A mountain landslide monitoring method based on remote sensing data

Abstract

This invention discloses a landslide monitoring method based on remote sensing data, comprising: acquiring high-resolution optical remote sensing images and synthetic aperture radar images of the study area; performing registration processing on the optical and radar remote sensing images to obtain remote sensing images with accurate spatial location and geometric shape; generating a high-spectral-resolution fused image of the landslide area; obtaining matched high-resolution DEM data; calculating slope parameters of the study area based on the matched high-resolution DEM data, and determining a slope threshold according to slope distribution characteristics and prior knowledge; extracting slope aspect parameters of the study area using a grid-and-vector-based slope aspect algorithm, and filtering out landslide-prone areas in conjunction with the slope threshold; obtaining a landslide hazard distribution map; obtaining the probability and hazard level of landslide occurrence, and obtaining the landslide situation based on the probability and hazard level of landslide occurrence. This invention achieves accurate identification and dynamic monitoring of landslide disasters.

Description

Technical Field

[0001] This invention belongs to the field of landslide monitoring technology, and in particular relates to a landslide monitoring method based on remote sensing data. Background Technology

[0002] Landslide disaster early warning is a complex technical problem involving the fusion and analysis of multi-source remote sensing data. The core challenge lies in effectively integrating the advantages of optical and radar remote sensing imagery to achieve high-precision and timely landslide risk assessment. First, images acquired by different sensors exhibit geometric distortion and spatial offset; ensuring accurate registration of multi-source data is a fundamental challenge. Second, optical and radar imagery each have their own characteristics; designing a reasonable fusion strategy to generate a comprehensive image with high spatial resolution, high spectral resolution, and multi-polarization characteristics is crucial. Third, the spatial matching between the fused imagery and DEM data directly affects the accuracy of subsequent topographic parameter extraction. Based on this, extracting key topographic parameters from massive amounts of remote sensing data and determining appropriate threshold conditions to identify potential landslide areas is a technical difficulty. Finally, facing a dynamically changing surface environment, establishing a sensitive and reliable landslide early warning mechanism to promptly capture abnormal changes in topographic parameters and make accurate judgments in conjunction with other factors is a major challenge for landslide disaster early warning systems. These interconnected technical problems constitute the core technical contradiction in the field of landslide disaster early warning. Summary of the Invention

[0003] To address the aforementioned technical problems, this invention proposes a landslide monitoring method based on remote sensing data, which enables accurate identification and dynamic monitoring of landslide disasters, providing effective technical support for landslide early warning and disaster prevention and mitigation.

[0004] To achieve the above objectives, the present invention provides a method for monitoring landslides based on remote sensing data, comprising: acquiring high-resolution optical remote sensing images and synthetic aperture radar images of the study area, performing registration processing on the optical and radar remote sensing images, and obtaining remote sensing images with accurate spatial location and geometric shape.

[0005] Based on the texture clarity of optical images and the penetration of radar images, pixel-level and feature-level fusion strategies are designed, and combined with multi-scale fusion algorithms, high-spectral-resolution fused images of landslide areas are generated.

[0006] The high-spectral-resolution fused image of the landslide area is registered and cropped to extract the region of interest that is consistent with the spatial range of the DEM of the study area, and the matched high-resolution DEM data is obtained.

[0007] Based on the matched high-resolution DEM data, the slope parameters of the study area are calculated, and the slope threshold is determined according to the slope distribution characteristics and prior knowledge.

[0008] A slope aspect algorithm based on grids and vectors is used to extract slope aspect parameters of the study area, and combined with the slope threshold, landslide-prone areas are screened out.

[0009] The curvature parameters of the selected landslide-prone areas are calculated. By setting the curvature threshold, high-risk areas prone to landslides are extracted to obtain a landslide hazard distribution map.

[0010] Real-time acquisition of new remote sensing image data detects changes in high-risk landslide areas and, in conjunction with the landslide hazard distribution map, obtains the probability and hazard level of landslide occurrence, and based on the probability and hazard level of landslide occurrence, obtains information on the mountain landslide situation.

[0011] Optionally, obtaining remote sensing images with precise spatial location and accurate geometry includes:

[0012] Feature information of each temporal image is extracted from optical and radar remote sensing images, and spatiotemporal registration of remote sensing images of different temporal phases is completed through feature matching of multi-temporal images.

[0013] Based on the registered optical remote sensing image and synthetic aperture radar remote sensing image, the imaging parameters of the image are obtained by solving the image imaging geometric model, and a unified spatial coordinate system is established by combining the ground control point information to complete the geometric correction of the remote sensing image.

[0014] Image resampling technology is used to eliminate geometric distortion in geometrically corrected optical remote sensing images and synthetic aperture radar remote sensing images, thereby obtaining remote sensing images with accurate spatial location and geometric shape.

[0015] Optionally, generating high-spectral-resolution fused images of landslide areas includes:

[0016] Based on the penetration of radar imagery, the contours of ground features are extracted using a threshold segmentation method, and the segmentation results are optimized using a region growing algorithm to obtain the penetration feature map of the radar imagery.

[0017] The pixel brightness values ​​of the optical image and the backscattering coefficients of the penetration feature map of the radar image are normalized, and the two are fused using IHS transformation to obtain an initial fused image of the optical image texture and the radar image penetration.

[0018] Based on the initial fused image, wavelet transform is used to extract multi-scale features. By setting fusion weights for different frequency bands, feature-level fusion is performed to generate a high-spectral-resolution fused image of the landslide area.

[0019] Optionally, obtaining matched high-resolution DEM data includes:

[0020] The high-spectral-resolution fused images of the landslide area are registered and cropped. Image cropping techniques are used to extract the region of interest images and DEM subsets that are consistent with the spatial range of the study area.

[0021] Spatial resolution is resampled and unified for the cropped region of interest image and the DEM subset. Topographic parameters within the study area are extracted using terrain analysis algorithms to obtain a terrain factor layer corresponding to the remote sensing image space.

[0022] The extracted topographic factor layers are overlaid with the image of the region of interest to obtain the correlation between topographic factors and the spectral and textural features of land features;

[0023] By utilizing the correlation between topographic factors and the spectral and textural features of land features, high-resolution DEM data after matching is obtained.

[0024] Optionally, based on the matched high-resolution DEM data, the slope parameters of the study area are calculated, including:

[0025] Acquire matched high-resolution DEM data within the study area, preprocess the DEM data to remove outliers and noisy data, and ensure data quality;

[0026] The maximum value method was used to process the DEM data, and the maximum elevation difference of each pixel in its neighborhood was calculated to obtain preliminary slope parameters.

[0027] The preliminary slope parameters are processed using the finite difference method to calculate the elevation gradient of each pixel in the horizontal and vertical directions, thereby obtaining accurate slope parameters.

[0028] Optionally, landslide-prone areas can be filtered to include:

[0029] Based on the geographic information data of the study area, extract the slope aspect parameters of each grid or vector unit within the study area;

[0030] Statistical analysis was performed on the extracted aspect parameters to obtain the distribution characteristics and variation patterns of the aspect parameters within the study area.

[0031] Preset slope threshold and slope aspect conditions, perform cluster analysis on raster or vector units that simultaneously meet the slope and slope aspect conditions, and aggregate spatially adjacent units that meet the combination conditions into a region;

[0032] The aggregated regions are classified, and landslide-prone areas are selected based on their characteristic parameters.

[0033] Optionally, obtaining a landslide hazard distribution map includes:

[0034] Based on the digital elevation model data of the selected landslide-prone areas, the planar curvature and profile curvature values ​​of each grid are calculated.

[0035] Based on the geological conditions and historical landslide data of the selected landslide-prone areas, the danger thresholds for planar curvature and profile curvature were determined through statistical analysis.

[0036] Determine whether the planar curvature and profile curvature of each grid cell exceed the threshold. If they exceed the threshold, classify the grid cell as a high-risk landslide area.

[0037] Obtain the spatial distribution of high-risk landslide areas and generate a distribution map of high-risk landslide areas;

[0038] Based on the distribution map of high-risk landslide areas and combined with the topographic features of the study area, a landslide hazard prediction model was established using machine learning algorithms.

[0039] The landslide hazard prediction model is used to predict and assess the landslide hazard of each grid in the landslide-prone area, and obtain the landslide hazard index.

[0040] Based on the landslide hazard index, a tiered mapping method was used to obtain a landslide hazard distribution map.

[0041] Optionally, obtaining the probability and hazard level of a landslide includes:

[0042] Topographic parameters within the landslide area are extracted using image interpretation technology and compared with preset thresholds to determine whether abrupt changes have occurred.

[0043] If the terrain parameters exceed the preset threshold, the images are further interpreted to obtain vegetation and soil information in the landslide area, and the abrupt changes in terrain parameters and vegetation and soil factors are analyzed.

[0044] Based on abrupt changes in topographic parameters and variations in vegetation and soil factors, a comprehensive assessment of landslide risk is conducted to determine the likelihood and risk level of landslides.

[0045] Technical Effects of this Invention: This invention discloses a landslide monitoring method based on remote sensing data. First, high-resolution optical and radar remote sensing images are geometrically corrected and spatially registered. A multi-scale fusion algorithm is then used to generate a fused image of the landslide area, possessing high spatial resolution, high spectral resolution, and multi-polarization characteristics. Next, based on matched DEM data, topographic parameters such as slope, aspect, and curvature are extracted, and threshold conditions are set to screen potential landslide-prone areas. When newly acquired remote sensing data detects abrupt changes in topographic parameters in a high-risk area that exceed a preset threshold, a landslide early warning signal is triggered, and the hazard assessment is dynamically updated based on image interpretation results. This invention achieves accurate identification and dynamic monitoring of landslide disasters, providing effective technical support for landslide early warning and disaster prevention and mitigation. Attached Figure Description

[0046] The accompanying drawings, which form part of this application, are used to provide a further understanding of this application. The illustrative embodiments and descriptions of this application are used to explain this application and do not constitute an undue limitation of this application. In the drawings:

[0047] Figure 1 This is a flowchart illustrating a landslide monitoring method based on remote sensing data, according to an embodiment of the present invention. Detailed Implementation

[0048] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0049] It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases the steps shown or described may be executed in a different order than that shown here.

[0050] like Figure 1 As shown, this embodiment provides a landslide monitoring method based on remote sensing data, including: acquiring high-resolution optical remote sensing images and synthetic aperture radar images of the study area, performing registration processing on the optical and radar remote sensing images, and obtaining remote sensing images with accurate spatial location and geometric shape.

[0051] Based on the texture clarity of optical images and the penetration of radar images, pixel-level and feature-level fusion strategies are designed, and combined with multi-scale fusion algorithms, high-spectral-resolution fused images of landslide areas are generated.

[0052] The high-spectral-resolution fused images of the landslide area were registered and cropped to extract the region of interest that is consistent with the spatial range of the DEM of the study area, and the matched high-resolution DEM data was obtained.

[0053] Based on the matched high-resolution DEM data, the slope parameters of the study area are calculated, and the slope threshold is determined according to the slope distribution characteristics and prior knowledge.

[0054] A slope aspect algorithm based on grids and vectors was used to extract slope aspect parameters of the study area, and combined with slope thresholds, landslide-prone areas were screened out.

[0055] The curvature parameters of the selected landslide-prone areas are calculated. By setting the curvature threshold, high-risk areas prone to landslides are extracted to obtain a landslide hazard distribution map.

[0056] Real-time acquisition of new remote sensing image data detects changes in high-risk landslide areas and, combined with landslide hazard distribution maps, determines the probability and hazard level of landslides, and based on the probability and hazard level, obtains information on landslide conditions.

[0057] Furthermore, obtaining remote sensing images with precise spatial location and accurate geometry includes:

[0058] Feature information of each temporal image is extracted from optical and radar remote sensing images, and spatiotemporal registration of remote sensing images of different temporal phases is completed through feature matching of multi-temporal images.

[0059] Based on the registered optical remote sensing images and synthetic aperture radar remote sensing images, the imaging parameters of the images are obtained by solving the image imaging geometric model, and a unified spatial coordinate system is established by combining the ground control point information to complete the geometric correction of the remote sensing images.

[0060] Image resampling technology is used to eliminate geometric distortion in geometrically corrected optical remote sensing images and synthetic aperture radar remote sensing images, thereby obtaining remote sensing images with accurate spatial location and geometric shape.

[0061] Specifically, based on the registered optical and radar remote sensing images, imaging geometry modeling was performed using RFM and Range-Doppler models respectively. Elevation control points were used to optimize the model, establishing a unified spatial coordinate system. Geometric correction was then applied to the remote sensing images, ensuring the average deviation between the corrected images and the actual ground positions was less than 1 meter. Bilinear interpolation resampling technology was used to resample the corrected remote sensing images, removing geometric distortions and ensuring accurate spatial correspondence between the images and actual ground features. Building upon this, image fusion algorithms such as IHS and wavelet transform were used to fuse the spectral, textural, and geometric information of the optical and radar remote sensing images, generating remote sensing images with both high spatial and spectral resolution. Orthorectification was then performed on the fused remote sensing images, using DEM data to remove the influence of terrain undulations, ensuring the geometric projection of the images aligns with the map coordinate system. The generated orthorectified remote sensing images can be directly used for GIS analysis and mapping.

[0062] Furthermore, generating high-spectral-resolution fused images of the landslide area includes:

[0063] Based on the penetration of radar images, the threshold segmentation method is used to extract the contours of ground features, and the segmentation results are optimized by the region growing algorithm to obtain the penetration feature map of the radar images.

[0064] The pixel brightness values ​​of the optical image and the backscattering coefficients of the penetration feature map of the radar image are normalized, and the two are fused using IHS transformation to obtain an initial fused image of the optical image texture and the radar image penetration.

[0065] Based on the initial fused image, wavelet transform is used to extract multi-scale features. By setting fusion weights for different frequency bands, feature-level fusion is performed to generate a high-spectral-resolution fused image of the landslide area.

[0066] Specifically, regarding the penetration of the radar image, a threshold segmentation method is used to extract the contours of ground features, and a region growing algorithm is used to optimize the segmentation results to obtain the penetration feature map of the radar image. The pixel brightness values ​​of the optical image and the backscattering coefficients of the radar image are normalized, and the two are fused using IHS transform to obtain an initial fused image that retains the texture of the optical image and the penetration of the radar image. For the initial fused image, wavelet transform is used to extract multi-scale features, and feature-level fusion is achieved by setting fusion weights for different frequency bands to generate a landslide area fused image with multi-scale features. If the spatial resolution of the fused image of the landslide area does not meet the requirements, an interpolation algorithm is used to improve the spatial resolution. If the spectral resolution of the fused image of the landslide area is insufficient, hyperspectral data is introduced for spectral enhancement. Based on the polarization scattering characteristics of the landslide area features, a combination of polarization channels is selected to generate the multi-polarization features of the fused image of the landslide area. By combining visual interpretation and quantitative evaluation, it is determined whether the fused image of the landslide area has achieved the expected effect. If it is not ideal, the wavelet transform step is returned to adjust the fusion strategy until a target fused image of the landslide area with high spatial, hyperspectral, and multi-polarization characteristics is generated.

[0067] Furthermore, obtaining the matched high-resolution DEM data includes:

[0068] The landslide area fusion images were registered and cropped using high spectral resolution images. Image cropping techniques were used to extract the region of interest images and DEM subsets that are consistent with the spatial extent of the study area.

[0069] Spatial resolution was resampled and unified for the cropped region of interest image and DEM subset. Topographic parameters within the study area were extracted using terrain analysis algorithms to obtain a terrain factor layer corresponding to the spatial location of the remote sensing image.

[0070] The extracted topographic factor layers are overlaid with the image of the region of interest to obtain the correlation between topographic factors and the spectral and textural features of land features;

[0071] By utilizing the correlation between topographic factors and the spectral and textural features of land features, high-resolution DEM data after matching is obtained.

[0072] Specifically, to acquire high-resolution DEM data within the study area, this embodiment employs aerial photogrammetry. A high-precision camera mounted on a drone is used to capture aerial images with 80% forward overlap and 60% lateral overlap, obtaining 1-meter resolution orthophotos. A 1-meter resolution DEM is then generated through stereo matching. For the fused 5-meter resolution remote sensing image, the SIFT algorithm is used to extract feature points. A nearest neighbor distance ratio matching strategy is employed to achieve accurate registration with the reference DEM, with the root mean square error after registration controlled within 5 pixels. Through masking and cropping techniques, the region of interest (ROI) image and DEM subset within the study area are extracted and resampled to 5-meter resolution. Terrain analysis algorithms, such as slope, aspect, and curvature, are used to extract terrain parameters within the study area, generating a terrain factor layer spatially consistent with the ROI image. Correlation analysis reveals a negative correlation between vegetation index and slope (correlation coefficient -67), and a positive correlation between bare rock distribution and aspect (correlation coefficient 58). Finally, the imagery of the region of interest, the DEM, and the topographic parameter layers were overlaid to construct a geographic dataset containing multi-source information such as spectrum, texture, elevation, and topography, providing important data support for subsequent landslide identification and risk assessment.

[0073] Furthermore, based on the matched high-resolution DEM data, the slope parameters of the study area were calculated, including:

[0074] Acquire matched high-resolution DEM data within the study area, preprocess the DEM data to remove outliers and noisy data, and ensure data quality;

[0075] The maximum value method was used to process the DEM data, and the maximum elevation difference of each pixel in its neighborhood was calculated to obtain preliminary slope parameters.

[0076] The finite difference method is used to process the initial slope parameters and calculate the elevation gradient of each pixel in the horizontal and vertical directions to obtain accurate slope parameters.

[0077] Specifically, the DEM data is preprocessed to remove outliers and noise, ensuring data quality. The maximum value method is used to process the DEM data, calculating the maximum elevation difference of each pixel within its neighborhood to obtain preliminary slope parameters. The finite difference method is then used to process the DEM data, calculating the elevation gradient of each pixel in both horizontal and vertical directions to obtain precise slope parameters. The slope parameters calculated by the maximum value method and the finite difference method are then fused, taking into account the advantages and disadvantages of both methods, to obtain more accurate and reliable slope distribution information. Based on the topographic features of the study area and known landslide distribution, the slope distribution patterns of potential landslide areas are summarized, forming prior knowledge. The slope distribution information is compared and analyzed with the prior knowledge to determine the slope threshold conditions for potential landslide areas; areas meeting certain slope conditions are identified as potential landslide areas. The slope threshold conditions for potential landslide areas are verified and optimized. Through field investigations and comparisons with historical landslide data, the threshold conditions are continuously adjusted to improve the accuracy of potential landslide area identification.

[0078] Furthermore, landslide-prone areas were identified as including:

[0079] Based on the geographic information data of the study area, extract the slope aspect parameters of each grid or vector unit within the study area;

[0080] Statistical analysis was performed on the extracted aspect parameters to obtain the distribution characteristics and variation patterns of the aspect parameters within the study area.

[0081] Preset slope threshold and aspect conditions, perform cluster analysis on raster or vector cells that simultaneously meet the slope and aspect conditions, and aggregate spatially adjacent cells that meet the combination conditions into a region;

[0082] The aggregated regions are classified, and landslide-prone areas are selected based on their characteristic parameters.

[0083] Specifically, using DEM and land use data of the study area, the slope aspect parameters of each 30m × 30m grid cell were extracted using the raster-based D8 algorithm. Statistical analysis revealed that the slope aspect in the study area was mainly concentrated in the south and southeast directions, accounting for approximately 65% ​​of the total area. Based on existing landslide distribution data, a slope threshold of 20° was set, and the slope of each grid cell was judged; those exceeding the threshold were marked as meeting the slope condition. Within the grid cells meeting the slope condition, the slope aspect was further determined to be between south and southeast; if so, it was marked as meeting the slope aspect condition. For grid cells that simultaneously met both conditions, the DBSCAN clustering algorithm was used for spatial aggregation, and the clustered areas were used as candidate landslide-prone areas. Using data from 100 known landslide locations, characteristic parameters such as slope, aspect, elevation, and lithology of each landslide location were extracted as training samples, and a support vector machine algorithm was used to construct a landslide-prone area discrimination model. The characteristic parameters of the candidate regions are input into a discriminant model for classification, resulting in the distribution of landslide-prone areas within the study area. Using the visualization function of ArcGIS software, a landslide-prone area distribution map is generated and compared with known landslide locations. The results show that 85% of the landslide locations are located within the predicted prone areas, demonstrating the effectiveness of the predictions and providing important reference for local landslide disaster prevention and control. Furthermore, a landslide hazard distribution map is obtained, including:

[0084] Based on the digital elevation model data of the selected landslide-prone areas, the planar curvature and profile curvature values ​​of each grid are calculated.

[0085] Based on the geological conditions and historical landslide data of the selected landslide-prone areas, the danger thresholds for planar curvature and profile curvature were determined through statistical analysis.

[0086] Determine whether the planar curvature and profile curvature of each grid cell exceed the threshold. If they exceed the threshold, classify the grid cell as a high-risk landslide area.

[0087] Obtain the spatial distribution of high-risk landslide areas and generate a distribution map of high-risk landslide areas;

[0088] Based on the distribution map of high-risk landslide areas and combined with the topographic features of the study area, a landslide hazard prediction model was established using machine learning algorithms.

[0089] The landslide hazard prediction model is used to predict and assess the landslide hazard of each grid in the landslide-prone area, and obtain the landslide hazard index.

[0090] Based on the landslide hazard index, a tiered mapping method was used to obtain a landslide hazard distribution map.

[0091] Specifically, based on the digital elevation model data of the study area, the curvature calculation tool in ArcGIS software was used to calculate the planar curvature and profile curvature values ​​of each 30m × 30m raster. Statistical analysis of the planar and profile curvature of 527 landslide sites within the study area identified rasters with a planar curvature greater than 15 or a profile curvature greater than 2 as high-risk landslide areas. Using a raster calculator, a threshold was applied to the planar and profile curvature of each raster; rasters exceeding the threshold were assigned a value of 1, indicating they were high-risk landslide areas, while other rasters were assigned a value of 0. The raster-to-vector tool was used to convert the high-risk landslide area rasters into vector surfaces, obtaining the spatial distribution of these areas and generating a high-risk landslide area distribution map.

[0092] Furthermore, obtaining the probability and hazard level of a landslide includes:

[0093] Topographic parameters within the landslide area are extracted using image interpretation technology and compared with preset thresholds to determine whether abrupt changes have occurred.

[0094] If the terrain parameters exceed the preset threshold, the images are further interpreted to obtain vegetation and soil information in the landslide area, and the abrupt changes in terrain parameters and vegetation and soil factors are analyzed.

[0095] Based on abrupt changes in topographic parameters and variations in vegetation and soil factors, a comprehensive assessment of landslide risk is conducted to determine the likelihood and risk level of landslides.

[0096] Specifically, in ArcGIS, a landslide hazard prediction model is established using a distribution map of high-risk landslide areas, combined with topographic features such as terrain relief and roughness of the study area, and employing a support vector machine (SVM) algorithm. The model is applied to each grid cell in the study area to obtain a landslide hazard index for each cell, with values ​​ranging from 0 to 1. The natural discontinuity method is used to classify the landslide hazard index into five levels, representing very low, low, medium, high, and very high hazard, generating a landslide hazard distribution map for the study area, providing a reference for landslide prevention and risk management. Based on abrupt changes in topographic parameters and vegetation and soil factors, the SVM algorithm is used to comprehensively assess landslide hazard, obtaining the probability and hazard level of landslide occurrence. If the assessment results indicate a high landslide hazard, an early warning signal is triggered, and relevant departments and personnel are promptly notified via SMS, application push notifications, etc., prompting them to take preventative measures. In subsequently acquired new image data, changes in various parameters and factors within the landslide area are continuously monitored, and incremental learning methods are used to dynamically update and optimize the hazard assessment model. Based on dynamically updated assessment results, the level and frequency of early warning signals are adjusted to provide continuous and effective decision support for landslide prevention and control.

[0097] The above are merely preferred embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for monitoring landslides based on remote sensing data, characterized in that, include: High-resolution optical remote sensing images and synthetic aperture radar images of the study area are acquired, and the optical remote sensing images and synthetic aperture radar images are registered to obtain remote sensing images with accurate spatial location and geometric shape. Based on the texture clarity of optical remote sensing images and the penetration of synthetic aperture radar images, pixel-level and feature-level fusion strategies are designed, and combined with multi-scale fusion algorithms, high-spectral-resolution fused images of landslide areas are generated. The high-spectral-resolution fused image of the landslide area is registered and cropped to extract the region of interest that is consistent with the spatial range of the DEM of the study area, and the matched high-resolution DEM data is obtained. Based on the matched high-resolution DEM data, the slope parameters of the study area are calculated, and the slope threshold is determined according to the slope distribution characteristics and prior knowledge. A slope aspect algorithm based on grids and vectors is used to extract slope aspect parameters of the study area, and combined with the slope threshold, landslide-prone areas are screened out. The curvature parameters of the selected landslide-prone areas are calculated. By setting the curvature threshold, high-risk areas prone to landslides are extracted to obtain a landslide hazard distribution map. Real-time acquisition of new remote sensing image data detects changes in high-risk landslide areas and, in conjunction with the landslide hazard distribution map, obtains the probability and hazard level of landslide occurrence, and based on the probability and hazard level of landslide occurrence, obtains information on the mountain landslide situation.

2. The landslide monitoring method based on remote sensing data as described in claim 1, characterized in that, Obtaining remote sensing images with precise spatial location and accurate geometry includes: Feature information of each temporal image is extracted from optical remote sensing image and synthetic aperture radar image, and spatiotemporal registration of remote sensing images of different temporal phases is completed by feature matching of multi-temporal images. Based on the registered optical remote sensing image and synthetic aperture radar remote sensing image, the imaging parameters of the image are obtained by solving the image imaging geometric model, and a unified spatial coordinate system is established by combining the ground control point information to complete the geometric correction of the remote sensing image. Image resampling technology is used to eliminate geometric distortion in geometrically corrected optical remote sensing images and synthetic aperture radar remote sensing images, thereby obtaining remote sensing images with accurate spatial location and geometric shape.

3. The landslide monitoring method based on remote sensing data as described in claim 1, characterized in that, Generating high-spectral-resolution fused images of landslide areas includes: Based on the penetration of synthetic aperture radar imagery, the contours of ground features are extracted using a threshold segmentation method, and the segmentation results are optimized using a region growing algorithm to obtain the penetration feature map of the synthetic aperture radar imagery. The pixel brightness values ​​of the optical remote sensing image and the backscattering coefficients of the penetration feature map of the synthetic aperture radar image are normalized, and the two are fused by IHS transformation to obtain an initial fused image of the texture of the optical remote sensing image and the penetration of the synthetic aperture radar image. Based on the initial fused image, wavelet transform is used to extract multi-scale features. By setting fusion weights for different frequency bands, feature-level fusion is performed to generate a high-spectral-resolution fused image of the landslide area.

4. The landslide monitoring method based on remote sensing data as described in claim 1, characterized in that, The obtained high-resolution DEM data after matching includes: The high-spectral-resolution fused images of the landslide area are registered and cropped. Image cropping techniques are used to extract the region of interest images and DEM subsets that are consistent with the spatial range of the study area. Spatial resolution is resampled and unified for the cropped region of interest image and the DEM subset. Topographic parameters within the study area are extracted using terrain analysis algorithms to obtain a terrain factor layer corresponding to the remote sensing image space. The extracted topographic factor layers are overlaid with the image of the region of interest to obtain the correlation between topographic factors and the spectral and textural features of land features; By utilizing the correlation between topographic factors and the spectral and textural features of land features, high-resolution DEM data after matching is obtained.

5. The landslide monitoring method based on remote sensing data as described in claim 1, characterized in that, Based on the matched high-resolution DEM data, the slope parameters of the study area are calculated as follows: Acquire matched high-resolution DEM data within the study area, preprocess the high-resolution DEM data to remove outliers and noisy data, and ensure data quality; The maximum value method is used to process high-resolution DEM data, calculate the maximum elevation difference of each pixel in its neighborhood, and obtain preliminary slope parameters. The preliminary slope parameters are processed using the finite difference method to calculate the elevation gradient of each pixel in the horizontal and vertical directions, thereby obtaining accurate slope parameters.

6. The landslide monitoring method based on remote sensing data as described in claim 1, characterized in that, The landslide-prone areas identified include: Based on the geographic information data of the study area, extract the slope aspect parameters of each grid or vector unit within the study area; Statistical analysis was performed on the extracted aspect parameters to obtain the distribution characteristics and variation patterns of the aspect parameters within the study area. Preset slope threshold and slope aspect conditions, perform cluster analysis on raster or vector cells that simultaneously meet the slope threshold and slope aspect conditions, and aggregate spatially adjacent cells that meet the slope threshold and slope aspect conditions into a region; The aggregated regions are classified, and landslide-prone areas are selected based on their characteristic parameters.

7. The landslide monitoring method based on remote sensing data as described in claim 1, characterized in that, Obtaining landslide hazard distribution maps includes: Based on the digital elevation model data of the selected landslide-prone areas, the planar curvature and profile curvature values ​​of each grid are calculated. Based on the geological conditions and historical landslide data of the selected landslide-prone areas, the danger thresholds for planar curvature and profile curvature were determined through statistical analysis. Determine whether the planar curvature and profile curvature of each grid cell exceed the threshold. If they exceed the threshold, classify the grid cell as a high-risk landslide area. Obtain the spatial distribution of high-risk landslide areas and generate a distribution map of high-risk landslide areas; Based on the distribution map of high-risk landslide areas and combined with the topographic features of the study area, a landslide hazard prediction model was established using machine learning algorithms. The landslide hazard prediction model is used to predict and assess the landslide hazard of each grid in the landslide-prone area, and obtain the landslide hazard index. Based on the landslide hazard index, a tiered mapping method was used to obtain a landslide hazard distribution map.

8. The landslide monitoring method based on remote sensing data as described in claim 1, characterized in that, The probability and hazard level of a landslide include: Topographic parameters within the landslide area are extracted using image interpretation technology and compared with preset thresholds to determine whether abrupt changes have occurred. If the terrain parameters exceed the preset threshold, the images are further interpreted to obtain vegetation and soil information in the landslide area, and the abrupt changes in terrain parameters and vegetation and soil factors are analyzed. Based on abrupt changes in topographic parameters and variations in vegetation and soil factors, a comprehensive assessment of landslide risk is conducted to determine the likelihood and risk level of landslides.

Images