A Train Operation Status Monitoring Method Based on DAS Spatiotemporal Sequence Imagery

Abstract

This invention relates to the field of signal source monitoring, specifically to a train operation status monitoring method based on DAS spatiotemporal sequence imaging. The invention employs a time-domain frame-segmented wavelet packet feature extraction method to preprocess the train vibration signal, removing trailing vibration signal interference. Then, based on image binarization and morphological filtering methods in image processing algorithms, the train trajectory is feature-enhanced. Image segmentation is used to detect local straight lines, k-means clustering is used to classify the corresponding straight lines of the train, and polynomial fitting is used to extract the train trajectory. Finally, based on the obtained train trajectory, the wavelet packet energy feature value matrix is ​​truncated, and an adaptive threshold method is used to obtain the train's forward and backward trajectories, thereby achieving accurate monitoring of train length and speed, effectively solving the accuracy problem of existing DAS systems in monitoring train operation status.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of signal source monitoring, in particular to a train operation state monitoring method based on DAS space-time sequence imaging. BACKGROUND

[0002] At present, the railway is one of the main forces in the transportation industry in the national economic construction, and the real-time monitoring of the train operation state has great significance for maintaining railway safety and ensuring national economy. DAS system is widely used in structure monitoring, pipeline security, perimeter pipeline and other fields due to its advantages of anti-electromagnetic radiation, corrosion resistance, long service life and low power consumption. It has very broad application prospects to apply DAS system to the real-time monitoring of train operation state.

[0003] In recent years, researchers have introduced DAS technology into railway track detection, which has achieved certain results. In 2017, Z. Wang et al. used a railway vibration detection scheme based on phase-sensitive OTDR, proposed a multi-dimensional comprehensive analysis to identify train signals and illegal intrusions during high-speed rail operation, successfully monitored the train operation signals, and calculated the length and running speed of the train, and identified some illegal disturbances along the railway. In 2019, H. Meng et al. used phase-sensitive optical time domain reflectometer (Ф-OTDR) and improved Canny to monitor running trains in real time and accurately, and extracted the vibration signals of running trains through difference and threshold.

[0004] However, in the process of using DAS system for real-time monitoring of train operation state, the "tail" signal of train vibration before and after the train will seriously interfere with the effective extraction of train vibration signal, and it is impossible to accurately track a specific train in the case of multiple trains crossing, so as to cause the real-time position, train length and real-time speed of the train cannot be accurately monitored. This poses a major challenge to the application of DAS system in real-time monitoring of train operation state. SUMMARY

[0005] In view of the above problems and deficiencies, in order to solve the accuracy problem of DAS system in monitoring the train operation state, the present application provides a train operation state monitoring method based on DAS space-time sequence imaging.

[0006] A train operation state monitoring method based on DAS space-time sequence imaging, comprising the following steps:

[0007] Step 1, for the train vibration space-time intensity image obtained by the DAS system, a time domain framing method is adopted, and the time domain framing length L (≤ total time domain length L TThe train vibration signal is processed by frame segmentation, and then wavelet packet decomposition is performed on each frame signal to obtain the wavelet packet energy eigenvalue matrix. The wavelet packet energy eigenvalue matrix is ​​used to remove the interference of the "tailed" vibration signal.

[0008] Step 2: The wavelet packet energy eigenvalue matrix obtained in Step 1 is visualized, and the train trajectory in the image is enhanced using image binarization and morphological filtering methods in image processing algorithms. A matrix of x×x (x≤L) pixels is selected for image segmentation after feature enhancement. Local line detection is then performed on each segmented x×x image to obtain the (ρ,θ) parameters in the polar coordinate system. Finally, k-means clustering is used to classify the corresponding lines of the trains based on the (k,b) parameters of all lines, and the curve trajectories of multiple trains are extracted using a polynomial fitting method to obtain the train curve trajectory equation. Here, (ρ,θ) are the parameters of the lines in the image coordinate system, k is the slope parameter of the line, and b is the intercept parameter of the line.

[0009] Step 3: Map the train trajectory equation obtained in Step 2 onto the wavelet packet energy eigenvalue matrix obtained in Step 1, and truncate it according to the wavelet packet energy eigenvalue matrix. Use an adaptive threshold method to obtain the train's forward and backward trajectories. Subtract the forward and backward trajectories to obtain the train length variation curve over time. After eliminating the interference of train intersection terms, the train length is obtained. Finally, the corresponding real-time train speed curve is obtained based on the train length, thereby achieving accurate monitoring of train length and speed.

[0010] Furthermore, in step 1, the train vibration signal is processed by time-domain framing, and the train vibration signal is preprocessed using wavelet packet energy feature extraction. Specific steps include:

[0011] Step 1-1: Select time window as L (time domain frame length), and divide the train vibration spatiotemporal intensity map obtained by the DAS system into frames along the time direction to obtain the framed train vibration data.

[0012] Step 1-2: Select the wavelet packet decomposition level N, perform wavelet packet decomposition on the train vibration data of each frame, and calculate the wavelet packet energy feature vector for each wavelet packet coefficient obtained by decomposition according to equation (1).

[0013]

[0014] Among them, C j (m) are the decomposed wavelet packet coefficients, and N is the number of wavelet packet decomposition layers;

[0015] Steps 1-3: Select the wavelet packet energy eigenvectors in which the difference between the train vibration signal and the residual vibration signal is greater than 20% of the wavelet packet energy eigenvectors of the residual vibration signal, sum them, and obtain the wavelet packet energy eigenvalue matrix. Use this matrix as the eigenvalue matrix of the "tailed" vibration signal to distinguish between the train vibration signal part and the residual vibration signal part.

[0016] Furthermore, the specific steps of step 2, which uses image processing algorithms to extract the train trajectory, are as follows:

[0017] Step 2-1: The wavelet packet energy eigenvalue matrix obtained in Step 1 is imaged using grayscale mapping, and then binarized using the OTSU thresholding method. Finally, morphological filtering is performed using the closing operation in morphological filtering to enhance the train trajectory image.

[0018] Step 2-2: For the enhanced image in Step 2-1, segment it into units of x×x, (x≤L) pixels. Perform Hough line detection on each segmented x×x image to obtain the (ρ,θ) parameters of the image in the polar coordinate system. Select q (q≤Q) lines from the Q line detection results. Classify the lines up and down according to the positive and negative values ​​of the θ parameters. Use the parameter mapping relationship (2) from the polar coordinate system (ρ,θ) to the linear coordinate system line y=kx+b to obtain the expression of the line detection result. Calculate the mean value of the lines classified up and down to obtain the line of each x×x image.

[0019]

[0020] Where (ρ,θ) are the parameters of the line in the image coordinate system, k is the slope parameter of the line, and b is the intercept parameter of the line;

[0021] Steps 2-3: After completing the line detection for each image block, combine all x×x images into the original train trajectory feature image. In the original feature image, the trajectory of a train can be approximated as a combination of multiple straight lines, and the (k,b) parameters of these lines are similar. Using the (k,b) parameters of each line as its feature, the k-means clustering algorithm is used to cluster all the lines in the image, thus dividing all the lines into several classes, each corresponding to a train in the image.

[0022] Steps 2-4: For the straight lines that have completed k-means clustering, use the fourth-order polynomial fitting method to extract the train curve trajectory and obtain the train curve trajectory equation.

[0023] Furthermore, the specific steps for accurately monitoring train length and speed in step 3 include:

[0024] Step 3-1: Map the train curve trajectory equation obtained in Step 2 to the feature matrix obtained in Step 1. The specific mapping relationship is as follows:

[0025] trace ts =D-trace i (3)

[0026] Where D is the width of the image, i.e., the maximum number of frames in step 1 for signal framing, trace i The equation of the trajectory curve in the image coordinate system is called trace. ts The equation of the trajectory curve in the spacetime coordinate system;

[0027] Step 3-2: Initialize an all-zero train matrix TrainMat, with the same size as the wavelet packet eigenvalue matrix CharMat. Select trace as the starting point. ts Centered on the time domain, segments are truncated before and after the time band, with a truncation width of Ti, where Ti is the time domain length after removing the "tail" vibration signal. The truncation formula is expressed as:

[0028]

[0029] Where M is the spatial length of the characteristic matrix.

[0030] Step 3-3: Along the time direction of the wavelet packet energy eigenvalue matrix, obtain the adaptive threshold according to Equation (5), and obtain the front and rear trajectory points of the train for each frame of the train feature space distribution.

[0031] thread(i)=mean(TrainMat(:,i))+max(TrainMat(:,i)) / 3,i=1,2,...,D (5)

[0032] In the formula, mean(.) represents the mean of the signal, and TrainMat(:,i) represents the spatial distribution of the signal in the i-th frame.

[0033] When determining the front and rear trajectory points of the train feature spatial distribution, for each frame of spatial distribution, the first spatial point greater than the threshold thread(i) is found from front to back as the front trajectory point of the train, and the first spatial point greater than the threshold thread(i) is found from back to front as the rear trajectory point of the train.

[0034] After obtaining the trajectory points before and after each frame's spatial distribution, since there is a certain error between different points, in order to eliminate the influence of random error on the length judgment result, a moving average is used to smooth the change curve of the trajectory before and after. This can be used to obtain the change curve of the trajectory before and after, which can be mapped to the train feature value matrix, fully reflecting the change trend of the train's trajectory before and after.

[0035] Steps 3-4: After obtaining the train's forward and backward trajectories, subtracting the forward and backward trajectories yields the spatial vibration length of the train at each frame, thus allowing us to calculate the actual length of the train.

[0036] length(i)=BackTrace(i)-FrontTrace(i),i=1,2,...,D(6)

[0037] After obtaining the relationship between spatial vibration length and time, the influence of the intersection on the length calculation is eliminated by using a threshold judgment method. The average length of the remaining part of the train is then multiplied by the spatial resolution of the system to obtain the calculated train length.

[0038] Steps 3-5: After accurately calculating the train length, use the train length to calculate the real-time operating speed of the train across the entire line. For the extracted train eigenvalue matrix along the spatial axis, calculate the preceding and following trajectory points for the time signal of each spatial point, and calculate the time t for the train to pass each spatial point. train Through average velocity V train =S train / t train The calculation yields the real-time average speed of the train over the entire fiber optic link, where S train It is the length of the train.

[0039] Furthermore, the train speed curves obtained in steps 3-5 are smoothed using a moving average method to eliminate random errors.

[0040] This invention employs a time-domain frame-segmented wavelet packet feature extraction method to preprocess train vibration signals, removing interference from "tailing" vibration signals. It enhances the train trajectory features using image binarization and morphological filtering algorithms, performs local line detection using image segmentation, classifies the corresponding lines of the train using k-means clustering, and extracts the train trajectory using polynomial fitting. Based on the obtained train trajectory, it truncates the wavelet packet energy feature value matrix and uses an adaptive thresholding method to obtain the train's forward and backward trajectories, thereby achieving accurate monitoring of train length and speed.

[0041] In summary, this invention uses time-domain frame-segmented wavelet packet feature extraction, image processing trajectory fitting extraction, and adaptive thresholding to calculate the preceding and following trajectories to achieve accurate monitoring of train length and speed, solving the problems of DAS systems in monitoring the operating status of railway trains and realizing precise monitoring of train operating status. Attached Figure Description

[0042] Figure 1This is the overall flowchart of the algorithm of this invention;

[0043] Figure 2 This is a diagram showing the signal preprocessing results of the present invention.

[0044] Figure 3 This is a diagram showing the result of the image processing algorithm of this invention.

[0045] Figure 4 This is a diagram showing the results of the image segmentation trajectory extraction algorithm proposed in this invention.

[0046] Figure 5 This is a diagram showing the train length calculation results proposed in this invention.

[0047] Figure 6 This is a diagram showing the train speed calculation results proposed in this invention. Detailed Implementation

[0048] The present invention will now be described in further detail with reference to the accompanying drawings and embodiments.

[0049] A method for monitoring train operation status based on DAS spatiotemporal sequence imaging (e.g.) Figure 1 As shown), the specific steps are as follows:

[0050] Step 1: For the train vibration spatiotemporal intensity map obtained by the DAS system, the time-domain framing method is adopted. The time-domain framing length is selected as L = 25ms to process the train vibration signal in frames. Then, each frame signal is decomposed into 3 layers of wavelet packets to obtain a wavelet packet energy feature vector of length 8. The 2nd to 8th energy feature values ​​in the vector are summed to complete the extraction of wavelet packet energy features. The train vibration signal is then preprocessed to obtain a wavelet packet energy feature value matrix of size 1440*2600 to remove the "tail" vibration signal interference.

[0051] Step 2: The wavelet packet energy feature value matrix obtained in Step 1 is visualized, and the train trajectory in the image is enhanced based on the image binarization and morphological filtering methods in the image processing algorithm. The enhanced image is segmented by selecting a size of 200*200 pixels, and local line detection is performed on each small image. Then, the k-means clustering algorithm is used to classify the corresponding lines of the trains using the (k,b) parameters of all lines, and the curve trajectory of multiple trains is extracted by the polynomial fitting method.

[0052] Step 3: Map the train trajectory equation obtained in Step 2 onto the wavelet packet energy eigenvalue matrix obtained in Step 1, and truncate it. Use an adaptive thresholding method to obtain the train's forward and backward trajectories. Subtract the forward and backward trajectories to obtain the train length variation curve over time. After eliminating the interference of train intersection terms, the train length is obtained. Further, the corresponding real-time train speed curve is obtained based on the train length, thereby achieving accurate monitoring of train length and speed.

[0053] In this embodiment, the data sample used is vibration signal data from multiple trains on a high-speed railway. Figure 2 The following are the signal preprocessing results of this embodiment: (a) is the original spatiotemporal intensity map of the train collected using the DAS system for 6 minutes, (b) wavelet packet energy vector, (c) wavelet energy feature matrix, and (d) wavelet energy feature comparison.

[0054] In step 1 of the above method, for Figure 2 (a) shows the train's spatiotemporal intensity. After dividing the time domain into 25ms frames, the wavelet packet energy vectors of the train vibration signal and the train "tail" signal are compared using a decomposition layer of N=3. Figure 2 As shown in (b). From Figure 2 (b) The comparison of wavelet packet energy features shows that the energy of the first wavelet packet of the train "tail" signal is much greater than that of the train vibration signal, while the energy vectors of the other wavelet packets are all lower than those of the train vibration signal. The sum of the second to eighth wavelet features of the wavelet packet energy features is used as the feature value of the signal to complete the feature extraction of the train spatiotemporal intensity map. Figure 2 (d) compares the train vibration signals at 16km and 17km with the wavelet packet feature extraction results. The figure shows that using wavelet packet energy features effectively eliminates the "tailing" effect. The wavelet packet energy feature matrix after wavelet energy feature extraction is as follows: Figure 2 As shown in (c), the effect of "tailing" is well eliminated after wavelet packet energy feature extraction.

[0055] In step 2 of the above method, for Figure 2 The result of image visualization, binarization, and closing operations on the wavelet energy feature matrix shown in (d) is as follows: Figure 3 As shown, (a) is the image binarization result, (b) is the binarization result, and (c) is the closing operation result. The image binarization process enhances the difference between the train trajectory and background noise. Furthermore, the morphological filtering algorithm effectively fills in image breaks in the train trajectory after image binarization. Here, we segment the image after the closing operation into 200×200 pixel units and perform line detection on each small image.

[0056] Figure 4Here is a diagram showing the results of the image segmentation trajectory extraction algorithm in this embodiment: Figure 4 (a) shows the results of line detection using the Hough transform for each small image, where (i), (ii) and (iii) represent the cases of single upline, single downline, and upline-downline intersection, respectively. Figure 4 (b) shows the final stitched image. Figure 4 (c) shows the classification results of the k-means clustering algorithm on the lines, Figure 4 (b) The (k,b) parameter of each line is used as the feature of the line. The k-means clustering algorithm is used to cluster all the lines in the image, which can divide all the lines into five categories, corresponding to the five trains in the image. Figure 4 (d) shows all the straight lines for each train, using the results of a 4th-order polynomial curve fitting.

[0057] In step 3 of the above method, Figure 5 This is a diagram showing the train length calculation results in this embodiment: Figure 5 (a) is the extracted feature matrix of train ②, which fully includes the vibration signal of the train while retaining the intersection of the train's operation with other trains. Then, according to equation (5), the threshold of each frame is adaptively calculated, and the front and rear trajectory points of the train are obtained from the train feature space distribution of each frame. Figure 5 (b) shows the train spatial distribution at frame 1000. The positions of the preceding and following trajectory points at this time can be obtained using an adaptive threshold. After obtaining the positions of the preceding and following trajectory points for all frames, a moving average method with a window length of 10 is used to smooth the change curves of the preceding and following trajectories, resulting in the following: Figure 5 The trajectory change curves shown in (c) can be mapped to the train feature matrix, fully reflecting the trend of the train's trajectory changes. Subtracting the train's trajectories from its trajectories yields the spatial vibration length of the train at each frame. Figure 5 As shown in (d), the length of the train can be calculated to be 400.549m, which is only 1.249m less than the actual train length of 399.3m.

[0058] After accurately calculating the train length, the train's real-time operating speed across the entire line is calculated using this length. Figure 6 This is a diagram showing the train speed calculation results in this embodiment. For the train speed, along the spatial axis direction, the time signal of each spatial point is used to calculate the preceding and following trajectory points, thereby obtaining the time variation of the train passing through the spatial point as shown below. Figure 6 As shown in (a), this includes two train crossings, which have been replaced by a straight line. Using the calculated train length of 400.549m, the formula V... train =Strain / t train The change of train speed with space was calculated as follows: Figure 6 As shown in (b), smoothing is performed using a moving average with a window length of 50, which can obtain the speed variation of the high-speed train along the fiber optic line.

[0059] As can be seen from the above embodiments, this invention discloses a train operation status monitoring method based on DAS spatiotemporal sequence imaging. The main steps are as follows: (1) The train vibration signal is preprocessed using the temporal-domain frame-segmented wavelet packet feature extraction method to remove the interference of "tail" vibration signals; (2) The train trajectory is enhanced by image binarization and morphological filtering methods in image processing algorithms, local straight line detection is completed using image segmentation, the train corresponding straight line is classified using k-means clustering algorithm, and the train trajectory is extracted using polynomial fitting method; (3) The wavelet packet energy feature value matrix is ​​truncated based on the train trajectory, and the front and rear trajectories of the train are obtained using the adaptive threshold method, thereby realizing accurate monitoring of train length and speed. This effectively solves the problem of effective extraction of railway train vibration signals and real-time monitoring and positioning of operation status, and can realize accurate monitoring of railway train operation status.

Claims

1. A train operation state monitoring method based on DAS spatiotemporal sequence imaging, characterized in that, The method comprises the following steps: Step 1, for the train vibration space-time intensity map obtained by the DAS system, a time-domain frame division method is adopted, a time-domain frame division length L is selected to perform frame processing on the train vibration signal, then wavelet packet decomposition is performed on each frame signal to obtain a wavelet packet energy eigenvalue matrix, and the wavelet packet energy eigenvalue matrix is used to remove the interference of the trailing vibration signal, L≤total time-domain length L T ; Step 2, the wavelet packet energy eigenvalue matrix obtained in step 1 is subjected to image operation, and the train track in the image is subjected to feature enhancement based on image binarization and morphological filtering methods in the image processing algorithm; an x x pixel point matrix is selected to perform image segmentation on the feature-enhanced image, then local straight line detection is performed on each segmented image with a size of x x, (ρ, θ) parameters of the image in the polar coordinate system are obtained, finally, the (k, b) parameters of all straight lines are subjected to k-means clustering algorithm to complete classification of the straight lines corresponding to the train, and polynomial fitting method is used to complete extraction of the curve track of the train, so that a train curve track equation is obtained; wherein (ρ, θ) is a parameter of the straight line in the image coordinate system, k is a slope parameter of the straight line, b is an intercept parameter of the straight line, and x≤L; Step 3, the train curve track equation obtained in step 2 is mapped into the wavelet packet energy eigenvalue matrix obtained in step 1, and the wavelet packet energy eigenvalue matrix is subjected to truncation, and an adaptive threshold method is used to obtain the front and rear tracks of the train; the front and rear tracks are subtracted to obtain a curve of the length of the train along time, and the length of the train is obtained after interference of train cross terms is excluded; Finally, the corresponding real-time speed curve of the train is obtained according to the length of the train, so that accurate monitoring of the length and speed of the train is realized.

2. The train operation state monitoring method based on DAS space-time sequence imaging according to claim 1, characterized in that, The step 1 specifically comprises: Step 1-1, selecting a time window L, the train vibration spatio-temporal intensity diagram obtained by the DAS system is subjected to frame division along the time direction, so that the train vibration data after frame division is obtained; Step 1-2, selecting the wavelet packet decomposition layer number N, decomposing the train vibration data of each frame by wavelet packet, and calculating the wavelet packet energy according to formula (1) to form a wavelet packet energy feature vector where C j (m) is the wavelet packet coefficient after decomposition, N is the number of wavelet packet decomposition layers; Step 1-3, the energy features with a difference between the train vibration signal and the residual vibration signal greater than 20% of the wavelet packet energy feature of the residual vibration signal are summed up, so that a wavelet packet energy eigenvalue matrix is obtained and used as an eigenvalue matrix of the trailing vibration signal for distinguishing the train vibration signal part and the residual vibration signal part.

3. The train operation state monitoring method based on DAS space-time sequence imaging according to claim 1, characterized in that, The step 2 specifically comprises: Step 2-1, the wavelet packet energy eigenvalue matrix obtained in step 1 is subjected to image operation using a gray scale mapping method, then OTSU threshold method is used to complete binarization of the image, and then a closing operation in morphological filtering is used to complete morphological filtering of the image, so that the train track is subjected to image enhancement; Step 2-2, the image after enhancement in step 2-1 is segmented in units of x x pixels, Hough straight line detection is performed on each x x image after segmentation to obtain (ρ, θ) parameters of the image in the polar coordinate system, q straight lines are selected from Q straight line detection results, the up and down lines of the straight lines are classified according to the positive and negative of the θ parameter, and a parameter mapping relationship (2) from the polar coordinate system (ρ, θ) to the straight line coordinate system y=kx+b is used to obtain an expression of the straight line detection result, and the average values of the straight lines in the up and down line classification are obtained to obtain a straight line of each x x image; Wherein (ρ, θ) is a parameter of the straight line in the image coordinate system, k is a slope parameter of the straight line, and b is an intercept parameter of the straight line; Step 2-3, after the straight line detection of each image is completed, all the x x images are combined into an original train track feature image. Using the (k, b) parameters of each straight line as the feature of the straight line, all straight lines in the image are clustered by a k-means clustering algorithm, so as to divide all straight lines into several categories, each category corresponding to a train in the image; In step 2-4, for the straight lines completed by the k-means clustering, the extraction of the train curve track is completed by using a fourth-order polynomial fitting method, so as to obtain a train curve track equation.

4. The train operation state monitoring method based on DAS space-time sequence imaging according to claim 1, characterized in that, The step 3 specifically comprises: In step 3-1, the train curve track equation obtained in step 2 is mapped into the feature matrix obtained in step 1, and the specific mapping relationship is as follows: trace ts = D-trace i (3) where D is the width of the image, i.e. the maximum number of frames in which the signal is framed in step 1, trace i is the trajectory curve equation in the image coordinate system, trace ts is the trajectory curve equation in the space-time coordinate system; Step 3-2, initialize a full zero train matrix TrainMat, the matrix size is the same as the wavelet packet characteristic value matrix CharMat, select trace ts as the center, cut in front and back in the time domain, the cutting width is Ti in front and back, Ti is the time domain length of removing the tail vibration signal, and the cutting formula is represented as: Wherein, M is the space length of the feature matrix; In step 3-3, along the time direction of the wavelet packet energy feature value matrix, the adaptive threshold is calculated according to formula (5), and the front track point and the rear track point of the train are calculated for each frame of the train feature space distribution. thread(i)=mean(TrainMat(:, i))+max(TrainMat(:, i)) / 3,i=1,2,...,D (5) In formula (5), mean(.) represents the mean value of a signal, and TrainMat(:, i) represents the space distribution of the i-th frame of signal. In the calculation of the front track point and the rear track point of the train feature space distribution, for each frame of space distribution, the first space point greater than the threshold thread(i) is found from front to back as the front track point of the train, and the first space point greater than the threshold thread(i) is found from back to front as the rear track point of the train. In step 3-4, after the front track and the rear track of the train are calculated, the vibration length of the train in space at each moment is obtained by subtracting the front track from the rear track of the train, so as to obtain the actual length of the train as follows: length(i)=BackTrace(i)-FrontTrace(i),i=1,2,...,D(6) In step 3-5, after the accurate calculation of the length of the train is completed, the real-time running speed of the train on the entire line is calculated using the length of the train. The intercepted train characteristic value matrix is along the space axis direction, and the time signal of each space point is calculated to obtain the time t of the train passing through each space point train The real-time average speed of the train on the entire optical fiber link is obtained by calculating the average speed V train = S train / t train , wherein S train is the train length.

5. The train operation state monitoring method based on DAS space-time sequence imaging according to claim 4, characterized in that, The train speed curve obtained in step 3-5 is also smoothed by using a moving average method.

6. The train operation state monitoring method based on DAS space-time sequence imaging according to claim 5, characterized in that, The train speed curve is smoothed by using a moving average with a window length of 50.

Images