm system and signal processing method |
Non-Patent Citations (2)
| Title |
|---|
| 吴耿: "基于微结构光纤分布式传感系统的铁路安全监测研究", 《中国优秀硕士学位论文全文数据库 工程科技I辑》 * |
| 文常保 等: "《人工神经网络理论及应用》", 31 March 2019 * |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112883802A (en) * | 2021-01-20 | 2021-06-01 | 刘信 | Method for identifying destructive event of pipeline optical fiber vibration safety early warning system |
| CN116204784A (en) * | 2022-12-30 | 2023-06-02 | 成都天仁民防科技有限公司 | DAS-based subway tunnel external hazard operation intrusion recognition method |
| CN116204784B (en) * | 2022-12-30 | 2023-09-08 | 成都天仁民防科技有限公司 | DAS-based subway tunnel external hazard operation intrusion recognition method |
| CN118243626A (en) * | 2024-05-21 | 2024-06-25 | 西安科技大学 | Photothermal spectrum gas detection device and method based on micro-nano optical fiber array |
| CN118243626B (en) * | 2024-05-21 | 2024-08-02 | 西安科技大学 | Photothermal spectrum gas detection device and method based on micro-nano optical fiber array |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Liu et al. | Vehicle detection and classification using distributed fiber optic acoustic sensing | |
| Shiloh et al. | Efficient processing of distributed acoustic sensing data using a deep learning approach | |
| Shao et al. | Data-driven distributed optical vibration sensors: a review | |
| Catalano et al. | Automatic traffic monitoring by ϕ-OTDR data and Hough transform in a real-field environment | |
| Meng et al. | Research on recognition method of railway perimeter intrusions based on Φ-OTDR optical fiber sensing technology | |
| US11783452B2 (en) | Traffic monitoring using distributed fiber optic sensing | |
| CN110660184A (en) | Adaboost-based railway perimeter early warning method of fiber laser radar | |
| Shiloh et al. | Deep learning approach for processing fiber-optic DAS seismic data | |
| CN111854921A (en) | Distributed optical fiber deceleration strip vibration early warning system and method | |
| He et al. | A method for real-time monitoring of running trains using Ф-OTDR and the improved Canny | |
| JP7539582B2 (en) | Distributed Intelligent SNAP Informatics | |
| US12000729B2 (en) | Perpendicular distance prediction of vibrations by distributed fiber optic sensing | |
| He et al. | Application of distributed acoustic sensor technology in train running condition monitoring of the heavy-haul railway | |
| Tian et al. | Intrusion signal recognition in OFPS under multi-level wavelet decomposition based on RVFL neural network | |
| CN105608822A (en) | Railway perimeter intrusion vibration fiber alarm method and device | |
| Jia et al. | Event Identification by F-ELM Model for $\varphi $-OTDR Fiber-Optic Distributed Disturbance Sensor | |
| Li et al. | An anti-noise ϕ-OTDR based distributed acoustic sensing system for high-speed railway intrusion detection | |
| KR20220146670A (en) | Traffic anomaly detection methods, devices, devices, storage media and programs | |
| CN112032575B (en) | Pipeline safety monitoring method and system based on weak grating and storage medium | |
| Timofeev et al. | Classification of seismoacoustic emission sources in fiber optic systems for monitoring extended objects | |
| Wang et al. | Employing fiber sensing and on-premise AI solutions for cable safety protection over telecom infrastructure | |
| CN111539393B (en) | Oil-gas pipeline third-party construction early warning method based on EMD decomposition and LSTM | |
| US20240003717A1 (en) | Novel ai-driven approach to detect adn localize vehicle emergency stops based on vehicle trajectory using distributed fiber optic sensing (dfos) data | |
| He et al. | An Efficient Separation and Identification Algorithm for Mixed Threatening Events Applied in Fiber-Optic Distributed Acoustic Sensor | |
| US12051326B2 (en) | Road traffic extraction for unknown anomaly detection using distributed fiber optic sensing |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20200107 |