A distributed acoustic sensing road cavity monitoring method based on vehicle-induced vibration signals

Abstract

The application discloses a kind of distributed acoustic sensing road cavity monitoring methods based on vehicle-induced vibration signal.The method is based on distributed acoustic sensing technology, first parameter initialization and obtain the distributed acoustic sensing signal sampling data sequence to be processed;Then the data is preprocessed and vehicle signal is enhanced;Vehicle trajectory parameters are extracted;On this basis, the correlation characteristics between distributed acoustic sensing channels are calculated using the extracted vehicle trajectory parameters to determine whether there is a cavity.This method makes full use of the all-weather perception and high resolution accuracy features of the distributed acoustic sensing array, and accurately monitors the road cavity with low computational complexity, suitable for large-scale real-time monitoring engineering applications.

Description

Technical Field

[0001] This invention belongs to the field of distributed acoustic sensing technology, and particularly relates to a distributed acoustic sensing method for monitoring road cavities based on vehicle-induced vibration signals. Background Technology

[0002] With the acceleration of urbanization and the increasing density of transportation infrastructure, urban road subsidence and collapse accidents are gradually increasing due to factors such as aging underground facilities, changes in the geological environment, frequent construction activities, and frequent natural disasters. Therefore, the monitoring and early warning of underground cavities in roads is of great significance for ensuring the safety of road users and the long-term stable operation of transportation infrastructure. Existing methods for monitoring road cavities mainly include: (1) Intrusive monitoring methods, which can obtain detailed information on the characteristics of cavities by drilling cores or local excavation, but are costly and time-consuming, making it difficult to meet the needs of large-scale rapid monitoring; (2) Ground penetrating radar methods, which identify cavities by analyzing the reflection of electromagnetic waves in discontinuous media, have the advantages of being non-destructive and highly accurate, and significantly improve monitoring efficiency, but their results are easily affected by external electromagnetic interference, and the post-processing and interpretation of data mainly depend on the experience of operators, making it difficult to accurately distinguish road cavities; (3) Active source imaging methods, which analyze artificially excited seismic wave signals to invert underground physical parameters, thereby realizing underground structure imaging, have the advantages of being spatiotemporally controllable and having a high signal-to-noise ratio, but are costly to implement due to the need to add an extra excitation source; (4) Background noise imaging methods, which obtain effective information on the characteristics of underground structures through environmental noise interference, have the advantages of being low-cost, non-invasive, and requiring no extra excitation source, but are costly to calculate because the background noise intensity is usually low and usually requires several days or even weeks of noise recording to obtain convergent cross-correlation results. Summary of the Invention

[0003] The purpose of this invention is to provide a distributed acoustic sensing method for monitoring road cavities based on vehicle-induced vibration signals, in order to solve the technical problem of large-scale, high-density, and long-term monitoring of cavities in urban roads.

[0004] Technical Solution: To solve the above-mentioned technical problems, this invention proposes a distributed acoustic sensing method for monitoring road cavities based on vehicle-induced vibration signals, comprising the following steps:

[0005] Step 1: Parameter initialization. Obtain the distributed acoustic sensor signal sampling data sequence x(k, l) to be processed, where k = 0, 1, … , K-1, l = 0, 1, … , L-1, k is the channel number, l is the discrete-time index, K is the number of sensor channels, and L is the number of time sampling points. K and L are both positive integers and satisfy K≥10 and L≥2. 14 ;

[0006] Step 2: Preprocess the distributed acoustic sensor signal sampling data sequence x(k, l) to be processed;

[0007] Step 3: Perform vehicle signal enhancement on the preprocessed data to obtain enhanced data;

[0008] Step 4: Extract vehicle trajectory parameters using the enhanced data;

[0009] Step 5: Extract the correlation features of the distributed acoustic sensing channel using vehicle trajectory parameters to determine whether there are any gaps.

[0010] Furthermore, in step 1, the parameters are initialized using the following method to obtain the distributed acoustic sensor signal sampling data sequence x(k, l), k = 0, 1, … , K-1, l = 0, 1, … , L-1, specifically including the following steps:

[0011] Step 1.1: Initialize the sensing parameters for cavity monitoring, specifically including the initialization of the following parameters:

[0012] sampling frequency f s Initialized as: 2 < f s Real numbers < 8 kHz;

[0013] lower bandpass filter frequency f L Initialized to 0.05 < f L Real numbers < 5 Hz;

[0014] upper limit frequency f of bandpass filter H Initialize to 10 < f H Real numbers < 50 Hz;

[0015] Minimum value of signal amplitude ε x Initialize to 10 -6 < ε x < 10 -4 real numbers;

[0016] The vehicle driving direction flag F is initialized to 0 or 1. 0 indicates that the vehicle is traveling in the direction of increasing distributed sound sensor channel number, and 1 indicates that the vehicle is traveling in the direction of decreasing distributed sound sensor channel number.

[0017] The sampling interval Δk of the correlation feature channel is initialized to an integer of 1 ≤ Δk ≤ 5;

[0018] The sliding window length W0 for correlation feature extraction is initialized to 0.2f. s ≤ W0 ≤ 2f s Integers;

[0019] Narrowband feature extraction sliding window length W1 is initialized to f s ≤ W1 ≤ 3f s Integers;

[0020] The split window length W2 for narrowband feature extraction is initialized to an integer of 0.6W1 ≤ W2 ≤ 0.98W1;

[0021] Relevance feature weights Initialize to 0.2 ≤ Real numbers ≤ 0.7;

[0022] The vehicle signal detection energy threshold γ is initialized to a real number 2 ≤ γ ≤ 10;

[0023] The vehicle trajectory detection slope step size Δa is initialized to 10. -4 ≤ Δa ≤ 10 -1 real numbers;

[0024] The vehicle trajectory detection intercept step size Δb is initialized to 0.01 f. s ≤ Δb ≤ 2f s real numbers;

[0025] Upper limit of vehicle trajectory detection slope n a Initialize to 10 ≤ n a ≤ 10 4 Integers;

[0026] Vehicle trajectory detection intercept upper limit n b Initialize to n b = , To round down;

[0027] The bandwidth ω of the vehicle trajectory detection integral region is initialized to an integer of 2 ≤ ω ≤ 10;

[0028] vehicle signal extraction start time offset l s Initialized to f s ≤ l s ≤ 3f s real numbers;

[0029] Vehicle signal extraction termination time offset l e Initialize to l s + f s ≤ l e ≤ l s + 4f s real numbers;

[0030] Cavity monitoring correlation coefficient threshold Cth Initialize to 0 ≤ C th Real numbers ≤ 0.5;

[0031] Step 1.2: Obtain the distributed acoustic sensor signal sampling data sequence x(k, l) to be processed, where k = 0, 1, …, K-1, l = 0, 1, … , L-1, k is the channel number, and l is the discrete-time index. The real-time acquisition data of L sampling points simultaneously received by K sensors in an array is used as the data sequence x(k, l) to be processed, or the distributed acoustic sensor signal data containing L sampling points acquired by K sensors is extracted from the memory as the data sequence x(k, l) to be processed.

[0032] Furthermore, in step 2, the distributed acoustic sensor signal sampling data sequence x(k, l) is preprocessed using the following method, specifically including the following steps:

[0033] Step 2.1, Design frequency range is [f L , f H A bandpass filter is used to apply bandpass filtering to the sampled data sequence x(k, l) of the distributed acoustic sensor signal to obtain the bandpass-filtered data x. b (k, l);

[0034] Step 2.2: Apply bandpass filtering to the data x b The spatial frequency spectrum X(m, n) is obtained by performing a two-dimensional Fourier transform on (k, l):

[0035]

[0036] Where j is the imaginary sign, m = 0, 1, …, K-1 are discrete wavenumber indices, and n = 0, 1, …, L-1 are discrete frequency indices; fk filtering is performed to obtain the filtered spatial frequency spectrum. :

[0037]

[0038] Where m = 0, 1, … , K - 1, n = 0, 1, … , L – 1, M(m) is the spatial filtering mask, defined as:

[0039]

[0040] Where m = 0, 1, … , K – 1, and for any fixed m, M (m) takes the same value for all n;

[0041] The filtered spatial frequency spectrum Perform an inverse two-dimensional Fourier transform to obtain the data x after fk filtering. fk (k, l):

[0042]

[0043] Where k = 0, 1, … , K-1, l = 0, 1, … , L-1;

[0044] Step 2.3: Filter the data x fk (k, l) undergoes channel equalization to obtain preprocessed data x p (k, l):

[0045]

[0046] Where k = 0, 1, … , K-1, l = 0, 1, … , L-1, x m (k) represents x for all time frames l = 0, 1, … , L-1 of channel k. fk (k, l) represents the mean of the absolute values, and max(·) represents taking the maximum of the two.

[0047] Furthermore, in step 3, the vehicle signal is enhanced using the following method, specifically including the following steps:

[0048] Step 3.1: Extract the signal correlation feature parameter r between channels using a sliding window. (k, l):

[0049]

[0050] Where k = 0, 1, … , K-1, l = 0, 1, … , L-1, and |·| is the modulo operation. For sliding window Ω w Discrete-time index within (l). The offset channel for relevance feature extraction is calculated as follows:

[0051]

[0052] Ω w (l) is a sliding window for extracting correlation features of frame l, defined as:

[0053]

[0054] and For the kth and the Correlation feature extraction of channel 1 in frame l using a sliding window Ω w The average signal amplitude within (l) is calculated as follows:

[0055]

[0056] in, This indicates that for all channels k that satisfy the condition... of Summation, Indicates the first All conditions of channel number are met. of Perform summation;

[0057] Step 3.2: Extract narrowband feature parameters of the signal using a sliding window. :

[0058]

[0059] Where k = 0, 1, … , K-1, l = 0, 1, … , L-1, N l For narrowband feature extraction of frame l, a sliding window Ω is used. l Total number of frames within, Ω l The sliding window for narrowband feature extraction is defined as:

[0060] ;

[0061] Step 3.3: Extract the narrowband feature parameters of the signal. Normalization is performed to obtain normalized narrowband feature parameters. :

[0062]

[0063] Where k = 0, 1, … , K-1, l = 0, 1, … , L-1, Narrowband characteristic parameters of the signal The minimum value, Narrowband characteristic parameters of the signal The maximum value;

[0064] Step 3.4: Based on the inter-channel signal correlation coefficient r(k, l) and normalized narrowband feature parameters extracted by the sliding window... Calculate the gain function G(k, l):

[0065]

[0066] Where k = 0, 1, … , K-1, l = 0, 1, … , L-1;

[0067] Step 3.5: Process the preprocessed data x p (k, l) Signal enhancement is performed to obtain enhanced data x e (k, l):

[0068]

[0069] Where k = 0, 1, … , K-1, l = 0, 1, … , L-1.

[0070] Furthermore, in step 4, the vehicle trajectory parameters are extracted using the following method, specifically including the following steps:

[0071] Step 4.1: Calculate the enhanced data x e The energy integral E(a, b) of (k, l) in each zonal region:

[0072]

[0073] Where a = Δa, 2Δa… , n a Δa represents the slope of the trajectory, and b = Δb, 2Δb, ..., n b Δb represents the trajectory intercept, and Ω(a, b) is the zone of integration, defined as the region formed by all points (k, l) that satisfy the following condition:

[0074]

[0075] Where k = 0, 1, … , K-1, l = 0, 1, … , L-1;

[0076] Step 4.2: Detect vehicle trajectory based on energy integral E(a, b), and store the detection results in the vehicle trajectory parameter set D, i.e.: D = {(a, b)} i , b i ) | E(a i , b i ) > γE ave} , i = 1, 2, …, I, a i Take a value from a, b i It takes values ​​from b, where i is the vehicle signal index, I is the total number of vehicle signals, and E ave The average signal energy is calculated as follows:

[0077]

[0078] in, This indicates that the sequence a = Δa, 2Δa, ..., n is iterated sequentially. a Δa and b = Δb, 2Δb, ..., n b Sum all values ​​of Δb over E(a, b);

[0079] Step 4.3: For each detected vehicle trajectory segment, based on the extracted vehicle trajectory slope 'a'... i and intercept b i Let i = 1, 2, …, I, and calculate the midpoint of the vehicle trajectory in time. , To round down;

[0080] Step 4.4, Determine L m If the data set is empty, it is determined that there is no valid hole monitoring vehicle signal and the processing ends; otherwise, proceed to step 5.

[0081] Furthermore, in step 5, the following method is used to extract the correlation features of the distributed acoustic sensing channels using vehicle trajectory parameters to determine whether there are gaps, specifically including the following steps:

[0082] Step 5.1: Based on the midpoint L of the vehicle trajectory time. m Obtain the vehicle signal interval T = { [l + l] for correlation feature extraction. s , l + l e ]| l ∈L m And l + l e < L};

[0083] Step 5.2: Calculate the inter-channel correlation coefficient for each segment of vehicle signal data. :

[0084]

[0085] Among them, k = 0, 1, … , K-1, i = 1, 2, …, I, The offset channel is extracted based on the correlation features calculated in step 3.1, with the vehicle signal time range T. i ∈ T, i = 1, 2, …, I, and For the kth and the Channel 1 in vehicle signal time range T i The average signal amplitude within the range is calculated as follows:

[0086]

[0087] in, This represents all channels k that satisfy l ∈ T. i of Summation, Indicates the first All channels satisfy l ∈ T i of Perform summation;

[0088] Step 5.3: Calculate the inter-channel correlation coefficients for each segment of vehicle signals. The average correlation coefficient between channels is obtained by averaging. The calculation is as follows:

[0089]

[0090] Where k = 0, 1, … , K-1;

[0091] Step 5.4: Based on the cavity monitoring correlation coefficient threshold C th To satisfy all Channel k is determined to be a void channel and stored in the void channel set V, that is:

[0092] V = { k | }, where k = 0, 1, … , K-1.

[0093] Beneficial effects: Compared with the prior art, the technical solution of the present invention has the following beneficial technical effects:

[0094] 1. This invention introduces preprocessing techniques such as bandpass filtering, FK filtering, and channel equalization, as shown in step 2, to gradually separate the vehicle-induced vibration signals propagating in the same direction within the vehicle's frequency band. This effectively reduces the impact of interference between different distributed acoustic sensor data and inconsistent channel coupling, improves the quality of the cross-correlation results between vehicle signals and channels, and thus enhances the robustness of vehicle signal detection and road cavity identification.

[0095] 2. Based on the inter-channel correlation characteristics and narrowband characteristics of vehicle-induced vibration signals, this invention enhances the signal, as shown in step 3. The extracted feature parameters have good distinguishability between vehicle signals and background noise in the spatiotemporal domain, thereby further improving the robustness of the method in complex urban road environments.

[0096] 3. This invention makes full use of the waveform distortion of vehicle-induced vibration signals caused by cavities and achieves feature quantification through correlation coefficients, as shown in the corresponding processing steps in step 5. The extracted feature parameters can better reflect the structural features of underground space, thereby achieving efficient and accurate identification of road cavities. Attached Figure Description

[0097] Figure 1This is a schematic flowchart of the method of the present invention;

[0098] Figure 2 This is a time-channel diagram of the distributed acoustic sensor signal sampling data sequence for Example 1;

[0099] Figure 3 This is a time channel diagram of the signal preprocessing in Example 1;

[0100] Figure 4 This is a time channel diagram of the enhanced vehicle-induced signal in Example 1;

[0101] Figure 5 This is a graph showing the extraction results of the average correlation coefficient between vehicle signal channels in Example 1;

[0102] Figure 6 This is a time-channel diagram of the distributed acoustic sensor signal sampling data sequence for Example 2;

[0103] Figure 7 This is a time channel diagram of the signal preprocessing in Example 2;

[0104] Figure 8 This is a time channel diagram of the enhanced vehicle-induced signal in Example 2;

[0105] Figure 9 This is a graph showing the results of extracting the average correlation coefficient between vehicle signal channels in Example 2. Detailed Implementation

[0106] To better understand the purpose, structure, and function of this invention, the following detailed description of a distributed acoustic sensing method for monitoring road cavities based on vehicle-induced vibration signals is provided in conjunction with the accompanying drawings.

[0107] like Figure 1 As shown, this invention proposes a distributed acoustic sensing method for monitoring road cavities based on vehicle-induced vibration signals, comprising the following steps:

[0108] Step 1: Parameter initialization. Obtain the distributed acoustic sensor signal sampling data sequence x(k, l) to be processed, where k = 0, 1, … , K-1, l = 0, 1, … , L-1, k is the channel number, l is the discrete-time index, K is the number of sensor channels, and L is the number of time sampling points. K and L are both positive integers and satisfy K≥10 and L≥2. 14 ;

[0109] Step 2: Preprocess the distributed acoustic sensor signal sampling data sequence x(k, l) to be processed;

[0110] Step 3: Perform vehicle signal enhancement on the preprocessed data to obtain enhanced data;

[0111] Step 4: Extract vehicle trajectory parameters using the enhanced data;

[0112] Step 5: Extract the correlation features of the distributed acoustic sensing channel using vehicle trajectory parameters to determine whether there are any gaps.

[0113] Furthermore, in step 1, the parameters are initialized using the following method to obtain the distributed acoustic sensor signal sampling data sequence x(k, l), k = 0, 1, … , K-1, l = 0, 1, … , L-1, specifically including the following steps:

[0114] Step 1.1: Initialize the sensing parameters for cavity monitoring, specifically including the initialization of the following parameters:

[0115] sampling frequency f s Initialized as: 2 < f s Real numbers < 8 kHz;

[0116] lower bandpass filter frequency f L Initialized to 0.05 < f L Real numbers < 5 Hz;

[0117] upper limit frequency f of bandpass filter H Initialize to 10 < f H Real numbers < 50 Hz;

[0118] Minimum value of signal amplitude ε x Initialize to 10 -6 < ε x < 10 -4 real numbers;

[0119] The vehicle driving direction flag F is initialized to 0 or 1. 0 indicates that the vehicle is traveling in the direction of increasing distributed sound sensor channel number, and 1 indicates that the vehicle is traveling in the direction of decreasing distributed sound sensor channel number.

[0120] The sampling interval Δk of the correlation feature channel is initialized to an integer of 1 ≤ Δk ≤ 5;

[0121] The sliding window length W0 for correlation feature extraction is initialized to 0.2f. s ≤ W0 ≤ 2f s Integers;

[0122] Narrowband feature extraction sliding window length W1 is initialized to f s ≤ W1 ≤ 3f s Integers;

[0123] The split window length W2 for narrowband feature extraction is initialized to an integer of 0.6W1 ≤ W2 ≤ 0.98W1;

[0124] Relevance feature weights Initialize to 0.2 ≤ Real numbers ≤ 0.7;

[0125] The vehicle signal detection energy threshold γ is initialized to a real number 2 ≤ γ ≤ 10;

[0126] The vehicle trajectory detection slope step size Δa is initialized to 10. -4 ≤ Δa ≤ 10 -1 real numbers;

[0127] The vehicle trajectory detection intercept step size Δb is initialized to 0.01 f. s ≤ Δb ≤ 2f s real numbers;

[0128] Upper limit of vehicle trajectory detection slope n a Initialize to 10 ≤ n a ≤ 10 4 Integers;

[0129] Vehicle trajectory detection intercept upper limit n b Initialize to n b = , To round down;

[0130] The bandwidth ω of the vehicle trajectory detection integral region is initialized to an integer of 2 ≤ ω ≤ 10;

[0131] vehicle signal extraction start time offset l s Initialized to f s ≤ l s ≤ 3f s real numbers;

[0132] Vehicle signal extraction termination time offset l e Initialize to l s + f s ≤ l e ≤ l s + 4f s real numbers;

[0133] Cavity monitoring correlation coefficient threshold C th Initialize to 0 ≤ C th Real numbers ≤ 0.5;

[0134] Step 1.2: Obtain the distributed acoustic sensor signal sampling data sequence x(k, l) to be processed, where k = 0, 1, …, K-1, l = 0, 1, … , L-1, k is the channel number, and l is the discrete-time index. The real-time acquisition data of L sampling points simultaneously received by K sensors in an array is used as the data sequence x(k, l) to be processed, or the distributed acoustic sensor signal data containing L sampling points acquired by K sensors is extracted from the memory as the data sequence x(k, l) to be processed.

[0135] Furthermore, in step 2, the distributed acoustic sensor signal sampling data sequence x(k, l) is preprocessed using the following method, specifically including the following steps:

[0136] Step 2.1, Design frequency range is [f L , f H A bandpass filter is used to apply bandpass filtering to the sampled data sequence x(k, l) of the distributed acoustic sensor signal to obtain the bandpass-filtered data x. b (k, l);

[0137] Step 2.2: Apply bandpass filtering to the data x b The spatial frequency spectrum X(m, n) is obtained by performing a two-dimensional Fourier transform on (k, l):

[0138]

[0139] Where j is the imaginary sign, m = 0, 1, …, K-1 are discrete wavenumber indices, and n = 0, 1, …, L-1 are discrete frequency indices; fk filtering is performed to obtain the filtered spatial frequency spectrum. :

[0140]

[0141] Where m = 0, 1, … , K - 1, n = 0, 1, … , L – 1, M(m) is the spatial filtering mask, defined as:

[0142]

[0143] Where m = 0, 1, … , K – 1, and for any fixed m, M (m) takes the same value for all n;

[0144] The filtered spatial frequency spectrum Perform an inverse two-dimensional Fourier transform to obtain the data x after fk filtering. fk (k, l):

[0145]

[0146] Where k = 0, 1, … , K-1, l = 0, 1, … , L-1;

[0147] Step 2.3: Filter the data x fk (k, l) undergoes channel equalization to obtain preprocessed data x p (k, l):

[0148]

[0149] Where k = 0, 1, … , K-1, l = 0, 1, … , L-1, x m (k) represents x for all time frames l = 0, 1, … , L-1 of channel k. fk (k, l) represents the mean of the absolute values, and max(·) represents taking the maximum of the two.

[0150] Furthermore, in step 3, the vehicle signal is enhanced using the following method, specifically including the following steps:

[0151] Step 3.1: Extract the signal correlation feature parameter r between channels using a sliding window. (k, l):

[0152]

[0153] Where k = 0, 1, … , K-1, l = 0, 1, … , L-1, and |·| is the modulo operation. For sliding window Ω w Discrete-time index within (l). The offset channel for relevance feature extraction is calculated as follows:

[0154]

[0155] Ω w (l) is a sliding window for extracting correlation features of frame l, defined as:

[0156]

[0157] and For the kth and the Correlation feature extraction of channel 1 in frame l using a sliding window Ω w The average signal amplitude within (l) is calculated as follows:

[0158]

[0159] in, This indicates that for all channels k that satisfy the condition... of Summation, Indicates the first All conditions of channel number are met. of Perform summation;

[0160] Step 3.2: Extract narrowband feature parameters of the signal using a sliding window. :

[0161]

[0162] Where k = 0, 1, … , K-1, l = 0, 1, … , L-1, N l For narrowband feature extraction of frame l, a sliding window Ω is used. l Total number of frames within, Ω l The sliding window for narrowband feature extraction is defined as:

[0163] ;

[0164] Step 3.3: Extract the narrowband feature parameters of the signal. Normalization is performed to obtain normalized narrowband feature parameters. :

[0165]

[0166] Where k = 0, 1, … , K-1, l = 0, 1, … , L-1, Narrowband characteristic parameters of the signal The minimum value, Narrowband characteristic parameters of the signal The maximum value;

[0167] Step 3.4: Based on the inter-channel signal correlation coefficient r(k, l) and normalized narrowband feature parameters extracted by the sliding window... Calculate the gain function G(k, l):

[0168]

[0169] Where k = 0, 1, … , K-1, l = 0, 1, … , L-1;

[0170] Step 3.5: Process the preprocessed data x p (k, l) Signal enhancement is performed to obtain enhanced data xe (k, l):

[0171]

[0172] Where k = 0, 1, … , K-1, l = 0, 1, … , L-1.

[0173] Furthermore, in step 4, the vehicle trajectory parameters are extracted using the following method, specifically including the following steps:

[0174] Step 4.1: Calculate the enhanced data x e The energy integral E(a, b) of (k, l) in each zonal region:

[0175]

[0176] Where a = Δa, 2Δa… , n a Δa represents the slope of the trajectory, and b = Δb, 2Δb, ..., n b Δb represents the trajectory intercept, and Ω(a, b) is the zone of integration, defined as the region formed by all points (k, l) that satisfy the following condition:

[0177]

[0178] Where k = 0, 1, … , K-1, l = 0, 1, … , L-1;

[0179] Step 4.2: Detect vehicle trajectory based on energy integral E(a, b), and store the detection results in the vehicle trajectory parameter set D, i.e.: D = {(a, b)} i , b i ) | E(a i , b i ) > γE ave} , i = 1, 2, …, I, a i Take a value from a, b i It takes values ​​from b, where i is the vehicle signal index, I is the total number of vehicle signals, and E ave The average signal energy is calculated as follows:

[0180]

[0181] in, This indicates that the sequence a = Δa, 2Δa, ..., n is iterated sequentially. a Δa and b = Δb, 2Δb, ..., n b Sum all values ​​of Δb over E(a, b);

[0182] Step 4.3: For each detected vehicle trajectory segment, based on the extracted vehicle trajectory slope 'a'... i and intercept b i Let i = 1, 2, …, I, and calculate the midpoint of the vehicle trajectory in time. , To round down;

[0183] Step 4.4, Determine L m If the data set is empty, it is determined that there is no valid hole monitoring vehicle signal and the processing ends; otherwise, proceed to step 5.

[0184] Furthermore, in step 5, the following method is used to extract the correlation features of the distributed acoustic sensing channels using vehicle trajectory parameters to determine whether there are gaps, specifically including the following steps:

[0185] Step 5.1: Based on the midpoint L of the vehicle trajectory time. m Obtain the vehicle signal interval T = { [l + l] for correlation feature extraction. s , l + l e ]| l ∈L m And l + l e < L};

[0186] Step 5.2: Calculate the inter-channel correlation coefficient for each segment of vehicle signal data. :

[0187]

[0188] Among them, k = 0, 1, … , K-1, i = 1, 2, …, I, The offset channel is extracted based on the correlation features calculated in step 3.1, with the vehicle signal time range T. i ∈ T, i = 1, 2, …, I, and For the kth and the Channel 1 in vehicle signal time range T i The average signal amplitude within the range is calculated as follows:

[0189]

[0190] in, This represents all channels k that satisfy l ∈ T. i of Summation, Indicates the first All channels satisfy l ∈ T iof Perform summation;

[0191] Step 5.3: Calculate the inter-channel correlation coefficients for each segment of vehicle signals. The average correlation coefficient between channels is obtained by averaging. The calculation is as follows:

[0192]

[0193] Where k = 0, 1, … , K-1;

[0194] Step 5.4: Based on the cavity monitoring correlation coefficient threshold C th To satisfy all Channel k is determined to be a void channel and stored in the void channel set V, that is:

[0195] V = { k | }, where k = 0, 1, … , K-1.

[0196] Example 1:

[0197] Field verification of distributed acoustic sensing data collection and cavity monitoring was conducted on the southern section of the Second Ring Road in Wuhou District, Chengdu.

[0198] The following section selects a segment of distributed acoustic sensor data, consisting of 120 distributed acoustic sensor channels numbered 1380 to 1500, lasting 60 seconds, for cavity monitoring:

[0199] Based on step 1, set the sampling frequency f. s = 4 kHz, lower bandpass filter cutoff frequency f L = 0.1 Hz, upper limit frequency f of bandpass filter H = 25 Hz, minimum signal amplitude ε x = 10 -5 The vehicle direction marker F = 0, the correlation feature channel sampling interval Δk = 4, and the correlation feature extraction sliding window length W0 = f s Narrow-band feature extraction sliding window length W1 = 2f s The split window length for narrowband feature extraction is W2 = 0.95W1, and the relevance feature weight is α. r = 0.5, vehicle signal detection energy threshold γ = 5, vehicle trajectory detection slope step size Δa = 0.001, vehicle trajectory detection intercept step size Δb = 4000, vehicle trajectory detection slope upper limit n a = 50, upper limit of vehicle trajectory detection intercept n b= 60, vehicle trajectory detection integration region bandwidth ω = 3, vehicle signal extraction start time offset l s = f s The termination time offset l of vehicle signal extraction e = 2f s Cavity monitoring correlation coefficient threshold C th = 0.3, the time channel diagram of the distributed acoustic sensor signal sampling data sequence is as follows. Figure 2 As shown;

[0200] Based on step 2, the designed frequency range is [f L , f H The bandpass filter is used to perform bandpass filtering, fk filtering, and channel equalization on the sampled data sequence x(k, l) of the distributed acoustic sensor signal to obtain the preprocessed data x. p (k, l) such as Figure 3 As shown;

[0201] Based on step 3, the preprocessed data x p (k, l) Perform vehicle signal enhancement, and the enhanced data x e (k,l) such as Figure 4 As shown;

[0202] Based on step 4, using the enhanced data x e Vehicle trajectory parameters were extracted from (k, l). Two vehicle trajectory segments were detected, and the midpoint L of the vehicle trajectory time was calculated. m = {22.19 f s , 25.38 f s};

[0203] Based on step 5, obtain the vehicle signal interval T = { [23.19 f} for correlation feature extraction. s , 24.19f s ], [26.38 f s , 27.38 f s The average correlation coefficient between the extracted vehicle signal channels is calculated as follows: like Figure 5 As shown, all those that satisfy Channel k is determined to be a void channel and stored in the void channel set V, V = {1404, 1427, 1471}.

[0204] The data segment contains three road cavities in the experimental section, located in lanes 1401-1411, 1420-1430, and 1463-1473 respectively. This method successfully detected the lanes where the cavities were located.

[0205] Example 2:

[0206] Field verification of distributed acoustic sensing data collection and cavity monitoring was conducted on the southern section of the Second Ring Road in Wuhou District, Chengdu.

[0207] The following section selects a segment of distributed acoustic sensor data, consisting of 100 distributed acoustic sensor channels (channel numbers 1520 to 1620) for a duration of 60 seconds, for cavity detection:

[0208] Based on step 1, set the sampling frequency f. s = 4 kHz, lower bandpass filter cutoff frequency f L = 0.1 Hz, upper limit frequency f of bandpass filter H = 25 Hz, minimum signal amplitude ε x = 10 -5 The vehicle direction marker F = 0, the correlation feature channel sampling interval Δk = 4, and the correlation feature extraction sliding window length W0 = f s Narrow-band feature extraction sliding window length W1 = 2f s The split window length for narrowband feature extraction is W2 = 0.95W1, and the relevance feature weight is α. r = 0.5, vehicle signal detection energy threshold γ = 5, vehicle trajectory detection slope step size Δa = 0.001, vehicle trajectory detection intercept step size Δb = 4000, vehicle trajectory detection slope upper limit n a = 50, upper limit of vehicle trajectory detection intercept n b = 60, vehicle trajectory detection integration region bandwidth ω = 3, vehicle signal extraction start time offset l s = 2f s The termination time offset l of vehicle signal extraction e = 4f s Cavity monitoring correlation coefficient threshold C th = 0.3, the time channel diagram of the distributed acoustic sensor signal sampling data sequence is as follows. Figure 6 As shown;

[0209] Based on step 2, the designed frequency range is [f L , f H The bandpass filter is used to perform bandpass filtering, fk filtering, and channel equalization on the sampled data sequence x(k, l) of the distributed acoustic sensor signal to obtain the preprocessed data x. p (k, l) such as Figure 7 As shown;

[0210] Based on step 3, the preprocessed data x p(k, l) Perform vehicle signal enhancement, and the enhanced data x e (k,l) such as Figure 8 As shown;

[0211] Based on step 4, using the enhanced data x e Vehicle trajectory parameters were extracted from (k, l). A total of 3 vehicle trajectory segments were detected, and the midpoint L of the vehicle trajectory time was calculated. m = {18.00 f s , 35.14 f s , 51.38 f s};

[0212] Based on step 5, obtain the vehicle signal interval T = { [20.00 f} for correlation feature extraction. s , 22.00f s ], [37.14 f s , 39.14 f s ], [53.38 f s , 55.38 f s The average correlation coefficient between the extracted vehicle signal channels was obtained. like Figure 9 As shown, all those that satisfy Channel k is determined to be a void channel and stored in the void channel set V, V = {1557, 1594}.

[0213] The data segment contains two road cavities in the experimental section, located in lanes 1548-1558 and 1590-1600 respectively. This method successfully detected the lanes where the cavities are located.

[0214] It is understood that the present invention has been described through some embodiments, and those skilled in the art will recognize that various changes or equivalent substitutions can be made to these features and embodiments without departing from the spirit and scope of the invention. Furthermore, under the teachings of the present invention, these features and embodiments can be modified to adapt to specific situations and materials without departing from the spirit and scope of the invention. Therefore, the present invention is not limited to the specific embodiments disclosed herein, and all embodiments falling within the scope of the claims of this application are within the protection scope of the present invention.

Claims

1. A distributed acoustic sensing method for monitoring road cavities based on vehicle-induced vibration signals, characterized in that, Includes the following steps: Step 1: Parameter initialization. Obtain the distributed acoustic sensor signal sampling data sequence x(k, l) to be processed, where k = 0, 1, … , K-1, l = 0, 1, … , L-1, k is the channel number, l is the discrete-time index, K is the number of sensor channels, and L is the number of time sampling points. K and L are both positive integers and satisfy K≥10 and L≥2. 14 ; Step 2: Preprocess the distributed acoustic sensor signal sampling data sequence x(k, l) to be processed; Step 3: Perform vehicle signal enhancement on the preprocessed data to obtain enhanced data; Step 4: Extract vehicle trajectory parameters using the enhanced data; Step 5: Extract the correlation features of the distributed acoustic sensing channel using vehicle trajectory parameters to determine whether there are any gaps.

2. The distributed acoustic sensing method for monitoring road cavities based on vehicle-induced vibration signals according to claim 1, characterized in that, In step 1, the parameters are initialized using the following method to obtain the sampled data sequence x(k, l) of the distributed acoustic sensor signal to be processed, where k = 0, 1, … , K-1 and l = 0, 1, … , L-1. Specifically, this includes the following steps: Step 1.1: Initialize the sensing parameters for cavity monitoring, specifically including the initialization of the following parameters: sampling frequency f s Initialized as: 2 < f s Real numbers < 8 kHz; lower bandpass filter frequency f L Initialized to 0.05 < f L Real numbers < 5 Hz; upper limit frequency f of bandpass filter H Initialize to 10 < f H Real numbers < 50 Hz; Minimum value of signal amplitude ε x Initialize to 10 -6 < ε x < 10 -4 real numbers; The vehicle driving direction flag F is initialized to 0 or 1. 0 indicates that the vehicle is traveling in the direction of increasing distributed sound sensor channel number, and 1 indicates that the vehicle is traveling in the direction of decreasing distributed sound sensor channel number. The sampling interval Δk of the correlation feature channel is initialized to an integer of 1 ≤ Δk ≤ 5; The sliding window length W0 for correlation feature extraction is initialized to 0.2f. s ≤ W0 ≤ 2f s Integers; Narrowband feature extraction sliding window length W1 is initialized to f s ≤ W1 ≤ 3f s Integers; The split window length W2 for narrowband feature extraction is initialized to an integer of 0.6W1 ≤ W2 ≤ 0.98W1; Relevance feature weights Initialize to 0.2 ≤ Real numbers ≤ 0.7; The vehicle signal detection energy threshold γ is initialized to a real number 2 ≤ γ ≤ 10; The vehicle trajectory detection slope step size Δa is initialized to 10. -4 ≤ Δa ≤ 10 -1 real numbers; The vehicle trajectory detection intercept step size Δb is initialized to 0.01 f. s ≤ Δb ≤ 2f s real numbers; Upper limit of vehicle trajectory detection slope n a Initialize to 10 ≤ n a ≤ 10 4 Integers; Vehicle trajectory detection intercept upper limit n b Initialize to n b = , To round down; The bandwidth ω of the vehicle trajectory detection integral region is initialized to an integer of 2 ≤ ω ≤ 10; vehicle signal extraction start time offset l s Initialized to f s ≤ l s ≤ 3f s real numbers; Vehicle signal extraction termination time offset l e Initialize to l s + f s ≤ l e ≤ l s + 4f s real numbers; Cavity monitoring correlation coefficient threshold C th Initialize to 0 ≤ C th Real numbers ≤ 0.5; Step 1.2: Obtain the distributed acoustic sensor signal sampling data sequence x(k, l) to be processed, k = 0, 1, … ,K-1, l = 0, 1, … ,L-1, k is the channel number, l is the discrete time index. The real-time acquisition data of L sampling points simultaneously received by K sensors in an array is used as the data sequence x(k, l) to be processed, or the distributed acoustic sensor signal data containing L sampling points acquired by K sensors is extracted from the memory as the data sequence x(k, l) to be processed.

3. The distributed acoustic sensing method for monitoring road cavities based on vehicle-induced vibration signals according to claim 2, characterized in that, In step 2, the distributed acoustic sensor signal sampling data sequence x(k, l) is preprocessed using the following method, specifically including the following steps: Step 2.1, Design frequency range is [f L , f H A bandpass filter is used to apply bandpass filtering to the sampled data sequence x(k, l) of the distributed acoustic sensor signal to obtain the bandpass-filtered data x. b (k, l); Step 2.2: Apply bandpass filtering to the data x b The spatial frequency spectrum X(m, n) is obtained by performing a two-dimensional Fourier transform on (k, l): ； Where j is the imaginary sign, m = 0, 1, …, K-1 are discrete wavenumber indices, and n = 0, 1, …, L-1 are discrete frequency indices; fk filtering is performed to obtain the filtered spatial frequency spectrum. : ； Where m = 0, 1, … , K - 1, n = 0, 1, … , L – 1, M(m) is the spatial filtering mask, defined as: ； Where m = 0, 1, … , K – 1, and for any fixed m, M (m) takes the same value for all n; The filtered spatial frequency spectrum Perform an inverse two-dimensional Fourier transform to obtain the data x after fk filtering. fk (k,l): ； Where k = 0, 1, … , K-1, l = 0, 1, … , L-1; Step 2.3: Filter the data x fk (k, l) undergoes channel equalization to obtain preprocessed data x p (k,l): ； Where k = 0, 1, … , K-1, l = 0, 1, … , L-1, x m (k) represents x for all time frames l = 0, 1, … , L-1 of channel k. fk (k, l) represents the mean of the absolute values, and max(·) represents taking the maximum of the two.

4. The distributed acoustic sensing method for monitoring road cavities based on vehicle-induced vibration signals according to claim 3, characterized in that, Step 3 involves enhancing vehicle signals using the following method, specifically including the following steps: Step 3.1: Extract the signal correlation feature parameter r(k, l) between channels using a sliding window: ； Where k = 0, 1, … , K-1, l = 0, 1, … , L-1, and |·| is the modulo operation. For sliding window Ω w Discrete-time index within (l). The offset channel for relevance feature extraction is calculated as follows: ； Ω w (l) is a sliding window for extracting correlation features of frame l, defined as: ； and For the kth and the Correlation feature extraction of channel 1 in frame l using a sliding window Ω w The average signal amplitude within (l) is calculated as follows: ； in, This indicates that for all channels k that satisfy the condition... of Summation, Indicates the first All conditions of channel number are met. of Perform summation; Step 3.2: Extract narrowband feature parameters of the signal using a sliding window. : ； Where k = 0, 1, … , K-1, l = 0, 1, … , L-1, N l For narrowband feature extraction of frame l, a sliding window Ω is used. l Total number of frames within, Ω l The sliding window for narrowband feature extraction is defined as: ； Step 3.3: Extract the narrowband feature parameters of the signal. Normalization is performed to obtain normalized narrowband feature parameters. : ； Where k = 0, 1, … , K-1, l = 0, 1, … , L-1, Narrowband characteristic parameters of the signal The minimum value, Narrowband characteristic parameters of the signal The maximum value; Step 3.4: Based on the inter-channel signal correlation coefficient r(k, l) and normalized narrowband feature parameters extracted by the sliding window... Calculate the gain function G(k, l): ； Where k = 0, 1, … , K-1, l = 0, 1, … , L-1; Step 3.5: Process the preprocessed data x p (k, l) Signal enhancement is performed to obtain enhanced data x e (k, l): ； Where k = 0, 1, … , K-1, l = 0, 1, … , L-1.

5. The distributed acoustic sensing method for monitoring road cavities based on vehicle-induced vibration signals according to claim 4, characterized in that, In step 4, the vehicle trajectory parameters are extracted using the following method, specifically including the following steps: Step 4.1: Calculate the enhanced data x e The energy integral E(a, b) of (k, l) in each zonal region: ； Where a = Δa, 2Δa… , n a Δa represents the slope of the trajectory, and b = Δb, 2Δb, ..., n b Δb represents the trajectory intercept, and Ω(a, b) is the zone of integration, defined as the region formed by all points (k, l) that satisfy the following condition: ； Where k = 0, 1, … , K-1, l = 0, 1, … , L-1; Step 4.2: Detect vehicle trajectory based on energy integral E(a, b), and store the detection results in the vehicle trajectory parameter set D, i.e.: D = {(a, b)} i , b i ) | E(a i , b i ) > γE ave } , i = 1, 2, …, I, a i Take a value from a, b i It takes values ​​from b, where i is the vehicle signal index, I is the total number of vehicle signals, and E ave The average signal energy is calculated as follows: ； in, This indicates that the sequence a = Δa, 2Δa, ..., n is iterated sequentially. a Δa and b = Δb, 2Δb,…, n b Sum all values ​​of Δb over E(a, b); Step 4.3: For each detected vehicle trajectory segment, based on the extracted vehicle trajectory slope 'a'... i and intercept b i Let i = 1,2,…,I, calculate the midpoint of the vehicle trajectory in time. , To round down; Step 4.4, Determine L m If the data set is empty, it is determined that there is no valid hole monitoring vehicle signal and the processing ends; otherwise, proceed to step 5.

6. The distributed acoustic sensing method for monitoring road cavities based on vehicle-induced vibration signals according to claim 5, characterized in that, In step 5, the following method is used to extract the correlation features of the distributed acoustic sensing channels using vehicle trajectory parameters to determine whether there are gaps. Specifically, the steps include: Step 5.1: Based on the midpoint L of the vehicle trajectory time. m Obtain the vehicle signal interval T = { [l + l] for correlation feature extraction. s , l + l e ]| l ∈L m And l + l e < L}; Step 5.2: Calculate the inter-channel correlation coefficient for each segment of vehicle signal data. : ； Among them, k = 0, 1, … , K-1, i = 1, 2, …, I, The offset channel is extracted based on the correlation features calculated in step 3.1, with the vehicle signal time range T. i ∈ T, i = 1, 2, …, I, and For the kth and the Channel 1 in vehicle signal time range T i The average signal amplitude within the range is calculated as follows: ； in, This represents all channels k that satisfy l ∈ T. i of Summation, Indicates the first All channels satisfy l ∈ T i of Perform summation; Step 5.3: Calculate the inter-channel correlation coefficients for each segment of vehicle signals. The average correlation coefficient between channels is obtained by averaging. The calculation is as follows: ； Where k = 0, 1, … , K-1; Step 5.4: Based on the cavity monitoring correlation coefficient threshold C th To satisfy all Channel k is determined to be a void channel and stored in the void channel set V, that is: V = { k | }, where k = 0, 1, … , K-1.

Images