A method for InSAR tropospheric delay correction considering the random characteristics of the atmosphere

By constructing an atmospheric negative index time series model and 3D-ZTD network, the problem of ignoring spatiotemporal variability in the GAMs method is solved, and the precise estimation and correction of tropospheric delay is achieved, and the deformation solution accuracy of timing InSAR is improved.

CN118584431BActive Publication Date: 2024-11-22KUNMING UNIV OF SCI & TECH
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410597906.X
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-05-14
Publication Date
2024-11-22
Estimated Expiration
2044-05-14

AI Technical Summary

Technical Problem

The existing global atmospheric model (GAMs) correction method ignores atmospheric spatiotemporal variability, resulting in limited spatial and temporal accuracy, and the spatial and temporal randomness of tropospheric delay in timing InSAR cannot be accurately corrected.

Method used

Using a method that takes into account the random characteristics of the atmosphere, a negative atmospheric time series model was constructed through ERA-5 reanalysis data from the European Medium-term Numerical Forecast Center, combined with the Prophet time series prediction model fitting parameters, a 3D-ZTD network was established, and nuclear density estimation and The principle of minimum variance performs accurate estimation of tropospheric delay.

Benefits of technology

The vertical stratification and turbulent delay estimation accuracy of troposphere delay are improved, and the near-real-time estimation and accurate correction of troposphere delay is achieved, the phase standard deviation of the interference map is significantly improved, and the accuracy of timing InSAR deformation solution is enhanced.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118584431B_ABST
    Figure CN118584431B_ABST
Patent Text Reader

Abstract

The present invention discloses an InSAR tropospheric delay correction method taking into account the atmospheric random model, which is used to solve the tropospheric delay estimation bias problem caused by the atmospheric spatiotemporal variability. Firstly, an atmospheric negative exponential time series model is constructed based on the latest ERA‑5 reanalysis data released by the European Centre for Medium-Range Weather Forecasts ECMWF. The purpose is to use the Prophet time series prediction model to correct the negative exponential model parameters β and ZTD. r Fitting is performed in the time dimension to achieve near-real-time estimation of the total zenith delay ZTD of the grid points, rather than using a subjective linear interpolation scheme in the time dimension. The InSAR atmospheric delay correction method disclosed in the present invention can take into account the temporal and spatial variability of the atmosphere, rather than using a subjective inverse distance method, local spline function or other standard interpolation schemes. In addition, the InSAR atmospheric delay correction method disclosed in the present invention does not require prior deformation information of the surface, nor is it affected by other errors in the interferometric phase, and is applicable to PS, SBAS and other time-series InSAR methods.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to the technical field of synthetic aperture radar interferometry (InSAR), and in particular to an InSAR tropospheric delay correction method taking into account the random characteristics of the atmosphere. Background Art

[0002] Surface deformation monitoring is an important method for identifying geological hazards. Conventional methods include GNSS, Lidar, leveling and inclinometers, but due to the limitations of temporal and spatial resolution, manpower, material resources and financial resources, it is difficult to achieve large-scale and long-term surface deformation monitoring. Spaceborne synthetic aperture radar interferometry (InSAR) is widely used in various deformation monitoring due to its high resolution, all-day and all-weather characteristics. With the accumulation of SAR data and the development of time series InSAR (TS-InSAR) technology, it has the ability to obtain the long-term and spatial evolution characteristics of large-scale surface deformation, thus becoming an indispensable means for the study of the formation and deformation laws of disasters such as landslides, earthquakes, volcanoes and ground subsidence. However, TS-InSAR technology also has its inherent limitations and is easily affected by vegetation, atmospheric signals or terrain effects, resulting in unsatisfactory deformation analysis results. Among them, atmospheric delay is one of the main error sources of TS-InSAR surface deformation monitoring.

[0003] In order to overcome the influence of atmospheric delay, statistical or empirical methods such as interferogram stacking, filtering, and power law are currently used, as well as methods using external observation data such as GPS, GPS / MODIS, MODIS / MERIS, CRA40, ERA-5, and GACOS to correct tropospheric delay, and good results have been achieved. The empirical method takes into account the spatial variability of tropospheric delay, but this type of method often focuses on modeling vertical stratified delays related to terrain, and cannot accurately correct turbulent turbulence delays with stronger temporal and spatial randomness. Statistical methods overcome or weaken the influence of tropospheric delay based on the statistical characteristics of tropospheric delay in the time and space domain. Another solution is to treat it as an observation error in the TS-InSAR deformation estimation algorithm. However, most of these methods assume that the variance of atmospheric delay in the time series is constant, which does not meet the seasonal variation characteristics of atmospheric delay.

[0004] Recent studies have shown that global atmospheric models (GAMs) have the potential to mitigate TS-InSAR tropospheric delay. However, existing GAMs-based correction methods often ignore the modeling of atmospheric spatiotemporal variability, resulting in GAMs methods being limited in spatial and temporal accuracy. Summary of the invention

[0005] The present invention aims to solve the problem that the existing GAMs-based correction methods often ignore the modeling of atmospheric spatiotemporal variability, resulting in the GAMs method being limited by spatial and temporal accuracy. The present invention provides a time-series InSAR tropospheric delay correction method that takes into account the atmospheric random model, takes into account the 3D spatiotemporal variability characteristics of the tropospheric delay, and improves the accuracy of vertical stratification and turbulence delay estimation. This method is based on the temporal and spatial dimensional variation characteristics of atmospheric delay. First, based on the latest ERA-5 reanalysis data released by the European Centre for Medium-Range Numerical Weather Forecasts ECMWF, an atmospheric negative exponential time series model is constructed. The purpose is to use the Prophet time series prediction model to predict the negative exponential model parameters β and ZTD. r Fitting is performed in the time dimension to achieve near-real-time estimation of the total zenith delay (ZTD) of the grid points, rather than using a subjective linear interpolation scheme in the time dimension. Then, the elevation sampling step is adaptively segmented based on the kernel density estimation method according to the SAR satellite acquisition time and the three-dimensional spatial information provided by the InSAR coherent points, and the tropospheric delay samples on different altitude planes are calculated to form a 3D-ZTD network. Secondly, a random model of tropospheric delay is established based on the minimum variance principle to reconstruct the ZTD random field at the PS point, rather than directly using bilinear interpolation, inverse distance weighting, local average simple Kriging interpolation or other standard interpolation schemes. Finally, the delay is converted from the zenith direction to the radar line of sight direction according to the radar local incident angle parameters and pseudorange conversion coefficients, achieving accurate estimation of the tropospheric delay at the InSAR pixel level and assisting the time series InSAR in solving deformation displacement.

[0006] To achieve the above object, the present invention provides the following technical solution: an InSAR tropospheric delay correction method taking into account the random characteristics of the atmosphere, the method comprising the following steps:

[0007] S1. Data preprocessing: Collect the original RADARSAT-2 satellite data under N-scene repeated orbits in the study area, achieve sub-pixel-level precise matching of the same target point on the main and auxiliary images according to the intensity cross-correlation criterion, and obtain N-scene SAR data with the same spatial reference; obtain the ERA-5 meteorological reanalysis data released by the European Centre for Medium-Range Numerical Forecasts, including air pressure, temperature, and water vapor pressure; obtain 1 arc second SRTM DEM data;

[0008] S2. Interference phase calculation: Conjugate calculation is performed on the registered primary and secondary SAR complex images to obtain N(N-1) / 2 corresponding interference graphs. Then, the terrain phase is subtracted using DEM data to obtain N(N-1) / 2 differential interference phase graphs. Finally, the interference graphs with the same main image are grouped, and the phase connection algorithm is used to estimate the optimal phase of each common main image interference graph. Then, the time baseline 100 days and the space baseline 200m threshold are used to screen the qualified differential interference graphs.

[0009] S3, high coherence point phase unwrapping: mask low coherence points with coherence less than 0.5, and use polynomial fitting orbital plane trend for high coherence points to deduct orbit errors in differential interferograms. Finally, use Delaunay triangulation to phase unwrap high coherence points with coherence greater than 0.5 in differential interferograms to obtain unwrapped phase values ​​of high coherence points.

[0010] S4. Calculation of layered ZTD: Based on the atmospheric parameters of air pressure, temperature and relative humidity of each layer provided by ERA-5 meteorological reanalysis data, the layered zenith dry delay ZHD, layered wet delay ZWD and layered zenith total delay ZTD of all grid points in the study area with a sampling interval of 1 hour are calculated by the integration method;

[0011] S5. Construction of spatiotemporal vertical profile model: based on the total zenith delay (ZTD) r The vertical profile model is the starting point. A negative exponential function model is constructed for each grid point and the corresponding elevation value. The elevation decreasing factor β in the model is fitted using the least squares algorithm. Then, the Prophet time series prediction model is used to predict β and ZTD in the time dimension with a step size of 1 hour. r Fitting is performed to establish the Prophet_ZTD-NEF vertical profile model, which enables near-real-time estimation of the ZTD of grid points at different heights at any time;

[0012] S6, stratified ZTD sample selection criteria: Establish stratified ZTD sample selection criteria for the study area. Adaptively divide the elevation intervals according to the interquartile range (IQR) of the elevation data in the study area, the number of high coherence target points, and the elevation range of high coherence target points, and use the Prophet_ZTD-NEF vertical profile model established in S5 to predict the grid point tropospheric delay samples on the plane of different height intervals to form a 3D-ZTD network;

[0013] S7. Construction of random model of tropospheric delay at any high coherence point: For each high coherence point, calculate the three-dimensional spatial distance between the nearby ZTD samples and the test point, and use the ZTD samples within the distance threshold to establish a random model of the ZTD of the test point based on the minimum linear unbiased estimation principle to predict the ZTD of each high coherence point. Finally, convert the ZTD into tropospheric delay phase according to the local incident angle of each high coherence point, the RADARSAT-2 data acquisition time and the pseudorange conversion coefficient;

[0014] S8. Model performance evaluation: The performance of the model of the method of the present invention is comprehensively evaluated through simulation experiments and real experiments.

[0015] Preferably, in step S4, ZHD, ZWD and ZTD are calculated according to the following formula:

[0016]

[0017]

[0018] ZTD=ZHD+ZWD

[0019] Where, T represents Kelvin temperature; k1 = 0.776KPa -1 , k'2=0.233KPa -1 , k3=3.75×10 3 K 2 Pa -1 ; P represents the extracted stratified air pressure at each grid point, in hPa; e represents the water vapor pressure, in hPa.

[0020] Preferably, in step S5, the spatiotemporal vertical profile model can be specifically expressed as:

[0021] S51: Set the reference point ZTD r Constrained on the surface, a negative exponential function is established to describe the variation of stratified atmospheric delay with elevation at any grid point, and the fitting β is calculated based on the weighted least squares algorithm with a step size of one hour;

[0022]

[0023] Where: h r Indicates the elevation of the ground reference point; ZTD r Represents the surface ZTD.

[0024] S52: Use the Prophet time series forecasting model to predict ZTD r and β in the time dimension to establish the time series Prophet_ZTD r Model and Prophet_β model to achieve near real-time estimation of grid point ZTD;

[0025]

[0026] Where: ZTD r (t) represents the time series Prophet_ZTD r Model; ZTD r Saturated growth model, where C represents carrying capacity, k represents growth rate, and m represents offset; ZTD r Periodic function, where a n and b n represents the model coefficient; is a normally distributed error term.

[0027] Similarly: β(t) = g β (t)+s β (t)+ε β

[0028] Where: β(t) represents the time series Prophet_β model; g β (t) represents the β-saturated growth model; s β (t) represents the β periodic function; ε β is a normally distributed error term.

[0029] S53: Combine S51 and S52 to obtain the final Prophet_ZTD-NEF vertical profile model.

[0030]

[0031] Preferably, in step S6, the stratified ZTD sample selection criteria are specifically expressed as:

[0032] S61: Calculate the minimum interval h and the number of initial elevation interval groups bins according to the interquartile range IQR of the elevation data of the study area, the number of high-coherence target points n, and the elevation range of high-coherence target points [Min, Max];

[0033] IQR=Q3-Q1

[0034]

[0035]

[0036] In the formula, Q1 represents the lower quartile value of the elevation data; Q3 represents the upper quartile value of the elevation data; IQR represents the interquartile range; n represents the number of highly coherent target points; Max represents the maximum range value of the elevation; and Min represents the minimum range value of the elevation.

[0037] S62: Establishing the elevation probability density function, setting 1 / 2 of the maximum kernel density value as the interval threshold, and adaptively dividing different group intervals with the minimum interval unit h as the step size;

[0038] S63: Based on the calculated group intervals, the Prophet_ZTD-NEF vertical profile model is used to calculate the corresponding ZTD samples under each group interval.

[0039] Preferably, in step S7, the random model of tropospheric delay at any high coherence point is specifically:

[0040] Step S71: Input SAR time into Prophet_ZTD r The model and Prophet_β model calculate the best ZTD among the m grid ZTD models rand β, and then calculate the k sample ZTD values ​​according to the sample point elevation and ZTD grid space-time model;

[0041] Step S72: Calculate the three-dimensional spatial distance between the high coherence point to be tested and k ZTD samples, and select nearby ZTD samples that meet the distance threshold. The three-dimensional distance calculation formula is expressed as:

[0042]

[0043] Where Dist i Represents the distance samples between all ZTD samples and the point to be measured, (x o ,y o ,z o ) represents the three-dimensional coordinates of the point to be measured, (x i ,y i ,z i ) represents the three-dimensional coordinates of the ZTD sample, where i = 1, 2, ... k

[0044] Step S73: Calculate the semivariance based on the nearby ZTD samples and spatial coordinates, establish a variogram model, and then calculate the semivariance r between the high coherence point to be tested and the nearby ZTD according to the distance between the high coherence point to be tested and the nearby ZTD. io , expressed in matrix form as

[0045]

[0046] Where φ represents the Lagrange constant, w i represents the weight of the i-th ZTD sample, r ij represents the semivariance of the two target points i and j.

[0047] Step S74: Solve the equations in S73 according to the minimum variance principle to obtain the optimal coefficient w i ,Then weight the ZTD samples according to the optimal coefficient to obtain the ZTD of the high coherence point to be measured;

[0048] Step S75: repeating steps S71 to S74, performing pixel-by-pixel calculation for each high coherence point, and obtaining the ZTD of each high coherence point in the interference pattern;

[0049] Step S76: According to the satellite radar incident angle and the pseudorange conversion coefficient, the zenith ZTD is converted into the atmospheric delay phase in the radar line of sight direction for further time-series InSAR deformation solution.

[0050] In the above steps, the conversion formula is:

[0051]

[0052] Where θ represents the radar incident angle, Represents the pseudorange conversion coefficient.

[0053] Preferably, in step S8,

[0054] The accuracy of the simulation experiment is verified by using ERA-5 and DEM to simulate the temporal and spatial distribution of tropospheric delay, including qualitative and quantitative analysis of the correlation between the ZTD calculated by the model and DEM; verifying the model parameter ZTD r And verify the accuracy of the ZTD at the ERA-5 stratification height output by the model.

[0055] The real experiment uses the three-dimensional coordinate information provided by ERA-5 and time-series InSAR to simulate the spatial and temporal distribution of tropospheric delay at the time of SAR acquisition, including qualitative and quantitative evaluation of the reduction amplitude of the corrected interferometric phase standard deviation STD and the decorrelation performance of phase and elevation.

[0056] The present invention can effectively correct the vertical layering delay related to terrain undulation and the turbulence delay with stronger temporal and spatial random changes, and improve the accuracy of time-series InSAR deformation solution. The method is independent of the InSAR data itself and is applicable to time-series InSAR methods such as PS / SBAS or DS-InSAR.

[0057] Compared with the prior art, the present invention has the following beneficial effects:

[0058] The present invention utilizes the convenience of free global access to ERA-5 meteorological reanalysis data, and the constructed Prophet_ZTD-NEF atmospheric delay model can realize near-real-time estimation of tropospheric delay of ERA-5 grid points, and further constructs a tropospheric delay random field through tropospheric delay samples on planes at different altitudes (3D), so as to realize accurate estimation of tropospheric delay. The phase of the corrected interferogram has good decorrelation performance with the terrain, and the phase standard deviation of the interferogram is significantly improved. In addition, the method of the present invention is independent of the SAR data set and can be integrated into conventional time-series InSAR methods such as PS-InSAR, SBAS-InSAR and DS-InSAR. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 It is a technical flow diagram of the present invention;

[0060] Figure 2 It is a schematic diagram of the tropospheric delay of the main image predicted by other methods and the method of the present invention on May 28, 2020;

[0061] Figure 3 It is a schematic diagram of the tropospheric delay of the auxiliary image predicted by other methods and the method of the present invention on April 8, 2020;

[0062] Figure 4is a correlation diagram of tropospheric delays predicted by other methods and the method of the present invention;

[0063] Figure 5 Schematic diagram of the correlation between the ZTD values ​​of different layers predicted by the method of the present invention and the test values;

[0064] Figure 6 It is a schematic diagram of the density scatter of the interference phase value and elevation before and after correction by the four methods;

[0065] Figure 7 This is a schematic diagram of the relationship between the change in phase standard deviation after correction by the four methods and the local correlation coefficient between the interference phase and the topography. DETAILED DESCRIPTION

[0066] 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.

[0067] like Figure 1 As shown, an InSAR tropospheric delay correction method taking into account the random characteristics of the atmosphere comprises the following steps:

[0068] Step S1: Data preprocessing: Collect the original RADARSAT-2 satellite data under N-scene repeated orbits in the study area, achieve sub-pixel-level precise matching of the same target point on the primary and secondary images according to the intensity cross-correlation criterion, and obtain N-scene SAR data with the same spatial reference; obtain the ERA-5 meteorological reanalysis data released by the European Centre for Medium-Range Forecasts, including air pressure, temperature, and water vapor pressure; obtain 1 arc second SRTM DEM data;

[0069] Step S2: Interference phase calculation: Conjugate calculation is performed on the registered primary and secondary SAR complex images to obtain N(N-1) / 2 corresponding interference graphs. Then, the terrain phase is subtracted using DEM data to obtain N(N-1) / 2 differential interference phase graphs. Finally, the interference graphs with the same main image are grouped, and the phase connection algorithm is used to estimate the optimal phase of each common main image interference graph. Then, the time baseline 100 days and the space baseline 200m threshold are used to screen the qualified differential interference graphs.

[0070] Step S3: Phase unwrapping of high coherence points: mask low coherence points with coherence less than 0.5, and use polynomial fitting of orbital plane trend for high coherence points to deduct orbital errors in the differential interferogram. Finally, use Delaunay triangulation to phase unwrap high coherence points with coherence greater than 0.5 in the differential interferogram to obtain the unwrapped phase value of the high coherence point.

[0071] Step S4: stratified ZTD calculation: According to the atmospheric parameters of air pressure, temperature and relative humidity of each layer provided by ERA-5 meteorological reanalysis data, the stratified zenith dry delay ZHD, stratified wet delay ZWD and stratified zenith total delay ZTD of all grid points in the study area with a sampling interval of 1 hour are calculated by using the integration method;

[0072] Specifically, ZHD, ZWD and ZTD are calculated according to the following formula:

[0073]

[0074]

[0075] ZTD=ZHD+ZWD

[0076] Where, T represents Kelvin temperature; k1 = 0.776KPa -1 , k'2=0.233KPa -1 , k3=3.75×10 3 K 2 Pa -1 ; P represents the extracted stratified air pressure at each grid point, in hPa; e represents the water vapor pressure, in hPa.

[0077] Step S5: Construction of spatiotemporal vertical profile model: using the total zenith delay ZTD r The vertical profile model is the starting point. A negative exponential function model is constructed for each grid point and the corresponding elevation value. The elevation decreasing factor β in the model is fitted using the least squares algorithm. Then, the Prophet time series prediction model is used to predict β and ZTD in the time dimension with a step size of 1 hour. r Fitting is performed to establish the Prophet_ZTD-NEF vertical profile model, which enables near-real-time estimation of the ZTD of grid points at different heights at any time;

[0078] Specifically, step S5 includes the following steps:

[0079] Step S51: Set the reference point ZTD r Constrained on the surface, a negative exponential function is established to describe the variation of stratified atmospheric delay with elevation at any grid point, and the fitting β is calculated based on the weighted least squares algorithm with a step size of one hour;

[0080]

[0081] Where: h r Indicates the elevation of the ground reference point; ZTD r Represents the surface ZTD.

[0082] Step S52: Use the Prophet time series prediction model to predict ZTD r and β in the time dimension to establish the time series Prophet_ZTD r Model and Prophet_β model to achieve near real-time estimation of grid point ZTD

[0083]

[0084] Where: ZTD r (t) represents the time series Prophet_ZTD r Model; ZTD r Saturated growth model, where C represents carrying capacity, k represents growth rate, and m represents offset; ZTD r Periodic function, where a n and b n represents the model coefficient; is a normally distributed error term.

[0085] Similarly: β(t) = g β (t)+s β (t)+ε β

[0086] Where: β(t) represents the time series Prophet_β model; g β (t) represents the β-saturated growth model; s β (t) represents the β periodic function; ε β is a normally distributed error term.

[0087] Step S53: Combine S51 and S52 to obtain the final Prophet_ZTD-NEF vertical profile model.

[0088] Step S6: Layered ZTD sample selection criteria: Establish the layered ZTD sample selection criteria for the study area. According to the interquartile range (IQR) of the elevation data in the study area, the number of high coherence target points, and the elevation range of high coherence target points, the elevation intervals are adaptively divided, and the Prophet_ZTD-NEF vertical profile model established in S5 is used to predict the grid point tropospheric delay samples on the planes of different height intervals to form a 3D-ZTD network;

[0089] Specifically, step S6 includes the following steps:

[0090] S61: Calculate the minimum interval h and the number of initial elevation interval groups bins according to the interquartile range IQR of the elevation data in the study area, the number of highly coherent target points n, and the elevation range of highly coherent target points [Min, Max].

[0091] IQR=Q3-Q1

[0092]

[0093]

[0094] In the formula, Q1 represents the lower quartile value of the elevation data; Q3 represents the upper quartile value of the elevation data; IQR represents the interquartile range; n represents the number of highly coherent target points; Max represents the maximum range value of the elevation; Min represents the minimum range value of the elevation.

[0095] S62: Establishing the elevation probability density function, setting 1 / 2 of the maximum kernel density value as the interval threshold, and adaptively dividing different group intervals with the minimum interval unit h as the step size;

[0096] S63: Based on the calculated group intervals, the Prophet_ZTD-NEF vertical profile model is used to calculate the corresponding ZTD samples under each group interval.

[0097] Step S7: Construction of random model of tropospheric delay at any high coherence point: For each high coherence point, the three-dimensional spatial distance between the nearby ZTD samples and the test point is calculated, and the ZTD random model of the test point is established based on the minimum linear unbiased estimation principle for the ZTD samples within the distance threshold to predict the ZTD of each high coherence point. Finally, the ZTD is converted into the tropospheric delay phase according to the local incident angle of each high coherence point, the RADARSAT-2 data acquisition time and the pseudorange conversion coefficient.

[0098] Specifically, step S7 includes the following steps:

[0099] Step S71: Input SAR time into Prophet_ZTD r The model and Prophet_β model calculate the best ZTD among the m grid ZTD models r and β, and then calculate the k sample ZTD values ​​according to the sample point elevation and ZTD grid space-time model;

[0100] Step S72: Calculate the three-dimensional spatial distance between the high coherence point to be tested and k ZTD samples, and select nearby ZTD samples that meet the distance threshold. The three-dimensional distance calculation formula is expressed as:

[0101]

[0102] Where Dist iRepresents the distance samples between all ZTD samples and the point to be measured, (x o ,y o ,z o ) represents the three-dimensional coordinates of the point to be measured, (x i ,y i ,z i ) represents the three-dimensional coordinates of the ZTD sample, where i = 1, 2, …k.

[0103] Step S73: Calculate the semivariance based on the nearby ZTD samples and spatial coordinates, establish a variogram model, and then calculate the semivariance r between the high coherence point to be tested and the nearby ZTD according to the distance between the high coherence point to be tested and the nearby ZTD. io , expressed in matrix form as

[0104]

[0105] Where φ represents the Lagrange constant, w i represents the weight of the i-th ZTD sample, r ij represents the semivariance of the two target points i and j.

[0106] Step S74: Solve the equations in S73 according to the minimum variance principle to obtain the optimal coefficient w i , and then weight the ZTD samples according to the optimal coefficient to obtain the ZTD of the high coherence point to be measured

[0107] Step S75: repeating steps S71 to S74, performing pixel-by-pixel calculation for each high coherence point, and obtaining the ZTD of each high coherence point in the interference pattern;

[0108] Step S76: according to the satellite radar incident angle and pseudo-range conversion coefficient, the zenith ZTD is converted into the atmospheric delay phase in the radar line of sight direction for further time series InSAR deformation solution;

[0109] In the above steps, the conversion formula is:

[0110]

[0111] Where θ represents the radar incident angle, Represents the pseudorange conversion coefficient.

[0112] Step S8: Model performance evaluation: comprehensively evaluate the performance of the model of the method of the present invention through simulation experiments and real experiments.

[0113] Specifically, the simulation experiment uses DEM and ERA-5 to simulate the temporal and spatial distribution of tropospheric delay, including qualitative and quantitative analysis of the correlation between the tropospheric delay calculated by the model and the DEM elevation; verify the model parameter ZTD rThe actual experiment uses the three-dimensional coordinate information provided by ERA-5 and time-series InSAR to simulate the spatial and temporal distribution of tropospheric delay at the time of SAR acquisition, including qualitative and quantitative evaluation of the reduction of the corrected interferometric phase standard deviation STD and the decorrelation performance of phase and elevation.

[0114] We used the hourly ERA-5 meteorological reanalysis data and DEM modeling from 2019 to 2021, and predicted the spatial and temporal distribution of tropospheric delay in Lijiang area on May 20, 2020 and April 8, 2020, as shown in Figure 2. Figure 2 and Figure 3 shown. Figure 2 (ac) are the tropospheric delays predicted by PZTD-NEF, GACOS and the method of the present invention on May 20, 2020 in Lijiang area, respectively. Figure 2 (d) is the DEM covering the Lijiang area. Figure 3 (ac) are the tropospheric delays predicted by PZTD-NEF, GACOS and the method of the present invention on April 8, 2020 in Lijiang area, respectively. Figure 3 (d) is the DEM covering the Lijiang area. Figure 2 and Figure 3 It can be clearly seen that the PZTD-NEF and GACOS methods have overfitting in area A where the terrain changes strongly, and in area B where the elevation is relatively low, there is an obvious underfitting phenomenon. On the contrary, the tropospheric delay predicted by the method of the present invention is consistent with the terrain changes, without obvious overfitting and underfitting phenomena.

[0115] To further verify the performance of the method of the present invention, we also randomly selected 10,000 samples and cross-validated them with the GACOS and PZTD-NEF models. Figure 4 The correlation diagram of the tropospheric delay predicted by the method of the present invention and the PZTD-NEF and GACOS methods is shown. From the figure, we can see that on April 8, 2020, the tropospheric delay predicted by the method of the present invention has a high correlation with other methods, and the correlation coefficients of the three different sample sets are 0.9890 and 0.9906, respectively. On May 20, 2020, the correlation coefficients of the tropospheric delay predicted by the method of the present invention and other methods were 0.9977 and 0.9995, respectively, but the other two methods were underestimated. The main reason is that GACOS and PZTD-NEF do not take into account the horizontal non-stationarity of the troposphere when predicting the tropospheric delay of unknown points. Therefore, it is difficult to capture the turbulent delay with more randomness, resulting in a significant underestimation of the tropospheric delay. We also selected ERA-5 stratified ZTD as verification data, and made predictions using PZTD-NEF and the method of the present invention, respectively. We also drew the correlation diagram and calculated the root mean square error, as shown Figure 5 As shown in the figure, we can find that the RMSE of the actual data calculated by the method of the present invention is only 2.24 cm, which is much lower than the 5.73 cm calculated by PZTD-NEF, which shows that the method of the present invention can also estimate the ZTD at any time more accurately.

[0116] Further using the three-dimensional coordinate information provided by ERA-5 and time-series InSAR, we simulated the temporal and spatial distribution of tropospheric delay at the time of 18 SAR acquisitions. We first used the three mainstream atmospheric delay methods based on GAMs (PyAPS, GACOS and PZTD-NEF) for comparative analysis. We counted the corrections of all interferograms and found that the average phase standard deviation of the original interferograms after correction by the method of the present invention was reduced the most, improving 79.2% of the interferograms, which was better than 57.6% (PyAPS), 68% (GACOS) and 63.2% (PZTD-NEF) of the three atmospheric delay correction methods. This also shows that the method of the present invention can well alleviate the influence of tropospheric delay in the interferometric phase.

[0117] Table 1125 Phase standard deviation changes before and after correction of interferograms

[0118]

[0119]

[0120] Figure 6 The figure is a schematic diagram of the density scatter of the interference phase value and the elevation before and after correction by the three methods in Lijiang area. A larger coefficient value indicates that the atmospheric vertical stratification delay is significant in the original interference pattern. Obviously, there is an obvious terrain-related phase in the original interference pattern, which is mainly caused by the vertical stratification delay. After correction by PyAPS, GACOS, PZTD-NEF and the method of the present invention, the absolute correlation coefficient between the interference phase and the elevation dropped from 3.97rad / km to 0.99rad / km (PyAPS), 0.82rad / km (GACOS), 0.92rad / km (PZTD-NEF) and 0.5rad / km (the method of the present invention). We also counted the change amplitude of the standard deviation of the phase of all interference patterns and the distribution diagram of the absolute correlation coefficient of the interference phase and the elevation, as shown in the figure. Figure 7 As shown. Figure 5 It can be seen that all three methods show that there is a statistical correlation between the atmospheric vertical stratification delay and the reduction amplitude of the phase standard deviation. The effect of the method of the present invention is more obvious.

[0121] In summary, the method of the present invention can simply and efficiently correct the tropospheric delay effect in spaceborne InSAR deformation monitoring.

[0122] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants 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.

[0123] 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

1. An InSAR tropospheric delay correction method taking into account the random characteristics of the atmosphere, characterized by: The method comprises the following steps: S1. Data preprocessing: Collect the original RADARSAT-2 satellite data under N-scene repeated orbits in the study area, achieve sub-pixel-level precise matching of the same target point on the main and auxiliary images according to the intensity cross-correlation criterion, and obtain N-scene SAR data with the same spatial reference; obtain the ERA-5 meteorological reanalysis data released by the European Centre for Medium-Range Numerical Forecasts, including air pressure, temperature, and water vapor pressure; obtain 1 arc second SRTM DEM data; S2. Interferometric phase calculation: conjugate calculation is performed on the registered primary and secondary SAR complex images to obtain N(N-1) / 2 corresponding interferograms. Then, the terrain phase is subtracted using DEM data to obtain N(N-1) / 2 differential interferometric phase maps. Finally, the interferograms with a common primary image are grouped, and the phase connection algorithm is used to estimate the optimal phase of each common primary image interferogram. Then, the time baseline 100 days and the space baseline 200m threshold are used to screen the qualified differential interferograms. S3, high coherence point phase unwrapping: mask low coherence points with coherence less than 0.5, and use polynomial fitting orbital plane trend for high coherence points to deduct orbit errors in differential interferograms. Finally, use Delaunay triangulation to phase unwrap high coherence points with coherence greater than 0.5 in differential interferograms to obtain unwrapped phase values ​​of high coherence points. S4. Calculation of layered ZTD: Based on the atmospheric parameters of air pressure, temperature and relative humidity of each layer provided by ERA-5 meteorological reanalysis data, the layered zenith dry delay ZHD, layered wet delay ZWD and layered zenith total delay ZTD of all grid points in the study area with a sampling interval of 1 hour are calculated by the integration method; S5. Construction of spatiotemporal vertical profile model: based on the total zenith delay (ZTD) r The vertical profile model is the starting point. A negative exponential function model is constructed for each grid point and the corresponding elevation value. The elevation decreasing factor β in the model is fitted using the least squares algorithm. Then, the Prophet time series prediction model is used to predict β and ZTD in the time dimension with a step size of 1 hour. r Fitting is performed to establish the Prophet_ZTD-NEF vertical profile model, which enables near-real-time estimation of the ZTD of grid points at different heights at any time; S6, stratified ZTD sample selection criteria: Establish stratified ZTD sample selection criteria for the study area, adaptively divide the elevation interval according to the interquartile range (IQR) of the elevation data in the study area, the number of high coherence target points, and the elevation range of high coherence target points, and use the Prophet_ZTD-NEF vertical profile model established in S5 to predict the grid point tropospheric delay samples on the plane of different height intervals to form a 3D-ZTD network; S7. Construction of random model of tropospheric delay at any high coherence point: For each high coherence point, the three-dimensional spatial distance between the nearby ZTD samples and the test point is calculated, and the ZTD random model of the test point is established based on the minimum linear unbiased estimation principle for the ZTD samples within the distance threshold to predict the ZTD of each high coherence point. Finally, the ZTD is converted into the tropospheric delay phase according to the local incident angle of each high coherence point, the RADARSAT-2 data acquisition time and the pseudorange conversion coefficient; S8. Model performance evaluation: The performance of the model of this method is comprehensively evaluated through simulation experiments and real experiments.

2. The InSAR tropospheric delay correction method taking into account the random characteristics of the atmosphere according to claim 1, characterized in that: In step S4, ZHD, ZWD and ZTD are calculated according to the following formula: ZTD=ZHD+ZWD Where, T represents Kelvin temperature; k1 = 0.776KPa -1 , k′2=0.233KPa -1 , k3=3.75×10 3 K 2 Pa -1 ; P represents the extracted stratified air pressure at each grid point, in hPa; e represents the water vapor pressure, in hPa.

3. The InSAR tropospheric delay correction method taking into account the random characteristics of the atmosphere according to claim 1, characterized in that: The specific method in step S5 is: S51: Set the reference point ZTD r Constrained on the surface, a negative exponential function is established to describe the variation of stratified atmospheric delay with elevation at any grid point, and the fitting β is calculated based on the weighted least squares algorithm with a step size of one hour; Where: h r Indicates the elevation of the ground reference point; ZTD r represents the surface ZTD; S52: Use the Prophet time series forecasting model to predict ZTD r and β in the time dimension to establish the time series Prophet_ZTD r Model and Prophet_β model to achieve near real-time estimation of grid point ZTD; Where: ZTD r (t) represents the time series Prophet_ZTD r Model; represents the ZTDr saturated growth model, where C represents the carrying capacity, k represents the growth rate, and m represents the offset; ZTD r Periodic function, where a n and b n represents the model coefficient; is the error term of normal distribution; Similarly: β(t) = g β (t) + s β (t) + ε β Where: β(t) represents the time series Prophet_β model; g β (t) represents the β-saturated growth model; s β (t) represents the β periodic function; ε β is the error term of normal distribution; S53: Combine S51 and S52 to obtain the final Prophet_ZTD-NEF vertical profile model.

4. The InSAR tropospheric delay correction method taking into account the random characteristics of the atmosphere according to claim 1, characterized in that: The specific method in step S6 is: S61: Calculate the minimum interval h and the number of initial elevation interval groups bins according to the interquartile range IQR of the elevation data of the study area, the number of high-coherence target points n, and the elevation range of high-coherence target points [Min, Max]; IQR=Q3-Q1 In the formula, Q1 represents the lower quartile value of the elevation data; Q3 represents the upper quartile value of the elevation data; IQR represents the interquartile range; n represents the number of highly coherent target points; Max represents the maximum range value of the elevation; Min represents the minimum range value of the elevation; S62: Establishing the elevation probability density function, setting 1 / 2 of the maximum kernel density value as the interval threshold, and adaptively dividing different group intervals with the minimum interval unit h as the step size; S63: Based on the calculated group intervals, the Prophet_ZTD-ZEF vertical profile model is used to calculate the corresponding ZTD samples under each group interval.

5. The InSAR tropospheric delay correction method taking into account the random characteristics of the atmosphere according to claim 1, characterized in that: The specific method in step S7 is: Step S71: Input SAR time into Prophet_ZTD r The model and Prophet_β model calculate the best ZTD among the m grid ZTD models r and β, and then calculate the k sample ZTD values ​​according to the sample point elevation and ZTD grid space-time model; Step S72: Calculate the three-dimensional spatial distance between the high coherence point to be tested and k ZTD samples, and select nearby ZTD samples that meet the distance threshold. The three-dimensional distance calculation formula is expressed as: Where Dist i Represents the distance samples between all ZTD samples and the point to be measured, (x o ,y o , z o ) represents the three-dimensional coordinates of the point to be measured, (x i ,y i , z i ) represents the three-dimensional coordinates of the ZTD sample, where i = 1, 2, ... k; Step S73: Calculate the semivariance based on the nearby ZTD samples and spatial coordinates, establish a variogram model, and then calculate the semivariance r between the high coherence point to be tested and the nearby ZTD according to the distance between the high coherence point to be tested and the nearby ZTD. io , expressed in matrix form as Where φ represents the Lagrange constant, w i represents the weight of the i-th ZTD sample, r ij represents the semivariance of the two target points i and j; Step S74: Solve the equations in S73 according to the minimum variance principle to obtain the optimal coefficient w i , and then weight the ZTD samples according to the optimal coefficient to obtain the ZTD of the high coherence point to be measured; Step S75: repeating steps S71 to S74, performing pixel-by-pixel calculation for each high coherence point, and obtaining the ZTD of each high coherence point in the interference pattern; Step S76: According to the satellite radar incident angle and pseudorange conversion coefficient, the zenith ZTD is converted into the atmospheric delay phase in the radar line of sight direction for further time series InSAR deformation solution, wherein the conversion formula is: Where θ represents the radar incident angle, Represents the pseudorange conversion coefficient.

6. The InSAR tropospheric delay correction method taking into account the random characteristics of the atmosphere according to claim 1, characterized in that: The model evaluation in step S8 includes: comprehensive evaluation of the performance of the model through simulation experiments and real experiments; The simulation experiment uses DEM and ERA-5 to simulate the temporal and spatial distribution of tropospheric delay, including qualitative and quantitative analysis of the correlation between the tropospheric delay calculated by the model and the DEM elevation; verify the model parameter ZTD r and verify the accuracy of the ZTD at the ERA-5 stratification height output by the model; The real experiment uses the three-dimensional coordinate information provided by ERA-5 and time-series InSAR to simulate the spatial and temporal distribution of tropospheric delay at the time of SAR acquisition, including qualitative and quantitative evaluation of the reduction amplitude of the corrected interferometric phase standard deviation STD and the decorrelation performance of phase and elevation.