Vehicle passing signal extraction method and device based on multi-threshold combination
By adopting a multi-threshold combined crossing signal extraction method in rail transit monitoring, the problem of difficulty in accurately extracting train crossing signals in the prior art is solved, and the monitoring effect of high accuracy and low resource consumption is achieved.
Patent Information
- Application Number
- CN202411722307.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-28
- Publication Date
- 2025-03-14
AI Technical Summary
The prior art is difficult to accurately extract the passing signals of trains when passing through them in a large number of original monitoring signals, especially in monitoring large-length infrastructure in rail transit.
Using a multi-threshold joint cross signal extraction method, the collected original signal is downsampled and sliding window segmented, the characteristic values of multiple energy characteristics are calculated, and the first and second preset thresholds are configured for each energy characteristic to identify the position of the train vibration section.
It realizes fast and highly accurate vehicle-passing signal extraction, improves monitoring accuracy, and reduces the consumption of storage and computing resources, and improves monitoring sustainability.
Smart Images
Figure CN119622292A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of train signal processing, and in particular to a method and device for extracting a passing train signal based on multi-threshold combination. Background Art
[0002] Urban rail is the hub of contemporary urban transportation and plays an important role in improving transportation efficiency and economic development. To ensure the safety of train operation, it has become crucial to implement structural health monitoring technology in urban rail transit. Strain sensors have been extensively developed for large-scale infrastructure due to their sensitivity. Although it is effective, the amount of collected monitoring data will surge as the number of sensors increases. Due to the time sparsity of commuter trains, the stored data mainly consists of low-information environmental noise and interference signals, and the train signal needs to be accurately extracted from the original signal.
[0003] Traditionally, the trigger sampling technology is used to control the sampling process by starting and stopping the external sensor. This method is suitable for a small number of sensors and a short length of the monitored object. For a long-length infrastructure such as rail transit, the monitoring sensors are arranged continuously to obtain continuous and accurate monitoring signals. Since each sensor has a different time to detect the target signal, it is impractical to apply a unified start and stop time.
[0004] Therefore, a train signal recognition method is needed to recognize the passing signal of a train from a large number of original monitoring signals. Summary of the invention
[0005] The present invention provides a method and device for extracting a passing train signal based on a combination of multiple thresholds, so as to solve the defect in the prior art that there is a lack of a method for accurately extracting a train signal from an original signal, and realize a fast and highly accurate method for extracting a passing train signal.
[0006] The present invention provides a method for extracting a vehicle passing signal based on a combination of multiple thresholds, comprising: The collected original signal is downsampled and then segmented using a sliding window; Calculating characteristic values corresponding to a plurality of energy features of the signal data of each sliding window, and converting the signal data of each sliding window into characteristic value responses, wherein the energy feature represents the energy change of the signal data in a sliding window; A first preset threshold and a second preset threshold are configured for each energy feature, and the position of the train vibration segment is identified according to the first preset threshold, wherein the time period in which the characteristic value is greater than the second preset threshold belongs to the train vibration segment, and the first preset threshold is determined based on the second preset threshold, and the first preset threshold is less than the second preset threshold.
[0007] According to a vehicle passing signal extraction method based on multi-threshold combination provided by the present invention, the multiple energy features include short-time energy and short-time low-frequency energy.
[0008] According to a method for extracting a passing vehicle signal based on a combination of multiple thresholds provided by the present invention, before the step of configuring a first preset threshold and a second preset threshold for each energy feature, the method further includes: determining a second preset threshold for each energy feature based on the data analysis; The second preset threshold is searched using a grid search strategy, wherein the search step is a multiple of the second preset threshold and the search range is 0 to the second preset threshold; The search result with the highest intersection-over-union ratio of the evaluation index is determined as the first preset threshold.
[0009] According to a method for extracting a passing vehicle signal based on a combination of multiple thresholds provided by the present invention, the step of downsampling the collected original signal and then segmenting it using a sliding window specifically includes: Applying low-pass filtering to obtain a trend signal in the original signal; Subtracting the trend signal from the original signal to obtain an original signal without drift; The drift-free original signal is segmented using a sliding window.
[0010] According to a method for extracting a passing train signal based on a combination of multiple thresholds provided by the present invention, the step of identifying the position of the signal train vibration segment according to the first preset threshold specifically includes: For each energy feature, segment the feature value response using a first preset threshold to obtain a train vibration segment corresponding to each energy feature; The intersection of the train vibration segments corresponding to the multiple energy features is obtained, and the position of the train vibration segment in the original signal is determined based on the intersection.
[0011] According to a method for extracting a passing vehicle signal based on a multi-threshold combination provided by the present invention, the step of determining the position of the train vibration segment of the original signal based on the intersection specifically includes: Based on the intersection and downsampling processing, reverse deduction is performed to obtain the time period corresponding to the intersection in the original signal; Post-processing the original signal, wherein the post-processing includes fusing adjacent vehicle passing signals and eliminating short-time vehicle passing signals; The train vibration segment is determined based on the post-processed original signal.
[0012] The present invention also provides a vehicle passing signal extraction device based on multi-threshold combination, comprising: A preprocessing module is used to downsample the collected original signal and then segment it using a sliding window; A calculation module, used to calculate characteristic values corresponding to a plurality of energy features of the signal data of each sliding window, and convert the signal data of each sliding window into a characteristic value response, wherein the energy feature represents the energy change of the signal data in a sliding window; An identification module is used to configure a first preset threshold and a second preset threshold for each energy feature, and identify the position of the train vibration segment according to the first preset threshold, wherein the time period in which the characteristic value is greater than the second preset threshold belongs to the train vibration segment, and the first preset threshold is determined based on the second preset threshold, and the first preset threshold is less than the second preset threshold.
[0013] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, a vehicle passing signal extraction method based on multi-threshold combination as described above is implemented.
[0014] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the method for extracting a passing vehicle signal based on multiple threshold combinations as described above is implemented.
[0015] The present invention also provides a computer program product, comprising a computer program, wherein when the computer program is executed by a processor, the computer program implements any of the above-mentioned methods for extracting passing vehicle signals based on multi-threshold combination.
[0016] The method and device for extracting passing train signals based on multi-threshold combination provided by the present invention accurately extract train signals to improve the accuracy of monitoring by combining a dual-threshold segmentation method with multiple energy characteristics, while reducing the storage and computing resource consumption of the monitoring system and improving the sustainability of monitoring. BRIEF DESCRIPTION OF THE DRAWINGS
[0017] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0018] Figure 1 It is a flow chart of a vehicle passing signal extraction method based on multi-threshold combination provided by the present invention; Figure 2 It is a schematic diagram for displaying the original signal in the vehicle passing signal extraction method based on multi-threshold combination provided by the present invention; Figure 3It is a schematic diagram for showing the characteristic value representation in the vehicle passing signal extraction method based on multi-threshold combination provided by the present invention; Figure 4 It is a schematic diagram for showing the distribution of strain sensors in the vehicle passing signal extraction method based on multi-threshold combination provided by the present invention; Figure 5 It is a schematic diagram for demonstrating grid search in the vehicle passing signal extraction method based on multi-threshold combination provided by the present invention; Figure 6 It is a structural schematic diagram of a vehicle passing signal extraction device based on multi-threshold combination provided by the present invention; Figure 7 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION
[0019] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0020] Combine the following Figures 1 to 5 The present invention introduces a vehicle passing signal extraction method based on multi-threshold combination, such as Figure 1 As shown, including: Step 101, downsampling the collected original signal and segmenting it using a sliding window; The original signal is the strain monitoring signal collected by the strain sensing system.
[0021] Specifically, the strain sensing system includes multiple strain sensors continuously installed on the surface of the ballast parallel to the track. Each strain sensor has the same sampling frequency, and different strain sensors form a continuous measurement area. Sufficient monitoring strain data of each strain sensor when the ballast passes and when the train does not pass is collected to form an original signal data set.
[0022] Optionally, the original signal dataset is saved as a two-dimensional matrix X, X∈R m×n , the number of matrix columns represents the number of strain sensors, and the rows of the matrix represent the data sequence monitored by time.
[0023] On this basis, the original signal of any strain sensor can be expressed as x(t), where x(t) belongs to one of the columns of X.
[0024] Since there are multiple strain sensors placed in succession and the strain sensors are high-frequency sampling sensors, it is necessary to downsample the original data before identifying the train passing signal so as to process a large amount of data at the same time.
[0025] It is understandable that the original signal contains components such as train vibration, environmental noise and interference vibration. When extracting the passing signal, it is defined as the vibration response of the track under the direct wheel load, and the train vibration signal is marked as the signal to be extracted, and the rest of the signals are all defined as noise.
[0026] On this basis, a sliding window is used to divide the downsampled original signal into smaller segments.
[0027] Step 102, calculating characteristic values corresponding to a plurality of energy features of the signal data of each sliding window, and converting the signal data of each sliding window into a characteristic value response, wherein the energy feature represents the energy change of the signal data in a sliding window; Furthermore, a plurality of energy features are determined for each window, and a feature value of the signal data of each sliding window is calculated.
[0028] The energy feature is used to characterize the energy change of the signal data, calculate the characteristic value of the signal data in each sliding window, obtain the characteristic value response, and normalize the characteristic value data in each sliding window. Figure 2 The strain variation shown is converted to Figure 3 The energy shown varies with time.
[0029] The process can be expressed as: .
[0030] In the formula, the function F For feature extraction, is the eigenvalue signal after the original signal is converted.
[0031] Optionally, the multiple energy features include short-time energy and short-time low-frequency energy.
[0032] On this basis, the eigenvalues corresponding to multiple energy characteristics of the signal data of each sliding window are calculated, that is, the eigenvalues corresponding to the short-time energy and the eigenvalues corresponding to the short-time low-frequency energy of the signal data of each sliding window are calculated respectively, and the corresponding response spectrum is obtained.
[0033] Specifically, short-time energy can reflect the total energy of the train vibration signal within a time window. For monitoring signals, short-time energy can capture the strong energy part of the signal and distinguish the signals when the train passes and when it does not pass.
[0034] Short-time low-frequency energy further calculates the energy of low-frequency vibration signals, which is more suitable for the characteristics of train vibration signals mainly concentrated in low frequencies. By extracting short-time energy or short-time low-frequency energy, the characteristics of train vibration signals can be captured more accurately.
[0035] like Figure 3 As shown in Figure 3, after converting the original data into a series of short-time energy characteristic values, it is not difficult to find that the amplitude of the short-time energy characteristic increases with the increase of the amplitude of the original data, which well reflects the vibration response when the train passes.
[0036] Step 103, configure a first preset threshold and a second preset threshold for each energy feature, identify the position of the train vibration segment according to the first preset threshold, wherein the time period in which the characteristic value is greater than the second preset threshold belongs to the train vibration segment, the first preset threshold is determined based on the second preset threshold, and the first preset threshold is less than the second preset threshold.
[0037] A set of first preset threshold value (T1) and second preset threshold value (T2) are configured for each energy feature, and the configuration method is the same. Therefore, in this embodiment, Figure 3 The short-time energy characteristic value shown is used as an example for explanation.
[0038] like Figure 3 As shown, it is not difficult to find that the short-time energy characteristic value of the train passing signal is higher than the short-time energy characteristic value of the noise. The second preset threshold is used for preliminary screening, and the time period corresponding to the train passing signal is screened and determined as the vibration period, which can be an empirical value.
[0039] Specifically, the intersection points of the second preset threshold and the short-time energy characteristic curve are point B and point C. It can be clearly seen that the short-time energy characteristic values corresponding to point B and point C and the time period between them must be the short-time energy characteristic values corresponding to the train passing signal. Therefore, the time segment [B, C] obtained by filtering the second preset threshold must belong to the train vibration segment.
[0040] On this basis, the first preset threshold is determined according to the second preset threshold, and the first preset threshold is used to accurately find the exact boundary between the train vibration and the environmental noise.
[0041] Optionally, a grid search strategy is adopted to determine the first preset threshold based on the second preset threshold.
[0042] On this basis, the train passing signal is identified according to the first preset threshold. Figure 3The intersection points of the second preset threshold and the short-time energy characteristic curve are point A and point D. It can be clearly seen that the short-time energy characteristic values corresponding to point A and point D and the time period between them must be the short-time energy characteristic values corresponding to the train passing signal. Therefore, [A, D] obtained by screening by the first preset threshold is the train vibration segment corresponding to the characteristic value determined according to the short-time energy characteristic.
[0043] It should be noted that, since multiple energy features are used in the present embodiment, each energy feature can determine a set of train vibration segments corresponding to characteristic values. Therefore, it is also necessary to merge the multiple sets of train vibration segments to obtain train vibration segments that can be used to identify train passing signals in the original signal.
[0044] Optionally, in a feasible implementation, taking the short-time energy feature and the short-time low-frequency energy feature as examples, for the same train passing signal, the train vibration segment corresponding to the short-time energy feature is [M1, N1], and the train vibration segment corresponding to the short-time low-frequency energy feature is [M2, N2]. The mean of M1 and M2 is calculated as the left endpoint, the mean of N1 and N2 is calculated as the right endpoint, and [left endpoint, right endpoint] is used as the final train vibration segment.
[0045] The present invention combines a dual-threshold segmentation method with multiple energy features to accurately extract train signals to improve monitoring accuracy, while reducing the storage and computing resource consumption of the monitoring system and improving the sustainability of monitoring.
[0046] In the vehicle passing signal extraction based on multi-threshold combination of the present invention, before the step of configuring the first preset threshold and the second preset threshold for each energy feature, the method further includes: determining a second preset threshold for each energy feature based on the data analysis; The second preset threshold is determined based on data analysis, and it is sufficient to ensure that the time segment determined based on the second preset threshold is definitely a train vibration segment.
[0047] The second preset threshold is searched using a grid search strategy, wherein the search step is a multiple of the second preset threshold and the search range is 0 to the second preset threshold; The search result with the highest intersection-over-union ratio of the evaluation index is determined as the first preset threshold.
[0048] On the basis of the second preset threshold, the optimal first preset threshold is searched and obtained through a grid search strategy.
[0049] Specifically, the search interval is a multiple of the second preset threshold, that is, the search step is the product of k and T2, K is the search multiple. The search range is (0, T2), so that when T1 is the optimal value, the evaluation index of train signal segmentation is the highest.
[0050] On this basis, it should be noted that short-time energy and short-time low-frequency energy as eigenvalues are simple to calculate and easy to implement. This helps to reduce the amount of calculation during grid search and improve the search speed. Since the train vibration signal is mainly concentrated in the low-frequency range, short-time energy or short-time low-frequency energy as eigenvalues can accurately reflect the signal characteristics, thereby helping the grid search strategy to find the best threshold combination faster.
[0051] Based on the search results of each eigenvalue, T1 is selected as the optimal value when the intersection over union (IoU) value of the evaluation index is the largest. Finally, based on the division result of this optimal T1, the position of the train vibration segment in the signal is accurately identified. ,like Figure 5 shown.
[0052] ; In the formula, TP represents the number of true positive samples, indicating correctly predicted positive instances; FP represents a false positive sample, which refers to a negative instance that is mistakenly labeled as positive; FN Represents a false negative sample, which refers to a positive instance that is mistakenly identified as a negative.
[0053] In the vehicle passing signal extraction based on multi-threshold combination of the present invention, the step of downsampling the collected original signal and then segmenting it using a sliding window specifically includes: Applying low-pass filtering to obtain a trend signal in the original signal; Subtracting the trend signal from the original signal to obtain an original signal without drift; The drift-free original signal is segmented using a sliding window.
[0054] Specifically, to address the baseline drift problem in the original signal, a low-pass filter is first applied to obtain a trend signal, which is then subtracted from the original signal to obtain a drift-free original signal: ; In the formula, To filter out the drift-free signal after the trend signal, LPF is a low-pass filter.
[0055] Furthermore, the original signal is subdivided into smaller segments using a sliding serial port: , where W is the size of the sliding window.
[0056] In a specific embodiment, Figure 4 As shown, the distance between adjacent strain sensors is designed to be 5 meters and they are continuously arranged on the track bed for continuous sampling. The system samples the vibration response at a frequency of 1000Hz (including when the train passes and when the train does not pass).
[0057] The sampling frequency of the strain sensor system is 1000Hz. Usually, the maximum frequency component of the effective vibration response caused by the passing of a train does not exceed 100Hz. Therefore, according to Shannon's sampling theorem (in order to obtain effective information in the signal, the sampling frequency should be greater than or equal to twice the highest frequency in the analog signal spectrum), the original monitoring data is initially filtered and then downsampled, and the sampling frequency is readjusted to 200Hz. This can greatly reduce the amount of data and greatly improve the efficiency of subsequent data analysis.
[0058] The long-term monitoring data sample extraction adopts the sliding fixed window method in the time domain. The window size is set to 10 minutes, which is about twice the interval between train passages. This ensures that each sample covers at least one train passage event and contains 120,000 data points.
[0059] In the vehicle passing signal extraction based on multi-threshold combination of the present invention, the step of identifying the position of the signal train vibration segment according to the first preset threshold specifically includes: For each energy feature, segment the feature value response using a first preset threshold to obtain a train vibration segment corresponding to each energy feature; The intersection of the train vibration segments corresponding to the multiple energy features is obtained, and the position of the train vibration segment in the original signal is determined based on the intersection.
[0060] The position of the train signal in the characteristic value is obtained by segmentation using the first preset threshold value, and the train vibration segment corresponding to the energy characteristic is obtained.
[0061] Therefore, multiple energy features correspond to multiple train vibration segments corresponding to the energy features. In this implementation, the intersection of the train vibration segments corresponding to the multiple energy features is taken, and the intersection is used as the position of the train signal in the characteristic value sequence. Subsequently, the position of the train signal in the original signal can be inferred.
[0062] In the vehicle passing signal extraction based on multi-threshold combination of the present invention, the step of determining the position of the train vibration segment in the original signal based on the intersection specifically includes: Based on the intersection and downsampling processing, reverse deduction is performed to obtain the time period corresponding to the intersection in the original signal; Post-processing the original signal, wherein the post-processing includes fusing adjacent vehicle passing signals and eliminating short-time vehicle passing signals; The train vibration segment is determined based on the post-processed original signal.
[0063] After obtaining the position of the train passing signal in the feature value sequence (i.e., the segmentation result of the first preset threshold), the position of the train passing signal in the original signal can be inferred based on the time period corresponding to the intersection and the specific parameters during preprocessing (such as the size and step size of the sliding window), thereby obtaining the time period corresponding to the intersection in the original signal: ; Where G is the mapping function from eigenvalue to original signal.
[0064] On this basis, in order to remove the erroneous data in the original data, the original data must be post-processed to obtain accurate train signals without errors. .
[0065] The post-processing operation includes fusing adjacent vehicle passing signals and eliminating short-duration vehicle passing signals. Specifically, the merging distance for fusing adjacent vehicle passing signals is set to t1, which means that vehicle passing signals detected within an interval of less than t1 seconds are merged. The minimum duration parameter for eliminating short-duration vehicle passing signals is set to t2, which means that vehicle passing signals identified as having a duration of less than t2 seconds are discarded.
[0066] Through post-processing of the original signal, the time periods corresponding to adjacent process signals are integrated in the original signal, and the time periods corresponding to the incorrectly identified short-term process signals are eliminated in the original signal. The time periods finally sorted out after post-processing are used as train vibration segments to characterize the precise position of the train passing signal in the original signal.
[0067] The vehicle passing signal extraction device based on multi-threshold combination provided by the present invention is described below. The vehicle passing signal extraction device based on multi-threshold combination described below and the vehicle passing signal extraction method based on multi-threshold combination described above can be referred to each other.
[0068] like Figure 6 As shown, the vehicle passing signal extraction device based on multi-threshold combination includes a preprocessing module 601, a calculation module 602 and a recognition module 603: A preprocessing module 601 is used to perform downsampling processing on the collected original signal and then segment it using a sliding window; The original signal is the strain monitoring signal collected by the strain sensing system.
[0069] Specifically, the strain sensing system includes multiple strain sensors continuously installed on the surface of the ballast parallel to the track. Each strain sensor has the same sampling frequency, and different strain sensors form a continuous measurement area. Sufficient monitoring strain data of each strain sensor when the ballast passes and when the train does not pass is collected to form an original signal data set.
[0070] Optionally, the original signal dataset is saved as a two-dimensional matrix X, X∈Rm×n , the number of matrix columns represents the number of strain sensors, and the rows of the matrix represent the data sequence monitored by time.
[0071] On this basis, the original signal of any strain sensor can be expressed as x(t), where x(t) belongs to one of the columns of X.
[0072] Since there are multiple strain sensors placed in succession and the strain sensors are high-frequency sampling sensors, it is necessary to downsample the original data before identifying the train passing signal so as to process a large amount of data at the same time.
[0073] It is understandable that the original signal contains components such as train vibration, environmental noise and interference vibration. When extracting the passing signal, it is defined as the vibration response of the track under the direct wheel load, and the train vibration signal is marked as the signal to be extracted, and the rest of the signals are all defined as noise.
[0074] On this basis, a sliding window is used to divide the downsampled original signal into smaller segments.
[0075] A calculation module 602 is used to calculate characteristic values corresponding to multiple energy features of the signal data of each sliding window, and convert the signal data of each sliding window into a characteristic value response, wherein the energy feature represents the energy change of the signal data in a sliding window; Furthermore, a plurality of energy features are determined for each window, and a feature value of the signal data of each sliding window is calculated.
[0076] The energy feature is used to characterize the energy change of the signal data, calculate the characteristic value of the signal data in each sliding window, obtain the characteristic value response, and normalize the characteristic value data in each sliding window. Figure 2 The strain variation shown is converted to Figure 3 The energy shown varies with time.
[0077] The process can be expressed as: .
[0078] In the formula, the function F For feature extraction, is the eigenvalue signal after the original signal is converted.
[0079] The identification module 603 is used to configure a first preset threshold and a second preset threshold for each energy feature, and identify the position of the train vibration segment according to the first preset threshold, wherein the time period in which the characteristic value is greater than the second preset threshold belongs to the train vibration segment, and the first preset threshold is determined based on the second preset threshold, and the first preset threshold is less than the second preset threshold.
[0080] A set of first preset threshold value (T1) and second preset threshold value (T2) are configured for each energy feature, and the configuration method is the same. Therefore, in this embodiment, Figure 3 The short-time energy characteristic value shown is used as an example for explanation.
[0081] like Figure 3 As shown, it is not difficult to find that the short-time energy characteristic value of the train passing signal is higher than the short-time energy characteristic value of the noise. The second preset threshold is used for preliminary screening, and the time period corresponding to the train passing signal is screened and determined as the vibration period, which can be an empirical value.
[0082] Specifically, the intersection points of the second preset threshold and the short-time energy characteristic curve are point B and point C. It can be clearly seen that the short-time energy characteristic values corresponding to point B and point C and the time period between them must be the short-time energy characteristic values corresponding to the train passing signal. Therefore, the time segment [B, C] obtained by filtering the second preset threshold must belong to the train vibration segment.
[0083] On this basis, the first preset threshold is determined according to the second preset threshold, and the first preset threshold is used to accurately find the exact boundary between the train vibration and the environmental noise.
[0084] Optionally, a grid search strategy is adopted to determine the first preset threshold based on the second preset threshold.
[0085] On this basis, the train passing signal is identified according to the first preset threshold. Figure 3 The intersection points of the second preset threshold and the short-time energy characteristic curve are point A and point D. It can be clearly seen that the short-time energy characteristic values corresponding to point A and point D and the time period between them must be the short-time energy characteristic values corresponding to the train passing signal. Therefore, [A, D] obtained by screening by the first preset threshold is the train vibration segment corresponding to the characteristic value determined according to the short-time energy characteristic.
[0086] It should be noted that, since multiple energy features are used in the present embodiment, each energy feature can determine a set of train vibration segments corresponding to characteristic values. Therefore, it is also necessary to merge the multiple sets of train vibration segments to obtain train vibration segments that can be used to identify train passing signals in the original signal.
[0087] Optionally, in a feasible implementation, taking the short-time energy feature and the short-time low-frequency energy feature as examples, for the same train passing signal, the train vibration segment corresponding to the short-time energy feature is [M1, N1], and the train vibration segment corresponding to the short-time low-frequency energy feature is [M2, N2]. The mean of M1 and M2 is calculated as the left endpoint, the mean of N1 and N2 is calculated as the right endpoint, and [left endpoint, right endpoint] is used as the final train vibration segment.
[0088] The present invention combines a dual-threshold segmentation method with multiple energy features to accurately extract train signals to improve monitoring accuracy, while reducing the storage and computing resource consumption of the monitoring system and improving the sustainability of monitoring.
[0089] Figure 7 An example of a physical structure diagram of an electronic device is shown in FIG. Figure 7 As shown, the electronic device may include: a processor 710, a communication interface 720, a memory 730 and a communication bus 740, wherein the processor 710, the communication interface 720 and the memory 730 communicate with each other through the communication bus 740. The processor 710 may call the logic instructions in the memory 730 to execute a vehicle passing signal extraction method based on a multi-threshold combination, the method comprising: performing downsampling processing on the collected original signal and then segmenting it using a sliding window; calculating the characteristic values corresponding to multiple energy features of the signal data of each sliding window, and converting the signal data of each sliding window into a characteristic value response, wherein the energy feature characterizes the energy change of the signal data in a sliding window; configuring a first preset threshold and a second preset threshold for each energy feature, identifying the position of the train vibration section according to the first preset threshold, wherein the time period when the characteristic value is greater than the second preset threshold belongs to the train vibration section, the first preset threshold is determined based on the second preset threshold, and the first preset threshold is less than the second preset threshold.
[0090] In addition, the logic instructions in the above-mentioned memory 730 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on this understanding, the technical solution of the present invention can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.
[0091] On the other hand, the present invention also provides a computer program product, which includes a computer program. The computer program can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the vehicle passing signal extraction method based on multi-threshold combination provided by the above methods, the method including: downsampling the collected original signal and segmenting it using a sliding window; calculating the characteristic values corresponding to multiple energy characteristics of the signal data of each sliding window, and converting the signal data of each sliding window into a characteristic value response, wherein the energy characteristic characterizes the energy change of the signal data in a sliding window; configuring a first preset threshold and a second preset threshold for each energy characteristic, identifying the position of the train vibration segment according to the first preset threshold, wherein the time period in which the characteristic value is greater than the second preset threshold belongs to the train vibration segment, and the first preset threshold is determined based on the second preset threshold, and the first preset threshold is less than the second preset threshold.
[0092] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to execute the vehicle passing signal extraction method based on multi-threshold combination provided by the above-mentioned methods, the method comprising: down-sampling the collected original signal and segmenting it using a sliding window; calculating characteristic values corresponding to multiple energy characteristics of the signal data of each sliding window, and converting the signal data of each sliding window into a characteristic value response, wherein the energy characteristic characterizes the energy change of the signal data in a sliding window; configuring a first preset threshold and a second preset threshold for each energy characteristic, identifying the position of the train vibration segment according to the first preset threshold, wherein the time period in which the characteristic value is greater than the second preset threshold belongs to the train vibration segment, and the first preset threshold is determined based on the second preset threshold, and the first preset threshold is less than the second preset threshold.
[0093] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.
[0094] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.
[0095] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims
1. A vehicle passing signal extraction method based on multi-threshold combination, characterized in that: include: The collected original signal is downsampled and then segmented using a sliding window; Calculating characteristic values corresponding to a plurality of energy features of the signal data of each sliding window, and converting the signal data of each sliding window into characteristic value responses, wherein the energy feature represents the energy change of the signal data in a sliding window; A first preset threshold and a second preset threshold are configured for each energy feature, and the position of the train vibration segment is identified according to the first preset threshold, wherein the time period in which the characteristic value is greater than the second preset threshold belongs to the train vibration segment, and the first preset threshold is determined based on the second preset threshold, and the first preset threshold is less than the second preset threshold.
2. The vehicle passing signal extraction method based on multi-threshold combination according to claim 1 is characterized in that: The multiple energy features include short-time energy and short-time low-frequency energy.
3. The vehicle passing signal extraction method based on multi-threshold combination according to claim 1 is characterized in that: Before the step of configuring the first preset threshold and the second preset threshold for each energy feature, the method further includes: determining a second preset threshold for each energy feature based on the data analysis; The second preset threshold is searched using a grid search strategy, wherein the search step is a multiple of the second preset threshold and the search range is 0 to the second preset threshold; The search result with the highest intersection-over-union ratio of the evaluation index is determined as the first preset threshold.
4. The vehicle passing signal extraction method based on multi-threshold combination according to claim 1 is characterized in that: The step of downsampling the collected original signal and then segmenting it using a sliding window specifically includes: Applying low-pass filtering to obtain a trend signal in the original signal; Subtracting the trend signal from the original signal to obtain an original signal without drift; The drift-free original signal is segmented using a sliding window.
5. The vehicle passing signal extraction method based on multi-threshold combination according to claim 1 is characterized in that: The step of identifying the position of the signal train vibration segment according to the first preset threshold value specifically includes: For each energy feature, segment the feature value response using a first preset threshold to obtain a train vibration segment corresponding to each energy feature; The intersection of the train vibration segments corresponding to the multiple energy features is obtained, and the position of the train vibration segment in the original signal is determined based on the intersection.
6. The vehicle passing signal extraction method based on multi-threshold combination according to claim 5 is characterized in that: The step of determining the position of the train vibration segment in the original signal based on the intersection specifically includes: Based on the intersection and downsampling processing, reverse deduction is performed to obtain the time period corresponding to the intersection in the original signal; Post-processing the original signal, wherein the post-processing includes fusing adjacent vehicle passing signals and eliminating short-time vehicle passing signals; The train vibration segment is determined based on the post-processed original signal.
7. A vehicle passing signal extraction device based on multi-threshold combination, characterized in that: include: A preprocessing module is used to downsample the collected original signal and then segment it using a sliding window; A calculation module, used to calculate characteristic values corresponding to a plurality of energy features of the signal data of each sliding window, and convert the signal data of each sliding window into a characteristic value response, wherein the energy feature represents the energy change of the signal data in a sliding window; An identification module is used to configure a first preset threshold and a second preset threshold for each energy feature, and identify the position of the train vibration segment according to the first preset threshold, wherein the time period in which the characteristic value is greater than the second preset threshold belongs to the train vibration segment, and the first preset threshold is determined based on the second preset threshold, and the first preset threshold is less than the second preset threshold.
8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the method for extracting passing vehicle signals based on multi-threshold combination as described in any one of claims 1 to 6 is implemented.
9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for extracting passing vehicle signals based on multi-threshold combination as described in any one of claims 1 to 6 is implemented.
10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the method for extracting passing vehicle signals based on multi-threshold combination as described in any one of claims 1 to 6 is implemented.