Mine high and steep slope landslide machine learning prediction method based on unmanned aerial vehicle technology and multi-source data fusion

Abstract

The invention discloses a mine high and steep slope landslide machine learning prediction method based on an unmanned aerial vehicle technology and multi-source data fusion. The method comprises the steps that a high-resolution digital surface model is shot by an unmanned aerial vehicle; removing non-topographic features from the digital surface model data in ArcGIS software, then converting the digital surface model data into digital elevation model data, obtaining a surface refining model of the surface mine in the digital elevation model, taking the surface refining model as a constraint to trim the surface of the three-dimensional geologic model, and fusing factors of ArcGIS software and three-dimensional geologic model analysis by using Python to obtain a three-dimensional geologic model surface refining model; complete slope data are constructed; inputting the fused complete side slope data into a machine learning algorithm training model, and generating a side slope landslide susceptibility prediction model; the optimized model is used for carrying out landslide susceptibility prediction on new slope data. According to the method, various factors influencing the slope stability can be comprehensively analyzed, high efficiency can be kept during large-scale data processing, and the slope stability in mine engineering can be evaluated more accurately.

Description

Technical Field

[0001] The present invention relates to the technical field of landslide prediction for high and steep slopes in mines, and specifically to a machine learning prediction method for landslides for high and steep slopes in mines based on unmanned aerial vehicle technology and multi-source data fusion, which is used for intelligent monitoring and early warning of open-pit mine slopes. The method is based on high-resolution terrain information from unmanned aerial vehicle terrain-simulating flight, combined with 3D+GIS technology to perform real-time fusion data streams of multi-source data, and integrates multi-level data fusion and machine learning algorithms to efficiently predict and evaluate landslide hazard areas, thereby realizing dynamic monitoring of slope stability and quantitative analysis of disaster risks. Technical Background

[0002] With the rapid expansion of the global economy and industry, the scale and depth of open-pit mines are also increasing, especially deep open-pit mines. Although this mining method has increased the mining potential of mineral resources, it has also brought serious slope stability problems. The slopes of deep open-pit mines often have large slopes and complex geological conditions, which makes slope stability a major challenge in mining engineering. Traditional slope stability evaluation methods mainly rely on field surveys and manual measurements, but these methods are often not effective enough in deep open-pit mines. Due to the complex mining environment and harsh weather conditions, the acquisition of field data is not only limited by operations, but also often affects the accuracy of evaluation due to incomplete data. In addition, traditional methods rely on empirical formulas and rules for calculations, which show obvious limitations when faced with complex and changeable factors affecting slope stability.

[0003] In recent years, advances in drone technology and remote sensing technology have brought new solutions to mine slope monitoring. Drones can obtain digital surface models (DSM) and digital orthophotos of mines with high precision and high resolution, greatly improving the efficiency and accuracy of data acquisition. These high-resolution data help to more accurately assess the geological characteristics and stability of slopes. However, it is still a challenge to effectively integrate and deeply analyze the large amount of data obtained by drones. Current slope stability evaluation methods include qualitative analysis based on empirical formulas, quantitative evaluation based on statistical regression, and prediction models based on machine learning. Although these methods have improved prediction accuracy to a certain extent, they still have significant limitations when dealing with large amounts of data, complex influencing factors, and model complexity, especially when data samples are insufficient and data quality is uneven. Summary of the invention

[0004] In order to solve the problems existing in the prior art, the purpose of the present invention is to provide a machine learning prediction method for high and steep slope landslides in mines based on drone technology and multi-source data fusion, aiming to combine the high-precision data acquisition capabilities of drones with advanced machine learning algorithms to improve the prediction accuracy and analysis efficiency of slope landslide susceptibility. This method can not only comprehensively analyze the various factors affecting slope stability, but also maintain high efficiency when processing large-scale data, thereby achieving a more accurate assessment of slope stability in mining projects to ensure the safety and sustainable development of mining production.

[0005] In order to achieve the above-mentioned purpose, the present invention adopts the following technical measures: A machine learning prediction method for mine high and steep slope landslide based on drone technology and multi-source data fusion, comprising the following steps:

[0006] The first step is data collection: using drones to perform terrain simulation flights to collect multi-source data of the slope area, including high-resolution digital surface models taken by drones; building a three-dimensional geological model based on historical mine research reports and data, using mine geological planing maps and geological drilling data; cataloging the specific location and size of large landslides in ArcGIS software based on historical mine landslide data and field surveys;

[0007] The second step is data fusion: the drone digital surface model data is converted into digital elevation model data after removing non-topographic features in ArcGIS software, and the surface refinement model of the open-pit mine in the digital elevation model is obtained. The surface refinement model is used as a constraint to trim the surface of the three-dimensional geological model. The elevation, curvature and slope factors are analyzed using ArcGIS technology; the non-continuous factors affecting lithology and faults are analyzed in the three-dimensional geological model; at the same time, Python is used to fuse the factors analyzed by ArcGIS software and the three-dimensional geological model to construct complete slope data;

[0008] The third step is model training: select an appropriate machine learning algorithm for training, input the complete slope data fused in the second step into the training model, perform training and verification, and thus generate a prediction model for slope landslide susceptibility;

[0009] Step 4: Model optimization: Compare the precision-recall curves and ROC curves of different machine learning models, select the two models with the highest accuracy, and use soft voting to propose a high-precision coupled machine learning model to improve prediction accuracy.

[0010] Step 5. Result generation and application: The optimized model is used to predict the landslide susceptibility of new slope data. The output of the model includes the probability of landslide occurrence, the scope of impact and the time when it may occur. A landslide susceptibility assessment report is generated and the prediction results are visualized as maps or charts for reference by engineers and decision makers.

[0011] From the above, according to the assessment results, the implementation of corresponding slope stability management measures can reduce the risk of landslides.

[0012] Optionally, in the first step, a field survey is conducted on the mining area, and 17 photo control points are set, of which 3 are used for checkpoint verification. All control points use the WGS84 coordinate system and are based on the 114° central meridian.

[0013] Optionally, in the first step, when tracing the terrain-simulating flight route, the UAV maintains a height of 60 meters from the ground.

[0014] Optionally, in the first step, Context Capture is used to stitch the collected images to generate a high-resolution orthophoto model and a digital surface model.

[0015] Optionally, in the second step, ArcGIS software is used to extract and analyze elevation difference, slope, aspect, and curvature to form raster data.

[0016] Optionally, in the second step, the digital surface model data is imported into 3D Mine software for preliminary data processing and surface feature extraction. By defining the geological profile and performing software Boolean operations, the fault information is input into the software to form the fault plane and lithology distribution, and the preliminary three-dimensional solid model is obtained by software Boolean operations; then, the preliminary three-dimensional solid model is imported into Rhino software to establish a detailed three-dimensional geological model, and the refined digital surface model of the open-pit mine obtained by drone is used to constrain and repair the preliminary model, and the modeling tools are used to refine the operation to create the detailed geometry of the fault plane and integrate it with the surface curve; after the optimization is completed, the three-dimensional geological model is exported for mine design, geological analysis and risk assessment.

[0017] Optionally, in the second step, ArcGIS and Rhino are combined, and the spatial coordinate information of the raster data is extracted using ArcGIS API for Python. The coordinate transformation is performed through the pyproj.Transformer class, the XYZ coordinate data in ArcGIS is converted into a point object in Rhino, the geometric information is extracted, and the spatial distance between the landslide point and the fault point is calculated using the geopy library.

[0018] Optionally, in the third step, the extracted feature data is cleaned and standardized, including removing noise, filling missing values ​​and normalizing the data.

[0019] Optionally, in the third step, the algorithm includes random forest, logistic regression, K nearest neighbor algorithm, support vector machine or gradient boosting.

[0020] The beneficial effects of the present invention are: significantly improving the accuracy of slope stability assessment. By combining high-resolution drone data and advanced machine learning technology, the present invention can accurately extract key features of the slope, such as slope, curvature, slope direction, and geological fault and lithology information. This method improves the ability to predict landslide risks, making the prediction of landslide susceptibility on new data more reliable, thereby enhancing the prediction accuracy of the probability of landslide occurrence, the scope of impact, and the time when it may occur. Using real-time monitoring data and prediction models, the present invention realizes dynamic risk management, enabling engineering personnel to quickly respond and adjust based on the latest data, and effectively prevent potential landslide accidents. At the same time, by generating detailed reports and intuitive visualization tools such as maps and charts, it helps decision makers and engineers better understand and apply prediction information and optimize the decision-making process. In summary, the present invention provides a scientific basis to support more reasonable engineering design, construction and maintenance strategies, reduce landslide risks, ensure engineering safety and protect the environment. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the technical solution of the present invention, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0022] Figure 1 This is the UAV workflow diagram.

[0023] Figure 2 Schematic diagram of UAV terrain imitation flight.

[0024] Figure 3 Extract impact factors for ArcGIS, where:

[0025] (a) is the elevation difference of the UAV DEM;

[0026] (b) is the profile curvature distribution map within the mining area;

[0027] (c) is the slope distribution map within the mining area;

[0028] (d) is the slope distribution map within the mining area.

[0029] Figure 4This is a schematic diagram of the rock faults in the mining area.

[0030] Figure 5 Create a flow chart for the 3D geological model.

[0031] Figure 6 Compute models for machine learning.

[0032] Figure 7 Precision-recall curve for machine learning.

[0033] Figure 8 ROC curve for machine learning.

[0034] Fig. 9 This is a mining landslide prediction map. DETAILED DESCRIPTION

[0035] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0036] Example 1

[0037] A machine learning prediction method for high and steep slope landslides in mines based on UAV technology and multi-source data fusion includes the following steps:

[0038] The first step is data collection: drones are used for terrain simulation flights to collect multi-source data of the slope area, including high-resolution digital surface models (DSM) taken by drones. Based on the historical research reports and data of the mine, a three-dimensional geological model is established using the mine geological planing map and geological drilling data. Based on the historical landslide data of the mine and field investigations, the specific location and size of large landslide points are catalogued in ArcGIS software.

[0039] The second step is data fusion: the drone digital surface model (DSM) data is converted into digital elevation model (DEM) data after removing non-topographic features in ArcGIS software, and the surface refinement model of the open-pit mine in the digital elevation model (DEM) is obtained. The surface refinement model is used as a constraint to trim the surface of the three-dimensional geological model. ArcGIS technology is used to analyze factors such as elevation, curvature, and slope. Non-continuous factors such as lithology and fault effects are used in the three-dimensional geological model for analysis. At the same time, Python is used to fuse the factors analyzed by ArcGIS software and the three-dimensional geological model to construct complete slope data;

[0040] The third step is model training: select appropriate machine learning algorithms for training, including but not limited to random forest, logistic regression, K nearest neighbor algorithm, support vector machine or gradient boosting, etc. Input the complete slope data fused in the second step into the training model for training and verification, so as to generate a prediction model for slope landslide susceptibility;

[0041] Step 4: Model optimization: Compare the precision-recall curves and ROC curves of different machine learning models, select the two models with the highest accuracy, and use soft voting to propose a high-precision coupled machine learning model to improve prediction accuracy.

[0042] Step 5. Result generation and application: The optimized model is used to predict the landslide susceptibility of new slope data. The output of the model includes the probability of landslide occurrence, the scope of impact and the time when it may occur. A landslide susceptibility assessment report is generated and the prediction results are visualized as maps or charts for reference by engineers and decision makers.

[0043] Based on the assessment results, appropriate slope stability management measures are implemented to reduce the risk of landslides.

[0044] Example 2

[0045] The method described in Example 1 is applied to the slope stability assessment of a large open-pit mine, specifically including:

[0046] First, choose a drone platform equipped with high-precision GPS and advanced sensors, such as the DJI M300 RTK, equipped with a high-resolution Zenmuse P1 lens to meet the needs of high-precision data collection. Make sure the drone has sufficient batteries and perform a system check of all sensors and cameras before flight, including navigation, obstacle avoidance, and communication systems to ensure they are functioning properly.

[0047] like Figure 1 The figure shows the UAV workflow diagram, which illustrates the entire process of UAV data acquisition and processing from field to office. The process covers the stages of UAV field data collection and data processing, and finally generates high-precision geographic information products through post-processing and analysis in the office.

[0048] Specifically, during the field survey stage, 17 phase control points (3 of which are used for checkpoint verification) are laid out to ensure data accuracy. All control points use the WGS84 coordinate system, with the 114° central meridian as the benchmark. In the early processing stage, two-dimensional aerial surveys are performed using drones, aerial photography of the survey area is performed, and digital surface models (DSMs) in tif format are generated using M3D professional software. These digital surface model (DSM) data are then imported into the DJI GSR APP software to perform terrain simulation flight missions, thereby ensuring high accuracy of data collection.

[0049] During the flight planning and implementation, a detailed terrain-simulating flight route was designed based on the processed digital surface model (DSM) data. The drone maintained a ground altitude of 60 meters during the flight to ensure the consistency and accuracy of data collection. Figure 2 As shown, it is a schematic diagram of UAV terrain imitation flight, which shows that UAV terrain imitation flight requires initial terrain data and reflects the characteristics of dynamic adjustment of altitude.

[0050] Route planning was performed using DJI Pilot software, which automatically generated flight paths and optimized flight routes to ensure comprehensive data coverage. Before the flight, it was necessary to ensure that the RTK system was functioning properly and to perform a safety check on the flight environment. The drone automatically collected high-resolution images during flight to ensure coverage of all high and steep slope areas, and used overlapping shooting technology to maintain image continuity.

[0051] In terms of image processing, Context Capture is used to stitch the collected images and first build a camera model. This step restores the geometric information of the camera at the time of shooting by inputting the camera's intrinsic parameters (focal length, principal point position, distortion coefficient) and extrinsic parameters (camera position and posture at the time of shooting), thereby providing a basis for the subsequent correction process. After the camera model is established, the image is geometrically corrected by comparing the image with a known geospatial model (phase control points and DSM). This process eliminates the tilt and distortion in the image by introducing phase control points and digital surface models (DSM) to make it conform to the actual geographic location and shape, to ensure the accuracy of the data's geographic coordinates, correct possible errors, ensure the quality of the generated model, and generate a high-resolution orthophoto model and digital surface model (DSM).

[0052] As shown in Figures (3a) to (3d), ArcGIS software was used to extract and analyze elevation difference, slope, aspect, and curvature to form raster data. Specifically, in the feature extraction and analysis stage, the drone DSM data was first converted into digital elevation model (DEM) data in ArcGIS after removing non-topographic features. The elevation difference raster data was used to extract the area of ​​elevation drop and supplement the cataloged landslide data points. The slope, curvature, and aspect factors were then extracted from the generated DEM using ArcGIS software. The DEM data was imported into ArcGIS to ensure that its coordinate system was consistent with the project settings. The slope was calculated using the "Slope" tool in "Spatial Analyst Tools", the curvature was extracted using the "Curvature" tool, and the aspect was extracted using the "Aspect" tool. The processed slope, curvature, and aspect layers were displayed on the map, and the data was displayed through symbolization. Further analysis can be done using the "Identify" tool to view detailed data for a specific location, and using the "Zonal Statistics" tool to calculate regional statistics. Finally, the "Layout View" tool can be used to generate a map layout containing the analysis results.

[0053] like Figure 4 As shown, digital elevation model (DEM) data and geological profiles were collected and organized to ensure that all data were in a unified coordinate system. Through comprehensive analysis and utilization of historical data from open-pit mines, the lithological characteristics of the slope surface and the location of deep faults were accurately determined. This process combined historical geological exploration data with mining records, providing a key geological basis for slope stability evaluation.

[0054] like Figure 5 As shown in the figure, it is a flowchart for establishing a three-dimensional geological model, which shows the idea and process of building a specific three-dimensional geological model. Specifically, the digital surface model (DSM) is imported into the 3D Mine software to extract surface features and generate a preliminary model of the terrain surface (DTM). By defining the geological profile and performing Boolean operations in the software. The fault information is input into the software to form the fault surface and lithology distribution, and the preliminary three-dimensional solid model is obtained by Boolean operations in the software. Then, the preliminary three-dimensional solid model is imported into the Rhino software to establish a detailed three-dimensional geological model, and the open-pit mine refined surface model (DSM) obtained by the drone is used to constrain and repair the preliminary model. The creation of continuous surfaces (NURBS) and filling and trimming tools in the modeling tools are used for refined operations to create detailed geometric shapes of the fault surface and integrate them with the surface surface. After the optimization is completed, the three-dimensional geological model is exported to various formats such as .3DS, STL, OBJ, etc. for mine design, geological analysis and risk assessment.

[0055] In the present invention, the processed ArcGIS raster data is combined with the refined three-dimensional geological model, and high-precision data integration is achieved through the fusion of ArcGIS and Rhino. Specifically, the ArcGIS raster data involved in this embodiment includes variables such as elevation difference, aspect, slope and curvature. Using ArcGIS API for Python, the spatial coordinate information of these raster data is extracted to generate a raster data set containing XYZ coordinate attributes. Subsequently, the precise data conversion between the Rhino model coordinate system and the WGS84 coordinate system of ArcGIS is realized through the `pyproj.Transformer` class to ensure the high accuracy of the data. In the Rhino environment, the API is called through Python scripts and Rhino's Python editor to complete the data conversion and processing. The specific operation includes using the `rs.coerce3dpoint` function in the `rhinoscriptsyntax` library to convert the XYZ coordinate data extracted from ArcGIS into point objects in Rhino, and using the RhinoCommon library to extract geometric information, including fault effects and rock mass characteristics. In order to calculate the spatial geographic distance between the landslide point and the fault point, the `great_circle` function in the `geopy` library was applied. In the process of further integration and analysis of the data, the lithology category characteristics were converted into numerical form through one-hot encoding, and the fault impact and rock mass information were superimposed on the existing raster data using the RasterCalculator tool of the ArcGIS API. This integration process not only improved the accuracy and practicality of the data, but also provided detailed spatial information for geological and geographical research, and provided a high-quality sample set for the training of subsequent machine learning models. The newly generated raster dataset achieved an efficient fusion of geographic data and model data by combining XYZ coordinates with influencing factors (such as elevation difference, aspect, slope, curvature, lithology and fault information), thereby supporting the accurate prediction and assessment of slope landslide risks. When using these feature data for machine learning modeling, the data is first cleaned and standardized, including removing noise, filling missing values ​​and data normalization. Specifically, the noise in the data refers to those interfering data that are inconsistent with the actual situation or useless, such as the coordinates of the grid appearing to be greatly offset in a very small area, and the extracted grid values ​​are abnormal, etc. The abnormal values ​​are corrected by interpolation, and the missing values ​​are inferred from the adjacent values ​​before and after for supplementation. Finally, the data of different grid data is converted to the same scale through data normalization to reduce the impact of dimensional differences on model performance. These steps ensure the consistency and accuracy of the data.

[0056] like Figure 6As shown in the figure, it is a machine learning calculation model. Select an appropriate machine learning algorithm, such as random forest (RF), logistic regression (LR), K nearest neighbor algorithm (KNN), support vector machine (SVM) or gradient boosting (XGboos) to train the processed feature data. The model establishes a landslide susceptibility prediction model by analyzing the relationship between feature data and landslide historical records, and optimizes parameters to minimize the prediction error.

[0057] Figure 7 Build a precision-recall curve for machine learning. The precision-recall curve is a tool for evaluating the performance of classification models, especially for dealing with class imbalance problems. The curve helps understand the performance of the model at different decision thresholds by showing the trade-off between the model's precision and recall. Figure 8 ROC curve for machine learning. The ROC curve is an important tool for evaluating the performance of classification models, especially when dealing with binary classification problems. It helps analyze the classification ability of the model by showing the relationship between the true positive rate and the false positive rate of the model at different decision thresholds.

[0058] After the model training is completed, Figure 7 and Figure 8 As shown in the figure, according to the machine learning results, the results of random forest (RF) and gradient boosting (XGboost) are significantly better than other models. However, a single model inevitably has the risk of underfitting and insufficient generalization ability. Based on this, the integration method of soft voting in the voting classifier is used to couple the random forest and XGboost methods for analysis and prediction. Soft voting is an integrated learning method that makes the final decision by combining the prediction probabilities of multiple classifiers. The RF+XGBoost coupling model combines the advantages of random forest and XGBoost algorithms. By fusing the prediction results of the two models, the prediction value is significantly smoothed, and the bias and variance that may exist in a single model are reduced. Compared with five independent models, the RF+XGBoost coupling model achieved a prediction accuracy of 0.984 and an AUC value of 0.983, indicating that it performs well in accurately predicting the location of landslides.

[0059] like Fig. 9 As shown in the figure, the optimized model is used to predict the landslide susceptibility of new slope data. The output of the model includes the probability of landslide occurrence, the scope of impact, and the time when it may occur. The prediction results are presented in the form of reports and visualized through maps or charts. These visualization results help engineers and decision makers to intuitively understand the distribution of landslide risks and make scientific and reasonable slope management and risk control decisions.

[0060] The above description is only a specific implementation of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be understood by anyone familiar with the technology within the technical scope disclosed by the present invention should be included in the scope of the present invention.

Claims

1. A machine learning prediction method for high and steep slope landslides in mines based on drone technology and multi-source data fusion, characterized in that: The steps include: The first step is data collection: using drones to perform terrain simulation flights to collect multi-source data of the slope area, including high-resolution digital surface models taken by drones; building a three-dimensional geological model based on historical mine research reports and data, using mine geological planing maps and geological drilling data; cataloging the specific location and size of large landslides in ArcGIS software based on historical mine landslide data and field surveys; The second step is data fusion: the drone digital surface model data is converted into digital elevation model data after removing non-topographic features in ArcGIS software, and the surface refinement model of the open-pit mine in the digital elevation model is obtained. The surface refinement model is used as a constraint to trim the surface of the three-dimensional geological model. The elevation, curvature and slope factors are analyzed using ArcGIS technology; the non-continuous factors affecting lithology and faults are analyzed in the three-dimensional geological model; at the same time, Python is used to fuse the factors analyzed by ArcGIS software and the three-dimensional geological model to construct complete slope data; The third step is model training: select an appropriate machine learning algorithm for training, input the complete slope data fused in the second step into the training model, perform training and verification, and thus generate a prediction model for slope landslide susceptibility; Step 4: Model optimization: Compare the precision-recall curves and ROC curves of different machine learning models, select the two models with the highest accuracy, and use soft voting to propose a high-precision coupled machine learning model to improve prediction accuracy. Step 5. Result generation and application: The optimized model is used to predict the landslide susceptibility of new slope data. The output of the model includes the probability of landslide occurrence, the scope of impact and the time when it may occur. A landslide susceptibility assessment report is generated and the prediction results are visualized as maps or charts for reference by engineers and decision makers.

2. The machine learning prediction method for high and steep slope landslide in mines based on drone technology and multi-source data fusion according to claim 1 is characterized in that: In the first step, a field survey is conducted in the mining area, and 17 photo control points are set, of which 3 are used for checkpoint verification. All control points use the WGS84 coordinate system and are based on the 114° central meridian.

3. The machine learning prediction method for high and steep slope landslide in mines based on drone technology and multi-source data fusion according to claim 1 is characterized in that: In the first step, when following the terrain-simulating flight route, the UAV maintains a height of 60 meters from the ground.

4. The machine learning prediction method for mine high and steep slope landslide based on UAV technology and multi-source data fusion according to claim 1 is characterized in that: In the first step, Context Capture is used to stitch the collected images to generate a high-resolution orthophoto model and a digital surface model.

5. The machine learning prediction method for high and steep slope landslide in mines based on drone technology and multi-source data fusion according to claim 1 is characterized in that: In the second step, ArcGIS software is used to extract and analyze elevation difference, slope, aspect, and curvature to form raster data.

6. The machine learning prediction method for high and steep slope landslide in mines based on UAV technology and multi-source data fusion according to claim 1 is characterized in that: In the second step, the digital surface model data is imported into the 3D Mine software for preliminary data processing and surface feature extraction. By defining the geological profile and performing software Boolean operations, the fault information is input into the software to form the fault plane and lithology distribution, and the preliminary three-dimensional solid model is obtained by using software Boolean operations. Next, the preliminary three-dimensional solid model is imported into the Rhino software to establish a detailed three-dimensional geological model. The refined digital surface model of the open-pit mine obtained by the drone is used to constrain and repair the preliminary model. The modeling tool is used for refined operations to create the detailed geometry of the fault plane and integrate it with the surface curve. After optimization is completed, the 3D geological model is exported for mine design, geological analysis and risk assessment.

7. The machine learning prediction method for high and steep slope landslide in mines based on drone technology and multi-source data fusion according to claim 1 is characterized in that: In the second step, ArcGIS and Rhino were combined, and the spatial coordinate information of the raster data was extracted using ArcGIS API for Python. The coordinate transformation was performed through the pyproj.Transformer class, the XYZ coordinate data in ArcGIS was converted into point objects in Rhino, the geometric information was extracted, and the spatial distance between the landslide point and the fault point was calculated using the geopy library.

8. The machine learning prediction method for high and steep slope landslide in mines based on drone technology and multi-source data fusion according to claim 1 is characterized in that: In the third step, the extracted feature data is cleaned and standardized, including removing noise, filling missing values ​​and normalizing data.

9. The machine learning prediction method for high and steep slope landslide in mines based on drone technology and multi-source data fusion according to claim 1 is characterized in that: In the third step, the algorithm includes random forest, logistic regression, K nearest neighbor algorithm, support vector machine or gradient boosting.

Images