A data-driven monitoring method and system for intelligent modeling and analysis of a slope

Abstract

This invention discloses a monitoring data-driven intelligent slope modeling and analysis method and system, relating to the field of intelligent monitoring, analysis, and control of major slopes in water conservancy and hydropower projects. The method includes: extracting key control points of the slope from collected slope morphology information; generating grid node data based on an AI model generated from slope grid nodes; using the grid node data as a foundation, dividing the node data into grid cells using the Delaunay triangulation algorithm to generate a two-dimensional grid model, and then extending it to generate a three-dimensional grid model; and performing slope mechanical stability, deformation, and safety analysis based on the grid data output from the two-dimensional and three-dimensional network models. This invention integrates multi-source monitoring data with AI generation technology to achieve fully automated and lightweight slope grid division, significantly improving the timeliness of slope safety monitoring and analysis, and providing an efficient computational means for slope safety analysis and treatment in water conservancy and hydropower projects.

Description

Technical Field

[0001] This invention relates to the field of intelligent monitoring, analysis and control of major slopes in water conservancy and hydropower projects. Specifically, it relates to a monitoring data-driven intelligent modeling and analysis method and system for slopes. Background Technology

[0002] Slope safety monitoring and analysis is a crucial link in the prevention of geological disasters in water conservancy and hydropower projects. With the development and progress of science and technology, slope safety monitoring and analysis has entered the era of digitalization and intelligence. The rapid development of multi-source monitoring technologies such as laser point cloud, UAV imagery, and satellite remote sensing has made real-time data acquisition possible, while also providing massive amounts of large-scale, high-precision data for slope monitoring and safety analysis. This poses new challenges to traditional slope analysis methods and working modes, and issues such as efficiency and adaptability have become new research focuses for slope analysis methods.

[0003] Currently, slope analysis methods mainly rely on numerical simulation tools such as the finite element method, finite difference method, and discrete element method, which typically require digital models. Digital slope modeling usually involves multiple steps, including geometric modeling, shape correction, and mesh generation, based on prior monitoring data, requiring significant manual intervention and fine-tuning. Especially for major slopes with complex terrain and rock structures, repeated adjustments and modeling lead to inefficiency and introduce various errors, resulting in low efficiency and accuracy in model analysis, significantly prolonging the overall assessment cycle and hindering efficient response in slope safety monitoring and analysis. Since slope safety control requires timely safety analysis results before disasters occur, the aforementioned cumbersome traditional methods are insufficient to meet this need.

[0004] Therefore, how to automatically and scientifically integrate massive amounts of slope monitoring data with real-time on-site monitoring data and convert them into high-quality grid models suitable for engineering analysis is a key technical issue in the field of slope safety monitoring and analysis. Summary of the Invention

[0005] To address the aforementioned issues, the present invention aims to provide a data-driven intelligent slope modeling and analysis technology, which effectively improves the overall level of slope safety monitoring and early warning.

[0006] To achieve the above technical objectives, this application provides a monitoring data-driven intelligent slope modeling and analysis method, comprising the following steps:

[0007] Key control points of the slope are extracted by collecting slope morphology information, and grid node data is generated based on the AI ​​model generated from the slope grid nodes.

[0008] Based on grid node data, using two-dimensional node data as the foundation, the Delaunay triangulation algorithm is used to divide the node data into grid cells to generate a two-dimensional grid model, and then extended to generate a three-dimensional grid model.

[0009] Based on the mesh data output from the two-dimensional and three-dimensional network models, the mechanical stability, deformation, and safety of the slope are analyzed.

[0010] Preferably, in the process of acquiring the AI ​​model, a generative adversarial network structure is adopted. During adversarial training, the model learns to generate complete grid node data under given control point conditions to build the AI ​​model. During training, constraints such as the minimum spacing between nodes and the angle of the line connecting adjacent nodes are added to the loss function.

[0011] Preferably, when identifying key control points on a slope, the collected slope morphology data and height information are subjected to noise filtering, registration, and normalization to obtain slope morphology information. Through edge detection, curve fitting, and point cloud segmentation algorithms, combined with geological survey data, key control points of the slope's external contour and internal geological stratification lines are extracted.

[0012] Preferably, when extracting key control points of the slope, the RANSAC algorithm is used to verify the robustness of the control point data.

[0013] Preferably, when dividing the two-dimensional grid into units, each grid unit consists of three nodes, and each unit and its contained nodes are numbered to form a correspondence between node number and coordinate data, and between unit number and node number.

[0014] Preferably, when expanding and generating a three-dimensional mesh model, new node data is generated by sampling at equal intervals along a preset sweep path, or by rotating and sampling around a given axis to generate three-dimensional nodes; then, the three-dimensional node data is divided into units according to the sweep or rotation path to form a complete three-dimensional mesh model.

[0015] Preferably, when performing slope mechanical stability, deformation and safety analysis, the slope mechanical stability, deformation and safety analysis is performed based on the mesh data output by the three-dimensional network model through finite element analysis or finite difference analysis.

[0016] This invention discloses a monitoring data-driven intelligent slope modeling and analysis system, comprising:

[0017] The slope grid node generation AI model training module uses generative adversarial networks to train a slope grid node generation AI model that meets the requirements.

[0018] The data acquisition module is used to collect multi-source monitoring data such as laser point clouds, UAV images, and satellite data to obtain the real-time morphology of the slope;

[0019] The control point extraction module is used to extract important geological profile control points of the slope based on the real-time morphology and height information of the slope, forming node coordinate data on the outer contour and layer lines of the slope; the node automatic generation module uses pre-trained slope grid nodes to generate an AI model, generates complete grid node coordinate data based on the control point data, and numbers each node.

[0020] The triangulation module uses the Delaunay triangulation algorithm to divide the generated node data into cells and generate a correspondence between cell numbers and the numbers of the nodes they contain.

[0021] The data output module is used to output node numbers and coordinate data, as well as unit numbers and the numbers of the nodes they contain.

[0022] The three-dimensional expansion module is used to generate three-dimensional mesh data from the two-dimensional mesh data through sweeping and rotation expansion, and to perform corresponding node and cell division;

[0023] The calculation and analysis module inputs the output data into the calculation software via scripts for analysis and calculation.

[0024] The present invention discloses the following technical effects:

[0025] 1. This invention, by employing generative adversarial networks and a fully automated modeling method, achieves rapid conversion from multi-source real-time monitoring data to a high-quality slope mesh model, which can simulate the dynamic response of slopes under different working conditions, thereby improving the real-time performance of slope safety assessment and early warning, and providing an efficient computational means for slope stability analysis.

[0026] 2. This invention provides an automated slope modeling method based on monitoring data. This method uses the external contour of the slope and key control points on the layer lines as input conditions. The slope grid node generation AI model trained by the conditional generative adversarial network automatically generates complete grid nodes, which can fully reflect the actual shape of the slope and effectively avoid the subjectivity and error accumulation in traditional manual modeling.

[0027] 3. The modeling method provided by this invention realizes fully automated processing of two-dimensional mesh generation and three-dimensional mesh expansion. It not only greatly reduces the time required for manual intervention and cumbersome operations of traditional commercial software, but also significantly reduces the consumption of computing resources and engineering costs, providing reliable technical support for slope safety monitoring and emergency response.

[0028] 4. By bypassing the limitations of traditional bulky software, this invention achieves a lightweight and modular system architecture, fully integrating multi-source monitoring information such as laser point clouds, UAV imagery, and satellite data, enabling automated slope modeling and subsequent engineering analysis to be completed in a short time, thereby effectively improving the overall efficiency of real-time slope condition monitoring and early warning. Attached Figure Description

[0029] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0030] Figure 1 This is a schematic diagram of the method described in this invention;

[0031] Figure 2 This is a schematic diagram of the training method for the AI ​​model for generating slope grid nodes as described in this invention;

[0032] Figure 3 This is a sample image of the automatic modeling results described in this invention in post-processing calculation software;

[0033] Figure 4 This is a schematic diagram of the usage process described in this invention. Detailed Implementation

[0034] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. The components of the embodiments of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely represents selected embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0035] like Figures 1-4As shown, this invention provides a data-driven intelligent slope modeling and analysis method. This method achieves accurate acquisition of real-time slope morphology by fusing multi-source monitoring data; it uses key control points on the slope's external contour and layer lines as input conditions, and employs a Generative Adversarial Network (GAN) to automatically generate complete grid node data; it then combines the Delaunay triangulation algorithm to achieve two-dimensional grid cell division, and can extend it to generate a three-dimensional grid model through sweep and rotation techniques; finally, the generated node and cell data are output for engineering analysis software to perform slope mechanical stability and safety analysis. Specifically, it includes the following: training the slope grid node generation AI model using a generative adversarial network structure, learning to generate complete grid node data under given control point conditions during adversarial training. To ensure the engineering applicability of the generated results, constraints such as minimum spacing between nodes and the angle of adjacent node connections are added to the loss function during training to ensure that the generated nodes have good geometric properties.

[0036] Multi-source monitoring data acquisition and preprocessing: Real-time slope morphology information is collected using multi-source monitoring methods such as laser point cloud, UAV imagery, and satellite data. The collected data is then subjected to noise filtering, registration, and normalization to ensure accurate and consistent slope morphology data.

[0037] Key control points of slopes are extracted based on slope morphology data and height information from monitoring data. Image and point cloud processing algorithms are used to extract important geological profile control points of the slope, namely nodes on the outer contour and layer lines of the slope. Their coordinates are represented in the form of (x,y) and are normalized as necessary to serve as input conditions for subsequent steps.

[0038] The grid nodes are automatically generated, and the control point coordinates are processed and fused, serving as input conditional information into a pre-trained slope grid node generation AI model. The grid node data generated by the generator is simultaneously numbered to form a complete two-dimensional node set.

[0039] Two-dimensional mesh generation is performed using automatically generated node data and the Delaunay triangulation algorithm, automatically generating two-dimensional mesh elements containing triangular units. Each element consists of three nodes, and each element and its contained nodes are numbered, forming a correspondence between node numbers and coordinate data, and between element numbers and node numbers.

[0040] In engineering applications, such as simple 3D slope analysis, the acquired 2D mesh model can be expanded using sweep and rotation techniques. Specifically, new node data is generated by sampling at equal intervals along a preset sweep path, or by rotating and sampling around a given axis to generate 3D nodes. Subsequently, the 3D node data is divided into cells according to the sweep and rotation paths to form a complete 3D mesh model.

[0041] Data output and engineering applications: The generated 2D and 3D mesh data is output, including node numbers and coordinate information, as well as element numbers and the numbers of the nodes they contain. The output data can be directly imported into finite element and finite difference analysis software for slope mechanical stability, deformation, and safety analysis, enabling real-time monitoring and early warning.

[0042] Multi-source monitoring data acquisition and preprocessing, including but not limited to laser point clouds, UAV imagery, and satellite data.

[0043] The training of the AI ​​model for generating slope grid nodes employs methods that impose constraints on the loss function, including but not limited to adding constraints on the minimum spacing between nodes and the angle of the connection between adjacent nodes.

[0044] The training network used for the AI ​​model that generates slope grid nodes includes, but is not limited to, traditional GAN ​​networks, as well as other GAN network variants used to optimize training.

[0045] The training of the AI ​​model for generating slope grid nodes uses a generator structure that includes, but is not limited to, structures based on convolutional neural networks (CNN) or graph neural networks (GNN).

[0046] Three-dimensional mesh expansion methods include, but are not limited to, sweeping and rotation.

[0047] The output data can be directly imported into engineering analysis software, including but not limited to finite element analysis and finite difference analysis software.

[0048] This invention provides a training method for an AI model for generating slope grid nodes, comprising:

[0049] The GAN training dataset is constructed by creating a slope grid node generation dataset x, which contains the node coordinates of real and virtual slopes after gridding in commercial software.

[0050] The GAN network generator is constructed by generating a generator G, whose function is to generate grid node coordinates based on the input random slope profile points z.

[0051] The GAN network discriminator is constructed by comparing the generator's output G(z) with the dataset x to train the discriminator D.

[0052] The optimization and iteration of the GAN network generator involves backpropagating the discrimination results of the discriminator D to the generator G to achieve generator iteration.

[0053] The GAN network discriminator is optimized and iterated by repeating the construction process of the GAN network discriminator to achieve the iteration of discriminator D.

[0054] The AI ​​model for generating slope grid nodes was trained through several adversarial iterations to obtain a suitable AI model Gn for generating slope grid nodes.

[0055] Example: This invention provides a monitoring data-driven intelligent slope modeling and analysis method, comprising the following steps:

[0056] Step 1: AI model training for slope grid node generation;

[0057] This invention utilizes Generative Adversarial Network (GAN) technology.

[0058] a) Construct a slope grid node generation dataset x, which contains the node coordinates of real and virtual slopes after gridding in commercial software. The commercial software used in this embodiment of the invention is ANSYS, in which the real and virtual slopes are subjected to batch gridding operations, and the node coordinates after gridding are output in batches.

[0059] b) Construct a generator G, whose function is to generate grid node coordinates based on the input random slope contour points z. In this embodiment of the invention, Python programming is used to implement the generation and input of random slope contour points and the writing of the generator. The generator adopts a structure based on convolutional neural network (CNN) or graph neural network (GNN), whose multi-layer convolution and upsampling modules can fully capture the spatial features of the input and gradually generate a fine node distribution;

[0060] c) The result G(z) generated by the generator is compared with the dataset x to train the discriminator D. In this embodiment of the invention, the training of the discriminator D is implemented by Python programming.

[0061] d) The discrimination result of discriminator D is backpropagated to generator G to achieve generator iteration. In this embodiment of the invention, Python programming is used to implement the backpropagation of the discrimination result of discriminator D to generator G to achieve generator iteration.

[0062] e) In this embodiment of the invention, the process of c is repeated using Python programming to achieve the iteration of the discriminator D.

[0063] f) In this embodiment of the invention, the process of repeating step d is performed using Python programming to achieve adversarial iteration between the generator G and the discriminator D. During the training process, in addition to the adversarial loss, geometric constraints are added to the loss function, including the minimum spacing constraint between nodes and the angle constraint of the connection between adjacent nodes, so as to ensure that the generated nodes meet the quality requirements of the mesh unit in engineering.

[0064] g) After several adversarial iterations, an AI model Gn for generating slope grid nodes that meets the requirements is obtained. Figure 2 The training flowchart is shown.

[0065] Step 2: Multi-source monitoring data acquisition and preprocessing;

[0066] In practical engineering, slope morphology information is collected in real time using various sensors. In this embodiment, the following monitoring methods are employed: Laser point cloud acquisition: High-density point cloud data of the slope surface is acquired using a ground-based laser scanner to obtain a detailed three-dimensional profile of the slope; UAV image acquisition: Orthophotos and high-resolution video data of the slope are acquired through UAV aerial photography to provide auxiliary reference for the point cloud data; Satellite remote sensing data: High-resolution satellite imagery is used to acquire information on the surrounding environment and large-scale topographic changes of the slope.

[0067] The collected data undergoes preprocessing, which includes the following processes:

[0068] Noise filtering: Statistical filtering or clustering filtering methods are used to remove noise points caused by equipment errors or environmental interference;

[0069] Data registration: Unify the coordinate system of data from different devices and use relevant algorithms to achieve multi-source data registration;

[0070] Data normalization: The slope elevation and horizontal distance are normalized to ensure scale consistency between different data sources;

[0071] Step 3, Extraction of key control points for slope;

[0072] The preprocessed data contains rich geometric information about the slope. Using image processing algorithms such as edge detection, curve fitting, and point cloud segmentation, combined with geological survey data, key control points for the external contour and internal geological stratification lines of the slope are extracted. The specific steps are: a) Contour detection: Detect the slope contour and extract the contour boundary; b) Stratification line extraction: Based on the geological survey results and slope elevation changes, a piecewise fitting method is used to extract stratification lines reflecting different strata interfaces; c) Control point selection: Sample the detected contour and stratification lines, extract representative control points, and store their (x,y) coordinates. To eliminate local anomalies, the RANSAC algorithm is used to verify the robustness of the control point data.

[0073] Step 4: Automatic generation of grid nodes;

[0074] This embodiment uses the pre-trained slope grid node generation AI model Gn to automatically generate complete grid node data under control point conditions. The specific details are as follows:

[0075] The external contour of the slope and the coordinates of the control points of the layered lines obtained in step 3 are fused with a random noise vector of a certain dimension and input into the slope grid node to generate an AI model. Given the key control points of the slope, it can automatically complete the generation of all node coordinates inside the slope and number each generated node.

[0076] Step 5, Two-dimensional mesh generation:

[0077] Based on the two-dimensional node data output from the AI ​​model for generating slope grid nodes, the Delaunay triangulation algorithm is used to divide the node data into grid cells. The specific implementation steps are as follows: a) Node data input: Input the two-dimensional node data generated in step 4 into the Delaunay triangulation module; b) Cell division: The Delaunay algorithm is used to automatically construct triangular cells, ensuring that each generated triangular cell does not contain excessively small interior angles, thereby ensuring the geometric rationality of the grid cells; c) Numbering management: Assign a unique number to each triangular cell and the nodes it constitutes, forming a dataset that corresponds to the node number and coordinate data, and the cell number and the numbers of the nodes it contains.

[0078] Step 6, 3D mesh expansion;

[0079] In some engineering applications, two-dimensional mesh models need to be extended to generate three-dimensional mesh models to facilitate more complex slope mechanics analyses. Specific methods include: a) Sweep method: extending the two-dimensional section along a predetermined direction (such as the forward or vertical direction of the slope), and sampling new node data at equal intervals along the extension direction; b) Rotation method: for slopes with rotational symmetry, rotating the two-dimensional section around a specific axis to sample and generate new three-dimensional nodes; c) Three-dimensional element meshing: tracing the extended three-dimensional node data along the sweep and rotation paths to achieve element meshing and generate a three-dimensional mesh model that meets engineering requirements.

[0080] Step 7, Data Output and Engineering Applications;

[0081] Output the element node data generated in step 5 or step 6, specifically including: a) Data format: The output file contains node numbers and coordinate information, as well as the correspondence between each element number and the numbers of its constituent nodes; b) Engineering interface: The output data can be directly imported into finite element, finite difference analysis software, or other slope analysis software for subsequent engineering calculations such as analysis and stability assessment. Figure 3 This is the generated case; c) Real-time updates: In the slope monitoring system, by regularly collecting and processing monitoring data, the grid model can be dynamically updated to reflect the slope changes in real time and provide accurate data support for the early warning system.

[0082] See Figure 4 The following is a flowchart of the module operation of this method: A slope grid node generation AI model training module uses a generative adversarial network to train a slope grid node generation AI model that meets the requirements; a data acquisition module collects multi-source monitoring data such as laser point clouds, UAV imagery, and satellite data to obtain the real-time slope morphology; a control point extraction module extracts important geological profile control points of the slope based on the real-time slope morphology and height information, forming node coordinate data on the external contour and layer lines of the slope; an automated node generation module uses the pre-trained slope grid node generation AI model to generate complete grid node coordinate data based on the control point data and assigns a number to each node; a triangulation module uses the Delaunay triangulation algorithm to divide the generated node data into units, generating a correspondence between unit numbers and the numbers of the contained nodes; a data output module outputs the node numbers and coordinate data, as well as the unit numbers and the numbers of the contained nodes; a 3D expansion module expands the 2D grid data through sweeping and rotation to generate 3D grid data and performs corresponding node and unit division; and a calculation and analysis module inputs the output data into calculation software via scripts for analysis and calculation.

[0083] The multi-source monitoring data acquisition and preprocessing provided in this embodiment of the invention includes, but is not limited to, laser point clouds, UAV images, and satellite data.

[0084] The training of the slope grid node generation AI model provided in this embodiment of the invention employs methods for imposing constraints on the loss function, including but not limited to adding constraints such as the minimum spacing between nodes and the angle of the connection between adjacent nodes.

[0085] The training network used for the slope grid node generation AI model provided in this embodiment of the invention includes, but is not limited to, traditional GAN ​​networks, as well as other GAN network variants used to optimize training.

[0086] The AI ​​model for generating slope grid nodes provided in this embodiment of the invention uses a generator with a structure including but not limited to structures based on convolutional neural networks (CNN) or graph neural networks (GNN).

[0087] The three-dimensional mesh expansion methods provided in the embodiments of the present invention include, but are not limited to, sweeping and rotation.

[0088] The output data provided in the embodiments of the present invention can be directly imported into engineering analysis software, including but not limited to finite element analysis software and finite difference analysis software.

[0089] In summary, this invention performs precise geometric modeling based on multi-source real-time monitoring data of slopes, extracting key control points such as the external contour and layer lines of the slope. An artificial intelligence model trained using a generative adversarial network (GAN) is used to automatically generate grid node data across the entire slope area. The Delaunay triangulation algorithm is employed to divide the generated node data into grid cells, and sweeping and rotation techniques are used to expand the two-dimensional grid into a three-dimensional grid model as needed. The generated two-dimensional and three-dimensional grid data are then imported into engineering analysis software to achieve real-time calculation and evaluation of the slope's mechanical response (such as deformation, stress distribution, and stability) under different working conditions. Through this process, the dynamic changes of slopes under natural and construction conditions can be effectively simulated, and key engineering parameters required for slope safety evaluation can be obtained in real time. This method provides an efficient and intelligent solution for the automated modeling and analysis of major slopes in water conservancy and hydropower projects, significantly improving the real-time performance and accuracy of slope safety monitoring and early warning, while significantly reducing the time and economic costs of manual modeling.

[0090] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0091] In the description of this invention, it should be understood that the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0092] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A data-driven intelligent monitoring and modeling analysis method for slope, characterized in that, Includes the following steps: By collecting slope morphology information, key control points of the slope are extracted from the external contour of the slope and the internal geological stratification line. The key control points of the slope are input into the trained slope grid nodes to generate an AI model. The AI ​​model of the slope grid nodes automatically generates two-dimensional grid node data. Based on the two-dimensional mesh node data, the Delaunay triangulation algorithm is used to divide the node data into mesh cells to generate a two-dimensional mesh model; the two-dimensional mesh model is then extended to generate a three-dimensional mesh model. Based on the mesh data output from the two-dimensional and three-dimensional mesh models, the mechanical stability, deformation, and safety of the slope are analyzed. In the process of training the slope grid node generation AI model, a generative adversarial network structure is adopted. During the adversarial training process, the model learns to generate complete grid node data under given control point conditions. During the training process, constraints such as the minimum spacing between nodes and the angle of the line connecting adjacent nodes are added to the loss function.

2. The intelligent slope modeling and analysis method driven by monitoring data according to claim 1, characterized in that: When extracting key control points of the slope, noise filtering, registration and normalization are performed on the collected slope morphology data and height information to obtain the slope morphology information. Through edge detection, curve fitting and point cloud segmentation algorithms, combined with geological survey data, the key control points of the slope's external contour and internal geological stratification lines are extracted.

3. The intelligent slope modeling and analysis method driven by monitoring data according to claim 2, characterized in that: When extracting key control points for slopes, the RANSAC algorithm is used to verify the robustness of the control point data.

4. The intelligent slope modeling and analysis method driven by monitoring data according to claim 3, characterized in that: When dividing the grid into units, each grid unit consists of three nodes, and each unit and its contained nodes are numbered to form a correspondence between node number and coordinate data, and between unit number and node number.

5. The intelligent slope modeling and analysis method driven by monitoring data according to claim 4, characterized in that: When expanding and generating a 3D mesh model, new node data is generated by sampling at equal intervals along a preset sweep path, or by rotating and sampling around a given axis to generate 3D nodes; then, the 3D node data is divided into units according to the sweep or rotation path to form a complete 3D mesh model.

6. The intelligent slope modeling and analysis method driven by monitoring data according to claim 5, characterized in that: When performing slope mechanical stability, deformation and safety analysis, based on the grid data output by the two-dimensional grid model and the three-dimensional grid model, the slope mechanical stability, deformation and safety analysis is performed through finite element analysis or finite difference analysis.

7. A data-driven, intelligent slope modeling and analysis system for monitoring, comprising: This system is used to implement a monitoring data-driven intelligent slope modeling and analysis method as described in any one of claims 1-6. The system includes: The slope grid node generation AI model training module uses generative adversarial networks to train a slope grid node generation AI model that meets the requirements. The data acquisition module is used to collect multi-source monitoring data, including laser point clouds, UAV images, and satellite data, to obtain the real-time morphology of the slope. The control point extraction module is used to extract important geological profile control points of the slope based on the real-time shape and height information of the slope, and form node coordinate data on the outer contour and layer lines of the slope. The node automatic generation module uses pre-trained slope grid nodes to generate an AI model, generates complete grid node coordinate data based on the control point data generated by the control point extraction module, and numbers each node. The triangulation module uses the Delaunay triangulation algorithm to divide the node data generated by the automatic node generation module into units, generating data on the correspondence between unit numbers and the numbers of the nodes they contain. The three-dimensional expansion module is used to expand the two-dimensional mesh data generated by the triangulation module into three-dimensional mesh data through sweeping and rotation, and to perform corresponding node and cell division. The data output module is used to output the correspondence between node numbers and coordinate data, and between unit numbers and the node numbers they contain. The calculation and analysis module is used to input the data output by the data output module into the calculation software for analysis and calculation via a script.

Images