Three-dimensional investigation method for low-amplitude brittle deformation landslide area

Abstract

The invention relates to a three-dimensional investigation method for a low-amplitude brittle deformation landslide area. The three-dimensional investigation method comprises the steps of multi-stage data acquisition, point cloud data preprocessing, three-dimensional model construction, deformation quantitative analysis, mineral composition auxiliary verification and landslide disaster risk assessment and early warning. According to the scheme provided by the invention, an unmanned aerial vehicle laser radar system carrying RTK positioning is adopted, an improved progressive encryption triangulation network filtering algorithm and a multi-period data registration technology are combined, a high-resolution three-dimensional model is constructed, and short-period dynamic monitoring of mountain geological disasters is realized; through analysis of a high-resolution three-dimensional model and a digital elevation difference model, low-amplitude brittle deformation characteristics are captured, topographic deformation data and geological analysis are fused, topographic evolution characteristics and a physical and chemical mechanism of deformation instability are analyzed, and deformation quantification and mechanism analysis of a geological disaster area in a mountainous area are realized. And technical support is provided for short-period dynamic monitoring and risk assessment of geological disasters in mountainous areas.

Description

Technical Field

[0001] This application relates to the field of landslide investigation technology, and in particular to a three-dimensional investigation method for low-amplitude brittle deformation landslide areas. Background Technology

[0002] In related technologies, loose deposit slopes in mountainous areas are prone to low-amplitude brittle deformation. The early changes in surface morphology of such deformation are weak and below the threshold that can be discerned by the human eye, resulting in a blurred deformation mechanism and difficulty in early identification, which poses a great challenge to geological disaster early warning and risk prevention and control.

[0003] Among existing surface deformation monitoring technologies, interferometric synthetic aperture radar (InSAR) technology can achieve large-scale monitoring, but its spatial resolution is low (usually >30 meters), making it difficult to identify deformation features with centimeter-level precision, and it is easily affected by vegetation cover, leading to unwrapping failure. Traditional ground survey methods are inefficient and cannot achieve large-scale, non-contact monitoring. Conventional UAV photogrammetry technology lacks the ability to penetrate vegetation, and its point cloud accuracy and terrain characterization capabilities are insufficient, making it difficult to meet the requirements for fine detection of low-amplitude brittle deformation.

[0004] Therefore, there is an urgent need to develop a three-dimensional survey method that combines high precision, high efficiency, and wide coverage to achieve early identification, deformation quantification, and mechanism analysis of low-amplitude brittle deformation landslide areas, and to provide technical support for geological disaster risk assessment and early warning. Summary of the Invention

[0005] To address or partially address the problems existing in related technologies, this application provides a three-dimensional stereoscopic investigation method for low-amplitude brittle deformation landslide areas, aiming to solve the problems of difficulty in early identification of deformation on loose deposit slopes in mountainous areas, low accuracy of deformation quantification, and insufficient analysis of deformation mechanisms.

[0006] This application provides a three-dimensional stereoscopic investigation method for low-amplitude brittle deformation landslide areas, including: A dual-platform collaborative operation mode was adopted, using drones equipped with lidar systems and drones with integrated RTK modules to conduct at least three phases of aerial surveys of the target landslide area to acquire point cloud data and multi-view image data; An improved progressive densification triangulation filtering algorithm is used to filter the acquired raw point cloud data, removing non-ground noise points such as vegetation and artificial buildings, and retaining bare ground points. The improved progressive densification triangulation filtering algorithm includes multi-scale seed point optimization, terrain adaptive parameter adjustment, and edge protection mechanism. Based on the 3D modeling platform, motion restoration and 3D reconstruction of structure are performed on preprocessed point cloud data and image data. Combined with the spatial analysis module, a high-resolution spatial model is generated, including digital elevation model, digital surface model and orthophoto. Point cloud data is classified using point cloud data processing software and constructed with the 3D modeling platform. Deformation analysis involves spatial registration of multi-period digital elevation model data using the DEM difference analysis module in the spatial analysis module, elimination of system offset by least squares surface fitting algorithm, generation of digital elevation difference model, calculation of three-dimensional displacement vector, and quantification of settlement, displacement rate and topographic parameter evolution characteristics of the target area. Based on the generated high-resolution model, three types of brittle gravity deformation regions were identified: stepped scarps, tensile fracture networks, and erosion gully diversion. Combined with the coupling relationship between surface roughness and slope angle, active deformation regions were determined. Mineral composition was used to assist in verification. Soil samples were collected from the target area, and the clay mineral content was analyzed by X-ray diffraction. Combined with hydrological data and tectonic stress conditions, the physicochemical mechanism of deformation instability was revealed. Based on deformation thresholds and geological parameters, a landslide disaster risk assessment and early warning model is constructed, and monitoring reports and risk maps are output.

[0007] Optionally, in some implementations, acquiring point cloud data and multi-view image data includes: The flight plan adopts a terrain-following flight mode, based on professional aerial survey software, and is equipped with multi-angle aerial photography. The forward-looking camera has an angle of -60° and a path direction of 60°, the rear-looking camera has an angle of -60° and a path direction of 150°, the left-looking camera has an angle of -60° and a path direction of 330°, and the right-looking camera has an angle of -60° and a path direction of 240°. The terrain-following flight mode is set at an altitude of 100m, with a heading overlap rate of ≥80% and a lateral overlap rate of ≥80%. It maintains a constant relative altitude between the aircraft and the ground surface in real time and generates the optimal three-dimensional trajectory scheme based on terrain features. After completing the flight path planning, the flight path parameters and elevation data are synchronized to the cloud database and then directly imported into the flight control system for execution via terminal devices; For complex terrain scenarios, it integrates global 1 arc-second resolution elevation data to dynamically construct a 3D terrain model. When a sudden change in elevation is detected, the flight control system adjusts the flight altitude in real time according to a preset safety threshold, effectively avoiding flight risks caused by sudden changes in terrain while ensuring the accuracy of terrain-following flight.

[0008] Optionally, in some implementations, determining the active deformation region includes: Based on the generated digital elevation model, digital surface model, and orthophoto, and combined with geographic information system software tools, three types of brittle gravity deformation indicators were identified: For stepped steep slopes, elevation sections are extracted using linear profile tools, and the vertical displacement of the steep slopes is measured. Tensile fracture network, based on three-dimensional curvature calculation and image texture analysis, identifies dense fracture distribution areas; Traces of erosion gully diversion were captured by comparing multiple images to identify changes in the flow direction and morphological adjustments of the erosion gullies. By combining the coupling relationship between surface roughness and slope angle obtained from deformation analysis, the active deformation areas and their spatial distribution patterns can be determined.

[0009] Optionally, in some implementations, mineral composition-assisted verification includes: Soil samples were collected from areas of active deformation, and mineral composition was analyzed using X-ray diffraction. The relative contents of illite, kaolinite, muscovite, quartz, and chlorite were quantified using the Rietveld full-spectrum refinement method. When the total amount of clay minerals exceeds 60% and the amount of stone is 30% to 50%, the target area is determined to have the physical and chemical basis for landslide instability. Combining hydrological data and tectonic stress conditions, the coupling mechanism of deformation instability is revealed.

[0010] Optionally, in some implementations, a landslide disaster risk assessment and early warning model is constructed, including: The model integrates multi-source monitoring and survey data to construct five types of input parameters, including deformation parameters, geological parameters, hydro-meteorological parameters, topographic parameters, and human activity parameters. Deformation parameters, including vertical deformation, horizontal displacement, deformation rate, deformation spatial gradient, and deformation direction angle, are extracted using a multi-period digital elevation difference (DEM) model. Geological parameters, including clay mineral content, rock weathering index, distance from fault zones, joint density, and soil / rock type, are obtained through field sampling, X-ray diffraction analysis, or geological maps. Hydro-meteorological parameters, including cumulative rainfall, rainfall intensity, extreme rainfall events, soil moisture content, and runoff path density, are obtained through meteorological station data and hydrological model analysis. Topographic parameters, including slope, aspect, surface roughness, coefficient of variation of elevation, and curvature, are obtained based on a DEM model. Human activity parameters, including slope cutting index, building density, road distance, and vegetation destruction index, are obtained through orthophoto interpretation and field surveys. The deformation threshold is dynamically calibrated by using historical landslide case inversion and numerical simulation verification to determine the initial deformation threshold. The primary threshold is an annual vertical deformation ≥10cm; the intermediate threshold is a monthly deformation rate ≥5cm / month; and the advanced threshold is a daily deformation rate ≥2cm / day, or a sudden change in the spatial gradient of deformation. The initial threshold is dynamically adjusted based on regional geological conditions, and regional adaptive corrections are performed. (1) In the formula, This indicates the adjusted threshold. Indicates the basic threshold. This represents the normalized value of clay mineral content. Indicates the frequency of extreme rainfall. , This represents the regional correction coefficient, obtained by fitting historical data. The risk assessment index is calculated using a weighted superposition model. Influencing factors are normalized, and their weights are determined by combining the analytic hierarchy process (AHP) with random forest feature importance analysis, resulting in the weighted superposition model for the risk assessment index. (2) In the formula, Indicates a risk assessment index; , , , , These represent the weights of deformation parameters, geological parameters, hydrometeorological parameters, topographic parameters, and human activity parameters, respectively; D, G, H, T, and A represent the normalized values ​​of deformation parameters, geological parameters, hydrometeorological parameters, topographic parameters, and human activity parameters, respectively. Risk levels are classified according to the risk assessment index: low risk (LRI < 0.3, weak deformation, stable geological conditions); medium risk (0.3 ≤ LRI < 0.6, significant deformation, requiring regular monitoring); and high risk (LRI ≥ 0.6, strong deformation, intervention recommended). The monitoring report and risk map output include monitoring status, deformation analysis results, geological and hydrological information, and risk assessment results; the risk map consists of digital elevation model, digital surface model, orthophoto, digital terrain model, digital elevation difference model, and orthophoto map of marked deformation active areas.

[0011] Optionally, in some implementations, the improved progressive encryption triangular mesh filtering algorithm includes: Select initial seed points and improve the uniformity of seed point distribution through a multi-scale seed point optimization algorithm; Based on the terrain adaptive parameter adjustment mechanism, an initial triangular network is constructed, and the triangular network structure is optimized by iterative densification. An edge protection mechanism is introduced to remove non-ground noise points from vegetation and man-made structures, while retaining bare ground points. Point cloud optimization was performed using LiDAR 360 and EPS software, controlling the point cloud density to ≥50 points per square meter to ensure the accuracy of ground point extraction.

[0012] Optionally, in some implementations, generating a spatially high-resolution model includes: The preprocessed point cloud data and multi-view image data are imported into the 3D modeling platform to perform motion recovery structure 3D reconstruction and generate the initial point cloud model. By combining ArcGIS Pro's spatial analysis module, we completed joint adjustment and dense matching of multi-view images, generating digital elevation models, digital surface models, and orthophotos with a spatial resolution of 0.1m. Point cloud data processing software was used to classify point clouds, and a digital terrain model was constructed using a 3D modeling platform. The vertical accuracy of the digital terrain model was ±5cm, which meets the requirements of the 1:500 terrain mapping specification.

[0013] Optionally, in some implementations, deformation analysis includes: The DEM difference analysis module of the spatial analysis module is used to spatially register the DEM data from multiple periods, forcibly unify them to a 0.1-meter resolution grid system, and use the least squares surface fitting algorithm to eliminate system offset; Interference zone filtering: Based on the NDVI index and building vector layer, vegetation and artificial building interference zones are automatically filtered out. At least four bedrock exposure feature points are selected to construct a stable reference network. Among them, the NDVI index threshold is >0.6 and the building area of ​​the building vector layer is >20m². Perform pixel-level elevation difference calculations to generate millimeter-level digital elevation difference models. Solve the three-dimensional displacement vectors using the Rodriguez matrix to quantify the settlement, displacement rate, slope angle changes, and surface roughness evolution characteristics of the target area.

[0014] The technical solution provided in this application may include the following beneficial effects: By employing an UAV lidar system equipped with RTK positioning, combined with an improved progressive densification triangular mesh filtering algorithm and multi-period data registration technology, a high-resolution 3D model is constructed to achieve short-cycle dynamic monitoring of geological disasters in mountainous areas. Through analysis of the high-resolution 3D model and the digital elevation difference model, low-amplitude brittle deformation characteristics are captured. By integrating terrain deformation data and geological analysis, the physical mechanisms of terrain evolution and deformation instability are analyzed, realizing the regional deformation quantification and mechanism analysis of geological disasters in mountainous areas, and providing technical support for short-cycle dynamic monitoring and risk assessment of geological disasters in mountainous areas.

[0015] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and do not limit this application. Attached Figure Description

[0016] The above and other objects, features and advantages of this application will become more apparent from the more detailed description of exemplary embodiments thereof in conjunction with the accompanying drawings, wherein the same reference numerals generally represent the same components in the exemplary embodiments thereof.

[0017] Figure 1 This is a flowchart illustrating the three-dimensional investigation method for low-amplitude brittle deformation landslide areas as shown in the embodiments of this application. Detailed Implementation

[0018] Embodiments of this application will now be described in more detail with reference to the accompanying drawings. While embodiments of this application are shown in the drawings, it should be understood that this application may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided to make this application more thorough and complete, and to fully convey the scope of this application to those skilled in the art.

[0019] The technical solutions of the embodiments of this application are described in detail below with reference to the accompanying drawings.

[0020] Figure 1 This is a flowchart illustrating the three-dimensional investigation method for low-amplitude brittle deformation landslide areas as shown in the embodiments of this application.

[0021] See Figure 1 A three-dimensional survey method for low-amplitude brittle deformation landslide areas includes: S101. Adopting a dual-platform collaborative operation mode, using drones equipped with lidar systems and drones with integrated RTK modules to conduct at least three phases of aerial surveys of the target landslide area to acquire point cloud data and multi-view image data. Specifically, the flight plan adopts a terrain-following flight mode, based on professional aerial survey software, and is equipped with multi-angle aerial photography: forward-looking camera angle -60°, path direction 60°, rear-looking camera angle -60°, path direction 150°, left-looking camera angle -60°, path direction 330°, right-looking camera angle -60°, path direction 240°. The terrain-following flight mode is set at an altitude of 100m, with a heading overlap rate of ≥80% and a lateral overlap rate of ≥80%. It maintains a constant relative altitude between the aircraft and the ground surface in real time and generates the optimal three-dimensional trajectory scheme based on terrain features. After completing the flight path planning, the flight path parameters and elevation data are synchronized to the cloud database and then directly imported into the flight control system for execution via terminal devices; For complex terrain scenarios, it integrates global 1 arc-second resolution elevation data to dynamically construct a 3D terrain model. When a sudden change in elevation is detected, the flight control system adjusts the flight altitude in real time according to a preset safety threshold, effectively avoiding flight risks caused by sudden changes in terrain while ensuring the accuracy of terrain-following flight.

[0022] S102. An improved progressive densification triangular mesh filtering algorithm is used to filter the acquired raw point cloud data, removing non-ground noise points such as vegetation and artificial buildings, and retaining bare ground points. Specifically, the improved progressive encryption triangular mesh filtering algorithm includes: Select initial seed points and improve the uniformity of seed point distribution through a multi-scale seed point optimization algorithm; Based on the terrain adaptive parameter adjustment mechanism, an initial triangular network is constructed, and the triangular network structure is optimized by iterative densification. An edge protection mechanism is introduced to remove non-ground noise points from vegetation and man-made structures, while retaining bare ground points. Point cloud optimization was performed using LiDAR 360 and EPS software, controlling the point cloud density to ≥50 points per square meter to ensure the accuracy of ground point extraction.

[0023] S103. Based on the 3D modeling platform, perform motion restoration and 3D reconstruction of the structure from the preprocessed point cloud data and image data. Combine with the spatial analysis module to generate a high-resolution spatial model, which includes a digital elevation model (DEM), a digital surface model (DSM), and an orthophoto (DOM). Use point cloud data processing software to classify the point cloud and, in conjunction with the 3D modeling platform, construct a digital terrain model (DTM). Specifically, generating a high-resolution spatial model includes: The preprocessed point cloud data and multi-view image data are imported into the 3D modeling platform to perform motion recovery structure 3D reconstruction and generate the initial point cloud model. By combining ArcGIS Pro's spatial analysis module, we completed joint adjustment and dense matching of multi-view images, generating a digital elevation model (DEM), a digital surface model (DSM), and an orthophoto (DOM) with a spatial resolution of 0.1m. Point cloud data processing software was used to classify point clouds, and a digital terrain model (DTM) was constructed using a 3D modeling platform. The vertical accuracy of the digital terrain model was ±5cm, which meets the requirements of the 1:500 terrain mapping specification.

[0024] S104. Deformation analysis: Through the DEM difference analysis module of the spatial analysis module, spatial registration is performed on the multi-period digital elevation model data. The least squares surface fitting algorithm is used to eliminate system offset, generate a digital elevation difference model, solve the three-dimensional displacement vector, and quantify the settlement, displacement rate and topographic parameter evolution characteristics of the target area. Specifically, deformation analysis includes: The DEM difference analysis module of the spatial analysis module spatially registers multiple DEM data in 8FDB rows, forcibly unifying them to a 0.1-meter resolution grid system. The least squares surface fitting algorithm is used to eliminate system offset. After completing the spatial registration of the multi-period digital elevation model, a set of reference points located in a stable region is selected as the fitting sample. The stable region is a bedrock exposed area or an artificially stabilized structure area that does not undergo deformation. Using the elevation difference of the reference points as the observation, a low-order polynomial surface model is constructed. The surface parameters are solved by the least squares method to obtain a trend surface reflecting the characteristics of systematic elevation offset. The trend surface is subtracted from the corresponding period of the digital elevation model to eliminate the overall system offset caused by positioning error, attitude error and system cumulative error, thereby improving the accuracy of the digital elevation difference model in depicting the real surface deformation.

[0025] Interference zone filtering: Based on the NDVI index and building vector layer, vegetation and artificial building interference zones are automatically filtered out. At least four bedrock exposure feature points are selected to construct a stable reference network. Among them, the NDVI index threshold is >0.6 and the building area of ​​the building vector layer is >20m². Pixel-level elevation difference calculations are performed to generate millimeter-level digital elevation difference models. Three-dimensional displacement vectors are calculated using the Rodrigues matrix to quantify the settlement, displacement rate, slope angle changes, and surface roughness evolution characteristics of the target area. After obtaining multi-period digital elevation difference models, a local terrain coordinate system is constructed based on the terrain surface normal vector and local slope attitude information. Pixel-level or grid-level elevation differences are considered as vertical displacement components, and the displacement components are transformed from the local terrain coordinate system to a unified three-dimensional spatial coordinate system using the Rodrigues rotation formula. This yields the three-dimensional displacement vector corresponding to each spatial location, including vertical and horizontal displacement components, thereby achieving a quantitative calculation of the true three-dimensional deformation characteristics of the target area.

[0026] S105. Based on the generated high-resolution model, three types of brittle gravity deformation regions are identified: stepped scarps, tensile fracture networks, and erosion gully diversion. Combined with the coupling relationship between surface roughness and slope angle, active deformation regions are determined. Specifically, determining the active deformation region includes: Based on the generated digital elevation model (DEM), digital surface model (DSM), and orthophoto map (DOM), combined with geographic information system software tools, three types of brittle gravity deformation indicators were identified: For stepped steep slopes, elevation sections are extracted using linear profile tools, and the vertical displacement of the steep slopes is measured. Tensile fracture network, based on three-dimensional curvature calculation and image texture analysis, identifies dense fracture distribution areas; Traces of erosion gully diversion were captured by comparing multiple images to identify changes in the flow direction and morphological adjustments of the erosion gullies. By combining the coupling relationship between surface roughness and slope angle obtained from deformation analysis, the active deformation areas and their spatial distribution patterns can be determined.

[0027] S106. Mineral composition auxiliary verification: Soil samples were collected from the target area, and the clay mineral content was analyzed by X-ray diffraction technology. Combined with hydrological data and tectonic stress conditions, the physicochemical mechanism of deformation instability was revealed. Specifically, mineral composition verification includes: Soil samples were collected from areas of active deformation, and mineral composition was analyzed using X-ray diffraction. The relative contents of illite, kaolinite, muscovite, quartz, and chlorite were quantified using the Rietveld full-spectrum refinement method. When the total amount of clay minerals exceeds 60% and the amount of stone is 30% to 50%, the target area is determined to have the physical and chemical basis for landslide instability. Combining hydrological data and tectonic stress conditions, the coupling mechanism of deformation instability is revealed.

[0028] S107. Based on deformation thresholds and geological parameters, construct a landslide disaster risk assessment and early warning model, and output monitoring reports and risk maps.

[0029] Specifically, a landslide disaster risk assessment and early warning model is constructed, including: The model integrates multi-source monitoring and survey data to construct five types of input parameters, including deformation parameters, geological parameters, hydro-meteorological parameters, topographic parameters, and human activity parameters. Deformation parameters, including vertical deformation, horizontal displacement, deformation rate, deformation spatial gradient, and deformation direction angle, are extracted using a multi-period digital elevation difference (DEM) model. Geological parameters, including clay mineral content, rock weathering index, distance from fault zones, joint density, and soil / rock type, are obtained through field sampling, X-ray diffraction analysis, or geological maps. Hydro-meteorological parameters, including cumulative rainfall, rainfall intensity, extreme rainfall events, soil moisture content, and runoff path density, are obtained through meteorological station data and hydrological model analysis. Topographic parameters, including slope, aspect, surface roughness, coefficient of variation of elevation, and curvature, are obtained based on a DEM model. Human activity parameters, including slope cutting index, building density, road distance, and vegetation destruction index, are obtained through orthophoto interpretation and field surveys. The deformation threshold is dynamically calibrated by using historical landslide case inversion and numerical simulation verification to determine the initial deformation threshold. The primary threshold is an annual vertical deformation ≥10cm; the intermediate threshold is a monthly deformation rate ≥5cm / month; and the advanced threshold is a daily deformation rate ≥2cm / day, or a sudden change in the spatial gradient of deformation. The initial threshold is dynamically adjusted based on regional geological conditions, and regional adaptive corrections are performed. (1) In the formula, This indicates the adjusted threshold. Indicates the basic threshold. This represents the normalized value of clay mineral content. Indicates the frequency of extreme rainfall. , This represents the regional correction coefficient, obtained by fitting historical data. The risk assessment index is calculated using a weighted superposition model. Influencing factors are normalized, and their weights are determined by combining the analytic hierarchy process (AHP) with random forest feature importance analysis, resulting in the weighted superposition model for the risk assessment index. (2) In the formula, Indicates a risk assessment index; , , , , These represent the weights of deformation parameters, geological parameters, hydrometeorological parameters, topographic parameters, and human activity parameters, respectively; D, G, H, T, and A represent the normalized values ​​of deformation parameters, geological parameters, hydrometeorological parameters, topographic parameters, and human activity parameters, respectively. Risk levels are classified according to the risk assessment index: low risk (LRI < 0.3, weak deformation, stable geological conditions); medium risk (0.3 ≤ LRI < 0.6, significant deformation, requiring regular monitoring); and high risk (LRI ≥ 0.6, strong deformation, intervention recommended). The monitoring report and risk map output include monitoring status, deformation analysis results, geological and hydrological information, and risk assessment results; the risk map consists of digital elevation model (DEM), digital surface model (DSM), orthophoto map (DOM), digital terrain model, digital elevation difference model, and orthophoto map with marked deformation active areas and risk levels.

[0030] The various embodiments of this application have been described above. These descriptions are exemplary and not exhaustive, nor are they limited to the disclosed embodiments. Many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the described embodiments. The terminology used herein is chosen to best explain the principles, practical application, or improvement of the technology in the market, or to enable others skilled in the art to understand the embodiments disclosed herein.

Claims

1. A three-dimensional stereoscopic investigation method for low-amplitude brittle deformation landslide areas, characterized in that, include: A dual-platform collaborative operation mode was adopted, using drones equipped with lidar systems and drones with integrated RTK modules to conduct at least three phases of aerial surveys of the target landslide area to acquire point cloud data and multi-view image data; An improved progressive densification triangulation filtering algorithm is used to filter the acquired raw point cloud data, removing non-ground noise points such as vegetation and artificial buildings, and retaining bare ground points. The improved progressive densification triangulation filtering algorithm includes multi-scale seed point optimization, terrain adaptive parameter adjustment, and edge protection mechanism. Based on the 3D modeling platform, motion restoration and 3D reconstruction of structure are performed on preprocessed point cloud data and image data, and a high-resolution spatial model is generated by combining the spatial analysis module; point cloud data classification is performed using point cloud data processing software, and digital terrain model is constructed in conjunction with the 3D modeling platform. Deformation analysis involves spatial registration of multi-period digital elevation model data using the DEM difference analysis module in the spatial analysis module, elimination of system offset by least squares surface fitting algorithm, generation of digital elevation difference model, calculation of three-dimensional displacement vector, and quantification of settlement, displacement rate and topographic parameter evolution characteristics of the target area. Based on the generated high-resolution model, three types of brittle gravity deformation regions were identified: stepped scarps, tensile fracture networks, and erosion gully diversion. Combined with the coupling relationship between surface roughness and slope angle, active deformation regions were determined. Mineral composition was used to assist in verification. Soil samples were collected from the target area, and the clay mineral content was analyzed by X-ray diffraction. Combined with hydrological data and tectonic stress conditions, the physicochemical mechanism of deformation instability was revealed. Based on deformation thresholds and geological parameters, a landslide disaster risk assessment and early warning model is constructed, and monitoring reports and risk maps are output.

2. The three-dimensional stereoscopic investigation method for low-amplitude brittle deformation landslide areas according to claim 1, characterized in that, The acquisition of point cloud data and multi-view image data includes: The flight plan adopts a terrain-following flight mode, is based on professional aerial survey software, and is equipped with multi-angle aerial photography; The terrain-following flight mode is set at an altitude of 100m, with a heading overlap rate of ≥80% and a lateral overlap rate of ≥80%.

3. The three-dimensional stereoscopic investigation method for low-amplitude brittle deformation landslide areas according to claim 1, characterized in that, The determination of the active deformation region includes: Based on the generated digital elevation model, digital surface model, and orthophoto, and combined with geographic information system software tools, three types of brittle gravity deformation indicators were identified: For stepped steep slopes, elevation sections are extracted using linear profile tools, and the vertical displacement of the steep slopes is measured. Tensile fracture network, based on three-dimensional curvature calculation and image texture analysis, identifies dense fracture distribution areas; Traces of erosion gully diversion were captured by comparing multiple images to identify changes in the flow direction and morphological adjustments of the erosion gullies. By combining the coupling relationship between surface roughness and slope angle obtained from deformation analysis, the active deformation areas and their spatial distribution patterns can be determined.

4. The three-dimensional stereoscopic investigation method for low-amplitude brittle deformation landslide areas according to claim 1, characterized in that, The mineral composition verification includes: Soil samples were collected from areas of active deformation, and mineral composition was analyzed using X-ray diffraction. The relative contents of illite, kaolinite, muscovite, quartz, and chlorite were quantified using the Rietveld full-spectrum refinement method. When the total amount of clay minerals exceeds 60% and the amount of stone is 30% to 50%, the target area is determined to have the physical and chemical basis for landslide instability. Combining hydrological data and tectonic stress conditions, the coupling mechanism of deformation instability is revealed.

5. The three-dimensional stereoscopic investigation method for low-amplitude brittle deformation landslide areas according to claim 1, characterized in that, The construction of the landslide disaster risk assessment and early warning model includes: The model integrates multi-source monitoring and survey data and constructs five types of input parameters, including deformation parameters, geological parameters, hydro-meteorological parameters, topographic parameters, and human activity parameters. The deformation threshold is dynamically calibrated by using historical landslide case inversion and numerical simulation verification to determine the initial deformation threshold. The primary threshold is an annual vertical deformation ≥10cm; the intermediate threshold is a monthly deformation rate ≥5cm / month; and the advanced threshold is a daily deformation rate ≥2cm / day, or a sudden change in the spatial gradient of deformation. The initial threshold is dynamically adjusted based on regional geological conditions, and regional adaptive corrections are made accordingly. （1） In the formula, This indicates the adjusted threshold. Indicates the basic threshold. This represents the normalized value of clay mineral content. Indicates the frequency of extreme rainfall. , This represents the regional correction coefficient, obtained by fitting historical data. The risk assessment index is calculated using a weighted superposition model. Influencing factors are normalized, and their weights are determined by combining the analytic hierarchy process (AHP) with random forest feature importance analysis, resulting in the weighted superposition model for the risk assessment index. （2） In the formula, Indicates a risk assessment index; , , , , These represent the weights of deformation parameters, geological parameters, hydrometeorological parameters, topographic parameters, and human activity parameters, respectively; D, G, H, T, and A represent the normalized values ​​of deformation parameters, geological parameters, hydrometeorological parameters, topographic parameters, and human activity parameters, respectively. Risk levels are classified according to the risk assessment index: low risk (LRI < 0.3, weak deformation, stable geological conditions); medium risk (0.3 ≤ LRI < 0.6, significant deformation, requiring regular monitoring); and high risk (LRI ≥ 0.6, strong deformation, intervention recommended). The monitoring report and risk map output include monitoring status, deformation analysis results, geological and hydrological information, and risk assessment results; the risk map consists of a digital elevation model, a digital surface model, orthophotos, a digital terrain model, a digital elevation difference model, and orthophotos of areas with marked deformation activity.

6. The three-dimensional stereoscopic investigation method for low-amplitude brittle deformation landslide areas according to claim 1, characterized in that, The improved progressive encryption triangular mesh filtering algorithm includes: Select initial seed points and improve the uniformity of seed point distribution through a multi-scale seed point optimization algorithm; Based on the terrain adaptive parameter adjustment mechanism, an initial triangular network is constructed, and the triangular network structure is optimized through iterative densification. An edge protection mechanism is introduced to remove non-ground noise points from vegetation and man-made structures, while retaining bare ground points. Point cloud optimization was performed using LiDAR 360 and EPS software, controlling the point cloud density to ≥50 points per square meter to ensure the accuracy of ground point extraction.

7. The three-dimensional stereoscopic investigation method for low-amplitude brittle deformation landslide areas according to claim 1, characterized in that, The generated high-resolution model includes: The preprocessed point cloud data and multi-view image data are imported into the 3D modeling platform to perform motion recovery structure 3D reconstruction and generate the initial point cloud model. By combining ArcGIS Pro's spatial analysis module, we completed joint adjustment and dense matching of multi-view images, generating a digital elevation model, a digital surface model, and orthophotos with a spatial resolution of 0.1m. Point cloud data processing software was used to classify point clouds, and a digital terrain model was constructed using a 3D modeling platform. The vertical accuracy of the digital terrain model was ±5cm, which meets the requirements of the 1:500 terrain mapping specification.

8. The three-dimensional stereoscopic investigation method for low-amplitude brittle deformation landslide areas according to claim 1, characterized in that, The deformation analysis includes: The DEM difference analysis module of the spatial analysis module is used to spatially register the DEM data from multiple periods, forcibly unify them to a 0.1-meter resolution grid system, and use the least squares surface fitting algorithm to eliminate system offset; Interference zone filtering: Based on the NDVI index and building vector layer, vegetation and artificial building interference zones are automatically filtered out. At least four bedrock exposure feature points are selected to construct a stable reference network. Among them, the NDVI index threshold is >0.6 and the building area of ​​the building vector layer is >20m². Perform pixel-level elevation difference calculations to generate millimeter-level digital elevation difference models. Solve the three-dimensional displacement vectors using the Rodriguez matrix to quantify the settlement, displacement rate, slope angle changes, and surface roughness evolution characteristics of the target area.

Images