Multi-platform SAR (Synthetic Aperture Radar) data and ground monitoring image fused earth surface deformation identification method

Abstract

The invention discloses a multi-platform SAR (Synthetic Aperture Radar) data and ground monitoring image fused earth surface deformation identification method, and belongs to the technical field of remote sensing monitoring and geological disaster early warning. According to the method, SAR images before and after deformation of a target area are obtained through a spaceborne, airborne or foundation SAR platform, and radar sight direction and azimuth direction deformation fields are extracted through preprocessing and offset tracking or interference processing; synchronously utilizing an unmanned aerial vehicle or ground photography equipment to obtain a high-resolution optical image, and calculating an east-west horizontal deformation field and a south-north horizontal deformation field through an image matching technology; after unifying the multi-source deformation fields to the same spatial resolution, constructing a deformation observation equation set based on a three-dimensional deformation decomposition model by combining radar geometric parameters, and inverting three-dimensional deformation in the vertical, east-west and north-south directions of the earth surface by adopting a weighted fusion algorithm (such as Helmert variance component estimation to determine a weight matrix); and finally, performing time sequence optimization on the deformation field by using Kalman filtering or a stress-strain model to generate a space-time continuous three-dimensional deformation product.

Description

Technical Field

[0001] This invention relates to remote sensing monitoring and geological disaster early warning, and in particular to a method for identifying surface deformation by fusing multi-platform SAR data and ground monitoring images. Background Technology

[0002] Surface deformation monitoring is a core component of environmental geology, engineering safety, and disaster early warning, playing a crucial role in ensuring infrastructure safety and preventing geological disasters. Traditional monitoring methods, such as leveling and GPS positioning, while capable of high-precision single-point measurements, have inherent limitations, including high workload, high cost, low spatial coverage, and difficulty in obtaining continuous spatiotemporal deformation fields. These methods cannot meet the needs for large-scale, all-weather, and high spatiotemporal resolution continuous monitoring, especially when dealing with high-altitude, concealed geological disaster hazards, where their efficiency and reliability are insufficient.

[0003] With the development of remote sensing technology, synthetic aperture radar interferometry (InSAR) technology, especially time-series InSAR technology (such as PS-InSAR and SBAS-InSAR), has become an important tool for monitoring regional surface deformation due to its advantages of wide coverage, high precision, and all-weather imaging. However, this technology still faces significant challenges in practical applications: First, SAR data from a single platform (spaceborne, airborne, or ground-based) can usually only provide one-dimensional deformation information along the radar line of sight, making it difficult to directly decompose and obtain the true three-dimensional deformation field of the surface, which greatly limits its application in accurately analyzing deformation mechanisms; second, InSAR technology is susceptible to interference from atmospheric delay and spatiotemporal decoherence (such as vegetation cover changes and large gradient deformation), affecting the accuracy and reliability of deformation extraction; in addition, single data sources have limitations in terms of spatial resolution and revisit period, making it difficult to simultaneously meet the needs of large-scale general surveys and detailed surveys of key areas.

[0004] To overcome the aforementioned limitations, existing research attempts to combine spaceborne SAR data with UAV photogrammetry (such as digital orthophoto maps, DOM), utilizing high-resolution optical images acquired by UAVs to calculate horizontal displacement, thereby assisting in solving three-dimensional deformation. Other studies explore fusing SAR data from multiple satellite platforms (such as ascending and descending orbits) to invert the three-dimensional deformation velocity field. However, these methods still have room for optimization in terms of multi-source data coordination: for example, data acquired from different platforms and sensors differ in spatiotemporal reference, resolution, and accuracy, making high-precision registration and effective fusion a significant challenge; existing fusion algorithms still have room for improvement in their adaptability to data differences and the accuracy and stability of deformation field reconstruction; simultaneously, how to fully utilize the complementarity of multi-platform data to achieve fully automated and intelligent processing from data preprocessing and fusion calculation to result optimization, thereby improving the integrity and reliability of the three-dimensional deformation field, is a key issue that urgently needs to be addressed in this technological field. Summary of the Invention

[0005] Purpose of the invention: The purpose of this invention is to provide a method for identifying surface deformation by fusing multi-platform SAR data with ground monitoring images. By effectively fusing multi-platform SAR data with ground monitoring images, the ability to identify surface deformation is improved.

[0006] Technical solution: A method for surface deformation identification by fusing multi-platform SAR data with ground monitoring images, including the following steps: S1. Multi-platform SAR data acquisition and preprocessing: Acquire two SAR images of the target area before and after deformation by spaceborne, airborne or ground-based SAR platforms. Preprocess the SAR images to generate single-view complex images. Extract the one-dimensional deformation field and azimuth deformation field of the target area along the radar line of sight based on offset tracking technology or interferometric synthetic aperture radar technology. S2. Ground monitoring image acquisition and registration: High-resolution optical images or digital orthophotos of the target area are acquired using UAVs or ground photography equipment within the same time period. The pixel-level offset between the two images in the east-west and north-south directions is calculated using an image matching algorithm. The pixel-level offset is then converted into actual physical displacement by combining the ground resolution, thereby obtaining the east-west horizontal deformation field and the north-south horizontal deformation field. S3. Data fusion and 3D deformation decomposition: The radar line-of-sight deformation field and azimuth deformation field obtained in step S1 are spatially resampled to a uniform resolution with the horizontal deformation field obtained in step S2. Based on the 3D deformation decomposition model, a set of deformation observation equations is constructed in conjunction with radar geometric parameters. The vertical deformation, east-west deformation and north-south deformation of the target area are inverted through least squares estimation or weighted fusion algorithm. S4. Deformation Result Optimization and Output: The preliminary three-dimensional deformation field is optimized in time using Kalman filtering or stress-strain model to remove noise and outliers, and a spatiotemporally continuous three-dimensional deformation field product of the land surface is generated based on the geographic information system platform. The product includes deformation contour map, deformation time series curve and deformation vector field animation. The multi-platform SAR data mentioned in step S1 includes, but is not limited to, Sentinel-1A, TerraSAR-X, or ground-based SAR system data, and the preprocessing process includes interferometric processing, orbit refinement, and deflating; the image matching algorithm mentioned in step S2 adopts normalized cross-correlation matching, least squares matching, or feature matching methods, and eliminates systematic errors through quadratic surface fitting after registration; the expression of the three-dimensional deformation decomposition model mentioned in step S3 is: Among them, D LOS For radar line-of-sight deformation, D U DE D N The deformations are vertical, east-west, and north-south, respectively, where θ is the radar incident angle and α is the satellite heading angle. In step S4, the optimization process introduces Helmert variance component estimation to determine the weight ratio between SAR and optical observations.

[0007] Furthermore, the offset tracking technique described in step S1 includes the following sub-steps: (a) Perform coarse and fine registration on two phases of SAR single-look complex images, and calculate the maximum cross-correlation value between pixels to determine the offset of radar line of sight and azimuth; (b) Improve the accuracy of offset estimation through spectral filtering and phase optimization, and correct residual terrain errors using ground control points or external DEM data; (c) Convert the offset into a deformation and use an adaptive filtering method to suppress the effects of atmospheric phase and speckle noise.

[0008] Furthermore, the image matching algorithm in step S2 further includes: Control points are selected in non-deformation areas, and system offset caused by differences in drone aerial photography attitude or lens distortion is eliminated by polynomial fitting. SIFT or ORB feature points are extracted using feature matching algorithms, and RANSAC algorithm is used to remove mismatched point pairs. Spatial interpolation is performed on the horizontal deformation field to ensure that the resolution is consistent with that of the SAR deformation field.

[0009] Furthermore, the weighted fusion algorithm described in step S3 is specifically as follows: The variance components are calculated based on the observation accuracy of SAR and optical data. The weight matrices of each data source are determined through Helmert variance estimation. The three-dimensional deformation field is then solved using weighted least squares, with the objective function being: Among them, W SAR and W opt The weighting matrices for SAR and optical data are respectively, L SAR and L op t is the observation vector, A SAR and A opt The design matrix is ​​given by D, which is the three-dimensional deformation vector to be determined.

[0010] Furthermore, the Kalman filtering process described in step S4 includes: Using the three-dimensional deformation field as the state variable, state transition equations and observation equations are constructed; A spatial constraint relationship for the deformation of adjacent pixels is established using a stress-strain model, and the state estimate is dynamically updated using Kalman gain. The temporal deformation data is smoothed to output the deformation time series of each pixel.

[0011] Furthermore, the method also includes a verification step for the deformation results: using GNSS measurement data or leveling measurement data as the true value, the mean absolute error and root mean square error of the three-dimensional deformation field are calculated, and a residual distribution map is generated to evaluate the accuracy; when the residual exceeds the threshold, it is automatically fed back to step S3 to readjust the data weight ratio.

[0012] A system for implementing the above method includes: SAR data processing module: used for registration, interferometric processing, and offset tracking of SAR data from multiple platforms; Optical image analysis module: used for matching and extracting horizontal deformation from UAV or ground optical images; Data fusion and calculation module: used to integrate SAR and optical deformation fields, and invert 3D deformation; Timing optimization module: used for Kalman filtering and stress-strain model optimization; Results output module: used for visualization and accuracy verification of deformed products.

[0013] Furthermore, the system also includes a user interface: users can use this interface to set the SAR data source, optical image resolution, deformation calculation cycle and output format. The system supports batch processing and automatically generates deformation monitoring reports.

[0014] Beneficial effects: (1) This invention effectively overcomes the limitations of a single data source in monitoring dimensions by integrating multi-source data acquired from spaceborne, airborne, and ground-based SAR platforms, as well as UAVs and ground photography equipment. It can not only acquire high-precision radar line-of-sight deformation, but also accurately calculate east-west and north-south horizontal displacements by combining optical images, thereby successfully reconstructing the complete three-dimensional deformation field of the Earth's surface. This multi-source information complementarity mechanism enhances the ability to detect and analyze complex deformation patterns, making the expression of deformation results more comprehensive and realistic.

[0015] (2) This invention introduces a weighted fusion algorithm and a Kalman filter / stress-strain model for cascaded optimization. The weighted fusion strategy can adaptively allocate weights according to the observation accuracy of SAR data and optical data, effectively balancing the accuracy differences between different source data, suppressing the impact of single data errors on the overall results, and improving the robustness of deformation calculation. Subsequent temporal optimization further eliminates noise and outliers, improving the temporal continuity and spatial consistency of the deformation field, and maintaining good monitoring performance, especially in low coherence regions or regions with large deformation gradients.

[0016] (3) This invention constructs a fully automated processing flow from data acquisition, preprocessing, registration, fusion calculation to time series optimization and result verification. This method reduces reliance on manual intervention, lowers operational complexity, and improves the efficiency of large-scale deformation monitoring. Its systematic design enables it to quickly respond to emergency monitoring needs and provides reliable and intuitive three-dimensional deformation products for safety monitoring and early warning of geological disasters in major projects such as bridges, slopes, mining areas, and oil and gas pipelines, with broad application prospects.

[0017] (4) This invention generates diverse products such as spatiotemporally continuous deformation contour maps, time series curves, and vector field animations through a geographic information system platform, making the deformation results more intuitive and easy to understand, and convenient for non-professionals to understand and use. At the same time, the verification process included in this method can evaluate the accuracy of the results and adjust the fusion parameters when necessary, forming a closed-loop processing system, which further ensures the accuracy and reliability of the final deformation monitoring results. Attached Figure Description

[0018] Figure 1 This is a schematic diagram of the method flow of the present invention. Detailed Implementation

[0019] To make the technical solution of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0020] Example This embodiment uses slope settlement monitoring in a mining area as an example to illustrate the specific implementation process of the present invention. Due to long-term mining activities, this mining area faces significant risks of surface deformation, necessitating high-precision three-dimensional deformation monitoring.

[0021] Multi-platform SAR data acquisition and preprocessing First, a coordinated data acquisition process was conducted using spaceborne and ground-based SAR platforms. The spaceborne platform utilized Sentinel-1A satellites (one on each of the ascending and descending orbits) and TerraSAR-X satellites to acquire 30 SAR images of the target area from January to December 2024. The ground-based SAR system employed the IBIS-L ground-based radar, deployed in a stable slope area, continuously acquiring high-frequency (one image per hour) radar data. All SAR images underwent preprocessing: for Sentinel-1A data, SBAS-InSAR technology was used to generate time-series interferograms, extracting the line-of-sight (LOS) deformation field; for TerraSAR-X data, offset tracking technology was used to calculate the maximum cross-correlation value between pixels in two image periods, obtaining the azimuth deformation field. Preprocessing included precise orbit correction, de-leveling, and adaptive filtering to eliminate atmospheric phase and speckle noise, ultimately generating a one-dimensional deformation field and azimuth deformation field with a uniform resolution of 10m × 10m.

[0022] Ground monitoring image acquisition and registration Meanwhile, using a DJI Matrice 300 RTK drone equipped with a Zenmuse P1 photogrammetric camera, high-resolution optical images (5 cm ground resolution) of the mining area were collected monthly. Digital orthophoto maps (DOMs) were generated using Pix4D software, and the coordinate offsets of corresponding pixels on the DOMs from two periods were calculated in the east-west and north-south directions using a normalized cross-correlation matching algorithm. To eliminate systematic errors caused by differences in drone aerial posture, 20 control points were selected in stable areas of the mining area, and the overall offset was corrected using quadratic surface fitting. Subsequently, the pixel offsets were converted into actual physical displacements based on the ground resolution of the DOMs, generating east-west and north-south horizontal deformation fields. The optical deformation fields were then resampled to a 10-meter resolution consistent with the SAR deformation field using bilinear interpolation.

[0023] Data fusion and 3D deformation decomposition The radar line-of-sight deformation field and azimuth deformation field extracted from SAR data are spatially registered with the horizontal deformation field acquired from optical data. Based on a three-dimensional deformation decomposition model, a set of deformation observation equations is constructed using radar geometric parameters (Sentinel-1A incident angle 34°, heading angle -12°; TerraSAR-X incident angle 40°, heading angle 190°). A weighted fusion algorithm is used for inversion: first, the weight matrix of SAR data and optical data is calculated by estimating the Helmert variance components (SAR weight is 0.7, optical weight is 0.3); then, the vertical, east-west, and north-south deformations of the target area are solved using the weighted least squares method. The core model equation is: the radar line-of-sight deformation equals the vertical deformation multiplied by the cosine of the incident angle minus the east-west deformation multiplied by the sine of the incident angle and the cosine of the heading angle, and then minus the north-south deformation multiplied by the sine of the incident angle and the sine of the heading angle.

[0024] Deformation result optimization and output The initially inverted 3D deformation field was temporally optimized. A Kalman filter algorithm was employed, using a stress-strain model as the state transition equation to establish spatial constraints on the deformation of adjacent pixels. The deformation estimate was dynamically updated using Kalman gain. After optimization, a spatiotemporally continuous 3D deformation product was generated using a geographic information system platform, including deformation contour maps (accuracy up to ±3 mm / year), deformation time series curves (monthly), and deformation vector field animation. Simultaneously, a verification step was introduced: 12 GNSS monitoring stations were deployed in the mining area as ground truth, and the mean absolute error (MAE ≤ 1.2 mm) and root mean square error (RMSE ≤ 2.1 mm) of the 3D deformation field were calculated. When the residual exceeded the threshold, the system automatically fed back to the data fusion step to readjust the weight matrix.

[0025] The system enables user interaction. The system in this embodiment includes a SAR data processing module (for SAR data registration and offset tracking), an optical image analysis module (for DOM matching and horizontal deformation extraction), a data fusion and calculation module (for weighted fusion and 3D inversion), a time-series optimization module (for Kalman filtering), and a results output module (for product visualization). Users can set the SAR data source (e.g., selecting Sentinel-1A or TerraSAR-X), optical image resolution (adjustable from 5-20 cm), deformation calculation cycle (monthly or quarterly), and output format (e.g., GeoTIFF or Shapefile) via a web-based interactive interface. The system supports batch processing and automatically generates deformation monitoring reports, including deformation statistics tables and early warning level assessments (e.g., blue warning for deformation rate <5 mm / year, red warning for ≥10 mm / year).

[0026] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention. Therefore, the scope of protection of this patent should be determined by the appended claims.

Claims

1. A method for surface deformation identification by fusing multi-platform SAR data with ground monitoring images, characterized in that, Includes the following steps: S1. Multi-platform SAR data acquisition and preprocessing: Acquire two SAR images of the target area before and after deformation by spaceborne, airborne or ground-based SAR platforms. Preprocess the SAR images to generate single-view complex images. Extract the one-dimensional deformation field and azimuth deformation field of the target area along the radar line of sight based on offset tracking technology or interferometric synthetic aperture radar technology. S2. Ground monitoring image acquisition and registration: High-resolution optical images or digital orthophotos of the target area are acquired using UAVs or ground photography equipment within the same time period. The pixel-level offset between the two images in the east-west and north-south directions is calculated using an image matching algorithm. The pixel-level offset is then converted into actual physical displacement by combining the ground resolution, thereby obtaining the east-west horizontal deformation field and the north-south horizontal deformation field. S3. Data fusion and 3D deformation decomposition: The radar line-of-sight deformation field and azimuth deformation field obtained in step S1 are spatially resampled to a uniform resolution with the horizontal deformation field obtained in step S2. Based on the 3D deformation decomposition model, a set of deformation observation equations is constructed in conjunction with radar geometric parameters. The vertical deformation, east-west deformation and north-south deformation of the target area are inverted through least squares estimation or weighted fusion algorithm. S4. Deformation Result Optimization and Output: The preliminary three-dimensional deformation field is optimized in time using Kalman filtering or stress-strain model to remove noise and outliers, and a spatiotemporally continuous three-dimensional deformation field product of the land surface is generated based on the geographic information system platform. The product includes deformation contour map, deformation time series curve and deformation vector field animation. The multi-platform SAR data mentioned in step S1 includes, but is not limited to, Sentinel-1A, TerraSAR-X, or ground-based SAR system data, and the preprocessing process includes interferometric processing, orbit refinement, and deflating; the image matching algorithm mentioned in step S2 adopts normalized cross-correlation matching, least squares matching, or feature matching methods, and eliminates systematic errors through quadratic surface fitting after registration; the expression of the three-dimensional deformation decomposition model mentioned in step S3 is: Among them, D LOS For radar line-of-sight deformation, D U D E D N The deformations are vertical, east-west, and north-south, respectively, where θ is the radar incident angle and α is the satellite heading angle. In step S4, the optimization process introduces Helmert variance component estimation to determine the weight ratio between SAR and optical observations.

2. The surface deformation identification method based on the fusion of multi-platform SAR data and ground monitoring images according to claim 1, characterized in that, The offset tracking technique described in step S1 includes the following sub-steps: (a) Perform coarse and fine registration on two phases of SAR single-look complex images, and calculate the maximum cross-correlation value between pixels to determine the offset of radar line of sight and azimuth; (b) Improve the accuracy of offset estimation through spectral filtering and phase optimization, and correct residual terrain errors using ground control points or external DEM data; (c) Convert the offset into a deformation and use an adaptive filtering method to suppress the effects of atmospheric phase and speckle noise.

3. The surface deformation identification method based on the fusion of multi-platform SAR data and ground monitoring images according to claim 1, characterized in that, The image matching algorithm in step S2 further includes: Control points are selected in non-deformation areas, and system offset caused by differences in drone aerial photography attitude or lens distortion is eliminated by polynomial fitting. SIFT or ORB feature points are extracted using feature matching algorithms, and RANSAC algorithm is used to remove mismatched point pairs. Spatial interpolation is performed on the horizontal deformation field to ensure that the resolution is consistent with that of the SAR deformation field.

4. The surface deformation identification method based on the fusion of multi-platform SAR data and ground monitoring images according to claim 1, characterized in that, The weighted fusion algorithm described in step S3 is as follows: The variance components are calculated based on the observation accuracy of SAR and optical data. The weight matrices of each data source are determined through Helmert variance estimation. The three-dimensional deformation field is then solved using weighted least squares, with the objective function being: Among them, W SAR and W opt The weighting matrices for SAR and optical data are respectively, L SAR and L op t is the observation vector, A SAR and A opt The design matrix is ​​given by D, which is the three-dimensional deformation vector to be determined.

5. The surface deformation identification method based on the fusion of multi-platform SAR data and ground monitoring images according to claim 1, characterized in that, The Kalman filtering process described in step S4 includes: Using the three-dimensional deformation field as the state variable, state transition equations and observation equations are constructed; A spatial constraint relationship for the deformation of adjacent pixels is established using a stress-strain model, and the state estimate is dynamically updated using Kalman gain. The temporal deformation data is smoothed to output the deformation time series of each pixel.

6. The surface deformation identification method based on the fusion of multi-platform SAR data and ground monitoring images according to claim 1, characterized in that, The method also includes a verification step for the deformation results: using GNSS measurement data or leveling measurement data as the true value, the mean absolute error and root mean square error of the three-dimensional deformation field are calculated, and a residual distribution map is generated to evaluate the accuracy; when the residual exceeds the threshold, it is automatically fed back to step S3 to readjust the data weight ratio.

7. A system for implementing the method according to any one of claims 1 to 6, characterized in that, include: SAR data processing module: used for registration, interferometric processing, and offset tracking of SAR data from multiple platforms; Optical image analysis module: used for matching and extracting horizontal deformation from UAV or ground optical images; Data fusion and calculation module: used to integrate SAR and optical deformation fields, and invert 3D deformation; Timing optimization module: used for Kalman filtering and stress-strain model optimization; Results output module: used for visualization and accuracy verification of deformed products.

8. The system according to claim 7, characterized in that, The system also includes a user interface: users can use this interface to set the SAR data source, optical image resolution, deformation calculation cycle and output format. The system supports batch processing and automatically generates deformation monitoring reports.

Images