Three-dimensional display method and system for landslide simulation and storage medium

Abstract

The invention relates to the technical field of three-dimensional simulation, in particular to a three-dimensional display method and system for landslide simulation and a storage medium. The method comprises the following steps: acquiring multi-source monitoring data of a landslide area; performing noise filtering and spatial interpolation on the multi-source monitoring data of the landslide area to obtain preprocessed landslide monitoring data; performing time-space reference unification processing on the preprocessed landslide monitoring data to obtain time-space alignment monitoring data; performing NURBS curved surface modeling based on the space-time alignment monitoring data to obtain an initial landslide three-dimensional structure model; performing finite element mesh adaptive subdivision on the initial landslide three-dimensional structure model to obtain a dynamic landslide mass mesh model; and acquiring current landslide displacement data. According to the method, the problem that updating of the three-dimensional model is delayed in a traditional method is effectively solved, and real-time performance and accuracy of landslide simulation are ensured.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of three-dimensional simulation, in particular to a landslide simulation three-dimensional display method, system and storage medium. BACKGROUND

[0002] Traditional landslide simulation display methods mostly rely on two-dimensional maps or simple three-dimensional models. These methods have obvious shortcomings in displaying the dynamic process and complex topography of landslides. For example, in mountainous road construction and maintenance work, engineers need to accurately assess the impact of potential landslides on roads in order to take preventive measures in advance. However, existing two-dimensional maps cannot intuitively display the three-dimensional spatial changes of landslides, making it difficult for engineers to accurately determine the scope and impact of landslides, thereby increasing the risk of engineering design and construction.

[0003] In practical applications, although existing three-dimensional display technologies have improved the visualization effect of landslide simulation to some extent, there are still some technical difficulties. For example, landslide monitoring systems usually need to combine real-time landslide data with three-dimensional models to dynamically display the evolution process of landslides. However, existing three-dimensional display technologies have a delay problem in data updating and model dynamic adjustment. Specifically, when the monitoring system obtains new landslide data, the update speed of the three-dimensional model often cannot keep up with the changes in the data, resulting in a deviation between the displayed landslide dynamic process and the actual situation. This delay not only affects the accuracy of disaster warning, but also misleads decision-makers and emergency response personnel, thereby delaying the implementation of rescue and protection measures. SUMMARY

[0004] Therefore, it is necessary to provide a landslide simulation three-dimensional display method, system and storage medium to solve at least one of the above technical problems.

[0005] To achieve the above purpose, a landslide simulation three-dimensional display method comprises the following steps: Step S1: Obtain multi-source monitoring data of the landslide area; perform noise filtering and spatial interpolation on the multi-source monitoring data of the landslide area to obtain preprocessed landslide monitoring data; perform spatiotemporal reference unification processing on the preprocessed landslide monitoring data to obtain spatiotemporally aligned monitoring data; Step S2: Based on the spatiotemporally aligned monitoring data, perform NURBS surface modeling to obtain an initial landslide three-dimensional structure model; perform finite element mesh adaptive subdivision on the initial landslide three-dimensional structure model to obtain a dynamic landslide body mesh model; Step S3: Obtain current landslide displacement data; perform landslide simulation based on the dynamic landslide body mesh model and the current landslide displacement data to obtain landslide motion simulation data; perform dynamic updating on the landslide motion simulation data to obtain real-time landslide evolution prediction data; Step S4: GPU accelerated rendering is performed on the real-time landslide evolution prediction data to obtain a landslide deformation visualization diagram; a heat map is superimposed on the landslide deformation visualization diagram, and a VR interactive interface is integrated to obtain an immersive landslide three-dimensional scene; Step S5: error inversion analysis is performed according to the immersive landslide three-dimensional scene and the current landslide displacement data to obtain landslide model optimization parameters; and early warning atlas generation is performed based on the landslide model optimization parameters to obtain closed-loop landslide risk early warning data.

[0006] Preferably, the present application provides a three-dimensional display system for landslide simulation, which is used to perform the three-dimensional display method for landslide simulation as described above, and the three-dimensional display system for landslide simulation comprises: A data preprocessing module is configured to acquire multi-source monitoring data of a landslide area, perform noise filtering and spatial interpolation on the multi-source monitoring data of the landslide area to obtain preprocessed landslide monitoring data, and perform spatiotemporal reference unification processing on the preprocessed landslide monitoring data to obtain spatiotemporally aligned monitoring data. A three-dimensional modeling module is configured to perform NURBS surface modeling based on the spatiotemporally aligned monitoring data to obtain an initial landslide three-dimensional structure model, and perform finite element grid adaptive subdivision on the initial landslide three-dimensional structure model to obtain a dynamic landslide body grid model. A motion simulation module is configured to acquire current landslide displacement data, perform landslide simulation according to the dynamic landslide body grid model and the current landslide displacement data to obtain landslide motion simulation data, and perform dynamic updating on the landslide motion simulation data to obtain real-time landslide evolution prediction data. A visualization rendering module is configured to perform GPU accelerated rendering on the real-time landslide evolution prediction data to obtain a landslide deformation visualization diagram, and perform heat map superposition and VR interactive interface integration on the landslide deformation visualization diagram to obtain an immersive landslide three-dimensional scene. An early warning module is configured to perform error inversion analysis according to the immersive landslide three-dimensional scene and the current landslide displacement data to obtain landslide model optimization parameters, and perform early warning atlas generation based on the landslide model optimization parameters to obtain closed-loop landslide risk early warning data.

[0007] Preferably, the present application further provides a computer readable storage medium storing a program capable of being loaded and executed by a processor to perform the three-dimensional display method for landslide simulation as described above.

[0008] The present application has the following beneficial effects: On the one hand, by acquiring multi-source monitoring data and performing noise filtering and spatial interpolation, spatiotemporal alignment of landslide area data is achieved, which provides accurate basic data for subsequent three-dimensional modeling, and effectively solves the modeling error problem caused by inconsistent data spatiotemporal reference in traditional methods.

[0009] On the other hand, by modeling based on NURBS surface and adaptive subdivision of finite element mesh, a three-dimensional structure model capable of dynamically reflecting the changes of the landslide body is generated, effectively solving the problem of delay in updating the three-dimensional model in the traditional method, and ensuring the real-time and accuracy of the landslide simulation.

[0010] On the other hand, by GPU accelerated rendering and heat map superimposition, combined with VR interactive interface integration, not only the efficiency of landslide deformation visualization is improved, but also the user's immersion is enhanced, so that the dynamic process of landslide can be more intuitively displayed, further improving the visualization effect of landslide simulation. BRIEF DESCRIPTION OF DRAWINGS

[0011] Other features, objects and advantages of the present application will become more apparent from the following detailed description when read in conjunction with the accompanying drawings: Fig. 1 A step flow diagram of a three-dimensional display method of landslide simulation of an embodiment is shown.

[0012] Fig. 2 A detailed step flow diagram of step S23 of an embodiment is shown.

[0013] Fig. 3 A three-dimensional scene display diagram of a landslide of an embodiment is shown. DETAILED DESCRIPTION

[0014] The technical method of the present application will be described in detail below in conjunction with the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0015] In addition, the accompanying drawings are only schematic illustrations of the present application, and are not necessarily drawn to scale. The same reference numerals in the drawings represent the same or similar parts, and thus repeated description thereof will be omitted. Some block diagrams shown in the drawings are functional entities, which do not necessarily correspond to physically or logically independent entities. The functional entities can be implemented in software form, or in one or more hardware modules or integrated circuits, or in different network and / or processor methods and / or microcontroller methods.

[0016] It should be understood that, although the terms "first", "second" or the like can be used herein to describe various elements, these elements should not be limited by these terms. These terms are only used to distinguish one element from another. For example, without departing from the scope of the example embodiments, a first element can be termed a second element, and similarly, a second element can be termed a first element. The term "and / or" as used herein includes any and all combinations of one or more of the associated associated items.

[0017] To achieve the above object, please refer to Figs. 1 to 3 The application provides a three-dimensional display method for landslide simulation, comprising the following steps: Step S1: acquiring multi-source monitoring data of a landslide area; performing noise filtering and spatial interpolation on the multi-source monitoring data of the landslide area to obtain preprocessed landslide monitoring data; performing time and space reference unification processing on the preprocessed landslide monitoring data to obtain time and space aligned monitoring data; Step S2: performing NURBS surface modeling based on the time and space aligned monitoring data to obtain an initial landslide three-dimensional structure model; performing finite element mesh self-adaptive subdivision on the initial landslide three-dimensional structure model to obtain a dynamic landslide body mesh model; Step S3: acquiring current landslide displacement data; performing landslide simulation according to the dynamic landslide body mesh model and the current landslide displacement data to obtain landslide motion simulation data; performing dynamic updating on the landslide motion simulation data to obtain real-time landslide evolution prediction data; Step S4: performing GPU accelerated rendering on the real-time landslide evolution prediction data to obtain a landslide deformation visualization graph; performing heat map superposition and VR interactive interface integration on the landslide deformation visualization graph to obtain an immersive landslide three-dimensional scene; Step S5: performing error inversion analysis according to the immersive landslide three-dimensional scene and the current landslide displacement data to obtain landslide model optimization parameters; generating a warning atlas based on the landslide model optimization parameters to obtain closed-loop landslide risk warning data.

[0018] Preferably, step S1 comprises: Step S11: performing three-dimensional scanning on the landslide area by a laser radar to obtain initial terrain point cloud data; In the embodiment of the application, a laser radar system, such as Riegl VZ-4000, is used to perform three-dimensional scanning on the landslide area. The system has high precision and long distance scanning capability, and can generate high-density point cloud data. During the scanning process, the laser radar is installed on a tripod, and the scanning range is set to cover the entire landslide area. By adjusting the scanning resolution parameters, such as setting the point spacing to 0.1 meters. After scanning is completed, the laser radar system outputs the initial terrain point cloud data, which contains the three-dimensional coordinate information of each scanning point.

[0019] Step S12: high-frequency noise filtering is performed on the initial terrain point cloud data to obtain denoised terrain point cloud data; and spatial completion is performed on the denoised terrain point cloud data to obtain complete terrain elevation data. In the embodiment of the present application, the initial terrain point cloud data is processed using the point cloud processing library PCL (Point Cloud Library). A statistical-based noise filtering method is adopted, the neighborhood search radius is set to 0.2 meters, the average distance of the neighborhood points of each point is calculated, and those points with a distance much greater than the average value are identified and removed, which are considered as noise points. After removing the noise, a K-nearest neighbor-based spatial interpolation method is used to perform spatial completion on the point cloud data. Specifically, for each missing elevation data point, the nearest 5 neighborhood points are found, the elevation value of the point is calculated by weighted average, and the weight is calculated according to the inverse of the distance. Through the above steps, complete and smooth terrain elevation data is generated.

[0020] Step S13: real-time landslide displacement data of the landslide area is collected through a ground displacement sensor network; and time series smoothing is performed on the real-time landslide displacement data to obtain optimized landslide displacement data. In the embodiment of the present application, a network composed of multiple GPS displacement sensors is used, and these sensors are installed at key positions of the landslide area, such as the front edge, middle and rear edge of the landslide body. Each sensor collects real-time landslide displacement data at a sampling frequency of 1 minute. A moving average filtering method is used to perform time series smoothing on the real-time landslide displacement data. Specifically, the moving average window size is set to 5 data points, the average value of the data points in each window is calculated, and the smoothed displacement data is obtained.

[0021] Step S14: multi-spectral image data of the landslide area is acquired. In the embodiment of the present application, a multi-spectral camera-equipped unmanned aerial vehicle, such as DJI Mavic 2Enterprise Advanced, is used to take aerial photographs of the landslide area. The multi-spectral camera can simultaneously collect image data of four wavebands of red light, green light, blue light and near-infrared light. During the aerial photography, the unmanned aerial vehicle takes photographs according to the preset flight path at a flight height of 10 meters and an image overlap rate of 50%. After the photographing is completed, the acquired multi-spectral image data is spliced and corrected by professional image processing software, such as Pix4Dmapper, to generate a multi-spectral image map covering the entire landslide area.

[0022] Step S15: spatial alignment is performed on the complete terrain elevation data, the optimized landslide displacement data and the multi-spectral image data to obtain spatially unified monitoring data. In the embodiment of the present application, GIS software QGIS is used to spatially align the complete terrain elevation data, optimized landslide displacement data and multispectral image data. The terrain elevation data is imported into QGIS and its coordinate system is set to WGS 84 / UTM zone 33N. The optimized landslide displacement data and multispectral image data are imported separately and their coordinate systems are unified to WGS 84 / UTM zone 33N through the coordinate conversion tool. Using the alignment tool of QGIS, the multispectral image data is spatially corrected by matching the feature points in the terrain elevation data and the ground feature points in the multispectral image data. Finally, the optimized landslide displacement data is spatially overlaid with the corrected multispectral image data to generate spatially unified monitoring data.

[0023] Step S16: Time series synchronization is performed on the spatially unified monitoring data to obtain spatio-temporally aligned monitoring data.

[0024] In the embodiment of the present application, Python programming language combined with Pandas library is used to perform time series synchronization processing on the spatially unified monitoring data. The timestamps of the terrain elevation data, optimized landslide displacement data and multispectral image data are extracted and stored in the Data Frame of Pandas. Through timestamp alignment operation, the timestamps of different data sources are unified to a common time reference. Specifically, the timestamp of the landslide displacement data is selected as the reference, and time interpolation is performed on other data sources to ensure that all data have corresponding values at the same time point. Through the above steps, spatio-temporally aligned monitoring data is generated.

[0025] Preferably, step S2 comprises: Step S21: Extracting terrain elevation feature points based on the spatio-temporally aligned monitoring data to obtain a landslide terrain feature point set; In the embodiment of the present application, GIS software QGIS is used to process the spatio-temporally aligned monitoring data. The terrain elevation data is imported into QGIS and its coordinate system is set to WGS 84 / UTM zone 33N. Using the "extract terrain feature points" tool of QGIS, the area with an elevation change rate greater than 10% is selected as the threshold for feature point extraction. By setting these parameters, QGIS can automatically identify and extract points with significant terrain elevation changes, which are usually located at the boundaries, cracks and steep slope areas of the landslide. Finally, the generated landslide terrain feature point set contains the coordinates and elevation information of these key positions.

[0026] Step S22: Three-dimensional surface modeling based on the landslide terrain feature point set to obtain an initial landslide surface model; In the embodiment of the present application, the three-dimensional modeling software Cloud Compare is used to perform three-dimensional surface modeling on the landslide topographic feature point set. The landslide topographic feature point set is imported into Cloud Compare, and the "surface reconstruction" function is selected. When setting parameters, the "Poisson reconstruction" algorithm is selected, which can generate a smooth three-dimensional surface according to point cloud data. The reconstruction resolution parameter is set to 0.5 meters. Through these operations, Cloud Compare generates an initial landslide surface model, which can intuitively show the topography of the landslide area.

[0027] Step S23: identifying the key topographic feature area of the initial landslide surface model to obtain a landslide feature area identifier; Step S24: encrypting the point cloud in the corresponding key area of the initial landslide surface model according to the landslide feature area identifier to obtain optimized landslide point cloud data; In the embodiment of the present application, the point cloud processing library PCL is used to encrypt the point cloud in the key area of the initial landslide surface model. The initial landslide surface model and the landslide feature area identifier are imported into PCL. The "voxel grid filtering" method is selected to encrypt the key area, and the voxel size is set to 0.1 meters. Finally, the generated optimized landslide point cloud data contains the encrypted key area point cloud.

[0028] Step S25: performing secondary surface reconstruction based on the optimized landslide point cloud data to obtain a landslide three-dimensional structure model; performing finite element mesh division on the landslide three-dimensional structure model to obtain a basic landslide mesh model; In the embodiment of the present application, the three-dimensional modeling software Cloud Compare is used to perform secondary surface reconstruction on the optimized landslide point cloud data. The optimized landslide point cloud data is imported into Cloud Compare, and the "surface reconstruction" function is selected. When setting parameters, the "Poisson reconstruction" algorithm is selected, and the reconstruction resolution parameter is set to 0.2 meters. Through these operations, Cloud Compare generates a landslide three-dimensional structure model. The finite element mesh division tool Gmsh is used to divide the mesh of the landslide three-dimensional structure model. The "tetrahedral mesh" type is selected, and the mesh size is set to 0.5 meters, which will not be too complex. Finally, the generated basic landslide mesh model contains the finite element mesh of the landslide area.

[0029] Step S26: encrypting the feature area of the basic landslide mesh model to obtain a dynamic landslide body mesh model.

[0030] In the embodiment of the present application, the basic landslide mesh model and the landslide feature area identifier are imported into Gmsh. The "local encryption" function is selected to encrypt the mesh of the key topographic feature area. The mesh size of the encrypted area is set to 0.2 meters. Through these operations, Gmsh generates a dynamic landslide body mesh model.

[0031] Preferably, step S23 comprises: Step S231: Perform surface curvature analysis on the initial landslide surface model to obtain landslide surface curvature distribution data; In the embodiment of the present application, the initial landslide surface model is imported into Cloud Compare. The "curvature analysis" function is selected, and the analysis radius is set to 0.5 meters to ensure that local curvature changes on the surface can be captured. Cloud Compare calculates the principal curvature value of each point and generates a curvature distribution map. The map displays areas with different curvature values in color-coded form, where high curvature areas are usually represented in red and low curvature areas are represented in blue. Finally, the generated landslide surface curvature distribution data contains the curvature value of each point.

[0032] Step S232: Extract high curvature areas in the initial landslide surface model based on the landslide surface curvature distribution data to obtain potential feature area data; In the embodiment of the present application, the landslide surface curvature distribution data is imported into PCL. The "threshold extraction" function is selected, and the curvature threshold is set to 0.1 to extract points with curvature greater than the threshold. These points are usually located in cracks, steep slopes and other areas with significant topographic changes of the landslide. Through the above steps, PCL can automatically identify and extract point cloud data of high curvature areas to generate potential feature area data. These data contain the point cloud coordinates and curvature values of high curvature areas.

[0033] Step S233: Perform edge enhancement on the potential feature area data to obtain enhanced feature edge data; and perform feature area clustering on the enhanced feature edge data to obtain a candidate feature area set; In the embodiment of the present application, the image processing library Open CV is used to perform edge enhancement and feature area clustering on the potential feature area data. The potential feature area data is converted into a grayscale image. The "edge enhancement" function of OpenCV is used, the Sobel operator is selected, and the kernel size is set to 3 to enhance the edge features in the image. The "region growing" algorithm is used to cluster the enhanced edge features to identify continuous feature areas. By setting the growth threshold to 0.05 and the minimum area to 50 square meters, Open CV can automatically identify and cluster multiple candidate feature areas. Finally, the generated candidate feature area set contains the location, range and area information of these areas.

[0034] Step S234: Based on the candidate feature area set, calculate the area and continuity index of each area to generate feature area evaluation data; In the embodiment of the present application, the candidate feature region set is imported into QGIS. Using the "Region Analysis" tool of QGIS, the area and continuity index of each region are calculated. The continuity index is evaluated by calculating the connectivity of points within the region, specifically by calculating the mean distance and standard deviation of points within the region. The area threshold is set to 100 square meters, and the continuity index threshold is set to 0.8, to filter out regions with sufficient area and good continuity. Through these operations, QGIS generates feature region evaluation data, which contains the area, continuity index and other related attributes of each candidate feature region.

[0035] Step S235: Determine the key feature region according to the feature region evaluation data, and obtain the landslide feature region identification.

[0036] In the embodiment of the present application, the feature region evaluation data is imported into the Data Frame of Pandas. By setting the filtering conditions, regions with an area greater than 100 square meters and a continuity index greater than 0.8 are selected as key feature regions. These regions are located at the boundaries, cracks and steep slope areas of the landslide, and have an important indication effect on the dynamic changes of the landslide. Finally, the generated landslide feature region identification contains the position, range and attribute information of these key regions.

[0037] Preferably, step S3 comprises: Step S31: Collect current landslide displacement data through the ground displacement sensor network; In the embodiment of the present application, a network composed of multiple high-precision GPS displacement sensors is used. These sensors are installed at key positions in the landslide area, such as the front edge, middle and rear edge of the landslide body. Each sensor collects real-time landslide displacement data at a sampling frequency of 1 minute. The sensors transmit data to the central data collection system through a wireless communication module, which automatically synchronizes data every 10 minutes to ensure real-time and integrity of the data. Through the above steps, real-time displacement data of the landslide area can be obtained.

[0038] Step S32: Data denoising is performed on the current landslide displacement data to obtain optimized current displacement data; In the embodiment of the present application, the collected displacement data is imported into the Data Frame of Pandas for data operation. The scipy.signal.savgol_filter function in the SciPy library is used to perform Savitzky-Golay filtering on the displacement data. The filter window length is set to 11 data points, and the polynomial order is set to 2 to smooth short-term fluctuations and noise in the data. Through the above steps, random noise in the data can be effectively removed, and the long-term trend of landslide displacement can be retained, thereby obtaining optimized current displacement data.

[0039] Step S33: Extracting grid node mechanical parameters of the dynamic landslide grid model to obtain landslide mechanical property data; In the embodiment of the present application, the dynamic landslide grid model is imported into Gmsh. The mechanical parameters of each grid node are extracted using the "grid node attribute extraction" function of Gmsh, including elastic modulus, Poisson's ratio, and density, etc. These parameters can be set according to the geological material properties of the landslide. For example, for a typical clay landslide, the elastic modulus is set to 10 MPa, the Poisson's ratio is set to 0.3, and the density is set to 1.8 g / cm³. Through these operations, Gmsh can generate a data file containing the mechanical parameters of each grid node.

[0040] Step S34: Based on the optimized current displacement data and the landslide mechanical property data, the kinematic equation is solved by the finite element method to obtain preliminary landslide motion field data; In the embodiment of the present application, the optimized current displacement data and the landslide mechanical property data are imported into Open FOAM. The "kinematic equation solver" of Open FOAM is used to set the time step to 0.1 seconds and the total simulation time to 100 seconds. The kinematic equation is solved by the finite element method to calculate the displacement, velocity, and acceleration of each grid node at different time steps. Finally, the generated preliminary landslide motion field data contains the dynamic change information of the landslide within the simulation time.

[0041] Step S35: Particle flow simulation based on the preliminary landslide motion field data to obtain landslide motion trajectory data; In the embodiment of the present application, the particle flow simulation software YADE (Yet Another Discrete Element library) is used to perform particle flow simulation on the preliminary landslide motion field data. The preliminary landslide motion field data is imported into YADE. The physical properties of the particles, such as diameter, density, and friction coefficient, are set. For example, for typical landslide particles, the diameter is set to 0.1 meters, the density is set to 2.5 g / cm³, and the friction coefficient is set to 0.5. The "particle flow simulator" of YADE is used to set the simulation time to 100 seconds and the time step to 0.01 seconds. Through simulation, the position and motion trajectory of each particle at different time steps are calculated. Finally, the generated landslide motion trajectory data can intuitively show the dynamic evolution process of the landslide.

[0042] Step S36: Based on the landslide motion trajectory data, dynamic extrapolation is performed to obtain real-time landslide evolution prediction data.

[0043] Preferably, step S36 includes: Step S361: Extracting time series displacement features based on the landslide motion trajectory data to obtain a landslide displacement time series feature set; In the embodiment of the present application, the landslide body motion trajectory data is imported into the Data Frame of Pandas, and the data includes the position coordinates of each particle at different time points. The displacement of each particle at each time step is calculated, and the Euclidean distance between the positions of adjacent time points is obtained. The displacement features of each particle are extracted, including the maximum displacement, the average displacement, and the standard deviation of the displacement. These features can reflect the dynamic changes of the landslide body at different time points. Finally, the generated landslide displacement time series feature set contains the displacement features of each particle at different time steps.

[0044] Step S362: short-term displacement prediction based on the landslide displacement time series feature set to obtain preliminary displacement prediction data; In the embodiment of the present application, the landslide displacement time series feature set is imported into the Data Frame of Pandas, and the first 10 data points of the time series are selected as training data and the last 5 data points are selected as test data. A linear regression model is used to predict the displacement of each particle, and the input features of the model include the time step, the maximum displacement, and the average displacement. Through training the model, the displacement in the future several time steps is predicted. Finally, the generated preliminary displacement prediction data contains the predicted displacement of each particle at the future time points.

[0045] Step S363: medium and long-term trend prediction according to the preliminary displacement prediction data to obtain landslide motion trend prediction data; In the embodiment of the present application, the preliminary displacement prediction data is imported into the Data Frame of Pandas, and the time step is set as the independent variable and the displacement is set as the dependent variable. The Prophet model is used to predict the displacement trend of each particle, and the prediction time range is set to the future 30 time steps. The seasonal parameters and trend parameters of the model are adjusted. Finally, the generated landslide motion trend prediction data contains the displacement trend of each particle in the future 30 time steps.

[0046] Step S364: prediction result correction on the landslide motion trend prediction data to obtain optimized displacement prediction data; In the embodiment of the present application, the landslide motion trend prediction data is imported into the Data Frame of Pandas. By comparing the differences between the prediction data and the actual observation data, the error of each time step is calculated. The moving average method is used to smooth the error, and the moving average window size is set to 5 time steps. By adjusting the prediction data, the influence of the error is reduced to obtain the optimized displacement prediction data. Finally, the generated optimized displacement prediction data is closer to the actual observation value.

[0047] Step S365: full-area displacement field reconstruction based on the optimized displacement prediction data to obtain landslide body displacement field prediction data; In the embodiment of the present application, the optimized displacement prediction data is imported into the Data Frame of Pandas. The interpolation function of NumPy is used to reconstruct the displacement field of the entire area. The interpolation method is set to cubic spline interpolation, and the displacement field of the entire landslide area at each time step is calculated through the known particle displacement data. Finally, the generated landslide displacement field prediction data can intuitively show the displacement distribution of the landslide at different time points.

[0048] Step S366: Calculate the motion speed of each grid point in the dynamic landslide grid model based on the landslide displacement field prediction data to obtain landslide velocity field prediction data; In the embodiment of the present application, the landslide displacement field prediction data is imported into the Data Frame of Pandas. The speed of each grid point in the adjacent time step is calculated by calculating the displacement difference of each grid point in the adjacent time step. The time step is set to 0.1 seconds, and the speed value of each grid point is obtained by dividing the displacement difference by the time step. Finally, the generated landslide velocity field prediction data contains the speed information of each grid point at different time steps.

[0049] Step S367: Perform spatio-temporal feature fusion on the landslide velocity field prediction data to obtain real-time landslide evolution prediction data.

[0050] In the embodiment of the present application, the landslide velocity field prediction data is imported into the Data Frame of Pandas. The spatio-temporal features of the velocity field, such as average speed, maximum speed and standard deviation of speed, are obtained by calculating the statistical features of the velocity field in each time step. The features are visualized using the Matplotlib library to generate the spatio-temporal distribution map of the velocity field.

[0051] Preferably, step S4 comprises: Step S41: Extract the displacement amount of each grid vertex of the real-time landslide evolution prediction data to obtain landslide deformation feature data; In the embodiment of the present application, the real-time landslide evolution prediction data is imported into the Data Frame of Pandas, and these data contain the displacement information of each grid point at different time steps. The displacement amount of each grid vertex is extracted by calculating the difference of the vertex position in adjacent time steps. Specifically, the time step is set to 0.1 seconds, and the landslide deformation feature data is obtained by calculating the displacement vector of each vertex at each time step.

[0052] Step S42: Update the vertex coordinates of the dynamic landslide grid model according to the landslide deformation feature data to obtain an updated landslide grid model; In the embodiment of the present application, the landslide deformation feature data is imported into Blender, which contains the displacement vector of each mesh vertex. Using the "vertex editing" function of Blender, the coordinates of each vertex are updated according to the displacement vector. Specifically, select the "vertex translation" tool, input the displacement vector of each vertex, and update the vertex position one by one. Through the above steps, Blender generates the updated landslide grid model.

[0053] Step S43: surface coloring is performed on the updated landslide grid model to obtain a landslide basic three-dimensional rendering map; In the embodiment of the present application, Blender is used to perform surface coloring on the updated landslide grid model. The updated landslide grid model is imported into Blender. Select the "material editing" function to add coloring materials to the model surface. Set the color mapping of the material, and gradually change the color from blue (small displacement) to red (large displacement) according to the size of the displacement. Specifically, use the "color gradient" tool of Blender to set the color mapping range to 0 to 10 meters of displacement. Through the above steps, Blender generates a landslide basic three-dimensional rendering map, which intuitively shows the deformation degree of the landslide body.

[0054] Step S44: real-time deformation animation rendering is performed according to the landslide basic three-dimensional rendering map to obtain landslide dynamic visualization data; In the embodiment of the present application, the landslide basic three-dimensional rendering map is imported into Blender. Using the "animation rendering" function of Blender, set the time range of the animation to 0 to 100 seconds, and the frame rate to 30 frames per second. By updating the vertex coordinates of the grid model frame by frame, the dynamic deformation process of the landslide body is generated. Specifically, use the "keyframe animation" tool of Blender to set keyframes for the grid model at each time step, and automatically generate transition animations for intermediate frames. Finally, Blender generates landslide dynamic visualization data.

[0055] Step S45: convert the displacement amount in the landslide dynamic visualization data into a thermal value to obtain thermal map superimposed data; In the embodiment of the present application, the landslide dynamic visualization data is imported into the Data Frame of Pandas, which contains the displacement information of the grid points at each time step. Using the imshow function of Matplotlib, the displacement amount is mapped to the color range of the thermal map. Set the color mapping to "viridis", and map the displacement amount from 0 to 10 meters to the color range. Through the above steps, thermal map superimposed data is generated.

[0056] Step S46: construct a virtual scene according to the thermal map superimposed data to obtain VR scene basic data; In the embodiment of the present application, the heat map superimposed data is imported into Unity3D, which contains the displacement distribution information of the landslide body. Using the "terrain editing" function of Unity3D, a virtual terrain is constructed according to the heat map data. The resolution and height range of the terrain are set, and the height and color of the terrain are adjusted according to the displacement. Specifically, using the "terrain texture" tool of Unity3D, the heat map data is mapped to the terrain surface to generate a virtual scene with heat map effect. Finally, Unity3D generates the VR scene basic data.

[0057] Step S47: Based on the VR scene basic data, gesture recognition and head tracking interaction function integration are carried out to obtain an immersive landslide three-dimensional scene.

[0058] In the embodiment of the present application, the VR scene basic data is imported into Unity3D. The Leap Motion plug-in is installed and configured for gesture recognition. The recognition range and sensitivity parameters of Leap Motion are set to ensure accurate recognition of user gestures. Then, the Oculus Rift plug-in is installed and configured for head tracking. The tracking range and refresh rate of Oculus Rift are set to ensure real-time tracking of the user's head position and direction. By integrating these devices, users can interact with the virtual scene through gestures and head movements to generate an immersive landslide three-dimensional scene.

[0059] Preferably, step S5 comprises: Step S51: Visual model displacement feature extraction is performed on the immersive landslide three-dimensional scene to obtain model predicted displacement data. In the embodiment of the present application, the immersive landslide three-dimensional scene is imported into Unity3D. Using the "model analysis" tool of Unity3D, the displacement features of each mesh vertex in the scene are extracted. Specifically, the analysis time step is set to 0.1 seconds, and the displacement vector of each time step is extracted. Through the above steps, Unity3D generates model predicted displacement data, which contains the displacement information of each mesh vertex at different time steps.

[0060] Step S52: Standardization processing is performed on the current landslide displacement data to obtain standard current displacement data. In the embodiment of the present application, the current landslide displacement data is imported into the Data Frame of Pandas. Using the scipy.stats.zscore function in the SciPy library, the displacement data is standardized. Specifically, the Z-score of each displacement data point is calculated, which is the difference between the data point and the average value divided by the standard deviation. Through the above steps, the displacement data is converted to standard normal distribution to obtain standard current displacement data. The mean of these data is 0 and the standard deviation is 1.

[0061] Step S53: error calculation is performed based on the model predicted displacement data and the standard current displacement data to obtain displacement error distribution data; In the embodiment of the present application, the model predicted displacement data and the standard current displacement data are imported into the DataFrame of Pandas. The difference between the predicted displacement and the actual displacement of each grid vertex in each time step is calculated to obtain the displacement error. Specifically, the Data Frame.subtract method of Pandas is used to calculate the point-by-point difference of the two data sets. Through the above steps, the displacement error distribution data reflecting the deviation between the model prediction and the actual observation is generated.

[0062] Step S54: model parameter inversion is performed on the dynamic landslide grid model according to the displacement error distribution data to obtain preliminary model correction parameters; In the embodiment of the present application, the displacement error distribution data is imported into the Data Frame of Pandas. The displacement error is used as the dependent variable and the parameters (such as elastic modulus and Poisson's ratio) of the grid model are used as the independent variable to perform regression analysis using a linear regression model. Specifically, the training data of the model is set to the first 80% of the time step data and the test data is set to the last 20% of the time step data. Through the training model, preliminary model correction parameters are obtained.

[0063] Step S55: parameter screening is performed on the preliminary model correction parameters to obtain an optimized model parameter set; and a preliminary warning map is obtained by generating a warning color scale according to the optimized model parameter set; In the embodiment of the present application, the preliminary model correction parameters are imported into the Data Frame of Pandas. According to the significance level (such as p value less than 0.05) and the error reduction amount, the effective correction parameters are screened out. Specifically, the Data Frame.query method of Pandas is used to screen out the parameters meeting the conditions. According to the optimized model parameter set, a warning color scale is generated. The displacement error is mapped to a color range using the color map function of Matplotlib to generate a preliminary warning map.

[0064] Step S56: global risk field reconstruction is performed based on the preliminary warning map to obtain closed-loop landslide risk warning data.

[0065] In the embodiment of the present application, the preliminary warning map is imported into the Data Frame of Pandas. The risk field of the entire landslide area is reconstructed using the interpolation function of NumPy. Specifically, the interpolation method is set to cubic spline interpolation, and the risk distribution of the entire area is calculated through the known risk point data. Through the above steps, the closed-loop landslide risk warning data is generated.

[0066] Therefore, the embodiments should be regarded, at any point, as being exemplary and not limiting, the scope of the application being defined by the appended claims and not by the above description, and all changes which come within the meaning and range of equivalency of the claims are therefore intended to be embraced therein.

[0067] The foregoing is considered as illustrative only of the principles of the application. Numerous modifications and changes will readily occur to those skilled in the art, and the generic principles defined herein can be applied to other embodiments without departing from the spirit or scope of the application. Therefore, the application is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A three-dimensional display method for landslide simulation, characterized in that, Includes the following steps: Step S1: Obtain multi-source monitoring data of the landslide area; Noise filtering and spatial interpolation were performed on the multi-source monitoring data of the landslide area to obtain preprocessed landslide monitoring data; The pre-processed landslide monitoring data is subjected to spatiotemporal benchmark unification processing to obtain spatiotemporally aligned monitoring data; Step S2: Perform NURBS surface modeling based on spatiotemporal aligned monitoring data to obtain the initial three-dimensional structure model of the landslide; The initial three-dimensional landslide structure model was adaptively meshed using the finite element method to obtain a dynamic landslide mesh model. Step S3: Obtain current landslide displacement data; perform landslide simulation based on the dynamic landslide mesh model and the current landslide displacement data to obtain landslide motion simulation data; dynamically update the landslide motion simulation data to obtain real-time landslide evolution prediction data; Step S4: GPU-accelerated rendering of real-time landslide evolution prediction data to obtain a landslide deformation visualization map; heat map overlay and VR interactive interface integration of the landslide deformation visualization map to obtain an immersive 3D landslide scene. Step S5: Perform error inversion analysis based on the immersive 3D landslide scene and the current landslide displacement data to obtain the optimized parameters of the landslide model; Early warning maps are generated based on optimized parameters of the landslide model, resulting in closed-loop landslide risk early warning data.

2. The three-dimensional display method for landslide simulation according to claim 1, characterized in that, Step S1 includes: Step S11: Perform a three-dimensional scan of the landslide area using lidar to obtain initial terrain point cloud data; Step S12: High-frequency noise is filtered out from the initial terrain point cloud data to obtain denoised terrain point cloud data; spatial completion is performed on the denoised terrain point cloud data to obtain complete terrain elevation data. Step S13: Collect real-time landslide displacement data in the landslide area through a ground displacement sensor network; perform time-series smoothing on the real-time landslide displacement data to obtain optimized landslide displacement data; Step S14: Acquire multispectral image data of the landslide area; Step S15: Spatially align the complete terrain elevation data, optimized landslide displacement data, and multispectral image data to obtain spatially unified monitoring data; Step S16: Perform time-series synchronization on the unified spatial monitoring data to obtain spatiotemporally aligned monitoring data.

3. The three-dimensional display method for landslide simulation according to claim 1, characterized in that, Step S2 includes: Step S21: Extract terrain elevation feature points based on spatiotemporal aligned monitoring data to obtain a landslide terrain feature point set; Step S22: Perform three-dimensional surface modeling based on the landslide topographic feature point set to obtain the initial landslide surface model; Step S23: Identify key terrain feature regions in the initial landslide surface model to obtain landslide feature region identifiers; Step S24: Based on the landslide feature area identifier, densify the point cloud of the corresponding key areas in the initial landslide surface model to obtain optimized landslide point cloud data; Step S25: Reconstruct the quadratic surface based on the optimized landslide point cloud data to obtain the three-dimensional structure model of the landslide; perform finite element mesh generation on the three-dimensional structure model of the landslide to obtain the basic landslide mesh model; Step S26: Refine the feature region of the basic landslide mesh model to obtain the dynamic landslide body mesh model.

4. The three-dimensional display method for landslide simulation according to claim 3, characterized in that, Step S23 includes: Step S231: Perform surface curvature analysis on the initial landslide surface model to obtain landslide surface curvature distribution data; Step S232: Extract the high curvature region from the initial landslide surface model based on the landslide surface curvature distribution data to obtain potential feature region data; Step S233: Perform edge enhancement on the potential feature region data to obtain enhanced feature edge data; perform feature region clustering on the enhanced feature edge data to obtain a candidate feature region set; Step S234: Based on the candidate feature region set, calculate the area and continuity index of each region to generate feature region evaluation data; Step S235: Determine key feature areas based on feature area assessment data to obtain landslide feature area identifiers.

5. The three-dimensional display method for landslide simulation according to claim 1, characterized in that, Step S3 includes: Step S31: Collect current landslide displacement data through a ground displacement sensor network; Step S32: Denoise the current landslide displacement data to obtain optimized current displacement data; Step S33: Extract the mechanical parameters of the grid nodes of the dynamic landslide mesh model to obtain the mechanical property data of the landslide body; Step S34: Based on the optimized current displacement data and landslide mechanical property data, the kinematic equations are solved using the finite element method to obtain preliminary landslide motion field data; Step S35: Perform particle flow simulation based on preliminary landslide motion field data to obtain landslide motion trajectory data; Step S36: Perform dynamic extrapolation based on the landslide trajectory data to obtain real-time landslide evolution prediction data.

6. The three-dimensional display method for landslide simulation according to claim 5, characterized in that, Step S36 includes: Step S361: Extract time-series displacement features based on landslide motion trajectory data to obtain a landslide displacement time-series feature set; Step S362: Perform short-term displacement prediction based on the landslide displacement time series feature set to obtain preliminary displacement prediction data; Step S363: Based on the preliminary displacement prediction data, perform medium- and long-term trend prediction to obtain landslide movement trend prediction data; Step S364: Correct the prediction results of the landslide movement trend prediction data to obtain optimized displacement prediction data; Step S365: Reconstruct the displacement field of the entire region based on the optimized displacement prediction data to obtain the landslide displacement field prediction data; Step S366: Calculate the velocity of each grid point in the dynamic landslide mesh model based on the landslide displacement field prediction data to obtain the landslide velocity field prediction data; Step S367: Perform spatiotemporal feature fusion on the landslide velocity field prediction data to obtain real-time landslide evolution prediction data.

7. The three-dimensional display method for landslide simulation according to claim 1, characterized in that, Step S4 includes: Step S41: Extract the displacement of the three-dimensional grid vertices from the real-time landslide evolution prediction data to obtain the landslide deformation characteristic data; Step S42: Update the vertex coordinates of the dynamic landslide mesh model based on the landslide deformation characteristic data to obtain the updated landslide mesh model; Step S43: Apply surface coloring to the updated landslide mesh model to obtain a 3D rendering of the landslide foundation; Step S44: Render a real-time deformation animation based on the 3D rendering of the landslide foundation to obtain dynamic visualization data of the landslide; Step S45: Convert the displacement data in the landslide dynamic visualization data into thermal values ​​to obtain the overlay data of the thermal map; Step S46: Construct a virtual scene based on the heatmap overlay data to obtain the basic data for the VR scene; Step S47: Integrate gesture recognition and head tracking interaction functions based on VR scene basic data to obtain an immersive landslide 3D scene.

8. The three-dimensional display method for landslide simulation according to claim 1, characterized in that, Step S5 includes: Step S51: Extract displacement features from the immersive 3D landslide scene visualization model to obtain the model's predicted displacement data; Step S52: Standardize the current landslide displacement data to obtain standard current displacement data; Step S53: Calculate the error between the model-predicted displacement data and the standard current displacement data to obtain the displacement error distribution data; Step S54: Perform model parameter inversion on the dynamic landslide mesh model based on the displacement error distribution data to obtain preliminary model correction parameters; Step S55: Filter the parameters of the preliminary model correction to obtain the optimized model parameter set; generate early warning color marks based on the optimized model parameter set to obtain the preliminary early warning map; Step S56: Reconstruct the entire risk field based on the preliminary early warning map to obtain closed-loop landslide risk early warning data.

9. A three-dimensional display system for landslide simulation, characterized in that, The three-dimensional display system for landslide simulation as described in claim 1, used to perform the landslide simulation three-dimensional display method, comprises: The data preprocessing module is used to acquire multi-source monitoring data of the landslide area; perform noise filtering and spatial interpolation on the multi-source monitoring data of the landslide area to obtain preprocessed landslide monitoring data; and perform spatiotemporal benchmark unification processing on the preprocessed landslide monitoring data to obtain spatiotemporally aligned monitoring data. The 3D modeling module is used to perform NURBS surface modeling based on spatiotemporal aligned monitoring data to obtain an initial 3D landslide structure model; the initial 3D landslide structure model is then subjected to adaptive finite element mesh generation to obtain a dynamic landslide body mesh model. The motion simulation module is used to acquire current landslide displacement data; perform landslide simulation based on the dynamic landslide mesh model and the current landslide displacement data to obtain landslide motion simulation data; and dynamically update the landslide motion simulation data to obtain real-time landslide evolution prediction data. The visualization rendering module is used to perform GPU-accelerated rendering of real-time landslide evolution prediction data to obtain a landslide deformation visualization map; the landslide deformation visualization map is overlaid with a heat map and integrated with a VR interactive interface to obtain an immersive 3D landslide scene. The early warning module is used to perform error inversion analysis based on the immersive 3D landslide scene and the current landslide displacement data to obtain optimized parameters for the landslide model; and to generate early warning maps based on the optimized parameters for the landslide model to obtain closed-loop landslide risk early warning data.

10. A computer-readable storage medium, characterized in that, The program contains a program capable of being loaded by a processor and executed as described in any one of claims 1 to 8 for the three-dimensional representation of landslide simulation.

Images