Railway intrusion location recognition monitoring method and related products

Abstract

The present application relates to the field of optical fiber sensing monitoring, and particularly relates to a railway intrusion limit identification monitoring method based on optical fiber vibration sensing and related products, the method lays sensing optical fiber along the railway line, collects back Rayleigh scattering light of the detection pulse and constructs an original data matrix; according to the distribution characteristics of signals at different spatial positions of the optical fiber, differentiated noise reduction intensity is determined for each position and filtering is performed; effective vibration segments are extracted from the noise-reduced signals, the corresponding intrusion limit event mileage is determined, a multi-dimensional feature vector is extracted to construct a feature matrix, an intrusion limit identification model is obtained by using a machine learning classifier for training, and the mileage and type of the intrusion limit event are determined for the real-time collected signals; compared with artificial patrol and video monitoring, the method is not limited by the patrol cycle, there is no monitoring blind area between adjacent cameras, long and large sections can be continuously covered, and the position of the intrusion limit event along the line can be directly given.

Description

Technical Field

[0001] This invention relates to the field of fiber optic sensing and monitoring technology, specifically to a railway encroachment positioning and identification monitoring method and related products based on fiber optic vibration sensing. Background Technology

[0002] Railways are vital transportation routes for passengers and freight, and the safety of the railway lines directly affects train operations. Common threats to railway tracks include landslides and rockfalls, construction machinery, and personnel climbing over fences to enter the tracks. Traditional monitoring of track encroachments relies on manual patrols and video surveillance.

[0003] Manual patrols rely on inspectors to conduct regular foot or vehicle patrols along the route. Due to limitations in manpower and patrol cycles, it is difficult to achieve 24 / 7, uninterrupted coverage of long sections. There is often a long time interval between the occurrence and discovery of an incident. At night, in inclement weather, or in sparsely populated areas, the timeliness of patrols is further reduced.

[0004] Video surveillance captures images through cameras, but the monitoring range of a single camera is limited. To cover long lines, a large number of points need to be deployed, along with power supply and transmission conditions. There are still blind spots between adjacent cameras. At night, when there is insufficient illumination, in rain, snow, fog, or when the target is obscured by vegetation or structures, the image quality and recognition effect will be significantly reduced. Moreover, video surveillance can usually only reflect the general picture of the event and cannot directly give the precise location of the intrusion event along the line.

[0005] Existing manual patrols and video surveillance methods are insufficient to achieve continuous coverage along railway lines. They cannot accurately locate the encroachment position in a short period of time after an incident occurs, nor can they reliably distinguish between different types of encroachment incidents such as landslides, rockfalls, mechanical construction, and personnel crossing over. They cannot meet the requirements of railway line safety monitoring for timeliness, continuity, and accurate location identification. Summary of the Invention

[0006] To address the aforementioned technical problems, this invention provides a railway encroachment positioning and identification monitoring method and related products based on fiber optic vibration sensing. It utilizes sensing optical fibers laid along the railway line to continuously collect vibration data. By applying differentiated noise reduction processing at different spatial locations along the fiber optic line, environmental noise is suppressed while retaining the vibration characteristics of the encroachment event. Furthermore, based on multidimensional features and a machine learning classifier, accurate identification of the mileage and type of the encroachment event is achieved.

[0007] This invention is achieved through the following technical solution:

[0008] A railway encroachment positioning and monitoring method based on fiber optic vibration sensing includes the following steps:

[0009] A sensing optical fiber is laid along the railway line, and the backscattered Rayleigh light of the detection pulse in the sensing optical fiber is collected to obtain the original reflection trajectory of each detection pulse.

[0010] An original data matrix is ​​constructed based on the original reflection trajectory of each of the probe pulses. The original data matrix is ​​then denoised, and effective vibration segments are extracted from the denoised signal.

[0011] Extract the multidimensional feature vector of the effective vibration segment, construct a feature matrix based on the multidimensional feature vector of multiple intrusion event samples, and train the feature matrix using a machine learning classifier to obtain an intrusion identification model;

[0012] The system processes the real-time acquired signals and extracts valid vibration segments, determines the mileage of the intrusion event corresponding to the valid vibration segment, and uses the intrusion identification model to classify the multidimensional feature vector of the valid vibration segment to obtain the type of intrusion event.

[0013] Optionally, the sensing fiber can be vibrated by actively applying multiple intrusion events along the railway line, and the backscattered Rayleigh light during the vibration can be collected as intrusion event samples. The multiple intrusion events include landslides, rockfalls, mechanical construction, and personnel crossing.

[0014] Optionally, the method for denoising the original data matrix includes:

[0015] Based on the distribution characteristics of the signals in the original data matrix at different spatial locations in the sensing fiber, the corresponding noise reduction intensity is determined for different spatial locations;

[0016] The original data matrix is ​​filtered according to the noise reduction intensity corresponding to each spatial location to obtain the noise-reduced signal matrix.

[0017] Optionally, methods for determining the corresponding noise reduction intensity include:

[0018] Calculate the local entropy of the signal at each spatial location. , ,in The column index corresponding to the spatial location. For position The probability density of the signal amplitude distribution at that location. To quantize the number of gray levels;

[0019] The adaptive diffusion intensity coefficient at this spatial location is determined based on the local entropy. , ,in As the reference diffusion intensity coefficient, and These represent the minimum and maximum values ​​of the local entropy at each spatial location, respectively.

[0020] A method for filtering the original data matrix includes:

[0021] An anisotropic diffusion equation is established based on the aforementioned adaptive diffusion intensity coefficient. ,in Let be the signal matrix to be filtered. The number of diffusion iterations. For the signal gradient, For edge stopping functions;

[0022] The original data matrix is ​​iteratively diffused and filtered to obtain the denoised signal matrix.

[0023] Optionally, a method for extracting effective vibration segments from the denoised signal includes:

[0024] Calculate the differential signal between the reflection trajectories of adjacent probe pulses after noise reduction processing. ;

[0025] The short-time energy of the differential signal within the sliding window is calculated. , ,in The starting pulse index for the sliding window. The length of the sliding window. This represents the total number of spatial locations.

[0026] short-term energy The signal segment corresponding to the sliding window that exceeds the energy threshold is determined as the effective vibration segment.

[0027] Optionally, the method for determining the intrusion event mileage corresponding to the effective vibration segment includes:

[0028] Determine the column index of the sampling point corresponding to the vibration peak value in the effective vibration segment in the original data matrix. ;

[0029] according to Determine the mileage of the intrusion event relative to the monitoring start point. ,in The speed of light in a vacuum The refractive index of the optical fiber. This represents the sampling time interval.

[0030] Optionally, the method for extracting the multidimensional feature vector of the effective vibration segment includes:

[0031] Multiple different categories of features are extracted from the effective vibration segments; the multiple different categories of features include time-domain features, frequency-domain features, time-frequency-domain features, and higher-order statistical features;

[0032] The features of the multiple different categories are concatenated and normalized to obtain the multidimensional feature vector of the effective vibration segment.

[0033] Optionally, the machine learning classifier is a support vector machine; the kernel function of the support vector machine is a radial basis function kernel function, and the parameters of the kernel function are determined by grid search;

[0034] A method for training an intrusion detection model using a machine learning classifier on the feature matrix includes:

[0035] A training sample set is constructed using the feature matrix and the type labels corresponding to each intrusion event sample;

[0036] The classification hyperplane is obtained by solving the soft-margin quadratic programming problem of the support vector machine.

[0037] The support vector set, support vector coefficients, bias terms, kernel function parameters, and event type mapping table of the support vector machine are saved as the intrusion detection model.

[0038] A railway encroachment positioning and identification monitoring system based on fiber optic vibration sensing includes a sensing fiber, an optical signal acquisition unit, and a data processing unit.

[0039] The sensing optical fiber is laid along the railway line;

[0040] The optical signal acquisition unit is connected to the sensing optical fiber and is used to transmit detection pulses to the sensing optical fiber and acquire the backscattered Rayleigh light of the detection pulses to obtain the original reflection trajectory of each detection pulse.

[0041] The data processing unit is connected to the optical signal acquisition unit and is used to construct an original data matrix based on each original reflection trajectory and perform noise reduction processing. It extracts effective vibration segments from the noise-reduced signal, extracts the multi-dimensional feature vector of the effective vibration segments, and uses the intrusion identification model to determine the mileage and type of the intrusion event.

[0042] A computer-readable storage medium having a computer program stored thereon, characterized in that, when the computer program is executed by a processor, it implements the steps of the railway encroachment positioning and identification monitoring method based on fiber optic vibration sensing as described above.

[0043] This invention utilizes sensing optical fibers laid along railway lines to continuously collect vibration data. It suppresses environmental noise and preserves the vibration characteristics of the events through location-adaptive noise reduction processing. Then, through multi-dimensional feature extraction and machine learning classification, it achieves mileage location and type identification of intrusion events. Compared with existing technologies, it has the following advantages:

[0044] By using sensing optical fibers laid along the railway line to continuously collect and monitor vibrations along the line, compared with manual patrols and video surveillance, it is not limited by the patrol cycle and there are no blind spots between adjacent cameras. It can continuously cover long sections and directly give the location of the intrusion event along the line.

[0045] Based on the local entropy of the signal at different spatial locations in the optical fiber, a differentiated noise reduction intensity is determined for that location. In event segments with cluttered signals and high entropy values, the diffusion intensity is reduced to preserve vibration details, while in noise segments with stable signals and low entropy values, the diffusion intensity is increased to suppress noise. This approach can suppress environmental noise while preserving the weak vibration characteristics of intrusion events.

[0046] The support vector machine, which is suitable for small samples and high dimensions, is used for training. The parameter model of the intrusion event is then used for the localization and identification of the actual event. This is in line with the application scenario of railway intrusion monitoring, where the sample collection volume is limited and the feature dimension is high, making the training classification and identification effect more accurate. Attached Figure Description

[0047] The accompanying drawings illustrate exemplary embodiments of the present invention and, together with the description thereof, serve to explain the principles of the invention. These drawings are included to provide a further understanding of the invention and are incorporated in and constitute a part of this specification, but do not constitute a limitation on the embodiments of the present invention.

[0048] Figure 1 This is a flowchart illustrating the railway encroachment positioning and monitoring method based on fiber optic vibration sensing provided in an embodiment of the present invention.

[0049] Figure 2 This is a schematic diagram of the fiber optic cable laying method provided in an embodiment of the present invention;

[0050] Figure 3 This is a comparative diagram of a certain signal segment before and after noise reduction in an event of personnel crossing the boundary in an embodiment of the present invention, wherein (a) is the original signal segment and (b) is the signal segment after noise reduction; Detailed Implementation

[0051] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the scope of the invention.

[0052] It should also be noted that, for ease of description, only the parts relevant to the present invention are shown in the accompanying drawings.

[0053] Where there is no conflict, the embodiments and features described in the present invention can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0054] Example 1

[0055] This embodiment combines Figure 1 This paper provides an overall description of the scheme for railway encroachment positioning and identification monitoring method based on fiber optic vibration sensing.

[0056] Using a single sensing optical fiber laid along the railway line as both the sensing medium and the signal transmission medium, the backscattered Rayleigh light of the probe pulse in the optical fiber is collected to obtain a signal that reflects the vibration at various spatial locations along the line. Environmental noise is first suppressed from this signal, and then segments containing intrusion events are extracted. Multidimensional features that can distinguish different events are extracted from these segments. An intrusion identification model is first trained using known event samples, and then used during actual monitoring to simultaneously provide the mileage location and event type of the collected events along the line.

[0057] like Figure 1 As shown, this method includes the following four steps in sequence:

[0058] The first step is to lay sensing optical fibers along the railway line within the monitoring range, inject detection pulses into the optical fibers, and collect the backscattered Rayleigh light of each detection pulse to obtain the original reflection trajectory of each detection pulse along the length of the optical fiber, providing raw data for subsequent processing.

[0059] The second step involves constructing an original data matrix based on the original reflection trajectories of each detection pulse, performing noise reduction on the original data matrix to suppress uneven environmental noise along the line, and extracting effective vibration segments containing intrusion events from the noise-reduced signal.

[0060] The third step is to extract multidimensional feature vectors of effective vibration segments. Multidimensional feature vectors are extracted from multiple known types of intrusion event samples, and feature matrices are constructed. A machine learning classifier is used to train the feature matrices to obtain an intrusion identification model.

[0061] The fourth step is to perform noise reduction and effective vibration segment extraction on the real-time acquired signals in the same way as described above, determine the mileage of the intrusion event corresponding to the effective vibration segment, and use the intrusion identification model to classify its multi-dimensional feature vector to obtain the type of intrusion event, thereby simultaneously determining the location and type of the intrusion event.

[0062] The first three steps described above constitute the training phase, used to build the intrusion detection model; the fourth step is the actual monitoring phase, where the trained model is used to process real-time signals. Through these steps, continuous monitoring along the railway line can be achieved using only a single sensing fiber optic cable laid along the line, providing the mileage location and event type simultaneously upon detecting an intrusion event.

[0063] Example 2

[0064] This embodiment describes the first step in Embodiment 1.

[0065] Sensor optical fibers are laid along the railway line within the monitoring range. Detection pulses are periodically injected into the sensor optical fibers through optical signal acquisition equipment, and the backscattered Rayleigh light generated by each detection pulse in the optical fiber is collected to obtain the original reflection trajectory of each detection pulse along the length of the optical fiber.

[0066] To obtain intrusion event samples with type labels, various intrusion events were actively applied along the railway line to induce vibration in the sensing fiber. Backscattered light during the vibration was collected as intrusion event samples. These events included landslides, rockfalls, mechanical construction, and personnel crossing. Probe pulses were injected into the sensing fiber, and the backscattered light from each pulse was collected to obtain the original reflection trajectory of each pulse along the fiber's length. The length of the sensing fiber can be determined based on the monitoring range, for example, from 1 km to 20 km; the sampling frequency... The rate can range from 10 MSa / s to 100 MSa / s; the number of probe pulses injected per second. The frequency can be from 1kHz to 10kHz.

[0067] Number of sampling points per probe pulse ,For example That is, the original reflection trajectory of each probe pulse contains 10,000 sampling points, and each sampling point corresponds to a spatial position on the optical fiber.

[0068] Example 3

[0069] This embodiment describes the second step in Embodiment 1, including noise reduction and extraction of effective vibration segments.

[0070] The background noise levels vary in different sections along the railway line. Therefore, instead of applying a uniform denoising intensity to the entire original data matrix, we determine the corresponding denoising intensity for different spatial locations based on the signal distribution characteristics at different spatial locations in the sensing fiber. Then, we filter the original data matrix according to the denoising intensity corresponding to each spatial location.

[0071] S2.1 Arrange the original reflection trajectories of each acquired detection pulse row by row according to the pulse sequence to construct the original data matrix. : ,in, For the first The original reflection trajectory of the probe pulse, For synchronization flag, ;matrix The row direction corresponds to time (different detection pulses), and the column direction corresponds to space (different positions on the optical fiber).

[0072] S2.2 Calculate the local entropy of the signal at each spatial location. : ,in, The column index corresponding to the spatial location. The column index corresponding to the effective sensing fiber end; For position The probability density of the signal amplitude distribution at a given location; To quantize the number of gray levels, the value can range from 256 to 4096.

[0073] Local entropy This reflects the degree of noise in the signal at that spatial location: the entropy value is lower in the segment where noise is dominant and the signal is stable, while the entropy value is higher in the segment where there is intrusive vibration and the signal changes drastically.

[0074] S2.3 Determine the adaptive diffusion intensity coefficient for this spatial location based on local entropy. : ,in, As the reference diffusion intensity coefficient, and These are the minimum and maximum local entropy values ​​at each spatial location, respectively. The local entropy at each location is normalized to... to Between, multiply by the base coefficient This results in a higher diffusion intensity at locations with higher entropy values ​​and a lower diffusion intensity at locations with lower entropy values.

[0075] S2.4. Establishing anisotropic diffusion equations based on adaptive diffusion intensity coefficients: ,in, Given the signal matrix to be filtered, the original data matrix is ​​initially used. ; This represents the number of diffusion iterations. For the signal gradient; It is an edge stopping function used to reduce diffusion and preserve signal abrupt changes at points where the gradient is large.

[0076] Iterative solution The noise-reduced signal matrix is ​​then obtained. Number of iterations The value can range from 5 to 50. Figure 3The image shows a comparison of a signal segment before and after noise reduction during a person crossing the boundary. (a) is the original signal segment and (b) is the signal segment after noise reduction. It can be seen that the environmental noise is suppressed while the vibration characteristics corresponding to the event are preserved.

[0077] By applying differentiated diffusion intensities to different spatial locations, weaker diffusion preserves vibration details in event segments with higher entropy values, while stronger diffusion suppresses noise in noise segments with lower entropy values, thus achieving a balance between noise suppression and event feature preservation.

[0078] S2.5. Intrusion events manifest as signal variations between adjacent detection pulses over time, while the static background remains largely unchanged between adjacent pulses. Therefore, the differential signal between the reflection trajectories of adjacent detection pulses after denoising is first solved on the time scale. : .

[0079] S2.6. Based on the differential signal, calculate the short-time energy within the window using a sliding window. : ,in, The starting pulse index for the sliding window. ; The length of the sliding window. The shortest duration of an intrusion incident; This represents the total number of spatial locations. The value can range from 0.1s to 2s. Sliding window length at 5kHz .

[0080] When short-time energy Greater than the energy threshold When the corresponding window is determined to be a valid vibration segment, the spatial submatrix of the valid vibration segment is: ,in, This represents the number of pulses contained in the segment. Energy threshold. It can be taken as background noise energy Add several times the noise standard deviation, that is ,in The value of can range from 2 to 5, that is... .

[0081] Example 4

[0082] This embodiment describes the third step in Embodiment 1.

[0083] S3.1 The mileage of the intrusion event is determined based on the position of the vibration on the optical fiber. First, determine the column index of the sampling point corresponding to the vibration peak in the effective vibration segment in the original data matrix. The column index can be obtained by superimposing the differential signals within the effective vibration segments and taking the maximum value; then, the mileage of the intrusion event relative to the monitoring starting point is determined by the following formula. : ,in, For the speed of light in a vacuum, take ; The refractive index of the optical fiber can range from 1.44 to 1.50. The sampling time interval is... When taking 50 MSa / s .

[0084] S3.2, Based on effective vibration segments Extract vibration location Time series vector at: .

[0085] S3.3 Extract multiple different categories of features from the time series. These features include time-domain features, frequency-domain features, time-frequency domain features, and higher-order statistical features.

[0086] Extracting temporal feature vectors Frequency domain eigenvectors Time-frequency domain feature vectors and higher-order statistics eigenvectors : Among them, time-domain features include the mean. Root mean square ,variance and peak factor ;

[0087] Frequency domain characteristics include the dominant frequency Center of gravity frequency Frequency band energy ratio and spectral flatness ;

[0088] Time-frequency domain features include the energy distribution of wavelet packets in each frequency band. and wavelet packet entropy ;

[0089] Higher-order statistical characteristics include skewness and kurtosis .

[0090] S3.4. Concatenate, merge, and normalize the feature vectors from the four dimensions mentioned above to construct a multi-dimensional feature vector for each vibration event. : ,in, The total dimension of all features across the four dimensions can range from 10 to 100.

[0091] S3.5. Extract multi-dimensional feature vectors from multiple intrusion event samples, normalize them, and then construct a feature matrix. : ; Construct label vectors : ,in, For the first The normalized multidimensional feature vector of each invasion event sample , The total number of samples; For the first The category label of each sample, with a value of , This represents the total number of infringement incident types.

[0092] S3.6 In the case of railway encroachment scenarios, the cost of collecting event samples is high and the number of available samples is limited. In addition, the dimensionality of multidimensional feature vectors is high. Therefore, support vector machines, which are suitable for small samples and high dimensions, are adopted as machine learning classifiers.

[0093] A training sample set is constructed using the feature matrix and the type labels corresponding to each invasion event sample. Support vector machines obtain the classification hyperplane by solving the following soft-margin quadratic programming problem:

[0094] ;

[0095] ;

[0096] in, The hyperplane normal vector; For bias terms; These are slack variables; This is the penalty coefficient (here, it is the penalty parameter for the soft margin of the support vector machine, and is unrelated to the diffusion intensity coefficient in Example 3). This is the mapping function that maps feature vectors to a high-dimensional space. The kernel function can be a radial basis function, and its parameters are determined through grid search.

[0097] Since intrusion events typically include multiple types, the event type mapping table is constructed based on a one-to-one strategy, that is, a binary classifier is constructed for each of the two intrusion event types.

[0098] After training is complete, save the set of all support vectors. Support vector coefficients and Bias terms The kernel function parameters and event type mapping table serve as the intrusion detection model and are stored in a structured file.

[0099] Thus, SVM is used to transform vibration events from feature space to category space.

[0100] Example 5

[0101] This embodiment describes the fourth step in Embodiment 1.

[0102] In actual monitoring, the real-time acquired signals are processed according to Examples 3 and 4 to obtain the multidimensional feature vector of the vibration event to be identified. Calculate the decision function value of each binary classifier using the following formula: ,in, This is the kernel function.

[0103] Based on the decision results of each binary classifier, the type of intrusion event is determined by voting, ultimately achieving real-time acquisition of the specific mileage and type of the intrusion event.

[0104] Example 6

[0105] This embodiment provides a specific application example.

[0106] S1. Fiber optic vibration sensors are used at a railway test site to monitor intrusion events within a 500m range (denoted as K0+000~K0+500).

[0107] First, fiber optic sensors are laid along the railway line within the monitoring range. The fiber optic laying method is as follows: Figure 2 As shown, where , , The system actively created multiple intrusion events that caused the optical fiber to vibrate, including 150 landslides and rockfalls, 150 mechanical constructions, and 150 personnel crossings. It also collected the backscattered Rayleigh light of the probe pulses in the optical fiber to obtain the original reflection trajectory.

[0108] S2. Using the original trajectory of the probe pulse from S1, construct the original data matrix. Apply an adaptive diffusion coefficient filtering algorithm based on signal local entropy to denoise the original data matrix, where the preset quantization grayscale level G=1024 and the number of iterations t=20. Then, extract the effective vibration segments. Take 0.5s, energy threshold Pick The comparison results of a certain segment of raw data before and after noise reduction in a personnel intrusion incident, and the effective vibration segments are as follows: Figure 3 As shown, sub-figure (a) is the original signal segment, sub-figure (b) is the denoised signal segment, and the area marked by the red dashed box in the figure is the detected effective vibration segment;

[0109] S3. Using the effective vibration segments from S2, calculate the multidimensional feature vector of the intrusion event, and construct a feature matrix based on this, where the dimension D=30; then train the feature matrix using a Support Vector Machine (SVM), and optimize the hyperparameters using a grid search method. The radial basis function (RBF) kernel is selected. Since the intrusion event includes three types of events: rockfall, machinery, and personnel, the event type mapping table has a total of... There are 3 binary classifiers, including [personnel and machinery], [personnel falling rocks], and [machine falling rocks]; the number of support vectors for each classifier is 14, 19, and 23 respectively; save all support vector sets, support vector coefficients, bias terms, kernel function parameters, and event type mapping tables;

[0110] S4. Deploy the intrusion parameter model trained in S3 into the railway intrusion monitoring system in the form of a JSON file; then conduct a real intrusion event test, test the impact of landslide rocks on the fence and the optical fiber near K0+210, and after the optical fiber vibrates, the system automatically calculates that the specific mileage of this intrusion event is about K0+210, the event type is landslide rocks, and the intrusion location identification and monitoring is achieved.

[0111] Example 7

[0112] This embodiment provides a railway encroachment positioning and identification monitoring system based on fiber optic vibration sensing. The system includes a sensing fiber, an optical signal acquisition unit, and a data processing unit. The sensing fiber is laid along the railway fence, and the top of the fence is equipped with barbed wire coils. When an encroachment event such as a landslide, falling rock, mechanical construction, or personnel crossing the fence touches the fence or barbed wire coils, the vibration is transmitted to the sensing fiber.

[0113] The optical signal acquisition unit is connected to the sensing fiber and is used to transmit probe pulses into the sensing fiber and acquire the backscattered Rayleigh light of the probe pulses to obtain the original reflection trajectory of each probe pulse.

[0114] The data processing unit is connected to the optical signal acquisition unit and is used to construct an original data matrix based on each original reflection trajectory and perform noise reduction processing. It then extracts effective vibration segments from the noise-reduced signal, extracts multi-dimensional feature vectors from these effective vibration segments, and uses an intrusion identification model to determine the mileage and type of the intrusion event. The specific implementation of the processing performed by each unit is the same as in Examples 1 to 6.

[0115] A computer program product comprising a computer program / instructions that, when executed by a processor, implement the steps of any of the methods described above.

[0116] Computer program products include computer programs or instruction sets used to perform specific tasks or achieve specific functions. These programs or instructions are designed to be executed by a processor to implement a series of predefined steps or operations. The program product may be stored in various forms of computer storage media, such as memory, hard disks, solid-state drives, optical discs, or other forms of digital storage devices. It may exist in the form of compiled binary code or in the form of scripts or bytecode that can be executed by an interpreter. Through carefully designed algorithms and logical instructions, the program product enables the processor to process data in a specific order and manner, performing various functions such as data analysis, user interaction, and device control.

[0117] In the description of this specification, the references to terms such as "one embodiment / mode," "some embodiments / modes," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment / mode or example is included in at least one embodiment / mode or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment / mode or example. Moreover, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments / modes or examples. Furthermore, without contradiction, those skilled in the art can combine and integrate the different embodiments / modes or examples described in this specification, as well as the features of different embodiments / modes or examples.

[0118] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this application, "multiple" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0119] Those skilled in the art should understand that the above embodiments are merely for illustrating the present invention and are not intended to limit the scope of the invention. Those skilled in the art can make other changes or modifications based on the above invention, and these changes or modifications still fall within the scope of the present invention.

Claims

1. A railway encroachment positioning and monitoring method based on fiber optic vibration sensing, characterized in that, Includes the following steps: A sensing optical fiber is laid along the railway line, and the backscattered Rayleigh light of the detection pulse in the sensing optical fiber is collected to obtain the original reflection trajectory of each detection pulse. An original data matrix is ​​constructed based on the original reflection trajectory of each of the probe pulses. The original data matrix is ​​then denoised, and effective vibration segments are extracted from the denoised signal. Extract the multidimensional feature vector of the effective vibration segment, construct a feature matrix based on the multidimensional feature vector of multiple intrusion event samples, and train the feature matrix using a machine learning classifier to obtain an intrusion identification model; The system processes the real-time acquired signals and extracts valid vibration segments, determines the mileage of the intrusion event corresponding to the valid vibration segment, and uses the intrusion identification model to classify the multidimensional feature vector of the valid vibration segment to obtain the type of intrusion event.

2. The railway encroachment positioning and monitoring method based on fiber optic vibration sensing according to claim 1, characterized in that, By actively applying various intrusion events along the railway line to cause the sensing optical fiber to vibrate, the backscattered Rayleigh light during the vibration is collected as intrusion event samples. The various intrusion events include landslides, rockfalls, mechanical construction, and personnel crossing.

3. The railway encroachment positioning and monitoring method based on fiber optic vibration sensing according to claim 1, characterized in that, A method for denoising the original data matrix includes: Based on the distribution characteristics of the signals in the original data matrix at different spatial locations in the sensing fiber, the corresponding noise reduction intensity is determined for different spatial locations; The original data matrix is ​​filtered according to the noise reduction intensity corresponding to each spatial location to obtain the noise-reduced signal matrix.

4. The railway encroachment positioning and monitoring method based on fiber optic vibration sensing according to claim 3, characterized in that, Methods for determining the appropriate noise reduction intensity include: Calculate the local entropy of the signal at each spatial location. , ,in The column index corresponding to the spatial location. For position The probability density of the signal amplitude distribution at that location. To quantize the number of gray levels; The adaptive diffusion intensity coefficient at this spatial location is determined based on the local entropy. , ,in As the reference diffusion intensity coefficient, and These represent the minimum and maximum values ​​of the local entropy at each spatial location, respectively. A method for filtering the original data matrix includes: An anisotropic diffusion equation is established based on the aforementioned adaptive diffusion intensity coefficient. ,in Let be the signal matrix to be filtered. For the number of diffusion iterations, For the signal gradient, For edge stopping functions; The original data matrix is ​​iteratively diffused and filtered to obtain the denoised signal matrix.

5. The railway encroachment positioning and monitoring method based on fiber optic vibration sensing according to claim 1, characterized in that, Methods for extracting effective vibration segments from denoised signals include: Calculate the differential signal between the reflection trajectories of adjacent probe pulses after noise reduction processing. ; The short-time energy of the differential signal within the sliding window is calculated. , ,in The starting pulse index of the sliding window. The length of the sliding window. This represents the total number of spatial locations. short-term energy The signal segment corresponding to the sliding window that exceeds the energy threshold is determined as the effective vibration segment.

6. The railway encroachment positioning and monitoring method based on fiber optic vibration sensing according to claim 1, characterized in that, The method for determining the intrusion event mileage corresponding to the effective vibration segment includes: Determine the column index of the sampling point corresponding to the vibration peak value in the effective vibration segment in the original data matrix. ; according to Determine the mileage of the intrusion event relative to the monitoring start point. ,in The speed of light in a vacuum The refractive index of the optical fiber. This represents the sampling time interval.

7. The railway encroachment positioning and monitoring method based on fiber optic vibration sensing according to claim 1, characterized in that, The method for extracting the multidimensional feature vector of the effective vibration segment includes: Multiple different categories of features are extracted from the effective vibration segments; the multiple different categories of features include time-domain features, frequency-domain features, time-frequency-domain features, and higher-order statistical features; The features of the multiple different categories are concatenated and normalized to obtain the multidimensional feature vector of the effective vibration segment.

8. The railway encroachment positioning and monitoring method based on fiber optic vibration sensing according to claim 1, characterized in that, The machine learning classifier is a support vector machine; the kernel function of the support vector machine is a radial basis function kernel function, and the parameters of the kernel function are determined by grid search. A method for training an intrusion detection model using a machine learning classifier on the feature matrix includes: A training sample set is constructed using the feature matrix and the type labels corresponding to each intrusion event sample; The classification hyperplane is obtained by solving the soft-margin quadratic programming problem of the support vector machine. The support vector set, support vector coefficients, bias terms, kernel function parameters, and event type mapping table of the support vector machine are saved as the intrusion detection model.

9. A railway encroachment positioning and monitoring system based on fiber optic vibration sensing, characterized in that, It includes sensing optical fiber, optical signal acquisition unit and data processing unit; The sensing optical fiber is laid along the railway line; The optical signal acquisition unit is connected to the sensing optical fiber and is used to transmit detection pulses to the sensing optical fiber and acquire the backscattered Rayleigh light of the detection pulses to obtain the original reflection trajectory of each detection pulse. The data processing unit is connected to the optical signal acquisition unit and is used to construct an original data matrix based on each original reflection trajectory and perform noise reduction processing. It extracts effective vibration segments from the noise-reduced signal, extracts the multi-dimensional feature vector of the effective vibration segments, and uses the intrusion identification model to determine the mileage and type of the intrusion event.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the railway encroachment positioning and identification monitoring method based on fiber optic vibration sensing as described in any one of claims 1 to 8.

Images