A vehicle signal extraction method and system based on distributed acoustic sensing
By combining bandpass filtering, frequency-wavenumber domain hard threshold filtering, sector filtering, and Hough transform, the problem of insufficient accuracy and anti-interference capability of distributed acoustic sensing technology in traffic monitoring is solved, and high-precision vehicle detection and real-time monitoring are achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- ANHUI TRANSPORTATION HLDG GRP CO LTD
- Filing Date
- 2026-02-11
- Publication Date
- 2026-06-09
AI Technical Summary
Existing distributed acoustic sensing technology in traffic monitoring suffers from problems such as low signal-to-noise ratio, complex signal composition, difficulty in accurately extracting vehicle signals, weak resistance to complex interference, and insufficient real-time processing.
By employing a combination of bandpass filtering, frequency-wavenumber domain hard threshold filtering, sector filtering, and Hough transform, irregular noise and directional noise are removed, thereby enhancing the spatiotemporal continuity of vehicle signals.
It significantly improves the signal-to-noise ratio and extraction accuracy of vehicle signals, enhances the accuracy and reliability of vehicle detection, and supports real-time traffic monitoring.
Smart Images
Figure CN122173774A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of intelligent traffic monitoring technology, and in particular to a method and system for extracting vehicle signals based on distributed acoustic sensing. Background Technology
[0002] Distributed acoustic sensing (DAS) technology is a continuous, distributed vibration sensing technology based on the Rayleigh scattering principle of optical fibers. By monitoring the phase changes of backscattered Rayleigh light distributed along an optical fiber link, it can acquire environmental vibration information distributed along the fiber with high spatial resolution and high sensitivity. This technology was initially widely used in fields such as oil and gas resource exploration, earthquake monitoring, pipeline safety, and perimeter security.
[0003] In recent years, with the development of intelligent transportation systems, applying DAS technology to traffic condition monitoring has gradually become a research hotspot. Its technological advantage lies in the fact that it can achieve large-scale, all-weather, real-time continuous monitoring of vibration fields along long-distance roads using existing communication optical cables or dedicated sensing optical fibers. Theoretically, it can provide multi-dimensional traffic information, including vehicle detection, classification, vehicle speed estimation, and abnormal event identification.
[0004] However, in practical road engineering applications, the raw signals acquired by DAS systems have low signal-to-noise ratios and extremely complex signal compositions. The vibration signals not only contain characteristic vibrations caused by passing vehicles, but also strongly couple various environmental background noises (such as wind noise, rain noise, and distant mechanical vibrations) as well as inherent system noises (such as coherent noise in the vertical and horizontal directions) introduced by changes in fiber optic cable laying conditions and coupling states. This complex noise background severely interferes with the accurate extraction and identification of target vehicle signals.
[0005] Currently, common techniques for processing DAS traffic monitoring signals include conventional time-domain filtering (such as bandpass filtering) and spectrum analysis. While these methods can initially suppress some out-of-band noise, their effectiveness is limited in handling noise overlapping with vehicle signal frequency bands and coherent noise with specific directions. Furthermore, because vehicle signals may exhibit discontinuities in the spatiotemporal domain, direct feature extraction is not highly accurate. Therefore, existing technologies generally suffer from low vehicle signal extraction accuracy, weak resistance to complex interference, and insufficient real-time processing, making it difficult to meet the practical needs of high-reliability real-time traffic monitoring and management. Summary of the Invention
[0006] To address the technical problems existing in the background art, this invention proposes a vehicle signal extraction method and system based on distributed acoustic sensing.
[0007] The present invention proposes a vehicle signal extraction method based on distributed acoustic sensing, comprising the following steps: S1. Acquire the raw spatiotemporal vibration signals collected along the road by the distributed fiber optic acoustic sensing system; S2. Perform bandpass filtering on the original spatiotemporal vibration signal to obtain a preliminary filtered signal. Then, perform superposition processing on the preliminary filtered signal in the time dimension with a preset time window to obtain a superimposed signal with signal enhancement. S3. Normalize the signal amplitude of each sensing channel in the superimposed signal to obtain a normalized vibration signal. Transform the normalized vibration signal to the frequency-wavenumber domain and apply a hard threshold filter to the frequency-wavenumber domain signal to remove irregular noise to obtain a first filtered signal. S4. Perform sector-area filtering on the first filtered signal in the frequency-wavenumber domain to remove structural noise in the vertical and horizontal directions, and obtain the second filtered signal. S5. Perform Hough transform on the second filtered signal to enhance the spatiotemporal continuity of the vehicle vibration signal and obtain the final processed signal for vehicle feature extraction.
[0008] Preferably, step S4 specifically includes: The first filtered signal is subjected to fan-shaped filtering in the frequency-wavenumber domain within the range of 30 to 60 degrees to remove structural noise in the vertical and horizontal directions, thereby obtaining the second filtered signal.
[0009] Preferably, the fan-shaped region with a wavenumber direction in the range of 30 to 60 degrees corresponds to the propagation direction of the ground vibration wave caused by vehicle movement.
[0010] Preferably, the preset time window is n seconds, where 1≤n≤2.
[0011] Preferably, the preset time window is 2 seconds.
[0012] Preferably, the preset threshold of the hard threshold filter is determined by calculating the statistical characteristic value of the noise component in the frequency-wavenumber domain signal.
[0013] Preferably, the frequency band of the bandpass filtering process is 40Hz to 60Hz.
[0014] Preferably, step S5 specifically includes: The second filtered signal is subjected to Hough transform processing to convert the discrete vehicle signal points in the second filtered signal into continuous spatiotemporal trajectory features, so as to enhance the spatiotemporal continuity of the vehicle vibration signal and obtain the final processed signal for vehicle feature extraction.
[0015] Preferably, it further includes: S6. Based on the final processed signal, perform vehicle type identification, traffic flow statistics, vehicle speed estimation, or traffic anomaly detection.
[0016] This invention proposes a vehicle signal extraction system based on distributed acoustic sensing, comprising: The signal acquisition module is used to acquire the raw spatiotemporal vibration signals collected by the distributed fiber optic acoustic sensing system along the road. The first processing module is used to perform bandpass filtering on the original spatiotemporal vibration signal to obtain a preliminary filtered signal, and to perform superposition processing on the preliminary filtered signal in the time dimension with a preset time window to obtain a superimposed signal with signal enhancement. The second processing module is used to normalize the signal amplitude of each sensing channel in the superimposed signal to obtain a normalized vibration signal, transform the normalized vibration signal to the frequency-wavenumber domain, and apply a hard threshold filter to the frequency-wavenumber domain signal to remove irregular noise to obtain a first filtered signal. The third processing module is used to perform sector-area filtering on the first filtered signal in the frequency-wavenumber domain to remove structural noise in the vertical and horizontal directions, and obtain the second filtered signal. The transformation output module is used to perform Hough transform processing on the second filtered signal to enhance the spatiotemporal continuity of the vehicle vibration signal and obtain the final processed signal for vehicle feature extraction.
[0017] This invention proposes a vehicle signal extraction method and system based on distributed acoustic sensing. By combining bandpass filtering and FK domain hard thresholding, it suppresses irregular environmental noise overlapping with the vehicle's frequency band. Through FK sector filtering, it directionally eliminates vertical and horizontal coherent noise introduced by the system and laying conditions, improving the signal-to-noise ratio of the target signal. Time superposition enhances signal strength, and Hough transform effectively correlates and enhances discrete vehicle signal points in the spatiotemporal domain, improving the continuity and integrity of the vehicle trajectory. This enhances the accuracy and reliability of vehicle detection and feature extraction. Attached Figure Description
[0018] Figure 1 This is a schematic diagram illustrating the workflow of a vehicle signal extraction method based on distributed acoustic sensing proposed in this invention. Figure 2 This is a schematic diagram of the original data in Embodiment 1 of the vehicle signal extraction method based on distributed acoustic sensing proposed in this invention; Figure 3 This is a schematic diagram of bandpass filtering in Embodiment 1 of the vehicle signal extraction method based on distributed acoustic sensing proposed in this invention. Figure 4 This is a superimposed normalization schematic diagram of Embodiment 1 of the vehicle signal extraction method based on distributed acoustic sensing proposed in this invention. Figure 5This is a schematic diagram of the Hough variation of an embodiment 1 of the vehicle signal extraction method based on distributed acoustic sensing proposed in this invention. Figure 6 This is a schematic diagram of the system architecture of a vehicle signal extraction system based on distributed acoustic sensing proposed in this invention. Detailed Implementation
[0019] Reference Figures 1-6 The present invention proposes a vehicle signal extraction method based on distributed acoustic sensing, comprising the following steps: S1. Acquire the raw spatiotemporal vibration signals collected along the road by the distributed fiber optic acoustic sensing system.
[0020] It should be noted that a communication-grade single-mode optical fiber for sensing, or an existing compliant communication optical cable, should be laid along the shoulder or central median of the road to be monitored. This optical fiber should be connected to a distributed acoustic sensing (DAS) demodulation device. The device parameters need to be optimized. Typically, the spatial sampling interval can be set to 1 to 10 meters, and the temporal sampling frequency must be higher than twice the highest frequency of the target signal, usually set to 500Hz or higher, to ensure complete capture of vehicle vibration signals. After the device starts continuous operation, it will output a two-dimensional data matrix. The rows represent the spatial channel number along the optical fiber, and the columns represent the temporal sampling points. Each element in the matrix represents the vibration amplitude or phase change at the corresponding spatial location at the corresponding time, which is the original spatiotemporal vibration signal.
[0021] S2. Bandpass filtering is performed on the original spatiotemporal vibration signal to obtain a preliminary filtered signal. The preliminary filtered signal is then superimposed on the signal in the time dimension within a preset time window to obtain the superimposed signal with signal enhancement.
[0022] In this embodiment, the frequency band of the bandpass filter is 40Hz to 60Hz.
[0023] In this embodiment, the preset time window is n seconds, where 1≤n≤2.
[0024] In a preferred embodiment, the preset time window is 2 seconds.
[0025] Specifically, the energy of ground vibration signals caused by vehicle movement is mainly concentrated in a specific frequency band (e.g., 40Hz to 60Hz), while most environmental noise (such as wind noise and low-frequency mechanical vibration) is distributed outside this frequency band. Therefore, a digital bandpass filter (such as an FIR or IIR filter) with a passband of 40Hz to 60Hz is first applied to each time series of the original spatiotemporal vibration signal. This step can initially filter out out-of-band noise, obtaining a preliminary filtered signal while retaining the core vehicle vibration component.
[0026] S3. Normalize the signal amplitude of each sensing channel in the superimposed signal to obtain a normalized vibration signal. Transform the normalized vibration signal to the frequency-wavenumber domain and apply a hard threshold filter to the frequency-wavenumber domain signal to remove irregular noise to obtain the first filtered signal.
[0027] In this embodiment, the preset threshold of the hard threshold filter is determined by calculating the statistical characteristic value of the noise component in the frequency-wavenumber domain signal.
[0028] Specifically, vehicle signals may be weak at a single moment. To enhance their significance, the initially filtered signal is superimposed and averaged over a time dimension. A time window T is set (e.g., T=2 seconds), and the continuous signal is divided into segments by window T. The signals at all time points within the window are superimposed or averaged. This operation can effectively enhance periodic or quasi-periodic vehicle signals, suppress random noise, and obtain a superimposed signal with improved signal-to-noise ratio.
[0029] Specifically, due to differences in coupling conditions and geological background, the signal baseline amplitudes of different sensing channels (fiber optic locations) may vary. To avoid these differences affecting subsequent global processing, channel normalization is performed on the superimposed signals. For example, the root mean square value or maximum value of each signal over the entire processing period is calculated, and then all data points of that channel are divided by this value, making the amplitude scale of the signals from each channel uniform and obtaining a normalized vibration signal. This improves the stability and fairness of subsequent processing.
[0030] It should be noted that the normalized vibration signal is converted from the spatiotemporal domain to the frequency-wavenumber domain using a two-dimensional Fourier transform. In this domain, the signal is represented by an energy distribution on a two-dimensional plane of frequency and wavenumber. Regularly propagating waves will appear as continuous straight lines or curves, while random noise will be diffusely distributed. Based on the statistical characteristics of noise in the FK domain, a global hard threshold is set. All points in the FK domain with amplitudes below this threshold are set to zero. This effectively filters out a large amount of non-directional random environmental background noise, retaining the higher-energy signal components, and outputting the first filtered signal.
[0031] S4. Perform sector-area filtering on the first filtered signal in the frequency-wavenumber domain to remove structural noise in the vertical and horizontal directions, and obtain the second filtered signal.
[0032] In this embodiment, step S4 specifically includes: The first filtered signal is subjected to fan-shaped filtering in the frequency-wavenumber domain within the range of 30 to 60 degrees to remove structural noise in the vertical and horizontal directions, thus obtaining the second filtered signal.
[0033] Specifically, the fan-shaped region with wavenumber direction in the range of 30 to 60 degrees corresponds to the propagation direction of ground vibration waves caused by vehicle movement.
[0034] Specifically, in the FK domain, the vibration waves generated by vehicles traveling along the road have a specific range in their propagation direction (wavenumber direction) (e.g., an angle between the wavenumber and the fiber optic axis of 30 to 60 degrees). The inherent coherent noise of the system, such as vertical and horizontal interference, typically manifests as energy bands close to 0 or 90 degrees. Therefore, a sector passband filter is designed to retain only the FK domain signal energy within the wavenumber direction range of 30 to 60 degrees, setting the energy outside this range to zero. This step can specifically filter out structural noise in both the vertical and horizontal directions, resulting in a cleaner second filtered signal.
[0035] S5. Perform Hough transform on the second filtered signal to enhance the spatiotemporal continuity of the vehicle vibration signal and obtain the final processed signal for vehicle feature extraction.
[0036] In this embodiment, step S5 specifically includes: The second filtered signal is processed by Hough transform to convert the discrete vehicle signal points in the second filtered signal into continuous spatiotemporal trajectory features, thereby enhancing the spatiotemporal continuity of the vehicle vibration signal and obtaining the final processed signal for vehicle feature extraction.
[0037] Specifically, after the aforementioned filtering, the vehicle signal in the spatiotemporal graph may still be an incompletely continuous trajectory composed of discrete points. The Hough transform is applied to detect straight or curved features in the spatiotemporal domain. Using the second filtered signal as input, the Hough transform parameterizes collinear discrete points into continuous straight lines, thus significantly enhancing the spatiotemporal continuity of the vehicle signal and directly outputting the parameterized trajectory information as the final processed signal. This step greatly facilitates subsequent vehicle counting, speed calculation, and other processes.
[0038] In this embodiment, it also includes: S6. Based on the final processed signal, perform vehicle type identification, traffic flow statistics, vehicle speed estimation, or traffic anomaly detection.
[0039] Specifically, based on the final processed signal output by the Hough transform, a wealth of traffic parameters can be extracted. For example, traffic flow statistics can be achieved by counting the number of trajectories per unit time; vehicle speed can be estimated by calculating the slope of the trajectory in the spatiotemporal graph; traffic anomalies (such as sudden braking and traffic accidents) can be identified by analyzing sudden interruptions, abrupt changes in slope, or specific vibration patterns of the trajectory; and preliminary vehicle type classification can be performed by combining the amplitude and spectral characteristics of the vibration signal.
[0040] Example 1
[0041] To verify the actual processing effect of the present invention, a distributed acoustic sensing (DAS) test system was built along a main road in a city using pre-buried communication optical cables. The system sampling frequency was set to 500Hz, the sensor channel spacing was 10 meters, and raw vibration data including multi-vehicle traffic scenarios were continuously collected.
[0042] Figure 2 The image shows the spatiotemporal distribution of the raw test data signal without any processing, namely the aforementioned "raw spatiotemporal vibration signal". As shown in the figure, the vehicle signal (displayed as diagonal stripes in the figure) is completely submerged in strong background noise, with only a partially blurred outline of the vehicle signal visible. The signal-to-noise ratio is extremely low, making it unsuitable for direct vehicle detection.
[0043] The original signal is input into the process of this invention. First, step S2 is executed to perform bandpass filtering processing with a passband of 40Hz to 60Hz. The processing result is as follows. Figure 3 The image shown is the "preliminary filtered signal". Compare... Figure 2 As can be seen, this step effectively removes most of the low-frequency environmental vibrations and high-frequency random noise, making the vibration signals originating from the vehicle stand out, becoming richer and clearer, but some noise still remains in the signal.
[0044] Subsequently, steps S3 and S4 are executed sequentially, namely, time window superposition and channel normalization processing are performed on the preliminary filtered signal to obtain the "normalized vibration signal". The result of this step is as follows: Figure 4 As shown, after superposition enhancement, the strength of the vehicle signal is significantly improved; and normalization processing makes the signal amplitude between different channels tend to be consistent. Nevertheless, residual vertical and horizontal stripe noise with specific directions can still be observed in the figure, and the continuity of the vehicle signal is still insufficient, with broken or incomplete trajectories.
[0045] Finally, steps S5 to S7 are performed on the aforementioned signal, namely, FK domain hard thresholding, sector filtering, and Hough transform. The final output "final processed signal" is as follows: Figure 5 As shown. After this series of targeted processing steps, most of the background noise was filtered out, and the vehicle signal was greatly purified and enhanced. For example... Figure 5 As clearly shown, the vehicle signal exhibits a continuous and complete oblique trajectory with good spatiotemporal continuity, which can be directly used for accurate traffic flow statistics, vehicle speed calculation, and trajectory tracking. This fully verifies the effectiveness of the method of the present invention in extracting vehicle signals with high precision in complex noise backgrounds.
[0046] Reference Figures 1-6 The present invention proposes a vehicle signal extraction system based on distributed acoustic sensing, comprising: The signal acquisition module is used to acquire the raw spatiotemporal vibration signals collected by the distributed fiber optic acoustic sensing system along the road. The first processing module is used to perform bandpass filtering on the original spatiotemporal vibration signal to obtain a preliminary filtered signal, and to perform superposition processing on the preliminary filtered signal in the time dimension with a preset time window to obtain a superimposed signal with signal enhancement. The second processing module is used to normalize the signal amplitude of each sensing channel in the superimposed signal to obtain a normalized vibration signal, transform the normalized vibration signal to the frequency-wavenumber domain, and apply a hard threshold filter to the frequency-wavenumber domain signal to remove irregular noise to obtain a first filtered signal. The third processing module is used to perform sector-area filtering on the first filtered signal in the frequency-wavenumber domain to remove structural noise in the vertical and horizontal directions, and obtain the second filtered signal. The transformation output module is used to perform Hough transform processing on the second filtered signal to enhance the spatiotemporal continuity of the vehicle vibration signal and obtain the final processed signal for vehicle feature extraction.
[0047] In this embodiment, it also includes: The signal processing module is used to perform vehicle type identification, traffic flow statistics, vehicle speed estimation, or traffic anomaly detection based on the final processed signal.
[0048] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.
Claims
1. A method for extracting vehicle signals based on distributed acoustic sensing, characterized in that, Includes the following steps: S1. Acquire the raw spatiotemporal vibration signals collected along the road by the distributed fiber optic acoustic sensing system; S2. Perform bandpass filtering on the original spatiotemporal vibration signal to obtain a preliminary filtered signal. Then, perform superposition processing on the preliminary filtered signal in the time dimension with a preset time window to obtain a superimposed signal with signal enhancement. S3. Normalize the signal amplitude of each sensing channel in the superimposed signal to obtain a normalized vibration signal. Transform the normalized vibration signal to the frequency-wavenumber domain and apply a hard threshold filter to the frequency-wavenumber domain signal to remove irregular noise to obtain a first filtered signal. S4. Perform sector-area filtering on the first filtered signal in the frequency-wavenumber domain to remove structural noise in the vertical and horizontal directions, and obtain the second filtered signal. S5. Perform Hough transform on the second filtered signal to enhance the spatiotemporal continuity of the vehicle vibration signal and obtain the final processed signal for vehicle feature extraction.
2. The vehicle signal extraction method based on distributed acoustic sensing according to claim 1, characterized in that, Step S4 specifically includes: The first filtered signal is subjected to fan-shaped filtering in the frequency-wavenumber domain within the range of 30 to 60 degrees to remove structural noise in the vertical and horizontal directions, thereby obtaining the second filtered signal.
3. The vehicle signal extraction method based on distributed acoustic sensing according to claim 2, characterized in that, The fan-shaped region with wavenumber direction in the range of 30 to 60 degrees corresponds to the propagation direction of ground vibration waves caused by vehicle movement.
4. The vehicle signal extraction method based on distributed acoustic sensing according to claim 1, characterized in that, The preset time window is n seconds, where 1≤n≤2.
5. The vehicle signal extraction method based on distributed acoustic sensing according to claim 4, characterized in that, The preset time window is 2 seconds.
6. The vehicle signal extraction method based on distributed acoustic sensing according to claim 1, characterized in that, The preset threshold of the hard threshold filter is determined by calculating the statistical characteristic value of the noise component in the frequency-wavenumber domain signal.
7. The vehicle signal extraction method based on distributed acoustic sensing according to claim 1, characterized in that, The frequency band of the bandpass filter is 40Hz to 60Hz.
8. The vehicle signal extraction method based on distributed acoustic sensing according to claim 1, characterized in that, Step S5 specifically includes: The second filtered signal is subjected to Hough transform processing to convert the discrete vehicle signal points in the second filtered signal into continuous spatiotemporal trajectory features, so as to enhance the spatiotemporal continuity of the vehicle vibration signal and obtain the final processed signal for vehicle feature extraction.
9. The vehicle signal extraction method based on distributed acoustic sensing according to claim 1, characterized in that, Also includes: S6. Based on the final processed signal, perform vehicle type identification, traffic flow statistics, vehicle speed estimation, or traffic anomaly detection.
10. A vehicle signal extraction system based on distributed acoustic sensing, characterized in that, include: The signal acquisition module is used to acquire the raw spatiotemporal vibration signals collected by the distributed fiber optic acoustic sensing system along the road. The first processing module is used to perform bandpass filtering on the original spatiotemporal vibration signal to obtain a preliminary filtered signal, and to perform superposition processing on the preliminary filtered signal in the time dimension with a preset time window to obtain a superimposed signal with signal enhancement. The second processing module is used to normalize the signal amplitude of each sensing channel in the superimposed signal to obtain a normalized vibration signal, transform the normalized vibration signal to the frequency-wavenumber domain, and apply a hard threshold filter to the frequency-wavenumber domain signal to remove irregular noise to obtain a first filtered signal. The third processing module is used to perform sector-area filtering on the first filtered signal in the frequency-wavenumber domain to remove structural noise in the vertical and horizontal directions, and obtain the second filtered signal. The transformation output module is used to perform Hough transform processing on the second filtered signal to enhance the spatiotemporal continuity of the vehicle vibration signal and obtain the final processed signal for vehicle feature extraction.