High and steep slope rockfall path analysis method and system based on refined modeling

Abstract

The application provides a high and steep slope rockfall path analysis method and system based on refined modeling, and belongs to the technical field of geotechnical engineering disaster prevention and reduction; the method comprises the following steps: surveying a high-risk slope and obtaining slope parameters; taking photos of the high-risk slope by using a drone to obtain point cloud data, and constructing a three-dimensional model of the high-risk slope; analyzing the structure surface of the three-dimensional model of the high-risk slope to identify a dangerous rock area, and performing refined three-dimensional scanning; constructing a dangerous rock mass model according to the scanning result, and importing the modeling information into a discrete element numerical simulation software to simulate a rockfall process; repeatedly performing multiple simulation calculations and performing statistical analysis, and finally obtaining the rockfall path of the high-risk slope dangerous rock. Through the refined modeling of the characteristics of the dangerous rock mass, the application can greatly reduce the deviation between the motion trajectory prediction and the actual rockfall path, and provides effective guidance for the construction and maintenance of a high and steep slope protection system.

Description

Technical Field

[0001] This invention belongs to the field of geotechnical engineering disaster prevention and mitigation technology, and particularly relates to a method and system for analyzing the path of falling rocks on steep slopes based on refined modeling. Background Technology

[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.

[0003] With the rapid advancement of transportation infrastructure construction, mountain highway tunnel projects are developing on a large scale, gradually forming a crisscrossing, three-dimensional transportation network in mountainous areas. However, in the process of this leapfrog development, the contradiction between complex geological conditions and engineering safety has become increasingly acute. Especially in some plateau and canyon areas, there are numerous steep slopes shaped by geological tectonic activity. These areas, due to the complex development of fractured rock layers and weak interlayers, have prominent conditions for landslides and rockfalls. Such disasters are characterized by high kinetic energy impact, uncertain movement paths, and instantaneous explosiveness. Conventional protection systems often fail due to insufficient understanding of the energy transfer patterns of falling rocks. Based on these problems, it is essential to systematically study the dynamic mechanisms of rockfall disasters on steep slopes and proactive prevention and control technologies.

[0004] Currently, significant progress has been made in the research on the stability of unstable rocks and the path of rolling stones on high slopes. Regarding the stability of unstable rocks, some scholars, based on studies of the geological environment and the causes of instability, have analyzed the distribution and morphological characteristics of unstable rock masses, and have quantitatively analyzed and evaluated the stability of individual unstable rock masses using the plane sliding method and engineering geological analysis methods, respectively. Other scholars have used a cantilever beam, generalized as being subjected to self-weight bending moment and external forces, as the object of study, and analyzed the failure mode of rock block toppling on a reverse-dip slope using a bending-tensile cracking model. Regarding the movement path of unstable rock masses, industry experts divide the process of unstable rockfall into four stages: the initial displacement stage, the collision stage, the sliding stage, and the rolling stage. Different types of unstable rocks exhibit different trajectory equations in these four stages. Related scholars have combined PFC 3D discrete element analysis with RockFall simulation software to construct a three-dimensional geological model for collapse simulation and to calculate and analyze the movement path of rolling stones.

[0005] It is evident that while existing methods have made some progress in identifying unstable rock masses and simulating their movement trajectories, they still fall short in refining the modeling of unstable rock mass characteristics, leading to discrepancies between the predicted movement trajectory and the actual rolling path. Summary of the Invention

[0006] To overcome the shortcomings of the existing technology, this invention provides a method and system for analyzing the path of falling rocks on steep slopes based on refined modeling. By refining the modeling of the characteristics of the unstable rock mass, the deviation between the predicted movement trajectory and the actual rolling path can be significantly reduced, providing effective guidance for the construction and maintenance of steep slope protection systems.

[0007] To achieve the above objectives, one or more embodiments of the present invention provide the following technical solutions: The first aspect of this invention provides a method for analyzing the path of falling rocks on steep slopes based on refined modeling.

[0008] Methods for analyzing the path of bouldering on steep slopes based on refined modeling include: The high-risk slopes are investigated and the slope and material composition are analyzed to obtain the slope parameters. UAV oblique photogrammetry was used to take multi-angle photos of the high-risk slope to obtain point cloud data, and a three-dimensional model of the high-risk slope was constructed based on the obtained point cloud data. Structural surface analysis was performed on the constructed 3D model of the high-risk slope to identify the dangerous rock areas; and a detailed 3D scan of the dangerous rock areas was conducted. Based on the refined 3D scanning results, a dangerous rock mass model is constructed. The modeling information of the constructed dangerous rock mass model is imported into discrete element numerical simulation software. The obtained slope parameters are used as input, and the rolling stone movement parameters are optimized using a machine learning model to simulate the process of dangerous stone rolling. By repeating the simulation calculations multiple times and statistically analyzing the results, the rolling path of the dangerous rocks on the high-risk slope was finally obtained.

[0009] Furthermore, structural surface analysis was performed on the constructed 3D model of the high-risk slope to identify the dangerous rock areas of the high-risk slope. This included: firstly, large-scale structural surfaces were initially extracted based on the improved RANSAC shape detection method; then, the segmentation parameters were corrected by using the point cloud normal vector difference and feature final value in combination with the KD-tree accelerated region growing algorithm to achieve accurate segmentation of the structural surfaces.

[0010] Furthermore, after accurately segmenting the structural planes, firstly, the orientation, spacing, and extension characteristics of each structural plane are extracted; then, based on the intersection relationship of each structural plane, the bare rock surface is divided into several independent rock mass units; finally, based on the block integrity, the degree of development of the free face, and the gravity driving force, the regions of dangerous rock masses and stable rock masses are labeled to identify the dangerous rock areas of high-risk slopes.

[0011] Furthermore, a refined 3D scan of the dangerous rock area is conducted, including: first, using a 3D laser scanner to create a refined model of the dangerous rock area and setting up photogrammetric control points for the identified dangerous rock masses; then, dynamically adjusting the laser emission frequency and scanning resolution based on the surface reflectivity of the dangerous rock masses and ambient lighting conditions.

[0012] Furthermore, when conducting a detailed 3D scan of the dangerous rock area, multiple scanning points were set up, and the spacing between each station and the overlap rate of adjacent stations were controlled within a preset range. In addition, by synchronously recording POS data, splicing errors caused by terrain undulations were eliminated.

[0013] Furthermore, based on the refined 3D scanning results, a model of the unstable rock mass is constructed, including: First, using non-uniform rational B-spline surface construction technology, combined with a point cloud density adaptive triangular mesh generation algorithm, a TIN model with millimeter-level accuracy is established while maintaining the angular characteristics of the unstable rock mass; then, potential joints and fissures in the structural surface development area are identified through curvature analysis; finally, a refined 3D mesh model of the unstable rock mass is constructed by combining the obtained detailed information, and the local model is fused with the overall slope model through the ICP algorithm to obtain the unstable rock mass model.

[0014] Furthermore, machine learning models are used to optimize the motion parameters of the rolling stones, including: first, using Gaussian process regression with a large amount of small-scale rolling stone simulation data as the training set to predict the horizontal movement distance and total kinetic energy of the rolling stones; then, using support vector machines to classify and regress the bounce height of the experimental rolling stones to extract the main factors affecting the bounce; finally, using neural networks to comprehensively predict the motion characteristic parameters of the rolling stones throughout the entire process.

[0015] The second aspect of this invention provides a system for analyzing the path of boulder rolling on steep slopes based on refined modeling.

[0016] A high-steep slope boulder path analysis system based on refined modeling includes: The survey and parameter acquisition module is configured to: survey high-risk slopes, analyze the slope gradient and material composition of the slopes, and obtain slope parameters; The high-risk slope 3D model construction module is configured to: use UAV oblique photogrammetry to take multi-angle photos of the high-risk slope to obtain point cloud data, and construct a 3D model of the high-risk slope based on the obtained point cloud data; The refined 3D scanning module is configured to: perform structural surface analysis on the constructed 3D model of the high-risk slope, identify the dangerous rock areas of the high-risk slope, and perform refined 3D scanning on the dangerous rock areas; The rockfall path simulation module is configured to: construct a dangerous rock mass model based on the refined 3D scanning results; import the modeling information of the constructed dangerous rock mass model into discrete element numerical simulation software; use the obtained slope parameters as input; optimize the rockfall motion parameters using a machine learning model; and simulate the rockfall process. The rockfall path analysis module is configured to repeatedly perform simulation calculations and statistically analyze the results to ultimately obtain the rolling path of dangerous rocks on high-risk slopes.

[0017] A third aspect of the present invention provides a computer-readable storage medium having a program stored thereon, which, when executed by a processor, implements the steps in the method for analyzing the rockfall path of steep slopes based on refined modeling as described in the first aspect of the present invention.

[0018] The fourth aspect of the present invention provides an electronic device including a memory, a processor, and a program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps in the method for analyzing the rockfall path of steep slopes based on refined modeling as described in the first aspect of the present invention.

[0019] The above one or more technical solutions have the following beneficial effects: This invention utilizes unmanned aerial vehicles (UAVs) to acquire point cloud data of high-risk slopes and constructs a 3D model of the slope. By analyzing the structural surfaces of the 3D model, it identifies and finely scans areas prone to falling rocks. Using the refined scanning data, it constructs a model of the unstable rock mass and finally simulates the rockfall path using discrete metadata simulation software. The entire simulation process of this invention is based on refined modeling data of the unstable rock mass characteristics in the rockfall area. Compared to existing technologies, it solves the problem of insufficient modeling precision, allowing for a more comprehensive and accurate understanding of the actual internal conditions of steep slopes. It significantly reduces the deviation between predicted movement trajectories and actual rolling paths, and uses simulation to predict the path of falling rocks, thus providing more scientific and effective guidance for the construction and maintenance of steep slope protection systems.

[0020] Advantages of additional aspects of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0021] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0022] Figure 1 This is a flowchart of the method for analyzing the path of rolling stones on steep slopes based on refined modeling in Embodiment 1 of the present invention.

[0023] Figure 2This is a schematic diagram of a high-risk slope and unstable rock mass in Embodiment 1 of the present invention.

[0024] Figure 3 This is a schematic diagram of the slope point cloud in Embodiment 1 of the present invention.

[0025] Figure 4 This is a schematic diagram of the identification of unstable rock masses on slopes in Embodiment 1 of the present invention.

[0026] Figure 5 This is a simplified schematic diagram of the rockfall movement on the slope in Embodiment 1 of the present invention. Detailed Implementation

[0027] It should be noted that the following detailed descriptions are exemplary and intended to provide further illustration of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0028] It should be noted that the terminology used herein is for the purpose of describing particular implementations only and is not intended to limit the exemplary implementations of the present invention.

[0029] Where there is no conflict, the embodiments and features in the embodiments of the present invention can be combined with each other.

[0030] Example 1 This embodiment discloses a method for analyzing the path of falling rocks on steep slopes based on refined modeling.

[0031] like Figure 1 As shown, the method for analyzing the path of falling rocks on steep slopes based on refined modeling includes: Step S1: Conduct an investigation of the high-risk slope, analyze the slope gradient and material composition to obtain slope parameters; Step S2: Use UAV oblique photogrammetry to take pictures of the high-risk slope from multiple angles to obtain point cloud data, and construct a three-dimensional model of the high-risk slope based on the obtained point cloud data; Step S3: Perform structural surface analysis on the constructed 3D model of the high-risk slope to identify the dangerous rock areas of the high-risk slope; and perform a refined 3D scan of the dangerous rock areas. Step S4: Based on the refined 3D scanning results, construct a dangerous rock mass model; import the modeling information of the constructed dangerous rock mass model into discrete element numerical simulation software, use the obtained slope parameters as input, and use a machine learning model to optimize the rolling stone movement parameters to simulate the dangerous rock rolling process. Step S5: Repeat the simulation calculation multiple times and perform statistical analysis on the results to finally obtain the rolling path of the dangerous rocks on the high-risk slope.

[0032] Based on the above process, this invention, through refined modeling of the characteristics of unstable rock masses, can significantly reduce the deviation between the predicted movement trajectory and the actual roll-off path, providing effective guidance for the construction and maintenance of high and steep slope protection systems. To facilitate understanding of the technical solution of this invention, the specific implementation methods of this invention will be further explained and described below.

[0033] In step S1, the high-risk slope is surveyed, and its slope gradient and material composition are analyzed to obtain slope parameters. This can be achieved through the following methods: Step S1-1: Conduct a geological survey of the high-risk slope.

[0034] Rock samples were collected from different lithological zones of the high-risk slope. Specifically, measuring instruments such as levels, theodolites, or total stations were used to conduct on-site measurements at different locations on the slope to obtain slope data. A geological compass was used to determine the occurrence and type of the rocks, and rock samples were taken by hammering or other tools.

[0035] After collecting rock samples, the physical and mechanical parameters of each rock sample were measured in the laboratory, and the rebound modulus (20-50 GPa) of the unstable rock mass was tested using a rebound hammer. The physical and mechanical parameters to be measured included density, elastic modulus, compressive strength, and coefficient of friction. Collecting the rebound modulus of the unstable rock mass was to use it as input parameters for subsequent numerical simulations. Based on this, the slope sample data collected earlier are collectively referred to as input parameters in the subsequent numerical simulations.

[0036] Step S1-2: Preliminary identification of dangerous rocks.

[0037] Based on the geological structure, joint development direction, and historical landslide records of the high-risk slope, potential unstable rock mass distribution areas were preliminarily delineated on the slope. A schematic diagram of unstable rocks on a high-risk slope is shown below. Figure 2 As shown in the figure, this diagram represents a longitudinal profile of a high-risk slope. The red line indicates the potential fracture surface of the dangerous rock mass, which is at risk of fracturing. The slope rock mass cut by the red line represents the dangerous rock mass, which may fall off the slope as the fracture surface breaks, becoming a rockfall.

[0038] In step S2, oblique photogrammetry using a drone is used to take multi-angle photos of the high-risk slope to obtain point cloud data, and a three-dimensional model of the high-risk slope is constructed based on the obtained point cloud data.

[0039] When using drones for oblique photogrammetry, a DJI Matrice 300 RTK drone (5-lens oblique photogrammetry system) was used. The drone's flight path was planned in advance, and a 60% forward overlap and a 70% lateral overlap were set. High-resolution images with a resolution of 5cm were selected to cover the slope and surrounding area. The scope of this invention is designed to collect point cloud data of high-risk slopes. Compared to existing technologies, the advantages of this design are: 1) Compared to lower overlap, setting 60% forward overlap and 70% lateral overlap can reduce matching gaps by approximately 20%-30%, increasing point cloud density by more than 50%, thereby supporting more accurate deformation monitoring; 2) Too low a resolution will ignore microcracks, leading to reduced simulation accuracy, while too high a resolution will increase data acquisition and simulation computation costs. A resolution of 5cm can avoid both problems to the greatest extent; 3) It covers the slope and surrounding area. A defined scope ensures the accuracy and comprehensiveness of the data.

[0040] Next, using Context Capture software, a 3D point cloud model was generated from the collected point cloud data and exported as OBJ or LAS format; the generated 3D point cloud model had a point cloud density of 10 pts / m² and an accuracy controlled within a certain range. 30cm. A schematic diagram of the 3D point cloud model of the slope is shown below. Figure 3 As shown in the figure, this diagram represents the point cloud data obtained after a 3D scan of a slope. The point cloud data is then used to create a 3D point cloud model in specific software. Figure 3 It is clearly a model of a slope, with different colors from top to bottom representing different heights.

[0041] In step S3, structural surface analysis is performed on the constructed three-dimensional model of the high-risk slope to identify the dangerous rock areas of the high-risk slope; and a refined three-dimensional scan of the dangerous rock areas is performed.

[0042] While existing methods have made some progress in identifying unstable rock masses and simulating their movement trajectories, they still fall short in refining the modeling of unstable rock mass features, leading to discrepancies between predicted trajectories and actual rolling paths. To overcome this limitation, this invention refines point cloud processing to accurately identify unstable rock masses. Specifically, this can be achieved through the following methods: Step S3-1: Structural surface identification.

[0043] The first step is to extract large-scale structural surfaces based on the improved RANSAC shape detection (RSD) method.

[0044] 1) The core formula of the improved RSD method includes: 1-1) Adaptive distance threshold.

[0045] Traditional fixed threshold Improved to a dynamic distance threshold function, combining point cloud density and local curvature: ; in, Point of Nearest neighbor average spacing is used to measure local density; This indicates local curvature estimation based on PCA (the high curvature region needs to be smaller). ); This represents the curvature weighting factor, which is typically set to 0.2-0.5.

[0046] 1-2) Model scoring with combined geological constraints. Geological consistency is improved by introducing attitude similarity weights: ; in, Point To model The Euclidean distance; Indicates the orientation of the model, i.e., the dip angle; This indicates the prior knowledge of the attitude of known structural planes in geological surveys; This represents the attitude constraint weight, which is generally set to 0.5 to 1.0, and can be adjusted according to the geological reliability.

[0047] 1-3) Probability-guided sampling strategy. By improving the sampling probability distribution, points in high-curvature regions are prioritized to enhance boundary detection capabilities: ; 2) Data preprocessing.

[0048] The acquired 3D point cloud data (such as laser scanning, photogrammetry, or borehole data) is used as input data, and isolated outliers and noise are removed through statistical filtering, radius filtering, etc. If the data density is too high, the data volume can be reduced and the computational efficiency improved by methods such as voxel grid.

[0049] 3) RANSAC improvement strategy.

[0050] First, the distance threshold is adaptively adjusted based on point cloud density or local curvature (e.g., using local point cloud density statistics as the threshold). Then, considering that structural surfaces are typically planar (e.g., rock mass structural surfaces), planar models are prioritized for fitting; if curved surfaces need to be detected, this can be extended to quadratic surfaces. Regional connectivity analysis is performed on the extracted candidate models, prioritizing models with large coverage areas and continuous internal point distribution. Finally, GPU acceleration or block processing is used to improve the computational efficiency of large-scale point clouds.

[0051] 4) Structural surface extraction.

[0052] Initial model generation: Randomly select the minimum point set (3 points are required for a planar model) to generate candidate models; calculate interior points (points falling within a certain distance of the model) based on dynamic thresholds, and count the number and distribution range of interior points.

[0053] Model scoring optimization: In addition to considering the number of inliers, it is also necessary to evaluate the spatial distribution of inliers (such as coverage area and continuity).

[0054] Iteration and Termination Conditions: Set a maximum number of iterations or an adaptive stopping criterion (such as not finding a higher-scoring model for a certain number of consecutive iterations); retain the model with the highest score as the current optimal structural surface.

[0055] Interior point removal and repeated fitting: Remove the interior points corresponding to the extracted structural surfaces from the point cloud. Repeat the above process for the remaining point cloud until the termination condition is met (e.g., the number of remaining points is less than a threshold).

[0056] 5) Post-processing and verification.

[0057] Structural plane merging: Merge spatially adjacent planes with similar normal vectors to eliminate oversegmentation.

[0058] Geological rationality verification: Based on the statistical regularity of geological occurrence (dip, dip angle), abnormal surfaces that do not conform to the geological background are eliminated.

[0059] Boundary optimization: Extract the precise boundaries of the structural surfaces using the convex hull algorithm or the Alpha Shape algorithm.

[0060] 6) Output of results: Structural surface parameters: Outputs the geometric equation, area, attitude (dip / dip angle), and other information for each structural surface.

[0061] Point cloud labels: Assign a structural surface label to each point to facilitate subsequent analysis.

[0062] The second step involves combining the KD-tree accelerated region growing algorithm with the point cloud normal vector difference and feature terminal value to correct the segmentation parameters, thereby achieving accurate segmentation of the structural surface.

[0063] KD-tree (K-Dimensional Tree) is a spatial partitioning data structure for efficiently storing and retrieving multidimensional data (such as 3D point clouds). Its core idea is to recursively partition the space along different dimensions to form a binary tree structure, thereby accelerating nearest neighbor search (such as K-nearest neighbor and radius search). In point cloud processing, KD-tree can optimize the time complexity of neighborhood queries from O(n) to O(logn).

[0064] 1) The region growing algorithm uses the difference in normal vectors and the final value of features as growth conditions. The specific formula is as follows: 1-1) Normal vector difference condition.

[0065] Setting points The normal vector is The current region's average normal vector is Then the condition for the difference of normal vectors is: ; in, This represents the normal vector consistency threshold.

[0066] 1-2) Characteristic terminal value condition.

[0067] Characteristic terminal value combined with curvature Local point density Features such as these dynamically adjust the growth threshold: ; in, For the regional centroid, , and These are the weighting coefficients; if Then the merging point To the region; Point The curvature is used to describe local concavity and convexity; Local density can be represented by the number of neighboring points or the distance variance. This represents the threshold for comprehensive features, which can be set based on training data or statistical experience.

[0068] 2) Method and process.

[0069] 2-1) Construct a KD-tree. Using the original point cloud... As input, Create a KD-tree index to support fast neighborhood queries.

[0070] 2-2) Initialize the seed point.

[0071] Seed source: large-scale structural surfaces or points with minimum curvature extracted by improved RSD; Storage: Add seed points to the queue .

[0072] 2-3) Regional growth cycle.

[0073] First, extract the seed point, that is, from... take out If the region already belongs to, skip it; otherwise, create a new region.

[0074] Then, perform a neighborhood search (accelerated by KD-tree): use KD-tree for querying. of Nearest neighbor or radius Point set within To find neighboring points and set the search radius. Avoid noise interference.

[0075] Next, the normal vector and feature final value are determined, that is: for each neighborhood point Calculate the difference in normal vectors ,like Proceed to the next step: Calculate the final value of the feature. If satisfied ,Will Join the region and added as a new seed. The termination condition is: queue Empty, or region Reaching maximum size Dynamic updates: based on Average curvature or density adjustment To improve adaptability.

[0076] Finally, the segmentation result is output, which is the generation of all regions. Each region represents a structural surface.

[0077] 3) Data preprocessing and feature calculation.

[0078] 3-1) Normal vector estimation.

[0079] Methods: Principal component analysis (PCA) or nearest neighbor search (radius) based on KD-tree was used. Calculate the point cloud normal vector. For each point... neighborhood point set Calculate the covariance matrix: ; The eigenvector corresponding to the smallest eigenvalue is the normal vector. Orientation is unified (oriented towards the viewpoint).

[0080] 3-2) Calculation of characteristic final value.

[0081] Curvature estimation: by As eigenvalues ​​of the covariance matrix, in low curvature regions (near the plane) When in high curvature regions (edges / noise) .

[0082] Local density calculation: ; in, The furthest point in the neighborhood is defined. Based on this, a minimum radius is set. density To prevent the denominator from being zero.

[0083] 4) Parameter dynamic correction mechanism.

[0084] 4-1) Dynamic adjustment of the normal vector difference threshold.

[0085] Objective: To avoid the accumulation of deviations in the direction of the normal vector due to the growth of the region.

[0086] method: Initialize threshold: Set the global initial angle threshold. .

[0087] Updated during regional growth: If the average normal vector of the current region Deviation from the initial seed point normal vector Then tighten the threshold: ; If the variance of the region's normal vector Relax the threshold (to compensate for noise): ; Ensure the threshold range is constrained within [ , ]Inside.

[0088] 4-2) Feature terminal value weight adaptive.

[0089] Objective: To dynamically adjust the weighting coefficients of curvature, density, and distance based on the region's geometric characteristics. ).step: Initial weights: , , (Plane priority is assumed by default).

[0090] Adjustments during regional growth: Curvature weight : If the current region has average curvature, increase the curvature weight to suppress oversegmentation: upper limit ; Density weight If the regional density standard deviation Reduce density weights to prevent missegmentation of sparse regions: lower limit ; Distance weight Mandatory constraints are This ensures that the weights are normalized.

[0091] 4-3) Optimization strategies in practical applications.

[0092] Adaptive termination condition: Automatic termination occurs when the region growth rate (number of new points per unit time) falls below a threshold to avoid merging noisy points. (If the feature final value...) If there is no significant decrease in 5 consecutive iterations, it is considered to have converged.

[0093] Noise suppression is achieved by introducing an outlier penalty term into the eigenvalue formula: ; in, This is the noise weight, with a default value of 0.3.

[0094] initial radius It can be dynamically adjusted according to the regional density: ; In high-density areas, the radius is reduced to improve accuracy, while in low-density areas, the radius is increased to avoid breakpoints.

[0095] The third step is to extract features such as the attitude (dipping angle, dip direction), spacing, and extension of each structural plane.

[0096] 1) Inclination and tilt angle.

[0097] 1-1) Plane Fitting: For each segmented structural surface point cloud, fit the plane equation using singular value decomposition (SVD) or least squares method. , thus obtaining the normal vector .

[0098] 1-2) Normal vector normalization: Calculate the unit normal vector It is necessary to ensure that the direction is towards the Earth's center (geological standard). If the normal vector is upward, it needs to be reversed. .

[0099] 1-3) Tendency Calculation: Extracting the Horizontal Projection Vector Calculate the dip azimuth. Convert to geographic azimuth ( (Starting from due north, increase clockwise).

[0100] 1-4) Inclination calculation: ; in, The absolute value of the perpendicular component of the normal vector reflects the degree of inclination of the plane. (Completely horizontal), tilt angle is 0 degrees; if (Vertical), with an inclination angle of 90 degrees.

[0101] 2) Spacing calculation.

[0102] 2-1) Grouping parallel structural surfaces: Use the angle between the normal vectors to filter parallel surfaces (e.g., the angle is less than 10 degrees).

[0103] 2-2) Projecting along the normal direction, that is, projecting the centroid onto the common normal vector. On a straight line in the direction: ; according to Sort by size from smallest to largest and calculate the distance between adjacent faces: By using projection, the three-dimensional distance is simplified to a one-dimensional calculation, avoiding errors caused by non-strict parallelism. Finally, the average spacing is calculated, and the spacing between all adjacent faces is taken. The median or mean is used as the final result.

[0104] 3) Calculation of elongation.

[0105] Projecting onto the fitting plane: Project the point cloud onto the fitting plane (two-dimensional coordinate system). Using the plane normal vector n as the Z-axis, choose two orthogonal tangent vectors (u, v) as the X and Y axes. Apply this to the projected two-dimensional point set... Perform PCA to obtain the principal direction. (Direction of maximum variance) and secondary direction Then, calculate the extension length (length along the primary direction and length along the secondary direction): Length along the main direction: ; Length along the secondary direction: ; The length in the primary direction reflects the maximum extension range of the structural surface, the length in the secondary direction reflects the lateral expansion range, and the area comprehensively represents the overall scale.

[0106] 4) Area estimation: Method 1: Convex hull area (most conservative estimate); Method 2: Area of ​​a rectangle ; Method 3: Using ellipse fitting parameters ( , (These are the major and minor semi-axis, respectively).

[0107] Step S3-2: Further identification of unstable rock masses.

[0108] Based on the intersection relationships of various structural planes, the bare rock surface is divided into several independent rock mass units. Based on block integrity, the degree of development of free faces, and gravity driving force, the regions of dangerous and stable rock masses are marked. These markings are then verified in conjunction with the results of the "preliminary identification of dangerous rocks" in steps S1-3 to identify the dangerous rock areas of high-risk slopes. A schematic diagram of dangerous rock markings for high-risk slopes is shown below. Figure 4 As shown in the figure, this diagram illustrates the process of identifying dangerous rocks in a 3D point cloud model, with the affected rock areas highlighted in yellow. Figure 5The diagram shown is a simplified illustration of the movement of a boulder rolling down a slope. The basic principle of boulder movement on high-risk slopes is the same.

[0109] Block integrity refers to the degree of integrity of the independent blocks formed after a rock mass is cut by structural planes (such as fissures, joints, faults, etc.). Rock masses with poor integrity are prone to sliding or collapsing along structural planes, and are an important indicator of unstable rock masses. In slope engineering, integrity assessment is often used to identify potentially unstable areas.

[0110] The block integrity was determined using the RANSAC algorithm to extract approximate planar regions (joints, faults, and other structural surfaces) from the point cloud. The planes were then segmented based on normal vector similarity to distinguish structural surface clusters with different attitudes. After dividing the block into units, the number of structural surfaces per unit volume was counted using the following formula: ; in, This represents the average spacing of the structural surfaces extracted from the point cloud.

[0111] The degree of free face development refers to the number and scale of freely exposed surfaces around a rock mass, i.e., the development status of unsupported surfaces formed by natural erosion or human excavation. The degree of development is quantified by parameters such as the number, steepness, height, and continuity of free faces. A higher degree of development (e.g., multiple steep free faces coexisting) indicates a more significant downward trend in the rock mass driven by gravity.

[0112] The degree of free surface development is determined through point cloud modeling and analysis. A continuous triangular mesh surface is generated using Poisson reconstruction, the normal vector of each triangular facet is calculated, and the slope is selected. In the region, the boundaries of free surfaces such as steep slopes and cliffs are identified by changes in curvature, and the depressions (potential free surfaces) formed by erosion or collapse are marked.

[0113] In this embodiment, the following three parameters are used: slope Histogram of slope distribution at the vertices of a point cloud triangular mesh; altitude : Calculate the elevation difference between the top and bottom of the free face; Continuity: The continuous length of the free face (the distance extended along the slope).

[0114] Furthermore, its classification rules are as follows: a) Low development: and Isolated, small-scale freefront; b) High development: and Continuous length .

[0115] Gravity-driven force, namely the sliding force generated by rock mass along potential sliding surfaces or structural planes under the action of gravity, is the core mechanical factor leading to the instability of unstable rock masses. When the gravity-driven force exceeds the sliding resistance of the structural plane (such as frictional resistance and cementation strength), the rock mass will slide or collapse. The process of obtaining gravity-driven force includes: Block gravity calculation: Calculate the closed volume based on the segmented independent block point cloud model. (Integral method or voxel accumulation method); Refer to laboratory rock sample tests or regional geological data to determine the rock mass unit weight. (e.g., sandstone) ): ; Sliding direction and dip angle analysis: Planes with a dip direction consistent with the slope aspect and a dip angle close to the slope angle are selected from the structural surface point cloud cluster; the dip angle is calculated based on the structural surface point cloud plane equation, i.e.: ; in, Let be the plane normal vector component. Based on this, the gravitational driving force can be expressed as: (Theoretical maximum value when friction and bonding forces are ignored).

[0116] Spatially overlay the preliminary unstable rock mass distribution map from step S1-2 with the refined unstable rock mass annotation map from step S3-2 (e.g., in GIS software or a 3D model). Check the consistency of the two in terms of location, extent, and shape: Consistent Areas: If the initially identified area highly overlaps with the finely labeled area, it is directly confirmed as a dangerous rock area, and the confidence level is increased.

[0117] Inconsistent areas: For areas initially identified but not covered by refined annotation, re-examine the structural surface analysis parameters (such as segmentation thresholds) or consider supplementary on-site investigation to confirm whether omissions are due to insufficient model accuracy or environmental factors (such as vegetation obstruction). For areas refinedly annotated but not initially identified: combine geological experience to judge, for example, check whether the area has hidden structural surfaces or newly formed fractures, and verify through on-site verification (such as manual investigation) if necessary.

[0118] In step S4, a dangerous rock mass model is constructed based on the refined three-dimensional scanning results. The modeling information of the constructed dangerous rock mass model is imported into discrete element numerical simulation software. The obtained slope parameters are used as input, and the rolling stone movement parameters are optimized using a machine learning model to simulate the dangerous rock rolling process.

[0119] Step S4-1: Detailed local scanning of the unstable rock mass.

[0120] First, a detailed model of the hazardous rock area was created using a RIEGL VZ-4000 3D laser scanner. Photogrammetric control points (ground targets or artificial reflective markers) were then established in the identified hazardous rock areas to ensure 3D scanning accuracy during a second high-density scan (point spacing ≤ 1mm). Subsequently, the laser emission frequency and scanning resolution were dynamically adjusted based on the surface reflectivity of the hazardous rock mass and ambient lighting conditions. The scanning resolution was selected as follows: Stepping accuracy.

[0121] Simultaneously, multiple scanning points are deployed, with the spacing between stations and the overlap rate of adjacent stations controlled within preset ranges. Furthermore, POS data is recorded synchronously to eliminate stitching errors caused by terrain undulations. As an optional embodiment, the station spacing can be controlled within the range of 50-150m, with an overlap rate of no less than 40% between adjacent stations. Positioning is achieved through built-in IMU and GNSS modules, and POS data is recorded synchronously to eliminate stitching errors caused by terrain undulations.

[0122] By deploying multiple scanning points, the spacing between stations and the overlap rate of adjacent stations are controlled within a preset range, achieving the following results: 1) It can ensure the integrity and seamless splicing of the 3D model.

[0123] Eliminating scanning blind spots: A single scanning station is limited by the field of view of the laser scanner and terrain obstruction, making it impossible to cover all surfaces of the unstable rock mass. By scanning from multiple different positions and angles, the field of view of each station can be complemented, effectively capturing the complex geometric features of the unstable rock mass, such as its various sides, grooves, and fissures, thus avoiding data loss.

[0124] Ensuring overlapping areas: Maintaining a sufficiently high overlap rate between adjacent stations provides ample common feature areas for subsequent point cloud stitching. This enables the precise alignment and fusion of scan data from individual stations using registration algorithms such as ICP, resulting in a complete, continuous, and seamless 3D model.

[0125] 2) Significantly improves the overall accuracy and detail reproduction of the model.

[0126] Reducing stitching errors: Multi-site scanning with strict control over spacing and overlap essentially utilizes redundant observations to improve accuracy. When multiple stations scan the same area, the stitching algorithm can use a large amount of overlapping point cloud data for adjustment calculations, effectively reducing random and systematic errors at individual stations, thereby keeping the overall registration error at an extremely low level.

[0127] Enhanced detail representation: The high overlap rate ensures that key structural features (such as microcracks, joint surfaces, and edges) are captured and cross-validated by multiple sites, enabling the final model to more realistically and precisely reproduce the millimeter-level details of the surface of the unstable rock mass, providing a reliable geometric basis for subsequent stability analysis and motion simulation.

[0128] Step S4-2: Construction of the unstable rock mass model.

[0129] A TIN model with millimeter-level accuracy was established by using non-uniform rational B-spline (NURBS) surface construction technology and combining it with point cloud density adaptive triangular mesh generation algorithm, while maintaining the angular characteristics of the unstable rock mass.

[0130] TIN (Triangulated Irregular Network) is a computer graphics model used to represent 3D surfaces. It accurately fits complex terrain or object surfaces by piecing together a large number of irregular triangles of varying density. Compared to regular mesh models, TIN models can adaptively adjust the density and size of triangles according to the terrain's undulations. Sparser, larger triangles are used in flat areas, while denser, smaller triangles are used in steep, fragmented, or detailed areas, thus optimizing data volume while maintaining accuracy.

[0131] The "point cloud density adaptive triangulation generation algorithm" is employed. This is the core step in constructing the TIN model. This algorithm automatically adjusts the structure of the triangulation based on the density of the point cloud. In areas with dense point clouds (such as the edges and fissures of unstable rock masses), the algorithm generates smaller, denser triangles to capture and preserve these key, subtle geometric features.

[0132] In areas with relatively sparse point clouds (such as relatively flat rock surfaces), larger triangles are generated, optimizing computational resources while ensuring model accuracy. This step directly generates an initial TIN model that can highly reproduce the macroscopic morphology of the unstable rock mass.

[0133] The model was further optimized in the later stages through steps such as surface optimization and feature preservation, detail enhancement and model integration.

[0134] The TIN model, using a dense and variable-sized triangular mesh, can realistically reproduce the complex surface morphology of unstable rock masses with millimeter-level geometric accuracy, including minute features such as micro-cracks, sharp edges, pits, and surface spalling. This modeling method can smooth the curved surface while deliberately preserving the angular characteristics of the unstable rock mass. These edges significantly affect the bounce trajectory and energy dissipation when rocks collide with the slope, providing essential geometric information for accurate simulation. The high-precision TIN model forms the basis for subsequent curvature analysis. By analyzing the curvature changes on the model surface, potential joints, cracks, and other structural surfaces can be effectively identified. These structural surfaces are crucial for determining rock mass stability and the potential location of rock detachment. Using the ICP algorithm, this detailed model of the local unstable rock mass, represented by the TIN model, is seamlessly integrated with the overall 3D slope model. This constructs a unified, high-precision digital scene from the macroscopic slope to the microscopic unstable rock mass. In step S4-3, the TIN model is converted into STL format and then imported into discrete element numerical simulation software (such as PFC3D) as the terrain boundary conditions for the boulder movement. The high accuracy of the model ensures that the simulation of the collision, friction, and other interactions between the boulder and the slope is more realistic, fundamentally improving the reliability of the motion trajectory and energy calculations.

[0135] For areas with well-developed structural planes, potential joints and fractures need to be identified through curvature analysis. After obtaining detailed information such as micro-fractures and surface spalling, a refined three-dimensional mesh model of the unstable rock mass (such as a triangular mesh or tetrahedral mesh) is constructed. The local model is then fused with the overall slope model using the ICP algorithm, with a registration error of <2mm. This can be achieved through the following methods: The first step is data preparation and input.

[0136] The target model and the source model are used as models to be fused. The target model (fixed model) is the overall three-dimensional model of the high-risk slope, which provides the macroscopic geographic coordinate system and the overall shape of the slope. The source model (model to be moved) is the detailed model of the local dangerous rock mass that needs to be fused.

[0137] The second step is initial registration.

[0138] Before performing precise ICP iterations, the two models are roughly aligned to avoid algorithm failure or getting trapped in local optima due to excessive initial position deviations. Specifically, using the POS data (derived from positioning and attitude information provided by the IMU and GNSS modules) recorded synchronously during scanning, the local unstable rock mass model is initially placed onto the corresponding approximate area on the overall slope model. This step is also known as "coarse registration".

[0139] The third step is iterative fine registration using ICP. This is the core of the fusion process, an automatic, iterative calculation process, as detailed below: First, search for the nearest point pair: For each point on the local unstable rock mass model (source model), search for the nearest point on the overall slope model (target model) to form a point pair.

[0140] Next, the optimal transformation is calculated: based on all the point pairs found in the previous step, an optimal spatial transformation (usually including a rotation matrix and a translation vector) is calculated that minimizes the overall distance error between all point pairs. This calculation typically uses optimization methods such as least squares.

[0141] Next, an application transformation is performed: the calculated rotation and translation transformations are applied to the entire local unstable rock mass model, bringing it closer to the overall slope model.

[0142] Finally, iteration and convergence checks are performed: steps one through three are repeated. After each iteration, the average distance or root mean square error between all point pairs is calculated. If the reduction in this error value compared to the previous iteration is less than a preset threshold, or if the preset maximum number of iterations is reached, the algorithm stops, indicating that it has "converged," meaning registration is complete.

[0143] Step S4-3: Discrete element simulation of the rockfall path on a high-risk slope.

[0144] First, the model is imported and parameters are assigned. Specifically, the modeling information obtained in the previous steps is converted to STL format and imported into the discrete element numerical simulation software PFC 3D. The particle size distribution and contact mechanics model are then set. Based on the physical and mechanical parameters measured in step S1, the bond strength (for rock mass integrity), equivalent damping coefficient (vegetation correction term), contact mechanics parameters, and boundary conditions are set in the software. As an optional embodiment, the particle size distribution can be set to 0.1-0.5 μm, and the contact mechanics model can be either Hertz-Mindlin or a linear bond model.

[0145] Subsequently, machine learning models were used to train and optimize motion parameters. Specifically, Gaussian process regression (GPR) was used as the training set with a large amount of simulated data from small-scale boulder rolling experiments (different heights and initial positions) to predict the horizontal distance and total kinetic energy of the boulder. Support vector machine (SVM) was used to classify and regress the bouncing height of the experimental boulder, extracting the main factors affecting the bouncing (slope angle, surface roughness, etc.). Finally, a neural network was used to input the outputs of GPR and SVM along with more environmental variables (vegetation residue, local unevenness) to comprehensively predict the motion characteristic parameters of the entire boulder rolling process.

[0146] Next, a stone-rolling process simulation is performed. Specifically, in PFC 3D, based on the results of the above machine learning model, the initial position of the rolling stone (rock detachment point), the gravity field (9.81 m / s²), and the terrain boundary (slope surface model) are defined. Large-scale Monte Carlo simulations are then conducted, recording information such as the path, energy decay, and number of collisions for each rolling stone during the simulation process. Monte Carlo simulation is a large-scale numerical computation method based on probability statistics and random sampling. Its core idea is to simulate the uncertainty or randomness existing in complex systems through repeated random trials and statistical analysis, thereby evaluating the possible evolution of the system. Using this method, all simulation results are statistically analyzed to obtain key parameters such as a heatmap of rolling stone path coverage probability, the probability distribution of energy peaks, and the statistical significance of the most dangerous collision area.

[0147] Finally, the stone-rolling path is extracted. This involves importing all simulated trajectories into a GIS or 3D visualization platform, performing heatmap analysis based on stone-rolling frequency, and calculating the most probable stone-rolling path. The stone-rolling process is then used to generate a 3D animation and contour heatmap, ultimately yielding the stone-rolling path of the target high-risk slope. As an optional implementation, the smallest area covering 95% of the stone-rolling trajectories can be selected as the most probable stone-rolling path.

[0148] Example 2 This embodiment discloses a rockfall path analysis system for steep slopes based on refined modeling.

[0149] A high-steep slope boulder path analysis system based on refined modeling includes: The survey and parameter acquisition module is configured to: survey high-risk slopes, analyze the slope gradient and material composition of the slopes, and obtain slope parameters; The high-risk slope 3D model construction module is configured to: use UAV oblique photogrammetry to take multi-angle photos of the high-risk slope to obtain point cloud data, and construct a 3D model of the high-risk slope based on the obtained point cloud data; The refined 3D scanning module is configured to: perform structural surface analysis on the constructed 3D model of the high-risk slope, identify the dangerous rock areas of the high-risk slope, and perform refined 3D scanning on the dangerous rock areas; The rockfall path simulation module is configured to: construct a dangerous rock mass model based on the refined 3D scanning results; import the modeling information of the constructed dangerous rock mass model into discrete element numerical simulation software; use the obtained slope parameters as input; optimize the rockfall motion parameters using a machine learning model; and simulate the rockfall process. The rockfall path analysis module is configured to repeatedly perform simulation calculations and statistically analyze the results to ultimately obtain the rolling path of dangerous rocks on high-risk slopes.

[0150] Example 3 The purpose of this embodiment is to provide a computer-readable storage medium.

[0151] A computer-readable storage medium having a computer program stored thereon that, when executed by a processor, implements the steps in the method for analyzing the rockfall path of steep slopes based on refined modeling as described in Embodiment 1 of this disclosure.

[0152] Example 4 The purpose of this embodiment is to provide an electronic device.

[0153] An electronic device includes a memory, a processor, and a program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps in the method for analyzing the rockfall path of steep slopes based on refined modeling as described in Embodiment 1 of this disclosure.

[0154] The steps and methods involved in the apparatuses of Embodiments 2, 3, and 4 above correspond to those in Embodiment 1. For specific implementation details, please refer to the relevant description section of Embodiment 1. The term "computer-readable storage medium" should be understood as a single medium or multiple media including one or more instruction sets; it should also be understood as including any medium capable of storing, encoding, or carrying an instruction set for execution by a processor and enabling the processor to perform any of the methods in this invention.

[0155] Those skilled in the art will understand that the modules or steps of the present invention described above can be implemented using general-purpose computer devices. Optionally, they can be implemented using computer-executable program code, thereby allowing them to be stored in a storage device for execution by a computer device, or they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. The present invention is not limited to any particular combination of hardware and software.

[0156] While the specific embodiments of the present invention have been described above in conjunction with the accompanying drawings, this is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art without creative effort based on the technical solutions of the present invention are still within the scope of protection of the present invention.

Claims

1. A method for analyzing the path of boulder shearing on steep slopes based on refined modeling, characterized in that, include: The high-risk slopes are investigated and the slope and material composition are analyzed to obtain the slope parameters. UAV oblique photogrammetry was used to take multi-angle photos of the high-risk slope to obtain point cloud data, and a three-dimensional model of the high-risk slope was constructed based on the obtained point cloud data. Structural surface analysis was performed on the constructed 3D model of the high-risk slope to identify the dangerous rock areas; and a detailed 3D scan of the dangerous rock areas was conducted. Based on the detailed 3D scanning results, a model of the unstable rock mass was constructed; The modeling information of the constructed dangerous rock mass model is imported into the discrete element numerical simulation software. The obtained slope parameters are used as input, and the rolling stone movement parameters are optimized using a machine learning model to simulate the rolling stone process. By repeating the simulation calculations multiple times and statistically analyzing the results, the rolling path of the dangerous rocks on the high-risk slope was finally obtained.

2. The method for analyzing the path of bouldering on steep slopes based on refined modeling as described in claim 1, characterized in that, Structural surface analysis was performed on the constructed 3D model of the high-risk slope to identify the dangerous rock areas of the high-risk slope. This included: firstly, large-scale structural surfaces were initially extracted based on the improved RANSAC shape detection method; then, the segmentation parameters were corrected by using the point cloud normal vector difference and feature final value in combination with the KD-tree accelerated region growing algorithm to achieve accurate segmentation of the structural surfaces.

3. The method for analyzing the path of bouldering on steep slopes based on refined modeling as described in claim 2, characterized in that, After accurately segmenting the structural planes, firstly, the orientation, spacing, and extension characteristics of each structural plane are extracted; then, based on the intersection relationship of each structural plane, the bare rock surface is divided into several independent rock mass units; finally, based on the block integrity, the degree of development of the free face, and the gravity driving force, the regions of dangerous rock masses and stable rock masses are marked to identify the dangerous rock areas of high-risk slopes.

4. The method for analyzing the path of bouldering on steep slopes based on refined modeling as described in claim 1, characterized in that, The process of conducting a detailed 3D scan of the dangerous rock area includes: first, using a 3D laser scanner to create a detailed model of the dangerous rock area and setting up photogrammetric control points for the identified dangerous rock masses; then, dynamically adjusting the laser emission frequency and scanning resolution based on the surface reflectivity of the dangerous rock masses and ambient lighting conditions.

5. The method for analyzing the path of bouldering on steep slopes based on refined modeling as described in claim 4, characterized in that, When conducting a detailed 3D scan of the dangerous rock area, multiple scanning points were set up, and the spacing between each station and the overlap rate of adjacent stations were controlled within a preset range. In addition, by synchronously recording POS data, splicing errors caused by terrain undulations were eliminated.

6. The method for analyzing the path of bouldering on steep slopes based on refined modeling as described in claim 1, characterized in that, Based on the refined 3D scanning results, a model of the unstable rock mass was constructed, including: First, using non-uniform rational B-spline surface construction technology and combined with a point cloud density adaptive triangulation generation algorithm, a TIN model with millimeter-level accuracy was established while maintaining the angular characteristics of the unstable rock mass; then, potential joints and fissures in the structural surface development area were identified through curvature analysis; finally, a refined 3D mesh model of the unstable rock mass was constructed by combining the obtained detailed information, and the local model was fused with the overall slope model using the ICP algorithm to obtain the unstable rock mass model.

7. The method for analyzing the path of bouldering on steep slopes based on refined modeling as described in claim 1, characterized in that, The optimization of rolling stone motion parameters using machine learning models includes: First, using Gaussian process regression with a large amount of small-scale rolling stone test simulation data as the training set, the horizontal movement distance and total kinetic energy of the rolling stone are predicted; then, support vector machine is used to classify and regress the bounce height of the experimental rolling stone to extract the main factors affecting the bounce; finally, neural network is used to comprehensively predict the motion characteristic parameters of the entire rolling stone process.

8. A system for analyzing the path of boulder rolling on steep slopes based on refined modeling, characterized in that: include: The survey and parameter acquisition module is configured to: survey high-risk slopes, analyze the slope gradient and material composition of the slopes, and obtain slope parameters; The high-risk slope 3D model construction module is configured to: use UAV oblique photogrammetry to take multi-angle photos of the high-risk slope to obtain point cloud data, and construct a 3D model of the high-risk slope based on the obtained point cloud data; The refined 3D scanning module is configured to: perform structural surface analysis on the constructed 3D model of the high-risk slope, identify the dangerous rock areas of the high-risk slope, and perform refined 3D scanning on the dangerous rock areas; The rockfall path simulation module is configured to: construct a dangerous rock mass model based on the refined 3D scanning results; import the modeling information of the constructed dangerous rock mass model into discrete element numerical simulation software; use the obtained slope parameters as input; optimize the rockfall motion parameters using a machine learning model; and simulate the rockfall process. The rockfall path analysis module is configured to repeatedly perform simulation calculations and statistically analyze the results to ultimately obtain the rolling path of dangerous rocks on high-risk slopes.

9. A computer-readable storage medium having a program stored thereon, characterized in that, When executed by the processor, the program implements the steps in the method for analyzing the rockfall path of steep slopes based on refined modeling as described in any one of claims 1-7.

10. An electronic device, comprising a memory, a processor, and a program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps in the method for analyzing the rockfall path of steep slopes based on refined modeling as described in any one of claims 1-7.

Images