A cross-correlation method for urban short-time micro-motion detection scenarios

Abstract

The application is suitable for the field of geophysical exploration technology, and provides a cross-correlation method for urban short-time microtremor detection scene, which sequentially comprises original data standardization preprocessing, adaptive windowing processing, unequal weight cross-correlation calculation processing, quality optimization, frequency dispersion curve picking and transverse wave velocity inversion and imaging. In the adaptive windowing processing, Hanning window, Hamming window and Blackman window are matched according to low frequency band, medium frequency band and high frequency band; in the unequal weight cross-correlation calculation processing, the comprehensive weight is calculated based on signal stationarity, signal-to-noise ratio and energy stability for weighted superposition; and the quality optimization automatically iteratively optimizes parameters with the signal-to-noise ratio of the cross-correlation function as an index. The method can suppress spectrum leakage, improve weak surface wave signal extraction capability and realize batch processing without manual intervention, and is suitable for urban underground space high-resolution microtremor detection.

Description

Technical Field

[0001] This invention belongs to the field of geophysical exploration technology, and in particular relates to a cross-correlation method for short-term micro-motion detection scenarios in cities. Background Technology

[0002] With the continuous development and utilization of urban underground space in my country, underground defects such as cavities, loose soil, and water-rich bodies pose significant risks to urban safety. Microseismic exploration techniques from geophysical methods are widely used in urban underground space exploration due to their advantages such as not requiring artificial seismic sources, non-destructive safety, and strong anti-interference capabilities. Among them, linear microseismic arrays have become the preferred array type for engineering exploration in narrow urban spaces due to their flexible deployment, small footprint, and ability to continuously collect data.

[0003] Current methods for cross-correlation calculations using linear micro-motion arrays employ fixed window functions to process and weighted superimpose signals within a time window. While this approach is feasible for linear micro-motion arrays under stable noise fields and long-term observation conditions, it suffers from the following drawbacks in short-term (5-30 minutes) micro-motion detection scenarios in urban environments:

[0004] (1) Loss of effective high-frequency information: Fixed window function cannot take into account both the frequency resolution requirements of low frequency band and the spectrum leakage suppression requirements of high frequency band, resulting in the loss of effective high-frequency signals corresponding to shallow strata in short-time data.

[0005] (2) Equal weight superposition leads to low signal-to-noise ratio: The urban noise distribution is uneven and the signal quality varies greatly in different time windows. Equal weight superposition causes low quality time windows to pollute the overall result, especially making it difficult to extract weak signals at long offsets in linear arrays.

[0006] (3) Unable to offset directional errors: Linear arrays are sensitive to the direction of noise sources. Short-term observations cannot weaken the directional influence through time accumulation, and the inversion results have strong ambiguity.

[0007] (4) Open-loop processing relies on manual parameter tuning: The existing process has no quality control mechanism, and the parameters need to be adjusted manually repeatedly, which is inefficient and inconsistent, and cannot meet the needs of batch processing.

[0008] Therefore, there is an urgent need for a cross-correlation method suitable for short-term micro-motion detection in urban areas and for linear arrays, in order to improve the quality of cross-correlation signals and the ability to extract effective surface wave signals. Summary of the Invention

[0009] The purpose of this invention is to provide a cross-correlation method for short-term micro-motion detection in urban environments, aiming to solve the problems mentioned in the background art.

[0010] The present invention is implemented as follows: a cross-correlation method for short-term micro-motion detection in urban scenarios includes the following steps:

[0011] Step 1: Standardize and preprocess the raw data, including resampling, bandpass filtering, mean removal, trend removal, instrument response removal, time window truncation and outlier removal, and output a single time window signal;

[0012] Step 2: Adaptive windowing processing; Perform fast Fourier transform on each single-time-window signal output from Step 1 to extract the main frequency band features and divide them into low-frequency, mid-frequency, and high-frequency bands according to frequency; Automatically match the corresponding window function and perform time-domain windowing processing according to the frequency band to which each single-time-window signal belongs; Perform time-domain normalization and frequency-domain spectral whitening on the windowed signal;

[0013] Step 3: Unequal weighted cross-correlation calculation and processing; calculate the cross-correlation function for each single-time-window signal output in Step 2; calculate the stationarity weight, signal-to-noise ratio weight, and energy stability weight for each single-time-window signal respectively, and multiply and fuse the three to obtain the comprehensive weight for each single-time-window; perform unequal weighted superposition of the cross-correlation functions of all single-time-windows based on the comprehensive weight to obtain the initial cross-correlation function;

[0014] Step 4: Quality optimization; Calculate the signal-to-noise ratio of the initial cross-correlation function and determine whether it meets the preset signal-to-noise ratio threshold; if it does, output the current cross-correlation function; if it does not, automatically adjust at least one processing parameter in Step 2 and Step 3, and repeat Step 2 to Step 4 until the output cross-correlation function meets the signal-to-noise ratio threshold or reaches the preset iteration termination condition.

[0015] Further technical solutions also include:

[0016] Step 5: Dispersion curve picking and shear wave velocity inversion; The phase velocity dispersion curve is picked up using the frequency domain-wavenumber analysis method, and the shear wave velocity is obtained by inversion using the least squares method;

[0017] Step 6: Generate a two-dimensional shear wave velocity profile based on the shear wave velocity structure.

[0018] The present invention provides a cross-correlation method for short-term micro-motion detection in urban scenarios, the beneficial effects of which are as follows:

[0019] (1) Suppress spectrum leakage and retain high-frequency information: By using the frequency band adaptive matching window function, the frequency resolution and side lobe attenuation capability are balanced, effectively retaining the high-frequency surface wave signal corresponding to the shallow strata in the short-time data and improving the shallow resolution.

[0020] (2) Improve weak signal extraction capability: Construct a three-dimensional weighted evaluation system of stationarity, signal-to-noise ratio and energy stability, and superimpose the cross-correlation results with unequal weights to make the high-quality time window dominate the results and weaken the directional dependence of the linear array.

[0021] (3) Achieve closed-loop automatic optimization: with As a quality indicator, it automatically iterates and adjusts key parameters without manual parameter tuning, significantly shortening convergence time, with high processing efficiency and good result consistency, making it suitable for large-scale batch detection in cities. Attached Figure Description

[0022] Figure 1 A flowchart of a cross-correlation method for short-term micro-motion detection in urban environments provided by an embodiment of the present invention;

[0023] Figure 2 The specific process for adaptive windowing;

[0024] Figure 3 The specific process for calculating and processing unequal weight cross-correlation;

[0025] Figure 4 Specific processes for quality optimization;

[0026] Figure 5 The experimental location scene is shown (where a and b are schematic diagrams of the survey line, and c and d are images of the data collection site).

[0027] Figure 6 The graph shows a comparison of dispersion curves (where a represents the method used in this study and b represents the traditional method).

[0028] Figure 7 This is a two-dimensional shear wave velocity profile using the method described in this paper.

[0029] Figure 8 This is a two-dimensional shear wave velocity profile obtained using traditional methods. Detailed Implementation

[0030] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0031] The specific implementation of the present invention will be described in detail below with reference to specific embodiments.

[0032] like Figures 1-4 As shown, an embodiment of the present invention provides a cross-correlation method for short-term micro-motion detection in urban environments, comprising the following steps:

[0033] Step 1: Standardize and preprocess the raw data;

[0034] ① Resampling: Reduce the original data sampling rate to 200 Hz.

[0035] ② Bandpass filtering: A Butterworth fourth-order zero-phase filter is used, with the effective frequency band set to 1-80 Hz.

[0036] ③ Remove mean and trend: Eliminate DC component and linear trend.

[0037] ④ Eliminate instrument response: Eliminate instrument-specific disturbances.

[0038] ⑤ Time window truncation and anomaly removal: A fixed window length of 2s is used to divide the preprocessed signal into multiple single time window signals; a sliding window (window length 0.1s, step size 0.05s) is used to detect strong interference. If the absolute value of the signal amplitude in the sliding window is greater than 5 times the overall standard deviation of the amplitude of all sampling points in the single time window, it is determined that there is strong interference; if there are no less than 2 strong interference windows in a single time window, the time window is removed; if there is only 1 strong interference window, it is repaired by interpolation and then retained.

[0039] Step 2: Adaptive windowing processing;

[0040] ① Single-time-window time-frequency analysis: Perform Fast Fourier Transform (FFT) on each single-time-window signal output in step 1 to extract the main frequency band features and divide them into three intervals according to frequency: low frequency band, mid frequency band and high frequency band. Among them, low frequency band: frequency < 10 Hz, mid frequency band: 10 Hz ≤ frequency < 40 Hz, high frequency band: frequency ≥ 40 Hz.

[0041] ② Automatic matching of frequency band window functions:

[0042] The Hanning window is used in the low-frequency band, with a narrow main lobe to ensure frequency resolution.

[0043]

[0044] in, For the Hanning window function in the th The values ​​of each sampling point This refers to the index number of the discrete sampling points within a single time window. This refers to the number of sampling points per time window.

[0045] The mid-frequency band uses the Hamming window:

[0046]

[0047] in, For the Hamming window function at the th The values ​​of each sampling point;

[0048] The Blackman window is used in the high-frequency band, which has strong sidelobe attenuation capability.

[0049]

[0050] in, For the Blackman window function in the th The values ​​of each sampling point;

[0051] ③ Time-domain windowing: Based on the optimal window function obtained from the matching, the single-time-window time-domain signal is windowed. The calculation formula is as follows:

[0052]

[0053] in, The time-domain signal after windowing. The original signal, For window functions;

[0054] ④ Time-domain normalization and frequency-domain spectral whitening: Peak value normalization is used to normalize the amplitude to the [-1, 1] interval. The calculation formula is as follows:

[0055]

[0056] in, The peak value normalized within a single time window The time-domain signal value of each sampling point The first time window after windowing processing within a single time window The absolute value of the amplitude of the time-domain signal at each sampling point. This is a function to find the maximum value.

[0057] Then, an FFT is performed to transform the signal to the frequency domain. A sliding window spectral whitening is used (window size 1 Hz, step size 0.5 Hz). The root mean square (RMS) value of all frequency components within each sliding window is calculated. Then, all amplitudes within the window are divided by their RMS values ​​to enhance the coherence of weak signals. Finally, an inverse FFT is performed to transform the signal back to the time domain.

[0058] Step 3: Calculation and processing of unequal weight cross-correlation;

[0059] ① Single-window cross-correlation calculation: For each single-window signal output in step 2, calculate the cross-correlation function, with the time delay range set as follows: That is, [-200, 200] sampling points (corresponding to a time delay of [-1s, 1s].

[0060] ② Multi-dimensional weight calculation for single time window: Construct a three-dimensional quality evaluation system of stationarity, signal-to-noise ratio and stability, and calculate the three weights for each single time window. In view of the characteristic of few effective time windows for short-term data, the weight of signal-to-noise ratio is strengthened.

[0061] stationarity weights The stationarity of a single-time-window signal is tested using the ADF unit root test method from econometrics; higher signal stationarity corresponds to larger weight values. Specifically, a regression equation is first constructed:

[0062]

[0063] in, For the first Within a single time window The first-order difference of the signal at each sampling point For the first Within a single time window The original signal values ​​of each sampling point For the first The regression coefficients of the first-order lag difference term, For the first Within a single time window The first-order difference of the signal at each sampling point For the intercept term, For trend items, The influence of historical conditions on current changes. The lag order is selected according to the AIC criterion, with the maximum lag order set to 10. This is the residual term.

[0064] According to the critical value table of ADF statistic (Table 1 below, where, For the first Regression coefficients in ADF unit root test of a single time window signal of The stationarity of the signal is divided into four levels using statistical measures, and corresponding stationarity weights are assigned. ;

[0065] Table 1 Critical Values ​​of ADF Statistics

[0066]

[0067] Signal-to-noise ratio weighting Calculate the ratio of effective signal power to noise power (i.e., signal-to-noise ratio) of a single time window signal. The higher the signal-to-noise ratio, the larger the weight value, and set the weight amplification factor.

[0068] Effective signal and noise frequency bands are defined as follows: the effective signal frequency band is greater than or equal to 1 Hz and less than or equal to 80 Hz, and the noise frequency band is greater than or equal to 0 Hz and less than 1 Hz (low-frequency drift), and greater than 80 Hz and less than or equal to 100 Hz (high-frequency noise).

[0069] The average power spectral density is calculated as follows:

[0070]

[0071]

[0072] in, For the first The average power spectral density of a single time window signal within the effective signal frequency band; The number of frequency points within the effective signal frequency band; It is the Fast Fourier Transform function; For the first Within a single time window The time-domain signal value of each sampling point; For the first The power spectral density values ​​at corresponding frequency points after a single-time-window time-domain signal undergoes a Fast Fourier Transform, where... Indicates the first Perform a Fast Fourier Transform on the time-domain signal consisting of all sampling points within a single time window; For the first The average power spectral density of a single time-window signal in the noise frequency band. This represents the number of frequency points within the noise band.

[0073] Signal-to-noise ratio calculation:

[0074]

[0075] in, For the first Signal-to-noise ratio of a single time window signal;

[0076] Weighting and Amplification Factors: Since short-term data has a limited effective time window, a signal-to-noise ratio weighting amplification factor is set. The calculation formula is:

[0077]

[0078] in, This represents the maximum signal-to-noise ratio across all valid time windows. This is a mathematical function that takes the minimum value of all elements within the parentheses.

[0079] Energy stability weight : Calculate the amplitude variance of a single time window signal. The smaller the amplitude fluctuation, the higher the signal stability, and the larger the weight value.

[0080] Divide the single-time-window signal into four equal sub-windows and calculate the energy of each sub-window:

[0081]

[0082] in, For the first In the single-time-window signal, the first Energy values ​​for each sub-window; For the first Within a single time window The time-domain signal value of each sampling point; This is the index number of the child window within a single time window, and ;

[0083] Calculate the standard deviation of the energy of the four sub-windows and mean The energy variation coefficient is obtained. ;

[0084] For energy stability weights The smaller the coefficient of variation of energy, the more stable the signal energy, and the weight increases exponentially. The calculation formula is as follows:

[0085]

[0086] in, For the first Energy stability weights for a single time window signal It is a natural exponential function; this function guarantees that when When energy is completely stable, ;when When it increases, Rapid decay.

[0087] ③ Single Time Window Overall Weight: The above three weights are normalized and merged to obtain the overall weight of each single time window. The calculation formula is as follows:

[0088]

[0089] in, For the first The comprehensive weight of each single time window signal;

[0090] ④ Unequal-weighted cross-correlation superposition: Based on the comprehensive weight of each single time window, the cross-correlation results of all single time windows are superimposed with unequal weights to obtain the initial cross-correlation function. The calculation formula is as follows:

[0091]

[0092] in, Let be the initial cross-correlation function. For the first Cross-correlation function for each time window, It is the discrete-time delay of the cross-correlation function. This represents the total number of valid single-time-window signals after anomaly removal. For the first Normalized composite weights for individual time-window signals.

[0093] Depend on The following formula is used to calculate:

[0094]

[0095] in, For the first The combined weight of each single time window signal.

[0096] Step 4: Quality Optimization;

[0097] ① Quality evaluation of cross-correlation results: For the initial cross-correlation function output in step 3, calculate the signal-to-noise ratio of the cross-correlation function (SNR). );

[0098] Effective signal range: The effective surface wave signal range is defined as the area within ±50 sampling points (corresponding to ±0.25s) near the main peak of the cross-correlation function.

[0099] Noise interval: The first and last 50 sampling points of the cross-correlation function constitute the pure noise interval;

[0100] Average power is calculated as follows:

[0101]

[0102]

[0103] in, The average power of the initial cross-correlation function within the effective surface wave signal range. This represents the average power of the initial cross-correlation function over the pure noise region;

[0104] Cross-correlation function signal-to-noise ratio The calculation is as follows:

[0105]

[0106] ② Judgment criteria: Set the threshold for the cross-correlation function signal-to-noise ratio to 6 dB. If dB is acceptable, and the current cross-correlation function meets the threshold requirement, it is considered acceptable and the cross-correlation function is output; otherwise, it enters automatic iterative optimization.

[0107] ③ Automatic Iterative Optimization: If the quality of the cross-correlation function does not meet the preset cross-correlation function signal-to-noise ratio threshold requirement, the parameters (effective frequency band range, window function threshold, weight coefficient, see Table 2 below) are automatically adjusted, and step 2 is returned to be executed again. The iteration continues until the quality meets the standard, and the cross-correlation function is output.

[0108] Table 2 Automatic Iterative Optimization Parameter Adjustment Table

[0109]

[0110] The iteration termination condition is as follows:

[0111] Normal iteration termination condition: After each iteration, calculate the signal-to-noise ratio (SNR) of the cross-correlation function obtained in this iteration; if the SNR is greater than or equal to a preset SNR threshold, the iteration process is terminated immediately, and the cross-correlation function obtained in this iteration is directly output.

[0112] Forced termination condition: The iteration count starts from 1, and the iteration process terminates when the number of completed iterations equals 10. Output the cross-correlation function corresponding to the iteration with the highest signal-to-noise ratio among all 10 completed iterations.

[0113] Step 5: Dispersion curve acquisition and shear wave velocity inversion;

[0114] ① Dispersion curve picking: For the cross-correlation function, the frequency domain-wavenumber analysis (FK analysis) method is used to calculate and pick the phase velocity dispersion curve;

[0115] ② Phase velocity inversion: The shear wave velocity is obtained by calculating the picked dispersion curve using the least squares method.

[0116] Step 6: Imaging;

[0117] A two-dimensional shear wave velocity profile is generated based on the shear wave velocity structure.

[0118] To verify the effectiveness of this method, relevant experiments were conducted on a specific plot of land. Specifically, on-site verification was carried out in front of the Hydraulic Engineering Building on the Chaoyang Campus of Jilin University. The location of the test survey line is as follows: Figure 5 a and Figure 5 As shown in b, the site layout is as follows Figure 5 c and Figure 5 As shown in d. The total length of the test survey line is 20m, with equal spacing and a track spacing of 1m; the data acquisition time is 20 minutes, and the survey line direction is from south to north; the survey line crosses two combined sewer and stormwater pipes and one water supply pipe, and the underground pipeline distribution is complex, which can effectively verify the detection effect of the present invention in complex urban environments.

[0119] Dispersion curves for example Figure 6 a and Figure 6 As shown in b, the results indicate that the proposed method exhibits extremely strong dispersion spectrum energy focusing, resulting in concentrated effective surface wave energy, a clean background, and no stray interference fringes. It retains all effective information, enabling accurate and unbiased dispersion curve picking, thus providing high-quality foundational data for subsequent inversion imaging. In contrast, traditional methods suffer from dispersion spectrum energy divergence, severe background noise interference, numerous stray fringes and spurious energy clusters, and significant masking of the effective surface wave signal by noise. They fail to achieve stable and accurate dispersion curve picking, directly leading to severe deviations in subsequent inversion imaging and complete loss of shallow stratigraphic resolution.

[0120] The two-dimensional shear wave velocity profile generated by this method is as follows: Figure 7 As shown, the shallow strata are continuously layered and the strata boundaries are clearly distinguishable. The low-velocity anomaly (blue area) corresponding to the underground pipeline is accurately focused and has a clear outline, which perfectly matches the known pipeline location. The strata structure is continuous and uniform, without any blurring or abrupt imaging defects.

[0121] The profile generated by traditional methods is as follows Figure 8 As shown, it cannot effectively distinguish between pipeline anomalies and normal strata, and is completely unable to support accurate geological interpretation and anomaly identification.

[0122] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A cross-correlation method for short-term micro-motion detection scenarios in urban areas, characterized in that, Includes the following steps: Step 1: Standardize and preprocess the raw data, including resampling, bandpass filtering, mean removal, trend removal, instrument response removal, time window truncation and outlier removal, and output a single time window signal; Step 2: Adaptive windowing processing; Perform fast Fourier transform on each single-time-window signal output from Step 1 to extract the main frequency band features and divide them into low-frequency, mid-frequency, and high-frequency bands according to frequency; Automatically match the corresponding window function and perform time-domain windowing processing according to the frequency band to which each single-time-window signal belongs; Perform time-domain normalization and frequency-domain spectral whitening on the windowed signal; Step 3: Unequal weighted cross-correlation calculation and processing; calculate the cross-correlation function for each single-time-window signal output in Step 2; calculate the stationarity weight, signal-to-noise ratio weight, and energy stability weight for each single-time-window signal respectively, and multiply and fuse the three to obtain the comprehensive weight for each single-time-window; perform unequal weighted superposition of the cross-correlation functions of all single-time-windows based on the comprehensive weight to obtain the initial cross-correlation function; Step 4: Quality optimization; Calculate the signal-to-noise ratio of the initial cross-correlation function and determine whether it meets the preset signal-to-noise ratio threshold; if it does, output the current cross-correlation function; if it does not, automatically adjust at least one processing parameter in Step 2 and Step 3, and repeat Step 2 to Step 4 until the output cross-correlation function meets the signal-to-noise ratio threshold or reaches the preset iteration termination condition. Step 3 includes the following steps: First, single-time-window cross-correlation calculation: For each single-time-window signal output in step 2, calculate the cross-correlation function, with the time delay range set as follows. That is, [-200, 200] sampling points; Then, multi-dimensional weight calculation for a single time window: construct a three-dimensional quality evaluation system of stationarity, signal-to-noise ratio, and stability, and calculate the three weights for each single time window; stationarity weights The stationarity of a single-time-window signal is tested using the ADF unit root test method from econometrics. First, a regression equation is constructed: in, For the first Within a single time window The first-order difference of the signal at each sampling point For the first Within a single time window The original signal values ​​of each sampling point For the first The regression coefficients of the first-order lag difference term, For the first Within a single time window The first-order difference of the signal at each sampling point For the intercept term, For trend items, The influence of historical conditions on current changes. The lag order is... For residual terms; Based on the ADF statistic critical value table, signal stationarity is divided into four levels, and corresponding stationarity weights are assigned. ; Signal-to-noise ratio weighting Calculate the ratio of effective signal power to noise power of a single time window signal, i.e., signal-to-noise ratio (SNR). The higher the SNR, the larger the weight value, and set the weight amplification factor. Effective signal and noise frequency bands are defined as follows: the effective signal frequency band is greater than or equal to 1 Hz and less than or equal to 80 Hz, and the noise frequency band is greater than or equal to 0 Hz and less than 1 Hz, and greater than 80 Hz and less than or equal to 100 Hz. The average power spectral density is calculated as follows: in, For the first The average power spectral density of a single time-window signal within the effective signal frequency band; The number of frequency points within the effective signal frequency band; It is the Fast Fourier Transform function; For the first Within a single time window The time-domain signal value of each sampling point; For the first The power spectral density values ​​at corresponding frequency points after a single-time-window time-domain signal undergoes a Fast Fourier Transform, where... Indicates the first Perform a Fast Fourier Transform on the time-domain signal consisting of all sampling points within a single time window; For the first The average power spectral density of a single time-window signal in the noise frequency band. This represents the number of frequency points within the noise band. Signal-to-noise ratio calculation: in, For the first Signal-to-noise ratio of a single time window signal; Weighting Calculation and Amplification Factor: Setting the signal-to-noise ratio weighting amplification factor The calculation formula is: in, This represents the maximum signal-to-noise ratio across all valid time windows. This is a mathematical function that takes the minimum value of all elements within the parentheses. Energy stability weight Divide the single-time-window signal into four equal sub-windows and calculate the energy of each sub-window: in, For the first In the single-time-window signal, the first Energy values ​​for each sub-window; For the first Within a single time window The time-domain signal value of each sampling point; This is the index number of the child window within a single time window, and ; Calculate the standard deviation of the energy of the four sub-windows and mean The energy variation coefficient is obtained. ; Energy stability weight The calculation formula is as follows: in, For the first Energy stability weights for a single time window signal It is a natural exponential function; Next, the single-window comprehensive weight is calculated by normalizing and fusing the above three weights to obtain the comprehensive weight for each single window. The calculation formula is as follows: in, For the first The comprehensive weight of each single time window signal; Finally, unequal-weighted cross-correlation superposition: Based on the comprehensive weight of each single time window, the cross-correlation results of all single time windows are superimposed with unequal weights to obtain the initial cross-correlation function, calculated as follows: in, Let be the initial cross-correlation function. For the first Cross-correlation function for each time window, It is the discrete-time delay of the cross-correlation function. This represents the total number of valid single-time-window signals after anomaly removal. For the first Normalized composite weights for individual time-window signals; Depend on The following formula is used to calculate: in, For the first The combined weight of each single time window signal.

2. The cross-correlation method for short-term micro-motion detection scenarios in urban areas according to claim 1, characterized in that, In step 1, the time window truncation and anomaly removal specifically include: taking a fixed window length of 2s, dividing the preprocessed signal into multiple single time window signals; using a sliding window with a length of 0.1s and a step size of 0.05s to detect strong interference; if the absolute value of the signal amplitude within the sliding window is greater than 5 times the overall standard deviation of the amplitude of all sampling points within the single time window, then strong interference is determined to exist; if there are no less than 2 strong interference windows within a single time window, the time window is removed; if there is only 1 strong interference window, it is repaired using interpolation and then retained.

3. The cross-correlation method for short-term micro-motion detection scenarios in urban areas according to claim 1, characterized in that, In step 2, the frequency bands are divided into low-frequency, mid-frequency, and high-frequency bands as follows: frequencies less than 10 Hz are low-frequency bands, frequencies greater than or equal to 10 Hz and less than 40 Hz are mid-frequency bands, and frequencies greater than or equal to 40 Hz are high-frequency bands.

4. The cross-correlation method for short-term micro-motion detection scenarios in urban areas according to claim 3, characterized in that, In step 2, automatically matching the corresponding window function specifically includes: Hanning window is selected for low frequency band: in, For the Hanning window function in the th The values ​​of each sampling point This refers to the index number of the discrete sampling points within a single time window. This refers to the number of sampling points per time window. The mid-frequency band uses the Hamming window: in, For the Hamming window function at the th The values ​​of each sampling point; Blackman window is selected for high-frequency bands: in, For the Blackman window function in the th The values ​​of each sampling point.

5. The cross-correlation method for short-term micro-motion detection scenarios in urban areas according to claim 4, characterized in that, In step 2, the calculation formula for time-domain windowing is as follows: in, The time-domain signal after windowing. The original signal, For window functions.

6. The cross-correlation method for short-term micro-motion detection in urban scenarios according to claim 5, characterized in that, In step 2, the time-domain normalization and frequency-domain spectral whitening of the windowed signal specifically include: Peak value normalization is used to normalize the amplitude to the [-1, 1] interval. The calculation formula is as follows: in, The peak value normalized within a single time window The time-domain signal value of each sampling point The first time window after windowing processing within a single time window The absolute value of the amplitude of the time-domain signal at each sampling point. This is a function to find the maximum value. Then, an FFT transformation is performed to the frequency domain, using a sliding window with a window size of 1 Hz and a step size of 0.5 Hz for spectral whitening. The root mean square value of all frequency components within each sliding window is calculated, and then all amplitudes within the window are divided by their root mean square values. Finally, an inverse FFT transformation is performed to transform the frequency domain back to the time domain.

7. The cross-correlation method for short-term micro-motion detection in urban scenarios according to claim 6, characterized in that, In step 4, the signal-to-noise ratio of the initial cross-correlation function is calculated as follows: The range of ±50 sampling points near the main peak of the cross-correlation function is taken as the effective surface wave signal range, and the range of 50 sampling points at the beginning and end of the cross-correlation function is taken as the pure noise range. The average power in the effective surface wave signal range and the pure noise range is calculated separately using the following formulas: in, The average power of the initial cross-correlation function within the effective surface wave signal range. This represents the average power of the initial cross-correlation function over the pure noise region; Cross-correlation function signal-to-noise ratio The calculation is as follows: 。 8. The cross-correlation method for short-term micro-motion detection scenarios in urban areas according to claim 7, characterized in that, In step 4, the preset signal-to-noise ratio threshold is 6 dB. If dB is acceptable, and the current cross-correlation function meets the threshold requirement, it is deemed acceptable, and the cross-correlation function is output. Otherwise, it will be deemed unqualified; The automatically adjusted processing parameters include the signal-to-noise ratio weighting amplification factor, the high-frequency window function threshold, and the high-frequency cutoff frequency of the bandpass filter; Iteration termination conditions include normal iteration termination and forced termination, where normal iteration termination is: When the threshold is reached, the cross-correlation function is output; forced termination: the maximum number of iterations is 10, and the iteration count starts from 1. When the number of completed iterations equals 10, the iteration process is terminated, and the cross-correlation function corresponding to the iteration with the highest cross-correlation function signal-to-noise ratio among all 10 completed iterations is output.

9. The cross-correlation method for short-term micro-motion detection scenarios in urban areas according to claim 1, characterized in that, Also includes: Step 5: Dispersion curve acquisition and shear wave velocity inversion; The phase velocity dispersion curve was picked up using frequency domain-wavenumber analysis, and the shear wave velocity was obtained by inversion using the least squares method. Step 6: Generate a two-dimensional shear wave velocity profile based on the shear wave velocity structure.

Images