CN118778036A - A high-precision landslide deformation monitoring method based on spatiotemporal variation coupling model - Google Patents

Abstract

The invention relates to the technical field of landslide prediction and forecast, and discloses a landslide deformation high-precision monitoring method based on a space-time variation coupling model, which comprises the following steps of deploying GPS positioning equipment to a target monitoring area to obtain ground displacement, inSAR data of a target monitoring area is acquired, a GPS pseudo-range observation equation is constructed, the atmospheric delay error of the InSAR data is corrected by using the GPS data, a displacement matrix of GPS three-dimensional displacement projection in the LOS direction is constructed, and the orbit error of the InSAR data is corrected by using the GPS data. According to the landslide deformation high-precision monitoring method based on the space-time variation coupling model, limitation caused by a single data source is avoided by fusing ground measured data and InSAR data of a GPS, landslide deformation monitoring precision is improved, a result shows that the long time sequence settlement monitoring precision of a target area reaches 7mm, the available GPS measured data is expanded by using a reinforcement learning model, and the ground surface settlement variation of a certain time window in the future is predicted, so that the landslide deformation condition of the target area is restored by using the space-time coupling reinforcement learning model is constructed.

Description

A high-precision monitoring method for landslide deformation based on a spatiotemporal coupling model

Technical Field

The invention relates to the technical field of landslide prediction and forecasting, and in particular to a high-precision landslide deformation monitoring method based on a time-space variation coupling model.

Background Art

Landslide refers to the deformation and destruction phenomenon of slope rock and soil under the action of gravity (or superimposed earthquake force, water pressure, etc.), which is mainly horizontal movement along a certain surface. It includes two main processes: deformation and destruction. The occurrence process is generally fast. At the same time, it will cause great harm to farmland, farmhouses, roads and other fields. Landslides will also trigger other secondary disasters such as mudslides, which will indirectly damage human safety and social operations. Therefore, disasters such as landslides need to be monitored and warned so that relevant departments can make disaster prevention preparations and reduce the loss of manpower and material resources caused by landslides.

GPS and InSAR technologies are often used to monitor landslides. However, in the current study of geological hazards, the research on predicting landslide deformation and extracting landslide deformation areas based on time-series InSAR technology is relatively niche. In addition, the InSAR technology mainly uses remote sensing earth observation data as the basis for experiments to monitor landslide deformation and predict landslide hazards. The earth observation data obtained will be affected by external conditions such as atmospheric conditions and weather conditions. In addition to environmental factors, equipment and observation sites will also have an impact on images. For example, due to the high density of stripes caused by high-rise buildings in high-resolution images, phase unwrapping is difficult. Due to the low resolution and accuracy of existing external DEM data, a large number of DEM phases are residual. Because the commonly used MT-InSAR technology assumes that the deformation at each PS/DS point is linear, and in many cities, deformation does not meet this condition, resulting in the problem of misestimation of deformation information. Although InSAR has achieved good application results in infrastructure deformation monitoring, there are still challenges in the actual application of InSAR. Due to the limitation of the minimum antenna area, traditional SAR sensors cannot meet the requirements of image resolution and width at the same time, so it is difficult to achieve deformation monitoring of ultra-large-scale high-resolution infrastructure, such as the Beijing-Guangzhou high-speed railway line in my country (more than a thousand kilometers long);

However, GPS technology is limited by the ground environment or cost, which results in the spatial resolution of the data monitored by GPS technology being limited. Therefore, for a large research area, the use of GPS technology will consume a lot of financial and material resources, which is far less advantageous than InSAR technology. Ground-measured data can obtain surface information with high precision, including horizontal displacement information and elevation displacement information. Whether integrating ground-measured data with InSAR technology, using ground-measured data to verify InSAR processing results, or coupling InSAR data with ground observation data, using ground-measured data to improve the phase unwrapping algorithm and atmospheric delay error correction of InSAR technology, the accuracy of the results obtained by data processing can be more in line with actual requirements, so that the prediction of geological disasters such as landslides based on the processing results can be more accurate, which is of great significance for geological disaster prevention. Therefore, a high-precision monitoring method for landslide deformation based on a spatiotemporal variation coupling model is proposed to solve the above problems.

Summary of the invention

1. Technical issues to be resolved

In view of the shortcomings of the prior art, the present invention provides a high-precision monitoring method for landslide deformation based on a spatiotemporal variation coupling model, which has the advantage of high precision in landslide deformation monitoring and solves the problem of single data of GPS ground measured data and InSAR data.

(II) Technical solution

In order to achieve the above-mentioned high-precision purpose of landslide deformation monitoring, the present invention provides the following technical solution: a high-precision landslide deformation monitoring method based on a spatiotemporal variation coupling model, comprising the following steps:

S1: Deploy the GPS positioning device to the target monitoring area to obtain the ground displacement and obtain the InSAR data of the target monitoring area;

S2: Construct GPS pseudo-range observation equation and use GPS data to correct the atmospheric delay error of InSAR data;

S3: Construct the displacement matrix of GPS 3D displacement projection in the LOS direction and use GPS data to correct the orbit error of InSAR data;

S4: Invert the landslide deformation using the InSAR data after correcting the atmospheric delay error and orbit error, and use the ground measured data to assist the InSAR data phase unwrapping to obtain the phase change of the target monitoring area;

S5: Use Kriging interpolation method to interpolate the settlement of multiple points in the target monitoring area, build a reinforcement learning model to fuse and correct the interpolation data, and predict the settlement of the target monitoring area at future time points and locations.

Preferably, the pseudorange observation equation in step S2 is:

;

in, is the sum of other deviations and residuals, is the tropospheric delay, and Represents different ionospheric frequencies, is the ionospheric delay, represents the pseudorange from receiver j to the first satellite, represents the pseudo-range from receiver j to the second dangerous satellite. and Represents the different geometric distances between the satellite antenna phase center at the time of signal transmission and the receiver antenna phase center at the time of signal reception in the inertial system.

Preferably, the specific steps of correcting the orbit error of InSAR data in step S3 are:

S3.1: Define the three-dimensional displacement matrix of GPS in the N, E, and U directions , where W is the three-dimensional displacement, is the displacement component of GPS in north-south, east-west and vertical directions;

S3.2: Construct the displacement matrix of GPS three-dimensional displacement projection in the LOS direction;

• ;

in, is the flight direction of the radar, This is the side view of the radar.

Preferably, the specific steps in step S4 are:

S4.1: Add Gaussian white noise to the signal for decomposition. The first modal component of the signal decomposed , the expression is:

;

in, is zero-mean Gaussian white noise, is the noise coefficient, M is the target initial signal, E ₁ represents the mathematical expectation value obtained in the first stage, and I represents the number of repeated decompositions;

S4.2: Calculate the first-order residual , the expression is:

;

S4.3: Continue to decompose the first-order residual , until the decomposition result meets the first IMF condition, redefine the second modal component , the expression is:

;

S4.4: First modal component Extract and get the overall average value. The expression is:

;

in, is the residual, is the signal-to-noise ratio selected when allowing the kth stage, is the zero-mean Gaussian white noise with unit variance in the kth stage, I represents the number of repeated decompositions, K is the total number of IMF modal components, is the mathematical expectation;

S4.5: Until the residual When there are no more than two extreme values of , the expression is:

;

Among them, x(t) is the target signal initially input;

S4.6: Calculate the exact reconstructed target signal (t), the expression is:

;

S4.7: The target signal after accurate reconstruction (t) is used to denoise the final signal by wavelet transform , the wavelet transform expression is:

;

Among them, a is the scale and v is the translation;

S4.8: Calculate the geodetic longitude, expressed as:

;

Calculate the latitude, the expression is:

;

Calculate the geodetic height using the expression:

;

Among them, XYZ is the three-dimensional coordinate value, N is the radius of curvature of the ellipse, is the eccentricity of the ellipsoid;

S4.9: Calculate the interferometric phase after ground deformation , the expression is:

;

in, is the main image slant distance, is the auxiliary image slant distance, is the ground point displacement caused by settlement, is the band length;

S4.10: Calculate the phase change of the target monitoring area using the InSAR data after correcting the atmospheric delay error and orbit error , the expression is:

;

in, is the flat ground phase, is the phase caused by terrain, is the interference phase after ground deformation, is the atmospheric delay phase, is the noise phase, is the integer ambiguity.

Preferably, the specific steps in step S5 are:

S5.1: Use Kriging interpolation to interpolate the settlement of the actual observation points, calculate the attribute similarity between the observation data, and use semivariogram The equivalent substitution expression is:

;

in, represents the attribute value of point i, Represents the attribute value of point j;

S5.2: To find a fitted distance and semivariance The curvilinear relationship (d), calculate the semivariance ;

;

in, and Represents the coordinates of two points in the X direction, and Represents the coordinates of two points in the Y direction, is the shortest distance between two points, represents the covariance between point i and point j, is the variance, (d) is the curve relationship between the fitting distance and semivariance;

S5.3: Solve the equations to obtain the optimal coefficient λ _i , which is expressed as:

;

in, represents the attribute similarity between point i and point j, represents the optimal weight coefficient of each point to be solved, and ρ is the Lagrange multiplier that introduces the unbiased constraint condition;

S5.4: Use the optimal coefficient to perform weighted summation on the attribute values of the known points to obtain the estimated value of the unknown point Z _O , expressed as:

;

in, is a point ), For point i to the target point The weight coefficient of influence, is the value corresponding to point i;

S5.5: Use the interpolation data to train the reinforcement learning model and obtain the predicted value of the target area. The expression is:

;

in, Indicates immediate reward, is the discount factor, For the next state, For the next action, are the parameters of the target network, is the target Q network in state For all possible actions The maximum value among the calculated Q values;

S5.6: Calculate the loss function Loss, the expression is:

;

Among them, s represents the current state, a represents the current action, and D is the experience playback buffer. Indicates immediate reward, are the parameters of the current Q network, Is the network in state and actions The value calculated when .

(III) Beneficial effects

Compared with the prior art, the present invention provides a high-precision landslide deformation monitoring method based on a spatiotemporal variation coupling model, which has the following beneficial effects:

This high-precision landslide deformation monitoring method based on the spatiotemporal variation coupling model avoids the limitations of a single data source by fusing GPS ground-based measured data with InSAR data, thereby improving the accuracy of landslide deformation monitoring. The results show that the long-term settlement monitoring accuracy of the target area reaches 7mm. The reinforcement learning model is used to expand the available GPS measured data and predict the surface settlement changes in a certain time window in the future, thereby constructing a spatiotemporal coupling reinforcement learning model to restore the landslide deformation in the target area.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG1 is a flow chart of a method for high-precision monitoring of landslide deformation based on a spatiotemporal variation coupling model proposed by the present invention;

FIG. 2 is a flow chart of CEEMD-WD processing INSAR data in a high-precision landslide deformation monitoring method based on a spatiotemporal variation coupling model proposed by the present invention.

DETAILED DESCRIPTION

The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

Please refer to Figure 1-2, a high-precision landslide deformation monitoring method based on the spatiotemporal variation coupling model includes the following steps:

S1: Deploy the GPS positioning device to the target monitoring area to obtain the ground displacement and obtain the InSAR data of the target monitoring area;

S2: Construct GPS pseudo-range observation equation and use GPS data to correct the atmospheric delay error of InSAR data;

S3: Construct the displacement matrix of GPS 3D displacement projection in the LOS direction and use GPS data to correct the orbit error of InSAR data;

S4: Invert the landslide deformation using the InSAR data after correcting the atmospheric delay error and orbit error, and use the ground measured data to assist the InSAR data phase unwrapping to obtain the phase change of the target monitoring area;

S5: Use Kriging interpolation method to interpolate the settlement of multiple points in the target monitoring area, build a reinforcement learning model to fuse and correct the interpolation data, and predict the settlement of the target monitoring area at future time points and locations.

InSAR is a remote sensing technology that measures surface deformation by comparing radar images. It uses radar images taken at different times and angles to generate interference patterns, thereby measuring small changes in the surface.

The pseudorange observation equation in step S2 is:

;

in, is the sum of other deviations and residuals, is the tropospheric delay, and Represents different ionospheric frequencies, is the ionospheric delay, represents the pseudorange from receiver j to the first satellite, represents the pseudo-range from receiver j to the second dangerous satellite. and Represents the different geometric distances between the satellite antenna phase center at the time of signal transmission and the receiver antenna phase center at the time of signal reception in the inertial system.

In the acquisition of InSAR data, there is inevitably an atmospheric delay error problem. According to experience, fixed parameters are generally used to correct it, but this method cannot adapt to the complex and changeable measurement environment. Therefore, dual-frequency GPS observation technology is introduced to eliminate the ionospheric delay effect, and then the GAMIT software is used to accurately calculate the tropospheric delay to obtain the total atmospheric delay of China's continuous GPS observation data station, thereby correcting the atmospheric delay error.

The specific steps of correcting the orbit error of InSAR data in step S3 are:

S3.1: Define the three-dimensional displacement matrix of GPS in the N, E, and U directions , where W is the three-dimensional displacement, is the displacement component of GPS in north-south, east-west and vertical directions;

S3.2: Construct the displacement matrix of GPS three-dimensional displacement projection in the LOS direction;

• ;

in, is the flight direction of the radar, This is the side view of the radar.

The specific steps in step S4 are:

S4.1: Add Gaussian white noise to the signal for decomposition. The first modal component of the signal decomposed , the expression is:

;

in, is zero-mean Gaussian white noise, is the noise coefficient, M is the target initial signal, E ₁ represents the mathematical expectation value obtained in the first stage, and I represents the number of repeated decompositions;

S4.2: Calculate the first-order residual , the expression is:

;

S4.3: Continue to decompose the first-order residual , until the decomposition result meets the first IMF condition, redefine the second modal component , the expression is:

;

S4.4: First modal component Extract and get the overall average value. The expression is:

;

in, is the residual, is the signal-to-noise ratio selected when allowing the kth stage, is the zero-mean Gaussian white noise with unit variance in the kth stage, I represents the number of repeated decompositions, K is the total number of IMF modal components, is the mathematical expectation;

S4.5: Until the residual When there are no more than two extreme values of , the expression is:

;

Among them, x(t) is the target signal initially input;

S4.6: Calculate the exact reconstructed target signal (t), the expression is:

;

S4.7: The target signal after accurate reconstruction (t) is used to denoise the final signal by wavelet transform , the wavelet transform expression is:

;

Among them, a is the scale, is the translation amount;

S4.8: Calculate the geodetic longitude, expressed as:

;

Calculate the latitude, the expression is:

;

Calculate the geodetic height using the expression:

;

Among them, XYZ is the three-dimensional coordinate value, N is the radius of curvature of the ellipse, is the eccentricity of the ellipsoid;

S4.9: Calculate the interferometric phase after ground deformation , the expression is:

;

in, is the main image slant distance, is the auxiliary image slant distance, is the ground point displacement caused by settlement, is the band length;

S4.10: Calculate the phase change of the target monitoring area using the InSAR data after correcting the atmospheric delay error and orbit error , the expression is:

;

in, is the flat ground phase, is the phase caused by terrain, is the interference phase after ground deformation, is the atmospheric delay phase, is the noise phase, is the integer ambiguity.

By coupling two different data sources, GPS and InSAR, the error can be significantly reduced and the accuracy of landslide deformation monitoring can be significantly improved.

The specific steps in step S5 are:

S5.1: Use Kriging interpolation to interpolate the settlement of the actual observation points, calculate the attribute similarity between the observation data, and use semivariogram The equivalent substitution expression is:

;

in, represents the attribute value of point i, Represents the attribute value of point j;

S5.2: To find a fitted distance and semivariance The curvilinear relationship (d), calculate the semivariance ;

;

in, and Represents the coordinates of two points in the X direction, and Represents the coordinates of two points in the Y direction, is the shortest distance between two points, represents the covariance between point i and point j, is the variance, (d) is the curve relationship between the fitting distance and semivariance;

S5.3: Solve the equations to obtain the optimal coefficient λ _i , which is expressed as:

;

in, represents the attribute similarity between point i and point j, represents the optimal weight coefficient of each point to be solved, and ρ is the Lagrange multiplier that introduces the unbiased constraint condition;

S5.4: Use the optimal coefficient to perform weighted summation on the attribute values of the known points to obtain the estimated value of the unknown point Z _O , expressed as:

;

in, is a point ), For point i to the target point The weight coefficient of influence, is the value corresponding to point i;

S5.5: Use the interpolation data to train the reinforcement learning model and obtain the predicted value of the target area. The expression is:

;

in, Indicates immediate reward, is the discount factor, For the next state, For the next action, are the parameters of the target network, is the target Q network in state For all possible actions The maximum value among the calculated Q values;

S5.6: Calculate the loss function Loss, the expression is:

;

Among them, s represents the current state, a represents the current action, and D is the experience playback buffer. Indicates immediate reward, are the parameters of the current Q network, Is the network in state and actions The value calculated when .

This high-precision landslide deformation monitoring method based on the spatiotemporal variation coupling model avoids the limitations of a single data source by fusing GPS ground-based measured data with InSAR data, thereby improving the accuracy of landslide deformation monitoring. The results show that the long-term settlement monitoring accuracy of the target area reaches 7mm. The reinforcement learning model is used to expand the available GPS measured data and predict the surface settlement changes in a certain time window in the future, thereby constructing a spatiotemporal coupling reinforcement learning model to restore the landslide deformation in the target area.

It should be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the existence of other identical elements in the process, method, article or device including the element.

Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims (5)

1. A landslide deformation high-precision monitoring method based on a space-time variation coupling model is characterized by comprising the following steps:

s1: deploying GPS positioning equipment to a target monitoring area to acquire ground displacement and InSAR data of the target monitoring area;

s2: constructing a GPS pseudo-range observation equation, and correcting the atmospheric delay error of InSAR data by utilizing GPS data;

s3: constructing a displacement matrix of the GPS three-dimensional displacement projection in the LOS direction, and correcting the orbit error of the InSAR data by utilizing the GPS data;

S4: inverting landslide deformation conditions by utilizing InSAR data after correcting the atmospheric delay error and the orbit error, and assisting InSAR data phase unwrapping by utilizing ground actual measurement data to obtain phase variation of a target monitoring area;

s5: interpolation is carried out on settlement amounts of a plurality of points in the target monitoring area by adopting a Kriging interpolation method, a reinforcement learning model is constructed to carry out fusion correction on the interpolation data, and settlement amounts of future time points and positions of the target monitoring area are predicted.

2. The high-precision landslide deformation monitoring method based on the space-time variation coupling model of claim 1, wherein the pseudo-range observation equation in the step S2 is as follows:

;

Wherein, For the sum of the other bias and residual terms,In order for the tropospheric delay to be sufficient,AndRepresenting the different ionospheric frequencies,In order to provide an ionospheric delay,Representing the pseudorange of receiver j to the first satellite,Representing the j receiver to the second dangerous pseudorange,AndRepresenting different geometric distances between the phase center of the satellite antenna at the time of signal transmission and the phase center of the receiver antenna at the time of signal reception under the inertia system.

3. The landslide deformation high-precision monitoring method based on the space-time variation coupling model according to claim 1, wherein the specific steps of correcting the track error of the InSAR data in the step S3 are as follows:

s3.1: defining a three-dimensional displacement matrix of GPS in N, E, U directions Wherein W is the three-dimensional displacement,The displacement component of the GPS in the north-south, east-west and vertical directions is shown;

s3.2: constructing a displacement matrix of the GPS three-dimensional displacement projection in the LOS direction;

•;

Wherein, In order to be the direction of flight of the radar,Is the side view of the radar.

4. The high-precision landslide deformation monitoring method based on the space-time variation coupling model of claim 1, wherein the specific steps in the step S4 are as follows:

s4.1: adding Gaussian white noise into the signal for decomposition, and decomposing a first modal component of the signal The expression is:

;

Wherein, Is zero-mean gaussian white noise,For the noise coefficient, M is a target initial signal, E ₁ represents a mathematical expected value obtained in the first stage, and I represents the number of repeated decomposition;

s4.2: calculating first order residuals The expression is:

;

S4.3: continuing to decompose first order residual Redefining to obtain a second modal component until the decomposition result meets the first IMF conditionThe expression is:

;

s4.4 performing a first modality component Extracting, and obtaining an overall average value, wherein the expression is as follows:

;

Wherein, As a residual error, the residual error is determined,To allow for a selected signal to noise ratio at the kth stage,For zero-mean gaussian white noise with unit variance in the kth stage, I represents the number of repeated decompositions, K is the total number of IMF modal components,Is a mathematical expectation;

S4.5: until residual error When the extremum of (a) is not more than two, calculating the final residualThe expression is:

;

wherein x (t) is the target signal originally input;

s4.6: calculating the target signal after accurate reconstruction (T) the expression:

;

S4.7: target signal after accurate reconstruction (T) denoising the wavelet transform to obtain a final signalThe wavelet transform expression is:

;

Wherein a is the scale of the device, Is the translation amount;

s4.8: calculating the geodetic longitude, and the expression is:

;

Calculating latitude, wherein the expression is:

;

The calculation is high, and the expression is:

;

wherein XYZ is a three-dimensional coordinate value, N is an elliptical curvature radius, Is the eccentricity of the ellipsoid;

S4.9: calculating interference phase after ground deformation The expression is:

;

Wherein, Is the oblique distance of the main image,In order to assist the image skew distance,In order for the ground point to be displaced by sedimentation,Is the band length;

s4.10: calculating the phase change amount of the target monitoring area by using InSAR data after correcting the atmospheric delay error and the orbit error The expression is:

;

Wherein, In order to achieve a level ground phase position,For the phase caused by the terrain,Is the interference phase after the deformation of the ground,For the phase delay of the atmosphere,For the phase of the noise it is,Is the integer ambiguity.

5. The high-precision landslide deformation monitoring method based on the space-time variation coupling model of claim 1, wherein the specific steps in the step S5 are as follows:

s5.1: interpolation is carried out on settlement of an actual observation point by adopting a Kriging interpolation method, attribute similarity between every two observation data is calculated, and half variance is used Equivalent substitution, the expression is:

;

Wherein, The attribute value representing the point i,An attribute value representing a point j;

s5.2: to find a fitting distance and half variance Is a curve relation of (2)(D) Calculating a half variance ;

;

Wherein, AndRepresenting the coordinates of the two points in the X direction,AndRepresenting the coordinates of the two points in the Y direction,Is the shortest distance between two points,Representing the covariance between points i and j,As a function of the variance of the values,(D) A curve relation between fitting distance and half variance;

S5.3: solving the equation set to obtain an optimal coefficient lambda _i, wherein the expression is as follows:

;

Wherein, Representing the similarity of properties between points i and j,Representing the weight optimal coefficient of each point to be solved, wherein ρ is a Lagrangian multiplier for introducing unbiased constraint conditions;

s5.4: and carrying out weighted summation on the attribute values of the known points by using the optimal coefficients to obtain an estimated value of the unknown point Z _O, wherein the expression is as follows:

;

Wherein, Is a point of%) An estimated value at which the position of the object is estimated,For point i to target pointThe weight coefficient of the influence is determined,The value corresponding to the point i;

s5.5: training the reinforcement learning model by using interpolation data, and obtaining a predicted value of a target area, wherein the expression is as follows:

;

Wherein, Indicating an instant prize is provided,As a discount factor, the number of times the discount is calculated,In order to be in the next state,For the next action to be taken,As a parameter of the target network,Is the target Q network in stateWhen for all possible actionsThe maximum value of the calculated Q values;

s5.6: and calculating a Loss function Loss, wherein the expression is as follows:

;

where s represents the current state, a represents the current action, D is the empirical playback buffer, Indicating an instant prize is provided,Is a parameter of the current Q-network,Is the network in stateAnd actionsA value calculated at that time.