Landslide emergency monitoring system and monitoring method

Abstract

The invention discloses a landslide emergency monitoring system, and the system comprises a ground-based interference synthetic aperture radar device which is used for monitoring the deformation condition of a landslide region in real time; the ground laser scanning equipment is used for acquiring three-dimensional space data of a landslide area; the unmanned aerial vehicle carries a high-resolution camera and is used for collecting high-resolution image information of the landslide area; and the data processing center is used for receiving data collected by the ground-based interferometric synthetic aperture radar, the ground laser scanning and the unmanned aerial vehicle and carrying out data fusion and analysis and early warning information generation. According to the emergency monitoring method, data collected by the ground-based interferometric synthetic aperture radar device, the ground laser scanning device and the unmanned aerial vehicle are transmitted to the data processing center, and multi-source data are matched and integrated through a data fusion method; and analyzing the fused data, evaluating the landslide risk, and generating early warning information. The method is suitable for geographic information monitoring in emergency.

Description

Technical Field

[0001] The present invention relates to the field of landslide monitoring technology, and in particular to a landslide emergency monitoring system and a monitoring method combining ground-based interferometric synthetic aperture radar (GB-InSAR), terrestrial laser scanning (TLS) and unmanned aerial vehicle (UAV). Background Art

[0002] Landslides, as a serious natural disaster, pose a significant threat to human life and property. To effectively prevent and mitigate landslide disasters, real-time monitoring and early warning of potential landslide areas are necessary. Traditional landslide monitoring methods often fail to meet the requirements of real-time performance and accuracy, making it difficult to respond promptly and effectively in emergencies. Summary of the Invention

[0003] Purpose of the invention: The purpose of the present invention is to provide a landslide emergency monitoring system and monitoring method, which can use multi-source remote sensing technology to monitor landslide areas in real time, provide accurate deformation data and risk assessment, monitor potential landslide areas in real time and accurately, and provide relevant departments with necessary monitoring data and early warning information so that emergency measures can be taken in a timely manner.

[0004] Technical solution: A landslide emergency monitoring system, including:

[0005] Ground-based interferometric synthetic aperture radar equipment, used to monitor deformation in the landslide area in real time;

[0006] Terrestrial laser scanning equipment to obtain three-dimensional spatial data of the landslide area;

[0007] Unmanned aerial vehicles equipped with high-resolution cameras to collect high-resolution image information of the landslide area;

[0008] The data processing center is used to receive data collected by ground-based interferometric synthetic aperture radar, ground-based laser scanning and unmanned aerial vehicles, and to perform data fusion, analysis and generation of early warning information.

[0009] Furthermore, the data processing center includes:

[0010] A data fusion module is used to match and fuse data collected by ground-based interferometric synthetic aperture radar equipment, ground-based laser scanning equipment, and unmanned aerial vehicles to generate comprehensive monitoring data for the landslide area;

[0011] Early warning analysis module, used to conduct landslide risk assessment based on comprehensive monitoring data and generate early warning information;

[0012] The visualization display module is used to present monitoring data and warning information to users in an intuitive form.

[0013] A landslide emergency monitoring method, applied to any of the above landslide emergency monitoring systems, comprises the following steps:

[0014] S1, using ground-based interferometric synthetic aperture radar equipment to continuously monitor landslide potential areas and obtain deformation data;

[0015] S2, using ground laser scanning equipment to perform three-dimensional scanning of the landslide area and establish a high-precision digital elevation model;

[0016] S3, using an unmanned aerial vehicle equipped with a high-resolution camera to take aerial photos of the landslide area and obtain high-resolution image information;

[0017] S4, transmits data collected by ground-based interferometric synthetic aperture radar equipment, ground-based laser scanning equipment, and unmanned aerial vehicles to the data processing center, and matches and integrates multi-source data through data fusion methods;

[0018] S5, analyze the fused data, assess the landslide risk, and generate early warning information.

[0019] Furthermore, in step S4, the steps for matching and integrating multi-source data are as follows:

[0020] S41, GB-InSAR image simulation using geometric correction method, including geometric simulation and grayscale simulation: Geometric simulation is to project the known DEM data into the GB-InSAR image coordinate system according to the geometric positioning model, and then generate an image coordinate lookup table in the given geographic coordinate system. The simulated GB-InSAR image coordinates corresponding to the geographic coordinates are stored in the form of complex numbers; for grayscale simulation, the area of ​​the ground scattering element is used as the pixel value of the simulated image;

[0021] S42, performs mismatch correction by geometrically mapping the image into three-dimensional space;

[0022] The distance and azimuth values ​​obtained by TLS are combined with the pixel values ​​of the radar image to form a new image; when the points of the ground laser scanning device are dense, the resampling method is used to remove the redundancy of the ground laser scanning device;

[0023] Then, two-dimensional feature points are selected from the radar image and three-dimensional feature points are selected from the texture point cloud of the UAV to form a feature point set with the same name; these feature point sets are input into the mismatch correction model, and the correction parameters of the deviation are obtained through iteration; when the deviation meets the requirements, the matching is completed.

[0024] Furthermore, in step S41, the original point cloud is used to perform GB-InSAR image simulation, the simulated image is used as a reference image, the real image is used as the image to be corrected, and then reference points for geometric correction are selected in the two images.

[0025] Compared with the prior art, the present invention has the following significant effects:

[0026] The present invention matches and fuses data collected by GB-InSAR, TLS and UAV to generate comprehensive monitoring data for landslide areas, achieving efficient processing of large-scale geographic data. The accuracy of multi-source information matching and fusion is improved through the training and optimization of machine learning models, effectively improving the matching accuracy of SAR images and TLS point cloud data, making it suitable for geographic information monitoring in emergency situations. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 This is the general schematic diagram of the landslide emergency monitoring system;

[0028] Figure 2 This is a flow chart of the landslide emergency monitoring method;

[0029] Figure 3 Flowchart of geometric correction of GB-InSAR images simulated for DEM and SAR images generated based on TLS;

[0030] Figure 4 (a) is a schematic diagram of the feature locations in the amplitude dispersion screen GB-InSAR image, and (b) is the amplitude dispersion image resampled by the geostatistical kriging method;

[0031] Figure 5 Schematic diagram of the geometric mapping process of the present invention. DETAILED DESCRIPTION

[0032] The present invention will be described in further detail below with reference to the accompanying drawings and specific implementations.

[0033] like Figure 1 As shown, a landslide emergency monitoring system includes a ground-based interferometric synthetic aperture radar (GB-InSAR), a terrestrial laser scanning (TLS) device, an unmanned aerial vehicle (UAV) and a data processing center;

[0034] GB-InSAR equipment, used to monitor deformation in the landslide area in real time;

[0035] TLS equipment, used to obtain three-dimensional spatial data of the landslide area;

[0036] The UAV, equipped with a high-resolution camera, was used to collect high-resolution image information of the landslide area;

[0037] The data processing center is used to receive data collected by ground-based interferometric synthetic aperture radar GB-InSAR, ground-based laser scanning TLS and unmanned aerial vehicles (UAVs), and to perform data fusion, analysis and generation of early warning information.

[0038] The data processing center includes:

[0039] The data fusion module is used to match and fuse the data collected by GB-InSAR, TLS and UAV to generate comprehensive monitoring data of the landslide area;

[0040] Early warning analysis module, used to conduct landslide risk assessment based on comprehensive monitoring data and generate early warning information;

[0041] The visualization display module is used to present monitoring data and warning information to users in an intuitive form.

[0042] In this embodiment, a landslide emergency monitoring system is deployed in potential landslide areas. GB-InSAR equipment monitors surface deformation in real time, TLS (terrestrial laser scanning) equipment acquires three-dimensional spatial data of the landslide area, and UAVs (unmanned aerial vehicles) collect high-resolution imagery of the landslide area. All collected data is transmitted to a data processing center. After processing by a data fusion module, the early warning analysis module assesses the landslide risk based on the fused data and generates early warning information. Finally, a visualization module presents the monitoring data and early warning information to the user in a graphical interface, enabling quick decision-making.

[0043] TLS and GB-InSAR are triggered simultaneously by a wireless clock. When TLS performs scanning and mapping tasks, the sensor scans in front of the radar. When the 3D laser scanner is in close proximity, the linear sliding trajectory of the center of the radar "reflector" can be accurately determined. GB-InSAR imaging is based on a distributed scattering target model. The geometric model of the distributed target can be constructed using DEM (digital elevation model) data generated by TLS. The original DEM data is measured at a fixed spatial sampling interval, which is far greater than the resolution of the SAR image. The accuracy of the point cloud obtained by 3D laser scanning meets the accuracy requirements of the geometric estimation.

[0044] In this embodiment, the unmanned aerial vehicle (UAV) is equipped with a three-axis, fixed-focus camera. This smaller-scale UAV utilizes a quadrotor fuselage. It features simple operation, low takeoff and landing requirements, the ability to hover within a certain accuracy range, and high-quality remote sensing imagery. A single flight duration of approximately 20 minutes satisfies the requirements for single-shot landslide image acquisition. For larger landslides, multiple flights can be completed. The flight control system is equipped with a precise position and positioning system (POS). This system includes an inertial measurement unit (IMU) and differential GPS (DGPS), forming an IMU / DGPS system. The POS ensures the UAV automatically follows the planned flight path and provides data for image processing. The aerial imaging sensor is a standard digital or SLR camera with over 18 megapixels, capable of acquiring centimeter-resolution images. A stable platform is added to minimize vibration during imaging. The flight path must be optimized based on image size, flight speed, altitude, and image interval to ensure accurate ground resolution, with image path overlap exceeding 80% and lateral overlap exceeding 60%. The flight altitude for a single landslide must be dynamically adjusted to suit the slope environment, with the drone typically remaining 100-200 meters above the landslide surface.

[0045] like Figure 2 As shown, the implementation process of a landslide emergency monitoring method includes the following steps:

[0046] Step 1: Use ground-based interferometric synthetic aperture radar equipment to continuously monitor the landslide potential area and obtain deformation data;

[0047] Step 2: Use terrestrial laser scanning equipment to perform a three-dimensional scan of the landslide area and establish a high-precision digital elevation model;

[0048] Step 3: Use an unmanned aerial vehicle equipped with a high-resolution camera to take aerial photos of the landslide area to obtain high-resolution image information;

[0049] Step 4: Transmit the data collected by the ground-based interferometric synthetic aperture radar equipment, ground-based laser scanning equipment, and unmanned aerial vehicles to the data processing center, and match and integrate the multi-source data through data fusion methods;

[0050] Step 5: Analyze the fused data, assess the landslide risk, and generate early warning information.

[0051] In step 4, multi-source data are matched and integrated through data fusion methods, including geometric correction assisted by SAR images simulated by DEM and mismatch correction of image geometric mapping to three-dimensional space:

[0052] (1) Geometric correction assisted by SAR images simulated by DEM

[0053] Using only geometric methods, GB-InSAR images cannot be correctly mapped to three-dimensional spatial data, making accurate matching difficult due to various error sources. Therefore, geometric correction is required to improve matching accuracy. Geometric correction methods for spaceborne SAR imagery typically roughly correct the image to eliminate geometric errors caused by terrain and generate geocoded orthophotos. Anomalous radiation effects may also occur due to changes in local incidence angles and range compression caused by terrain.

[0054] The most commonly used geometric correction methods for SAR images include those based on empirical models (such as polynomials), imaging models (such as the range-Doppler method and collinearity equations), and SAR image simulation. Other methods are also derived from the collinearity equations of traditional photogrammetry. Correction is performed based on the geometric characteristics of SAR imaging. GB-InSAR image simulation can be performed using TLS point clouds. In the absence of ground control points, feature points can be obtained from point clouds and GB-InSAR images.

[0055] The flow chart of the geometric correction method simulated using GB-InSAR images is as follows: Figure 3 As shown in the figure, in airborne SAR image processing, local terrain texture is poor, resulting in poor results due to the mismatch between the simulated image and the original GB-InSAR image. TLS can produce a DEM with better resolution than satellite-based image processing, allowing for precise correction of geometric parameters. GB-InSAR image simulation utilizes the TLS-generated SAR image and the incident angle information of the DEM to generate a topologically relevant grayscale image based on a specific scattering model.

[0056] GB-InSAR image simulation involves both geometric and grayscale simulation. Geometric simulation involves projecting known DEM data into the SAR image coordinate system based on a geometric positioning model. A lookup table of image coordinates is then generated in a given geographic coordinate system. The simulated GB-InSAR image coordinates corresponding to the geographic coordinates are stored as complex numbers. For grayscale simulation, the area of ​​the ground scattering element is used as the pixel value of the simulated image.

[0057] Assume that the GB-InSAR platform height is H and the proximal tilt range is S r0 , the sampling interval in the image is dr, and the size of the real image to be corrected is [N a ,N r ], then the tilt range S in the image r for:

[0058] S r =S r0 +[0:N r-1]*dr (1)

[0059] Among them, N a is the length of the crossover range, N r is the length in the range direction.

[0060] The range between each point and the aperture center can be determined by searching the TLS point cloud. The relative azimuth angle θ between each point and the radar aperture center can also be given by searching the TLS point cloud. i . Relative azimuth angle θ i The expression:

[0061]

[0062] Where, is the direction vector of the GB-InSAR antenna trajectory determined by TLS scanning, and the position vector of each point relative to the aperture center; is the direction angle. Using TLS point cloud, the incident angle of the point can be obtained according to the relationship between the point and the radar.

[0063] The magnitude of the reflection coefficient of the incident angle of the point in the simulated image can be calculated using the empirical Kirchhoff model:

[0064]

[0065] Where, σ i Represents the reflection coefficient of the i-th point in the image.

[0066] The difference in reflectance can be directly used as the image grayscale value for simulation.

[0067] The present invention uses a raw point cloud to simulate a SAR image, using the simulated image as the reference image and the real image as the image to be corrected. Reference points for geometric correction are then selected from these two images. Errors exist in the geometric positioning parameters between the real image and the simulated SAR image, and image matching is performed to determine the offset values ​​of corresponding image points. This DEM positioning process establishes a mapping relationship between points in the DEM, the simulated image, and the raw SAR image, from which reference points for calibrating the geometric positioning parameters can be extracted. The process for selecting reference points is as follows:

[0068] First, the simulated SAR image is divided into multiple grids at a fixed interval, and feature extraction operators are used to extract feature points within each grid. Once a valid feature extraction operator is identified, the maximum mutual information method (or another algorithm used for real image acquisition and image matching) is used to calculate the transformation parameters of the subregion corresponding to the center of the region with the same feature point. The simulated image is then matched with the original SAR image using the feature points as reference points.

[0069] It is difficult to directly and accurately determine the feature point pairs within the sub-search region. The image similarity of each sub-region is tested using a correlation metric and a search window. First, similar regions are uniformly identified throughout the entire region to obtain different sub-region maps. The sub-region maps are selected to cover as much of the image as possible. The coordinate transformation of the control points between the simulated GB-InSAR image and the image to be corrected is determined. The simulated GB-InSAR image and the image to be corrected are basically the same in structure and texture features. In principle, their coordinate transformation can be used to determine the linear transformation relationship:

[0070]

[0071] Among them, θ is the rotation angle, a0 and b0 are translation parameters; ρ is the scaling factor.

[0072] (2) Methods for geometrically mapping images to three-dimensional space, mismatch correction

[0073] like Figure 5 The figure shows a flowchart of the geometric mapping process. While point clouds can be used for geometric mapping, significant deviations still exist after geometric mapping. Each pixel in the GB-SAR image is encoded with a "range" and "azimuth" value, and each node in the TLSDEM also has relative range and azimuth values.

[0074] Since the two results don't correspond one-to-one, resampling and bilinear interpolation are used to combine the range and azimuth values ​​obtained from the TLS with the pixel values ​​of the radar image to form a new image. When the TLS points are too dense, resampling is used. GB-InSAR pixels correspond to a large number of points, and resampling is then used to remove redundancy in the TLS. This process results in a preliminary matching mapping table.

[0075] Next, we select 2D feature points from the radar image and 3D feature points from the drone's textured point cloud to form a set of feature points with the same name. These points are then fed into the mismatch correction model, and the deviation correction parameters are iterated to determine the corrected deviation. When the deviation meets the requirements, the match is complete.

[0076] Due to the harsh conditions at the emergency site, visual interpretation was primarily used. The matching error under the final limit condition met the requirement of within three spatial pixels. The equivalent azimuth spatial length was approximately 2*3=6m, and the range length was 0.3*3=0.9m.

[0077] This embodiment corrects these mismatches based on similar feature points in the 2D radar image and the 3D drone color point cloud after geometric mapping in an emergency. Although the synthetic aperture track parameters are very accurate, GB-InSAR images still exhibit discrepancies in planar position, azimuth, and scale. Radar parameters, elevation angle, and beam coverage were determined experimentally. During the monitoring process, radar parameters remain constant, and an image can be selected as the primary image.

[0078] The calculation results cannot directly reflect the distance and azimuth corresponding to the pixel grid of the GB-InSAR image. Therefore, bilinear resampling and interpolation are required. Here, "resampling" means inserting the slope distance and azimuth calculated for each TLS point into the grid as a new point in the 2D GB-InSAR grid. "Interpolation" means assigning a new value to a newly inserted node, which is calculated based on the values ​​of the existing grid nodes using the bilinear interpolation method. These nodes have a mapping relationship with the TLS points. Repeat the above steps on the new GB-InSAR image set. The resulting values ​​are RGB color mapped and visualized as a preliminary mapping in three-dimensional space:

[0079] R=f(Grey)=a1·log(b1·Grey+c1)+d1 (5)

[0080] Here, a1 represents the slope of the color mapping function, b1 represents the scaling factor of the input color value, c1 represents the horizontal shift of the color mapping function, and d1 represents the overall position of the color mapping function. The amplitudes of the three primary colors (red, green, and blue) stored at a point or pixel are typically reflected in this value. Differences in deformation are generally reflected in this value.

[0081] Coarse matching is performed using geometric constraints to extract features from 2D radar images. A rectangular window centered around a feature point is used as the target window. Based on prior knowledge, a larger window is selected from the image as the search window. The grayscale matrix of the target window is compared with the grayscale matrix of a subwindow of the same size within the search window. The centers of the most similar subwindows are the feature points with the same name ("homonymous") . The results of the coarse matching can be used as a control for subsequent fine matching. To eliminate any gross error, polynomial fitting is performed on synonymous points at different locations. Points with large fitting residuals should then be deleted. A coarse-to-fine matching strategy is adopted to improve reliability. Feature extraction and coarse matching are performed according to a hierarchical image pyramid structure, followed by distortion-corrected matching using ground slope, geometric constraints, and global relaxation matching.

[0082] When a radar beam traveling in a straight line is blocked by a tall object (such as a retaining wall in this case), the radar receiver cannot receive the electromagnetic waves returning from the back of the wall, and no radar echo is generated. A dark area (i.e., a radar shadow) appears at the corresponding location in the radar image. The radar shadow only appears in the direction behind the radar. The generation of a radar shadow depends on the degree of concavity of the radar beam. By comparing the spatial resolution of the GB-InSAR image and the size of the rock in the drone image, selected 3D ground feature points can be identified, thereby revealing the characteristic targets of the same name.

[0083] This example uses the coherence map amplitude dispersion method to automatically extract the coordinates of the retaining wall from a two-dimensional radar image. The coherence coefficient map is generated by "normalized coherence analysis," which is the complex correlation between the two complex images after registration. The coherence is derived as follows:

[0084]

[0085] Where E{·} represents the mathematical expectation. I1 and I2 represent the azimuth extension ranges of the two complex images, which can be calculated according to formula (7) as follows:

[0086]

[0087] Where R0 represents the center distance of the GB InSAR image, and β represents the incident angle width of the GB InSAR image.

[0088] The dispersion of the coherence amplitude (DA) was analyzed to extract retaining wall features from GB-InSAR images. The criterion for coherent scatterers is based on the amplitude of the pixels in the coherence map. The key parameter DA reflects the error accumulation caused by the screening of permanent scatterers. Lower DA values ​​indicate "good" pixels, i.e., higher signal-to-noise ratios. Given the Q number of the coherence map, DA can be calculated as follows:

[0089]

[0090]

[0091] Among them, Stdev[·] represents the standard deviation, A k (ij) is the amplitude of the pixel at (ij) in the kth coherence image, and MA(ij) is the average amplitude.

[0092] The threshold value of DA must be determined by weighting the phase quality and the density of available deformation measurements (e.g. Figure 4 (as shown in (a) and (b)).

[0093] The small-area differential correction algorithm extracts the necessary points from the target image as registration control points based on an interest operator. Dozens of pairs of feature points can be extracted from GB-InSAR images and drone aerial photography point clouds. These pairs represent characteristic lines of road retaining walls and landslide boundaries (or are themselves distinct feature points). These pairs form a corresponding triangular mesh. Due to the large number of points, the area of ​​the triangles is very small.

[0094] For each pair of triangles, P1P2P3 and P1'P2'P3', the coordinates of their three vertices (xi, yi), (x'i, y'i), i=1, 2, 3 are used to define the transformation:

[0095]

[0096] The coefficients a0, a1, a2, b0, b1, and b2 in the above equation can be obtained. Given these coefficients, the triangle P1', P2', and P3' on the image are then corrected to the corresponding triangle P1, P2, and P3 in the target image. The control points used in this method are distributed along the image features. Geometric distortions between different remote sensing images are accurately corrected, effectively correcting the remote sensing image of the slope. Using a weighted average of pixel grayscale values, GB-InSAR and UAV optical images of the area are fused to obtain the original image. The image data collected by the UAV in the landslide area is fused with radar data for 3D visualization.

[0097] use Figure 4 (b) The geostatistical kriging method resamples the amplitude dispersion image, creating strong line features, clear block targets, and easy-to-screen feature targets.

[0098] The results were tested using a false color method developed by the U.S. Naval Research Laboratory (NRL). Let IR represent the result of GB-InSAR fusion with point cloud data. It contains the deformation of the target, represented by grayscale differences. IV represents the image captured by the drone, which contains the RGB texture information of the target. NRL is represented as follows:

[0099]

[0100] The cumulative deformation map of the study area for one day is generated as a two-dimensional deformation map, which is then matched with the DEM generated by TLS. The grayscale expression of the image after RGB transformation is as follows:

[0101] I t =0.299*I R +0.578*I G +0.114*I B (12)

[0102] Where, It is the grayscale of the image after RGB conversion; I R , I G , I B Represent the red, green, and blue channel matrices of the image pixels respectively.

[0103] Apply a linear transformation to the grayscale to increase contrast:

[0104] I l =a*I t +b / 255 (13)

[0105] Among them, I l is the transformed grayscale image, a and b represent the coefficients.

[0106] The above method enhances the relative contrast of the deformation values ​​and can generate a deformation histogram. The distribution of the deformation histogram can be examined to determine whether the contrast has been sufficiently enhanced.

Claims

1. A landslide emergency monitoring system, characterized in that: include: Ground-based interferometric synthetic aperture radar equipment, used to monitor deformation in the landslide area in real time; Terrestrial laser scanning equipment to obtain three-dimensional spatial data of the landslide area; Unmanned aerial vehicles equipped with high-resolution cameras to collect high-resolution image information of the landslide area; The data processing center is used to receive data collected by ground-based interferometric synthetic aperture radar, ground-based laser scanning and unmanned aerial vehicles, and to perform data fusion, analysis and generation of early warning information.

2. The landslide emergency monitoring system according to claim 1, characterized in that: The data processing center includes: A data fusion module is used to match and fuse data collected by ground-based interferometric synthetic aperture radar equipment, ground-based laser scanning equipment, and unmanned aerial vehicles to generate comprehensive monitoring data for the landslide area; Early warning analysis module, used to conduct landslide risk assessment based on comprehensive monitoring data and generate early warning information; The visualization display module is used to present monitoring data and warning information to users in an intuitive form.

3. A landslide emergency monitoring method, applied to the landslide emergency monitoring system according to any one of claims 1-2, characterized in that: The following steps are involved: S1, using ground-based interferometric synthetic aperture radar equipment to continuously monitor landslide potential areas and obtain deformation data; S2, using ground laser scanning equipment to perform three-dimensional scanning of the landslide area and establish a high-precision digital elevation model; S3, using an unmanned aerial vehicle equipped with a high-resolution camera to take aerial photos of the landslide area and obtain high-resolution image information; S4, transmits data collected by ground-based interferometric synthetic aperture radar equipment, ground-based laser scanning equipment, and unmanned aerial vehicles to the data processing center, and matches and integrates multi-source data through data fusion methods; S5, analyze the fused data, assess the landslide risk, and generate early warning information.

4. The landslide emergency monitoring method according to claim 3, characterized in that: In step S4, the steps for matching and integrating multi-source data are as follows: S41, GB-InSAR image simulation using geometric correction method, including geometric simulation and grayscale simulation: Geometric simulation is to project the known DEM data into the GB-InSAR image coordinate system according to the geometric positioning model, and then generate an image coordinate lookup table in the given geographic coordinate system. The simulated GB-InSAR image coordinates corresponding to the geographic coordinates are stored in the form of complex numbers; for grayscale simulation, the area of ​​the ground scattering element is used as the pixel value of the simulated image; S42, performs mismatch correction by geometrically mapping the image into three-dimensional space; The distance and azimuth values ​​obtained by TLS are combined with the pixel values ​​of the radar image to form a new image; when the points of the ground laser scanning device are dense, the resampling method is used to remove the redundancy of the ground laser scanning device; Then, two-dimensional feature points are selected from the radar image and three-dimensional feature points are selected from the texture point cloud of the UAV to form a feature point set with the same name; these feature point sets are input into the mismatch correction model, and the correction parameters of the deviation are obtained through iteration; when the deviation meets the requirements, the matching is completed.

5. The landslide emergency monitoring method according to claim 4, characterized in that: In step S41, the original point cloud is used to simulate the GB-InSAR image, the simulated image is used as the reference image, the real image is used as the image to be corrected, and then the reference points for geometric correction are selected in the two images.

Images