Pavement settlement data fusion method based on multiple physical fields and computing equipment

Abstract

The invention discloses a pavement settlement data fusion method based on multiple physics fields and computing equipment, relates to the technical field of big data, and solves the problems that in the prior art, collected multi-source data are difficult to align in time and space, the dynamic evolution law of pavement settlement and a multi-physics field coupling mechanism cannot be captured, and the data fusion efficiency is high. Therefore, the fusion efficiency of a multi-physical field sensor is low, and the dynamic change of pavement settlement cannot be reflected in real time. The method comprises the following steps: acquiring multi-source heterogeneous data through a multi-source sensor, preprocessing the multi-source heterogeneous data, and performing feature extraction on the preprocessed multi-source heterogeneous data to obtain corresponding feature data; performing primary fusion on the feature data based on a space-time alignment mechanism; performing feature-level fusion on the feature data after the primary fusion based on a feature-level fusion mechanism; according to the method, useful information is extracted and integrated more effectively, and the information content of a data fusion result is improved.

Description

Technical Field

[0001] The present invention belongs to the field of big data and relates to big data fusion technology, specifically a road surface settlement data fusion method and computing equipment based on multi-physical fields. Background Art

[0002] Pavement subsidence is a common problem in the construction and operation of transportation infrastructure, especially in highways, urban roads and railway subgrades. Subsidence may cause unevenness, cracks and even collapse of the road surface, seriously affecting driving safety and road service life. In addition, subsidence problems in urban roads may also damage underground pipelines, affecting the normal operation of urban infrastructure.

[0003] The existing road subsidence monitoring methods mainly rely on leveling and GPS monitoring technology. Due to the sparse temporal and spatial sampling and single physical quantity output of these technologies, it is difficult to align multi-source data in time and space, and it is impossible to capture the dynamic evolution of road subsidence and the multi-physical field coupling mechanism, which makes the multi-physical field sensor fusion efficiency low. There is a technical problem that the dynamic changes of road subsidence cannot be reflected in real time. Summary of the Invention

[0004] The present invention aims to solve at least one of the technical problems existing in the prior art; to this end, the present invention proposes a road surface settlement data fusion method and computing equipment based on multi-physical fields, which is used to solve the technical problem that the multi-source data collected in the prior art are difficult to align in time and space, cannot capture the dynamic evolution law of road surface settlement and the multi-physical field coupling mechanism, thereby making the multi-physical field sensor fusion efficiency low and unable to reflect the dynamic changes of road surface settlement in real time.

[0005] To achieve the above objectives, the first aspect of the present invention provides a road surface settlement data fusion method based on multi-physics fields, comprising:

[0006] Acquire multi-source heterogeneous data through multi-source sensors, pre-process the multi-source heterogeneous data, and extract features from the pre-processed multi-source heterogeneous data to obtain corresponding feature data; wherein the multi-source heterogeneous data includes: micro-motion data, magnetotelluric field data, and self-potential monitoring data; and the corresponding feature data includes: micro-motion feature data, magnetotelluric field feature data, and self-potential feature data;

[0007] The feature data is initially fused based on the spatiotemporal alignment mechanism; and the feature data after the initial fusion is fused at the feature level based on the feature-level fusion mechanism.

[0008] Preferably, the micro-motion feature data includes: time domain features of micro-motion data, frequency domain features of micro-motion data, and spatial features of micro-motion data. The micro-motion feature data is obtained by extracting features from the pre-processed micro-motion data, including:

[0009] The preprocessed micro-motion data is used as input to construct the intrinsic mode function IMF i (t);

[0010] By formula The original signal x(t) in the time domain feature is calculated; where r n (t) is the residual term; n represents the number of components of the intrinsic mode function, t represents the time, and i represents the component of the i-th intrinsic mode function;

[0011] Extract the intrinsic mode function, perform Hilbert transform on the components of each intrinsic mode function, and obtain the instantaneous amplitude A corresponding to the components of the intrinsic mode function i (t);

[0012] By formula Calculate the instantaneous energy signal E of the component of the i-th eigenmode function i (t);

[0013] Divide the original signal into sliding windows of a preset fixed length N, obtain the maximum and minimum values ​​of the original signal; calculate the peak-to-peak value based on the maximum and minimum values ​​of the original signal;

[0014] Calculate and obtain sliding window statistics of the original signal; wherein the sliding window statistics include: window mean, window variance, and window root mean square; record the instantaneous amplitude, component instantaneous energy signal, peak-to-peak value, and sliding window statistics as time domain features of the micro-motion data;

[0015] The calculation to obtain the sliding window statistics of the original signal includes:

[0016] By formula Calculate the window root mean square RMS of the original signal; where k is the initial value of the preset sliding window length;

[0017] By formula Calculate the window mean μ of the original signal;

[0018] By formula Calculate the window variance σ of the original signal 2 ;

[0019] By formula The spectrum amplitude X(f) at frequency f is calculated; where x win (t) is the signal value of the original signal after windowing at time t, is the basis function of Fourier transform;

[0020] Within the preset full spectrum range, the formula Calculate the frequency point f with the maximum amplitude peak :

[0021] The spectrum amplitude and the frequency point with the maximum amplitude are recorded as the frequency domain features of the micro-motion data;

[0022] The predicted value and spatial gradient of the preset target point are obtained through Kriging interpolation method and semivariogram model;

[0023] The predicted value and spatial gradient of the preset target point are recorded as the spatial features of the micro-motion data.

[0024] The present invention uses the semivariogram to characterize the spatial autocorrelation of micro-motion signals and quantify spatial heterogeneity; Kriging interpolation uses the spatial covariance model to provide the optimal unbiased prediction value and uncertainty estimation of the target point, which is suitable for data completion of the large-scale multi-sensor monitoring network in the present invention.

[0025] Preferably, the method of obtaining the predicted value and spatial gradient of the preset target point by using the Kriging interpolation method and the semivariogram model includes:

[0026] Call the semivariogram model and obtain point s according to the semivariogram model i and s j The semivariance value γ(s i ,s j ) and point s i and s o The semivariance value γ(s i ,s0);i=1,2,3,…,j,…,n;

[0027] The Kriging equations are constructed as Where μ is the preset Lagrange multiplier, λ j is the sum of all weight coefficients;

[0028] When j is equal to i, the Kriging equations are converted into matrix form and the weight coefficient λ is obtained by linear algebra. i ;

[0029] By formula Calculate the predicted value at the preset target point s0 Where s i is a known point, z(s i ) is a known point s i The observed value at

[0030] By formula Calculate the spatial gradient and Where (x, y) is the point s i 's coordinates.

[0031] The present invention not only relies on the observed values ​​of known points by using the Kriging interpolation method, but also combines spatial location information. By constructing a set of Kriging equations and solving weight coefficients, spatial structure information is incorporated into the prediction process. This method can fully utilize the information of all known points, rather than just the few points closest to the target point, thereby more comprehensively reflecting the overall characteristics of spatial data.

[0032] Preferably, the method for obtaining the semivariogram model includes:

[0033] By formula The experimental semi-variance value γ(h)1 when the lag distance between the point pairs is h is calculated; where N(h) represents the number of point pairs when the distance is h;

[0034] By constructing a spherical model Obtain the theoretical semivariance value γ(h)2; where c0 is random noise, c is the preset structural variance, and α is the maximum correlation distance;

[0035] The spherical model is fitted with the experimental semivariogram value by the least squares method to obtain c0 and α, which are then substituted into the spherical model, marking the completion of the construction of the semivariogram model.

[0036] Preferably, the magnetotelluric field characteristic data includes: time domain characteristics of the magnetotelluric field, frequency domain characteristics of the magnetotelluric field, and spatial characteristics of the magnetotelluric field; and feature extraction is performed on the preprocessed magnetotelluric field data to obtain the magnetotelluric field characteristic data, including:

[0037] The pulse intensity is calculated by the formula = max(y(t)) - sliding mean(y(t)); where y(t) is the time domain data collected by the original signal sensor of the magnetotelluric field;

[0038] The signal trend of the original signal of the magnetotelluric field is fitted by a third-order polynomial;

[0039] The pulse intensity and signal trend of the original signal are used as the time domain characteristics of the magnetotelluric field;

[0040] By dividing the frequency bands and calculating the energy proportion of each frequency band, the energy proportion of each frequency band is used as the frequency domain characteristic of the magnetotelluric field;

[0041] By formula Calculate the eigenvalue Corr of the correlation matrix of the measurement points i，j ; In the formula, Cov(z(s i ), z(s j )) represents the covariance of the observation values ​​of measurement point i and measurement point j, is the standard deviation of the observation value at measuring point i; z(s i ) is the observation value of measuring point i, si and s j The spatial positions of measuring points i and j respectively;

[0042] The eigenvalues ​​of the correlation matrix of the measuring points are used as the spatial characteristics of the magnetotelluric field.

[0043] The present invention simplifies the complexity of data and improves the efficiency of data monitoring by outputting the eigenvalues ​​of the correlation matrix for dimensionality reduction and feature extraction.

[0044] Preferably, dividing the frequency bands and calculating the energy proportion of each frequency band includes:

[0045] Preset frequency band (A, B), preset sampling rate f s and FFT length M, by the formula Calculate the frequency index range (k start , k end );

[0046] The amplitude in each frequency band is recorded as X(k);

[0047] By formula Calculate the energy E of the frequency band (A, B) A-B ; Where c is the number of frequency band intervals.

[0048] Preferably, the spontaneous potential characteristic data includes: time domain characteristics of the spontaneous potential and spatial characteristics of the spontaneous potential; the spontaneous potential characteristic data is obtained by extracting the features of the preprocessed spontaneous potential data, including:

[0049] Retrieve the pre-processed spontaneous potential data, decompose the spontaneous potential data using time series to obtain the actual observation value, and extract the trend slope of the spontaneous potential data; the actual observation and trend slope are used as the time domain characteristics of the spontaneous potential;

[0050] Constructing the groundwater velocity field formula based on Darcy's law The fluid diffusion model interpolation is obtained; where K is the preset permeability coefficient, L is the water head height, and S is the preset water storage coefficient. represents the gradient value;

[0051] The fluid diffusion model interpolation is recorded as the spatial characteristics of the natural potential.

[0052] Preferably, the initial fusion of feature data based on the spatiotemporal alignment mechanism includes:

[0053] Retrieve characteristic data, unify the timestamp of characteristic data through the (GPS / NTP) setting protocol, downsample the high-frequency data (micro-motion) in the characteristic data to the preset frequency, and increase the low-frequency data (natural potential) to the preset frequency (10Hz) through interpolation;

[0054] The interpolation method is: to retrieve the original time point sequence t of the feature data raw , and the observation sequence x of the original signal raw ;

[0055] By formula x interp (t) = CubicSpline(t raw , x raw )(r) Calculate the estimated value x after interpolation interp (t); where CubicSpline is the cubic spline interpolation function, and r is the target interpolation time point;

[0056] Retrieve the spatial features that reach the preset frequency in the feature data, call the coordinate conversion tool to convert the plane coordinates of the spatial features into geographic coordinates, and mark it as the completion of the initial fusion of the feature data.

[0057] The present invention aligns feature data in time and space based on a spatiotemporal alignment mechanism, thereby ensuring spatial alignment of multi-source data and strongly supporting subsequent feature fusion and visualization.

[0058] Preferably, the feature-level fusion mechanism is used to perform feature-level fusion on the feature data after the initial fusion is completed, including:

[0059] Align the time domain, frequency domain, and spatial features corresponding to multi-source heterogeneous data according to the preset time window to form a multi-dimensional input matrix: X = [F 微动 , F 电磁 , F 电位 ]∈R T×D ; Where T is the preset number of time steps, D is the total feature dimension, and R represents the real number set; F 微动 is the micro-motion characteristic data, F 电磁 is the characteristic data of the magnetotelluric field, F 电位 The natural potential feature data is obtained, and each feature dimension is independently Z-Score normalized and marked as feature-level fusion of feature data.

[0060] The present invention performs feature splicing on the basis of a real number set through feature splicing and feature normalization, which is beneficial for making input data continuous and computable.

[0061] To achieve the above-mentioned purpose, the second aspect of the present invention provides a road surface settlement data fusion calculation device based on multiple physical fields, comprising: a memory and a processor, wherein the memory stores executable instructions of the processor; wherein the processor is configured to execute a road surface settlement data fusion method based on multiple physical fields provided in the first aspect by executing the executable instructions.

[0062] Compared with the prior art, the present invention has the following beneficial effects:

[0063] 1. The present invention uses transient electromagnetic method, natural potential method and micro-motion method, all of which are non-destructive detection technologies, to quickly and in real time obtain road surface settlement information and realize dynamic monitoring; the electromagnetic method can detect the internal structure of the roadbed, the location of voids and the water content; the natural potential method can monitor the migration of underground fluids in real time, reflecting the impact of groundwater flow on roadbed stability; the micro-motion method uses natural source surface wave information to quickly detect underground loose soil without the need for artificial seismic sources; by acquiring and combining multi-dimensional information, it can provide more comprehensive information, making up for the shortcomings of a single method.

[0064] 2. The present invention performs an initial fusion of feature data based on a spatiotemporal alignment mechanism, which can align and match data collected by different sensors at different times and spaces, ensuring the spatiotemporal consistency of the collected feature data; based on a feature-level fusion mechanism, the feature data after the initial fusion is further fused, which can more deeply explore the intrinsic connections and correlations between the data; feature-level fusion not only considers the surface features of the data, but also combines the semantic information of the data, which can more effectively extract and integrate useful information, thereby improving the information content and decision-making ability of the fusion results. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0066] Figure 1 Schematic diagram of the data fusion process of the present invention;

[0067] Figure 2 Schematic diagram of the specific steps of data fusion of the present invention;

[0068] Figure 3 A schematic diagram of the steps for obtaining characteristic data of the magnetotelluric field in the present invention; DETAILED DESCRIPTION

[0069] The technical solutions of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the embodiments described are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0070] See also Figure 1-Figure 2 The first embodiment of the present invention provides a road surface settlement data fusion method based on multi-physical fields, comprising:

[0071] Acquire multi-source heterogeneous data through multi-source sensors, pre-process the multi-source heterogeneous data, and extract features from the pre-processed multi-source heterogeneous data to obtain corresponding feature data; wherein the multi-source heterogeneous data includes: micro-motion data, magnetotelluric field data, and self-potential monitoring data; and the corresponding feature data includes: micro-motion feature data, magnetotelluric field feature data, and self-potential feature data;

[0072] The feature data is initially fused based on the spatiotemporal alignment mechanism; and the feature data after the initial fusion is fused at the feature level based on the feature-level fusion mechanism.

[0073] See also Figure 3 , the steps of obtaining the characteristic data of the magnetotelluric field are as follows: pulse intensity = max(y(t)) - sliding mean(y(t)); where y(t) is the time domain data collected by the magnetotelluric field original signal sensor;

[0074] The signal trend of the original signal of the magnetotelluric field is fitted by a third-order polynomial;

[0075] The pulse intensity and signal trend of the original signal are used as the time domain characteristics of the magnetotelluric field;

[0076] Preset frequency band (A, B), preset sampling rate f s and FFT length M, by the formula Calculate the frequency index range (k start , k end );

[0077] The amplitude in each frequency band is recorded as X(k);

[0078] By formula Calculate the energy E of the frequency band (A, B) A-B ; In the formula, c is the number of frequency bands; the energy proportion of each frequency band is used as the frequency domain characteristic of the magnetotelluric field;

[0079] By formula Calculate the eigenvalue Corr of the correlation matrix of the measurement points i，j ; In the formula, Cov(z(s i ), z(s j )) represents the covariance of the observation values ​​of measurement point i and measurement point j, is the standard deviation of the observation value at measuring point i; z(s i ) is the observation value of measuring point i, s i and s j The spatial positions of measuring points i and j respectively;

[0080] The eigenvalues ​​of the correlation matrix of the measuring points are used as the spatial characteristics of the magnetotelluric field.

[0081] For example, the frequency bands are divided into: 0-5Hz, 5-15Hz, 15-30Hz, according to the sampling rate f s And FFT length N, calculate the frequency index range of each frequency band:

[0082] 0-5Hz: k start =0,

[0083] 5-15Hz:

[0084] 15-30Hz:

[0085] The amplitude in each frequency band is recorded as X(k), and the square of the amplitude in each frequency band is summed:

[0086] By formula The energy of the 0-5Hz frequency band is calculated;

[0087] By formula The energy of the 5-15Hz frequency band is calculated;

[0088] By formula The energy of the 15-30Hz frequency band is calculated;

[0089] The total energy is calculated as: E total =E 0-5 +E 5-15 +E 15-30 .

[0090] The second aspect of the present invention provides a road surface settlement data fusion calculation device based on multiple physical fields, including: a memory and a processor, wherein the memory stores executable instructions of the processor; wherein the processor is configured to execute a road surface settlement data fusion method based on multiple physical fields provided in the first aspect by executing the executable instructions.

[0091] Some of the data in the above formula are calculated by removing the dimensions and taking their numerical values. The formula is a formula that is closest to the actual situation obtained by software simulation of a large amount of collected data; the preset parameters and preset thresholds in the formula are set by technical personnel in this field according to actual conditions or obtained through simulation of a large amount of data.

[0092] The working principle of the present invention is as follows: the present invention obtains multi-source heterogeneous data through multi-source sensors, pre-processes the multi-source heterogeneous data, extracts features from the pre-processed multi-source heterogeneous data to obtain corresponding feature data; performs initial fusion of the feature data based on a spatiotemporal alignment mechanism; and performs feature-level fusion of the feature data after the initial fusion based on a feature-level fusion mechanism.

[0093] The above embodiments are only used to illustrate the technical method of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical method of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical method of the present invention.

Claims

1. A road settlement data fusion method based on multi-physics fields, characterized in that: include: Acquire multi-source heterogeneous data through multi-source sensors, pre-process the multi-source heterogeneous data, and extract features from the pre-processed multi-source heterogeneous data to obtain corresponding feature data; wherein the multi-source heterogeneous data includes: micro-motion data, magnetotelluric field data, and self-potential monitoring data; and the corresponding feature data includes: micro-motion feature data, magnetotelluric field feature data, and self-potential feature data; Initial fusion of feature data based on spatiotemporal alignment mechanism; Based on the feature-level fusion mechanism, feature-level fusion is performed on the feature data after the initial fusion is completed.

2. The road surface settlement data fusion method based on multi-physical fields according to claim 1 is characterized in that: The micro-motion feature data includes: time domain features of micro-motion data, frequency domain features of micro-motion data, and spatial features of micro-motion data. The micro-motion feature data is obtained by extracting features from the pre-processed micro-motion data, including: The preprocessed micro-motion data is used as input to construct the intrinsic mode function IMF i (t); By formula The original signal x(t) in the time domain feature is calculated; where r n (t) is the residual term; n represents the number of components of the intrinsic mode function, t represents the time, and i represents the component of the i-th intrinsic mode function; Extract the intrinsic mode function, perform Hilbert transform on the components of each intrinsic mode function, and obtain the instantaneous amplitude A corresponding to the components of the intrinsic mode function i (t); By formula Calculate the instantaneous energy signal E of the component of the i-th eigenmode function i (t); Divide the original signal into sliding windows of a preset fixed length N, obtain the maximum and minimum values ​​of the original signal; calculate the peak-to-peak value based on the maximum and minimum values ​​of the original signal; Calculate and obtain sliding window statistics of the original signal; wherein the sliding window statistics include: window mean, window variance, and window root mean square; record the instantaneous amplitude, component instantaneous energy signal, peak-to-peak value, and sliding window statistics as time domain features of the micro-motion data; By formula The spectrum amplitude X(f) at frequency f is calculated; where x win (t) is the signal value of the original signal after windowing at time t, is the basis function of Fourier transform; Within the preset full spectrum range, the formula Calculate the frequency point f with the maximum amplitude peak : The spectrum amplitude and the frequency point with the maximum amplitude are recorded as the frequency domain features of the micro-motion data; The predicted value and spatial gradient of the preset target point are obtained through Kriging interpolation method and semivariogram model; The predicted value and spatial gradient of the preset target point are recorded as the spatial features of the micro-motion data.

3. The road surface settlement data fusion method based on multi-physical fields according to claim 2 is characterized in that: The method of obtaining the predicted value and spatial gradient of the preset target point by using the Kriging interpolation method and the semivariogram model includes: Call the semivariogram model and obtain point s according to the semivariogram model i and s j The semivariance value γ(s i ,s j ) and point s i and s o The semivariance value γ(s i ,s0);i=1,2,3,…,j,…,n; The Kriging equations are constructed as Where μ is the preset Lagrange multiplier, λ j is the sum of all weight coefficients; When j is equal to i, the Kriging equations are converted into matrix form and the weight coefficient λ is obtained by linear algebra. i ; By formula Calculate the predicted value at the preset target point s0 Where s i is a known point, z(s i ) is a known point s i The observation value at By formula Calculate the spatial gradient and Where (x, y) is point s i 's coordinates.

4. The road surface settlement data fusion method based on multi-physical fields according to claim 3 is characterized in that: The method for obtaining the semivariogram model comprises: By formula The experimental semi-variance value γ(h)1 when the lag distance between the point pairs is h is calculated; where N(h) represents the number of point pairs when the distance is h; By constructing a spherical model Obtain the theoretical semivariance value γ(h)2; where c0 is random noise, c is the preset structural variance, and α is the maximum correlation distance; The spherical model is fitted with the experimental semivariogram value by the least squares method to obtain c0 and α, which are then substituted into the spherical model, marking the completion of the construction of the semivariogram model.

5. The road surface settlement data fusion method based on multi-physical fields according to claim 1 is characterized in that: The magnetotelluric field characteristic data includes: time domain characteristics of the magnetotelluric field, frequency domain characteristics of the magnetotelluric field, and spatial characteristics of the magnetotelluric field; the magnetotelluric field characteristic data obtained by feature extraction of the preprocessed magnetotelluric field data includes: The pulse intensity is calculated by the formula = max(y(t)) - sliding mean(y(t)); where y(t) is the time domain data collected by the original signal sensor of the magnetotelluric field; The signal trend of the original signal of the magnetotelluric field is fitted by a third-order polynomial; The pulse intensity and signal trend of the original signal are used as the time domain characteristics of the magnetotelluric field; By dividing the frequency bands and calculating the energy proportion of each frequency band, the energy proportion of each frequency band is used as the frequency domain characteristic of the magnetotelluric field; By formula Calculate the eigenvalue Corr of the correlation matrix of the measurement points i，j Where Cov(z(s i ), z(s j )) represents the covariance of the observation values ​​of measurement point i and measurement point j, is the standard deviation of the observation value at measuring point i; z(s i ) is the observation value of measuring point i, s i and s j The spatial positions of measuring points i and j respectively; The eigenvalues ​​of the correlation matrix of the measuring points are used as the spatial characteristics of the magnetotelluric field.

6. The road surface settlement data fusion method based on multi-physical fields according to claim 5 is characterized in that: The method of dividing the frequency bands and calculating the energy proportion of each frequency band includes: Preset frequency band (A, B), preset sampling rate f s and FFT length M, by the formula Calculate the frequency index range (k start , k end ); The amplitude in each frequency band is recorded as X(k); By formula Calculate the energy E of the frequency band (A, B) A-B ; Where c is the number of frequency band intervals.

7. The road surface settlement data fusion method based on multi-physical fields according to claim 1 is characterized in that: The self-potential characteristic data includes: the time domain characteristics of the self-potential and the spatial characteristics of the self-potential; the self-potential characteristic data is obtained by extracting the characteristics of the pre-processed self-potential data, including: Retrieve the pre-processed spontaneous potential data, decompose the spontaneous potential data using time series to obtain the actual observation value, and extract the trend slope of the spontaneous potential data; the actual observation and trend slope are used as the time domain characteristics of the spontaneous potential; Constructing the groundwater velocity field formula based on Darcy's law The fluid diffusion model interpolation is obtained; where K is the preset permeability coefficient, L is the water head height, and S is the preset water storage coefficient. represents the gradient value; The fluid diffusion model interpolation is recorded as the spatial characteristics of the natural potential.

8. The road surface settlement data fusion method based on multi-physics fields according to claim 1 is characterized in that: The initial fusion of feature data based on the spatiotemporal alignment mechanism includes: Retrieve feature data, unify the timestamps of the feature data by setting a protocol, downsample the high-frequency data in the feature data to a preset frequency, and increase the low-frequency data to the preset frequency through interpolation; The interpolation method is: to retrieve the original time point sequence t of the feature data raw , and the observation sequence x of the original signal raw ; By formula x interp (t) = CubicSpline(t raw , x raw )(r) Calculate the estimated value x after interpolation interp (t); where CubicSpline is the cubic spline interpolation function, and r is the target interpolation time point; Retrieve the spatial features that reach the preset frequency in the feature data, call the coordinate conversion tool to convert the plane coordinates of the spatial features into geographic coordinates, and mark it as the completion of the initial fusion of the feature data.

9. The road surface settlement data fusion method based on multi-physical fields according to claim 1, characterized in that: The feature-level fusion mechanism is used to perform feature-level fusion on the feature data after the initial fusion, including: Align the time domain, frequency domain, and spatial features corresponding to multi-source heterogeneous data according to the preset time window to form a multi-dimensional input matrix: X = [F 微动 , F 电磁 , F 电位 ]∈R T×D ; Where T is the preset number of time steps, D is the total feature dimension, and R represents the real number set; F 微动 is the micro-motion characteristic data, F 电磁 is the characteristic data of the magnetotelluric field, F 电位 The natural potential feature data is obtained, and each feature dimension is independently Z-Score normalized and marked as feature-level fusion of feature data.

10. A road surface settlement data fusion calculation device based on multi-physics fields, characterized in that: include: A memory and a processor, wherein the memory stores executable instructions of the processor; wherein the processor is configured to execute a road surface settlement data fusion method based on multi-physical fields according to any one of claims 1 to 9 by executing the executable instructions.

Images