A method and system for identifying mine safety hazards based on drones

Abstract

The present invention relates to the technical field of mine safety hazard identification, and in particular to a mine safety hazard identification method and system based on drones. The method comprises the following steps: using a drone camera to obtain a global image of the mining area and construct a three-dimensional model; then, by restoring the weathering and erosion mechanism, combining the three-dimensional model and the global image, the probability of rock and soil landslides is estimated, and the landslide entrainment potential energy increment is calculated; then, according to the landslide probability and entrainment force increment data, the local area of ​​the mining area is assigned a risk level; finally, based on the risk level, a mine safety hazard identification optimization model is constructed, and the model is sent to a terminal for hazard identification. The present invention makes the mine safety hazard identification technology more perfect by optimizing the mine safety hazard identification technology.

Description

Technical Field

[0001] The present invention relates to the technical field of mine safety hazard identification, and in particular to a mine safety hazard identification method and system based on an unmanned aerial vehicle. Background Art

[0002] Drone technology has become an important tool for mine safety monitoring due to its high efficiency, flexibility, low cost and high precision. Equipped with high-definition cameras, laser radar (LiDAR), infrared imaging, thermal imaging and other sensors, drones can achieve large-scale, high-resolution real-time monitoring of mining areas, and can quickly complete the inspection of mining areas without being restricted by complex terrain and environment. The flight control system of drones can also realize automatic flight according to preset routes to ensure the continuity and systematicness of monitoring data. Using drones to identify potential safety hazards in mines, first of all, the mining area, slopes, stockpiles, etc. of the mine can be monitored in detail through real-time shooting and image recognition technology, so as to identify existing safety hazards, such as rock landslides, unstable slopes, and excessive mining depth. In addition, drone-based laser radar can measure the terrain changes of mines with high precision, and then evaluate the stability of the terrain structure during the mining process, and predict dangerous areas in advance. The application of thermal imaging technology can also help discover hidden dangers such as underground pipeline leakage and equipment overheating, further enhancing the comprehensiveness of safety monitoring. However, the traditional drone-based mine safety hazard identification method has the problems of large errors in geological risk assessment of open-pit non-metallic mines and low early warning accuracy for open-pit non-metallic mines. Summary of the invention

[0003] Based on this, it is necessary to provide a method and system for identifying mine safety hazards based on drones to solve at least one of the above technical problems.

[0004] To achieve the above purpose, a method for identifying safety hazards in mines based on drones is provided, the method comprising the following steps:

[0005] Step S1: using an airborne camera of a drone to collect a global image of the open-pit non-metallic mine to obtain a global image of the open-pit non-metallic mine; constructing a three-dimensional model of the open-pit non-metallic mine based on the global image of the open-pit non-metallic mine to obtain a three-dimensional model of the open-pit non-metallic mine;

[0006] Step S2: performing weathering erosion mechanism restoration on the three-dimensional model of the open-pit non-metallic mine to obtain weathering erosion mechanism data; performing rock and soil landslide probability estimation on the three-dimensional model of the open-pit non-metallic mine according to the weathering erosion mechanism data and the global image of the open-pit non-metallic mine to obtain rock and soil landslide probability data; performing landslide entrainment potential energy increment calculation on the rock and soil landslide probability data to obtain landslide entrainment potential energy increment data;

[0007] Step S3: assigning a local risk level to the three-dimensional model of the open-pit non-metallic mine according to the rock and soil landslide probability data and the landslide entrainment potential energy increment data to obtain the local risk level of the open-pit non-metallic mine;

[0008] Step S4: construct a mine safety hazard identification model based on the local risk level of the open-pit non-metallic mine to obtain a mine safety hazard identification optimization model; send the mine safety hazard identification optimization model to the terminal to execute the mine safety hazard identification method.

[0009] Preferably, step S1 comprises the following steps:

[0010] Step S11: using the airborne camera of the UAV to collect a global image of the open-pit non-metallic mine to obtain a global image of the open-pit non-metallic mine;

[0011] Step S12: performing color distribution ratio harmonization on the global image of the open-pit non-metallic mine to obtain a global color harmonized image of the open-pit non-metallic mine;

[0012] Step S13: marking the morphological points of the global color-harmonized image of the open-pit non-metallic mine to obtain the morphological point marking data of the open-pit non-metallic mine;

[0013] Step S14: constructing a three-dimensional model of the open-pit non-metallic mine according to the global color-harmonizing image of the open-pit non-metallic mine and the morphological point marking data of the open-pit non-metallic mine to obtain a three-dimensional model of the open-pit non-metallic mine.

[0014] Preferably, step S2 comprises the following steps:

[0015] Step S21: Analyze the mountain crack trend of the open-pit non-metallic mine three-dimensional model to obtain mountain crack trend data;

[0016] Step S22: based on the mountain crack trend data, the weathering and erosion mechanism of the open-pit non-metallic mine three-dimensional model is restored to obtain weathering and erosion mechanism data;

[0017] Step S23: estimating the probability of rock and soil landslide on the three-dimensional model of the open-pit non-metallic mine according to the weathering erosion mechanism data and the global image of the open-pit non-metallic mine, and obtaining rock and soil landslide probability data;

[0018] Step S24: Calculate the landslide entrainment potential energy increment for the rock and soil landslide probability data to obtain the landslide entrainment potential energy increment data.

[0019] Preferably, step S23 includes the following steps:

[0020] Step S231: performing a vertical distribution density analysis of vegetation in the mining area on the global image of the open-pit non-metallic mine to obtain the vertical distribution density of vegetation in the mining area;

[0021] Step S232: Decomposing the degree of wind erosion and wear of rock and soil in the vertical direction of the vertical distribution density of vegetation in the mining area and the three-dimensional model of the open-pit non-metallic mine according to the weathering erosion mechanism data, and obtaining the data of wind erosion and wear of rock and soil in the vertical direction;

[0022] Step S233: Based on the rock and soil wind erosion and wear degree data, the open-pit non-metallic mine three-dimensional model is evaluated for the lateral support loss of the free surface to obtain the lateral support loss data of the free surface;

[0023] Step S234: deduce the vertical spatial distribution risk difference coefficient for the lateral support loss data of the air-facing surface to obtain the vertical spatial distribution risk difference coefficient;

[0024] Step S235: Estimate the probability of rock and soil landslide based on the rock and soil wind erosion and wear degree data, the lateral support loss data of the free surface, and the vertical spatial distribution risk difference coefficient to obtain rock and soil landslide probability data.

[0025] Preferably, step S24 comprises the following steps:

[0026] Step S241: performing risk zoning processing on the rock and soil landslide probability data to obtain rock and soil landslide probability risk zoning data;

[0027] Step S242: Calculating the vertical drop between different risk areas based on the rock and soil landslide probability risk zoning data to obtain vertical drop data between different risk areas;

[0028] Step S243: Calculating the slope variance between different risk areas based on the rock and soil landslide probability risk zoning data to obtain the slope variance between different risk areas;

[0029] Step S244: simulating the kinetic energy of rock and soil carried by rainfall changes on the weathering erosion mechanism data according to the vertical drop data and the slope variance, and obtaining the kinetic energy of rock and soil carried by rainfall changes data;

[0030] Step S245: Calculate the incremental potential energy of the landslide entrainment force based on the kinetic energy data of rock and soil carried by rainfall changes to obtain the incremental potential energy data of the landslide entrainment force.

[0031] Preferably, step S244 includes the following steps:

[0032] Obtain historical rainfall data; analyze the spatial and temporal distribution of rainfall intensity in historical rainfall data to obtain spatial and temporal distribution data of rainfall intensity;

[0033] According to the vertical drop data and slope variance, the rock and soil looseness difference between different regions is analyzed for the weathering and erosion mechanism data, and the rock and soil looseness difference data between different regions are obtained;

[0034] According to the vertical drop data and slope variance, soil shear strength analysis is performed on the rock and soil looseness difference data to obtain soil shear strength data;

[0035] Based on the spatial and temporal distribution data of rainfall intensity, the rock looseness difference data and soil shear strength data between different regions were used to simulate and evaluate the erosion rock and soil carrying density, and the erosion rock and soil carrying density data were obtained;

[0036] Conduct vertical increment analysis on the scouring rock and soil carrying density data to obtain vertical increment data of carrying density;

[0037] The rainfall-variable rock-soil carrying kinetic energy data are obtained by simulating the rainfall-variable rock-soil carrying kinetic energy based on the vertical drop data and the slope variance and the vertical increment data of the carrying density.

[0038] Preferably, step S3 comprises the following steps:

[0039] Step S31: normalizing the landslide entrainment potential energy increment data to obtain normalized landslide entrainment potential energy data;

[0040] Step S32: performing convolution calculation on the normalized data of the landslide entrainment force potential energy to obtain the convolution data of the landslide entrainment force potential energy;

[0041] Step S33: assigning a local risk level to the three-dimensional model of the open-pit non-metallic mine according to the landslide entrainment potential energy convolution data and the rock and soil landslide probability data to obtain the local risk level of the open-pit non-metallic mine.

[0042] Preferably, step S33 includes the following steps:

[0043] Step S331: recalibrate the spatial coordinates of the open-pit non-metallic mine three-dimensional model to obtain the spatial recalibrated coordinates of the open-pit non-metallic mine;

[0044] Step S332: performing coordinate risk linkage analysis on the spatial recalibration coordinates of the open-pit non-metallic mine according to the landslide entrainment potential energy convolution data and the rock and soil landslide probability data to obtain coordinate risk linkage data;

[0045] Step S333: Calculate a multivariate local risk index based on the landslide entrainment potential energy convolution data, the rock and soil landslide probability data, and the coordinate risk linkage data to obtain a multivariate local risk index;

[0046] Step S334: assigning a local risk level to the open-pit non-metallic mine three-dimensional model according to the multivariate local risk index to obtain the local risk level of the open-pit non-metallic mine.

[0047] Preferably, step S4 comprises the following steps:

[0048] Step S41: constructing a mine safety hazard identification model based on the local risk level of the open-pit non-metallic mine to obtain a mine safety hazard identification model;

[0049] Step S42: performing deep optimization of the mine safety hazard identification model based on transfer learning to obtain a mine safety hazard identification optimization model;

[0050] Step S43: Send the mine safety hazard identification optimization model to the terminal to execute the mine safety hazard identification method.

[0051] Preferably, the present invention further provides a mine safety hazard identification system based on a drone, which is used to execute the mine safety hazard identification method based on a drone as described above. The mine safety hazard identification system based on a drone comprises:

[0052] The three-dimensional model building module is used to collect global images of the open-pit non-metallic mine through the airborne camera of the drone to obtain a global image of the open-pit non-metallic mine; and to build a three-dimensional model of the open-pit non-metallic mine based on the global image of the open-pit non-metallic mine to obtain a three-dimensional model of the open-pit non-metallic mine;

[0053] The entrainment potential energy increment calculation module is used to restore the weathering and erosion mechanism of the open-pit non-metallic mine three-dimensional model to obtain weathering and erosion mechanism data; estimate the probability of rock and soil landslides on the open-pit non-metallic mine three-dimensional model based on the weathering and erosion mechanism data and the global image of the open-pit non-metallic mine to obtain rock and soil landslide probability data; calculate the landslide entrainment potential energy increment on the rock and soil landslide probability data to obtain the landslide entrainment potential energy increment data;

[0054] The local risk level assignment module is used to assign a local risk level to the three-dimensional model of the open-pit non-metallic mine according to the rock and soil landslide probability data and the landslide entrainment potential energy increment data, so as to obtain the local risk level of the open-pit non-metallic mine;

[0055] The hidden danger identification model construction module is used to construct a mine safety hidden danger identification model based on the local risk level of the open-pit non-metallic mine, and obtain a mine safety hidden danger identification optimization model; the mine safety hidden danger identification optimization model is sent to the terminal to execute the mine safety hidden danger identification method.

[0056] The beneficial effect of the present invention is that the drone-mounted camera performs global image acquisition of open-pit non-metallic mines, and can quickly and comprehensively obtain high-definition images of the mining area, avoiding the time and labor costs brought by traditional manual surveys. Through these image data, the three-dimensional model of the open-pit mine can be accurately constructed to fully restore the spatial structure of the mining area. The three-dimensional model provides basic data support for subsequent mining area analysis, risk assessment, and safety hazard identification, and can accurately display the topography and mining status of the mining area. By restoring the weathering and erosion mechanism of the three-dimensional model, the evolution process of the mining area under different weathering conditions can be simulated, thereby identifying dangerous areas. The analysis of the weathering and erosion mechanism not only helps to evaluate the stability of rock and soil materials in the mining area, but also predicts the potential risks of disasters such as landslides and collapses according to the erosion process. In addition, combining the weathering and erosion mechanism data with the global image can realize the accurate estimation of the probability of rock and soil landslides in the mining area. Combined with the calculation of the potential energy increment of the landslide entrainment force, the probability and severity of the disaster can be further refined, providing accurate data support for safety management. The combination of geotechnical landslide probability data and landslide entrainment potential energy increment data makes the assessment of local risks in mining areas more scientific and accurate. By assigning specific risk levels to each local area, potential high-risk areas in the mining area can be effectively identified, thereby providing mine managers with a clear risk distribution map. This can help staff give priority to high-risk areas, strengthen monitoring and prevention measures in these areas, and reduce the threat of geological disasters such as landslides to mine safety. Based on the data of local risk levels, a mine safety hazard identification model can be constructed to achieve comprehensive monitoring and early warning of potential safety hazards in mines. The model can intelligently identify various safety hazards in mines, optimize the hazard investigation process, and make adjustments based on real-time data feedback. By sending the optimized identification model to the terminal device, mine staff can obtain the latest data on the safety status of the mining area at any time and take timely response measures. The establishment of this system helps to improve the automation and intelligence level of mine safety management, thereby effectively reducing the probability of accidents, ensuring the safety of miners' lives, and improving the overall safety and sustainable development capabilities of mines. Therefore, the present invention provides a method for identifying mine safety hazards based on drones, which solves the problems of large errors in geological risk assessment of open-pit non-metallic mines and low accuracy in early warning of open-pit non-metallic mines in the traditional method for identifying mine safety hazards based on drones, reduces the errors in geological risk assessment of open-pit non-metallic mines, and improves the accuracy of early warning of open-pit non-metallic mines. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 A schematic diagram of the steps of a method for identifying safety hazards in mines based on drones;

[0058] Figure 2 for Figure 1Detailed implementation steps of step S2 in the flowchart;

[0059] Figure 3 for Figure 1 Detailed implementation steps of step S3 in FIG. DETAILED DESCRIPTION

[0060] The technical method of the present invention is described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are 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 technicians in this field without creative work are within the scope of protection of the present invention.

[0061] In addition, the accompanying drawings are only schematic illustrations of the present invention and are not necessarily drawn to scale. The same reference numerals in the figures represent the same or similar parts, and their repeated description will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities and do not necessarily correspond to physically or logically independent entities. The functional entities can be implemented in software form, or implemented in one or more hardware modules or integrated circuits, or implemented in different networks and / or processor methods and / or microcontroller methods.

[0062] It should be understood that, although the terms "first", "second", etc. may be used herein to describe various units, these units should not be limited by these terms. These terms are used only to distinguish one unit from another unit. For example, without departing from the scope of the exemplary embodiments, the first unit may be referred to as the second unit, and similarly the second unit may be referred to as the first unit. The term "and / or" used herein includes any and all combinations of one or more of the listed associated items.

[0063] To achieve this, please refer to Figures 1 to 3 , a method for identifying safety hazards in mines based on drones, the method comprising the following steps:

[0064] Step S1: using an airborne camera of a drone to collect a global image of the open-pit non-metallic mine to obtain a global image of the open-pit non-metallic mine; constructing a three-dimensional model of the open-pit non-metallic mine based on the global image of the open-pit non-metallic mine to obtain a three-dimensional model of the open-pit non-metallic mine;

[0065] Step S2: performing weathering erosion mechanism restoration on the three-dimensional model of the open-pit non-metallic mine to obtain weathering erosion mechanism data; performing rock and soil landslide probability estimation on the three-dimensional model of the open-pit non-metallic mine according to the weathering erosion mechanism data and the global image of the open-pit non-metallic mine to obtain rock and soil landslide probability data; performing landslide entrainment potential energy increment calculation on the rock and soil landslide probability data to obtain landslide entrainment potential energy increment data;

[0066] Step S3: assigning a local risk level to the three-dimensional model of the open-pit non-metallic mine according to the rock and soil landslide probability data and the landslide entrainment potential energy increment data to obtain the local risk level of the open-pit non-metallic mine;

[0067] Step S4: construct a mine safety hazard identification model based on the local risk level of the open-pit non-metallic mine to obtain a mine safety hazard identification optimization model; send the mine safety hazard identification optimization model to the terminal to execute the mine safety hazard identification method.

[0068] In the embodiment of the present invention, reference Figure 1 FIG. 1 is a schematic diagram of a process flow of a method for identifying safety hazards in mines based on a drone according to the present invention. In this example, the method for identifying safety hazards in mines based on a drone includes the following steps:

[0069] Step S1: using an airborne camera of a drone to collect a global image of the open-pit non-metallic mine to obtain a global image of the open-pit non-metallic mine; constructing a three-dimensional model of the open-pit non-metallic mine based on the global image of the open-pit non-metallic mine to obtain a three-dimensional model of the open-pit non-metallic mine;

[0070] In the embodiment of the present invention, a drone equipped with a high-resolution camera is first used to collect global images of an open-pit non-metallic mine. The onboard camera of the drone must have automatic exposure adjustment and high dynamic range technology (HDR) to cope with the changes in illumination in the open-pit mine area. The resolution of the camera should be at least 4000×3000 pixels to ensure that the collected image has sufficient details. During the image acquisition process, multiple flight altitudes and flight trajectories are selected to cover every corner of the open-pit mine area to ensure that all-round image data is obtained. After the image acquisition is completed, the obtained global image is preprocessed by an image processing algorithm, including denoising, image enhancement and geometric correction. The denoising algorithm uses Gaussian filtering technology to effectively remove random noise in the image; the image enhancement uses a histogram equalization method to improve the contrast and brightness of the image and make the details more prominent; the geometric correction uses an image registration algorithm to splice multiple collected images so that images at different angles can be accurately aligned to eliminate errors caused by differences in camera viewing angles. Next, a three-dimensional modeling technology based on structured light reconstruction is used to construct a three-dimensional model of the open-pit non-metallic mine based on the global image. Structured light technology projects a known light pattern onto the target surface and calculates the depth information of each point by capturing the deformation of the light spot. This process is combined with image registration technology to accurately match the texture features in the image, ultimately generating a high-precision 3D model. The resulting 3D model can not only show the overall structure of the mining area, but also provide details such as rock layers, cracks, and soil layer distribution for subsequent analysis and judgment.

[0071] Step S2: performing weathering erosion mechanism restoration on the three-dimensional model of the open-pit non-metallic mine to obtain weathering erosion mechanism data; performing rock and soil landslide probability estimation on the three-dimensional model of the open-pit non-metallic mine according to the weathering erosion mechanism data and the global image of the open-pit non-metallic mine to obtain rock and soil landslide probability data; performing landslide entrainment potential energy increment calculation on the rock and soil landslide probability data to obtain landslide entrainment potential energy increment data;

[0072] In an embodiment of the present invention, the generated three-dimensional model of the open-pit non-metallic mine is restored to the mechanism of weathering and erosion. The weathering and erosion mechanism analysis is based on the physical properties, climatic conditions and terrain characteristics of the rocks in the mining area. The weathering and erosion coupling model is adopted, and the weathering degree of different parts of the mining area is calculated in combination with factors such as the weathering resistance of the rock, wind speed, and precipitation. This process is achieved by calculating the weathering index of the rock surface. The weathering index is quantitatively evaluated according to meteorological data such as precipitation, temperature, wind force, and the type of ore, and finally the weathering and erosion mechanism data is formed. Next, the weathering and erosion mechanism data and the global image of the open-pit non-metallic mine are used to estimate the probability of rock and soil landslides through a multi-factor model. The model uses probability statistics methods to evaluate the possibility of landslides in different areas based on factors such as the degree of weathering, the slope of the mining area, the distribution of cracks, and soil moisture. Through the multivariate regression analysis in statistics, combined with the weathering and erosion mechanism data, the probability of rock and soil landslides in each area is calculated. The estimation results are displayed in the form of probability values, specific to each mining area. Next, the landslide entrainment potential energy increment is calculated for the rock and soil landslide probability data. The entrainment potential energy increment refers to the additional potential energy obtained when the ore is pushed during the landslide. This calculation takes into account the slope, soil density and the area where the landslide occurs, and uses the law of conservation of energy combined with the potential energy calculation method to further analyze the energy increment of the landslide area. The formula is: entrainment potential energy increment = object mass × gravitational acceleration × height change. Finally, the landslide entrainment potential energy increment data is obtained as the basis for the next local risk assessment.

[0073] Step S3: assigning a local risk level to the three-dimensional model of the open-pit non-metallic mine according to the rock and soil landslide probability data and the landslide entrainment potential energy increment data to obtain the local risk level of the open-pit non-metallic mine;

[0074] In an embodiment of the present invention, firstly, the local risk level of the three-dimensional model of the open-pit non-metallic mine is assigned by the weighted average method according to the rock and soil landslide probability data and the landslide entrainment potential energy increment data. The specific weighting coefficient is set according to the historical landslide events, geological characteristics and environmental factors of the mining area. The landslide probability data and the potential energy increment data are respectively assigned different weights to reflect the degree of their respective impact on the landslide risk. In addition, the K-means clustering algorithm is used to partition the local risks and divide the mining area into multiple risk level areas. The risk level of each area is determined according to the weighted results of the landslide probability and the entrainment potential energy increment. Areas with higher risks will be marked as "high risk", while areas with lower risks will be marked as "low risk". This step provides basic data for the subsequent identification of mine safety hazards.

[0075] Step S4: construct a mine safety hazard identification model based on the local risk level of the open-pit non-metallic mine to obtain a mine safety hazard identification optimization model; send the mine safety hazard identification optimization model to the terminal to execute the mine safety hazard identification method.

[0076] In the embodiment of the present invention, firstly, based on the local risk level of the open-pit non-metallic mine obtained in step S3, a mine safety hazard identification model is constructed. This model mainly extracts the influencing factors of mine safety hazards by combining the historical landslide records of the mining area and the geographical and geological data of the mine through logistic regression analysis. The input variables of the model include the geological structure, weathering degree, landslide probability, potential energy of the entrainment force, etc. of the mining area, and the output result is the safety hazard level of each area. Next, the mine safety hazard identification model is deeply optimized through transfer learning. The transfer learning algorithm iteratively optimizes the identification model based on the existing mine data, so that the model can better adapt to different geological and climatic conditions in the new mining area data environment. The optimized model can provide more accurate safety hazard identification results when facing new open-pit mining area data. In the implementation process of transfer learning, pre-training is first performed through a small amount of labeled sample data, and then fine-tuning is performed using more unlabeled data to further improve the robustness of the model. Finally, the optimized mine safety hazard identification model is sent to the terminal device to execute the real-time identification method of mine safety hazards. This method uses the real-time data collected by the terminal and the transmitted model to automatically identify the safety hazards in the mining area and generate a detailed safety report.

[0077] Step S1 includes the following steps:

[0078] Step S11: using the airborne camera of the UAV to collect a global image of the open-pit non-metallic mine to obtain a global image of the open-pit non-metallic mine;

[0079] Step S12: performing color distribution ratio harmonization on the global image of the open-pit non-metallic mine to obtain a global color harmonized image of the open-pit non-metallic mine;

[0080] Step S13: marking the morphological points of the global color-harmonized image of the open-pit non-metallic mine to obtain the morphological point marking data of the open-pit non-metallic mine;

[0081] Step S14: constructing a three-dimensional model of the open-pit non-metallic mine according to the global color-harmonizing image of the open-pit non-metallic mine and the morphological point marking data of the open-pit non-metallic mine to obtain a three-dimensional model of the open-pit non-metallic mine.

[0082] In an embodiment of the present invention, a drone equipped with a high-resolution airborne camera is used to perform global image acquisition of an open-pit non-metallic mine. In order to ensure the comprehensiveness and accuracy of image acquisition, the drone needs to fly the mining area multiple times according to a predetermined flight path to cover various areas of the mining area. During each flight, the airborne camera should automatically adjust the focal length, aperture and shutter speed according to the geographical features of the mining area to cope with changes in illumination. When acquiring images, special attention should be paid to the overlap of the images to ensure that each area can be fully covered and data gaps should be avoided. The resolution of the image is required to reach at least 6000×4000 pixels to ensure that each image can provide sufficient detail information, especially when dealing with features such as rock formations, cracks and slopes in the mining area, details are particularly important. After the image acquisition is completed, all images need to be preliminarily geometrically corrected to eliminate image distortion caused by unstable drone flight paths or other factors. The color distribution ratio of the global image of the open-pit non-metallic mine is adjusted to obtain a visual image that is more in line with the actual mining area. Color distribution adjustment is mainly achieved through a color mapping algorithm, the goal of which is to balance the brightness, saturation and contrast of the colors in the image, thereby enhancing the detail performance of the image. First, the color histogram analysis of the original image is performed to calculate the distribution of each color channel (red, green and blue) in the image, calibrate the maximum and minimum values ​​of each channel, and adjust the color mapping range of the image based on these data. In order to make the adjusted image more consistent with the actual color characteristics of the open-pit mine area, the color equalization technology is used to make the color distribution of all pixels in the image more uniform. The key to this step is how to ensure that the natural visual effect of the image is not distorted after color harmony, and at the same time enhance the recognition of the image, especially when dealing with different categories such as rocks, vegetation and soil in the mining area, to ensure that the color distinction is clear and visible. After completing the color harmony, the output result is the global color harmony image of the open-pit non-metallic mine, ready for subsequent morphological analysis. The morphological point marking of the global color harmony image of the open-pit non-metallic mine is to provide basic data for the subsequent three-dimensional model construction. First, the edge detection algorithm is used to process the color-harmonized image to highlight the edges and significant features in the image. This process can accurately detect various boundary lines and change areas in the mining area image, especially the fault lines, cracks in the rock formations, and areas with uneven ground by applying the Canny edge detection algorithm. On the basis of edge detection, morphological operations such as expansion, erosion, and opening operations are further used to extract and enhance key points in the image. These morphological points include characteristic points, cracks, rock formation joints, and potential safety hazards in the mining area. The location and attributes of each marked point will be recorded and associated with the original image data. In this process, the region growing algorithm is used to identify different areas in the image based on edge detection and image segmentation, and feature points are marked in each area.The data of each feature point mark includes its spatial position, relative light intensity and pixel coordinates in the image. All the marked points form a certain structural layout in the image, which can reflect the overall characteristics of the mining area terrain and local abnormal points. The marked data will be output in the form of a data point set to provide necessary information for subsequent 3D modeling. Based on the global color harmony image of the open-pit non-metallic mine and the morphological point mark data of the open-pit non-metallic mine, the 3D model of the open-pit non-metallic mine is constructed. First, the color harmony image and the morphological point mark data are registered by image matching technology to ensure that the spatial position of each feature point corresponds to the image pixel. To achieve this, feature point matching algorithms such as SIFT (Scale Invariant Feature Transform) or SURF (Speeded Robust Features) are used to find and match corresponding feature points between multiple images. After image registration, the convolutional neural network (CNN) in deep learning is used to reconstruct the image in 3D to construct the 3D point cloud data of the mining area. Then, based on these point cloud data, the discrete point cloud data is converted into a continuous 3D surface through a triangular meshing algorithm (such as Delaunay triangulation). In this process, the point cloud needs to be denoised first to remove invalid points caused by image noise or flight errors. After denoising, the surface is smoothed using the least squares method or other fitting techniques to ensure that the generated 3D model fits the actual terrain as closely as possible. Finally, texture mapping technology is used to map the texture of the color-harmonized image onto the 3D surface model, so that the model not only has terrain morphological data, but also has a realistic visual effect. The resulting 3D model can accurately reflect the geomorphic structure of the mining area and provide detailed geographic information for subsequent risk assessment and safety hazard identification.

[0083] Step S2 includes the following steps:

[0084] Step S21: Analyze the mountain crack trend of the open-pit non-metallic mine three-dimensional model to obtain mountain crack trend data;

[0085] Step S22: based on the mountain crack trend data, the weathering and erosion mechanism of the open-pit non-metallic mine three-dimensional model is restored to obtain weathering and erosion mechanism data;

[0086] Step S23: estimating the probability of rock and soil landslide on the three-dimensional model of the open-pit non-metallic mine according to the weathering erosion mechanism data and the global image of the open-pit non-metallic mine, and obtaining rock and soil landslide probability data;

[0087] Step S24: Calculate the landslide entrainment potential energy increment for the rock and soil landslide probability data to obtain the landslide entrainment potential energy increment data.

[0088] As an example of the present invention, refer to Figure 2 As shown, in this example, step S2 includes:

[0089] Step S21: Analyze the mountain crack trend of the open-pit non-metallic mine three-dimensional model to obtain mountain crack trend data;

[0090] In the embodiment of the present invention, in step S21, the crack trend analysis of the mountain range of the three-dimensional model of the open-pit non-metallic mine is performed. First, by extracting cracks from the surface of the three-dimensional model of the open-pit non-metallic mine, the rock cracks in the mining area are systematically analyzed using digital terrain analysis technology. In order to ensure the accuracy of the crack trend analysis, the gray level co-occurrence matrix (GLCM) method is used to extract the texture features in the image, and the directionality of the rock formation cracks is analyzed by the texture features. The characteristic points of the cracks are accurately extracted by edge detection algorithms such as the Sobel operator, and on this basis, the Hough transform is used to detect and determine the main direction of the cracks. The Hough transform algorithm converts the image space into the parameter space, identifies the slope and intercept of each straight line in the image, and thus accurately obtains the main direction of the crack. Then, combined with the geographic information of the three-dimensional model, the vector analysis method is further used to calculate the azimuth of the crack, and combined with the slope, altitude and other geographical factors of the mining area to obtain the crack trend data of each area. These data include the directionality, length, distribution range and density of the cracks, forming a crack trend data set with spatial information, which provides a basis for subsequent weathering and erosion analysis.

[0091] Step S22: based on the mountain crack trend data, the weathering and erosion mechanism of the open-pit non-metallic mine three-dimensional model is restored to obtain weathering and erosion mechanism data;

[0092] In an embodiment of the present invention, the weathering erosion mechanism of the three-dimensional model of the open-pit non-metallic mine is restored based on the crack trend data of the mountain range. According to the crack trend data, the weathering degree and weathering mode of the rock strata in the mining area can be determined. First, the relationship between the crack trend and the geological structure of the mining area is used, combined with the mechanical properties of the rock strata, such as compressive strength and shear strength, to infer the weathering process of the rock through a mechanical model. The tribological principle in rock mechanics is used to analyze the expansion law of cracks in different directions, and further derive the geometric form of weathering erosion. During the weathering process, the expansion of cracks is affected by factors such as water flow, wind force and chemical substances. Therefore, these environmental factors need to be considered when restoring the weathering mechanism. According to the topography, climate and hydrological data of the mining area, a fluid dynamics model is used to simulate the penetration and erosion of water flow in the cracks, and then the weathering rate of the rock surface is inferred. By combining the specific geological environment data of the mining area, the weathering erosion mechanism of the open-pit mine is gradually restored, and the weathering erosion mechanism data is obtained, including parameters such as the extension depth of the cracks, the weathering rate, and the erosion morphology. These data are of great significance for predicting rock and soil landslides and other disasters, and provide basic data support for subsequent steps.

[0093] Step S23: estimating the probability of rock and soil landslide on the three-dimensional model of the open-pit non-metallic mine according to the weathering erosion mechanism data and the global image of the open-pit non-metallic mine, and obtaining rock and soil landslide probability data;

[0094] In an embodiment of the present invention, the probability of rock and soil landslide is estimated based on weathering erosion mechanism data and the global image of open-pit non-metallic mines. First, the stability of the rock and soil layers in different areas of the mining area is determined according to the weathering erosion mechanism data obtained in the previous step. The landslide stability analysis method in geotechnical engineering is used to consider the slope, soil thickness, rock and soil strength and hydrological conditions in different areas of the mining area to calculate the landslide potential of each area. Combined with the global image of the mining area, the image processing technology is used to extract the elevation change information of the mining area, and the digital elevation model (DEM) is used to analyze the slope distribution and slope morphology of the mining area. Then, based on the multi-factor landslide stability analysis model, a comprehensive evaluation is performed on each area, and the model includes multiple factors such as weathering erosion depth, crack extension, slope, soil type, precipitation, etc. On this basis, the probability of occurrence of rock and soil landslides in various areas of the mining area is calculated using a landslide probability estimation method, such as a probability analysis method or a fuzzy logic method. Finally, the rock and soil landslide probability data are obtained, which are specifically expressed as the landslide risk value of each area, which is convenient for subsequent risk assessment and safety hazard identification.

[0095] Step S24: Calculate the landslide entrainment potential energy increment for the rock and soil landslide probability data to obtain the landslide entrainment potential energy increment data.

[0096] In the embodiment of the present invention, the landslide entrainment potential energy increment is calculated for the rock and soil landslide probability data. First, the calculation of the landslide entrainment potential energy increment is based on the landslide probability data of the potential landslide area and the topography, rock and soil properties and hydrological conditions of the mining area. The entrainment potential energy refers to the potential energy stored due to gravity when the soil or rock slides along the slope during the landslide process. In order to calculate this increment, it is first necessary to calculate the mass center and sliding volume of each potential landslide area through the dynamic formula. This process uses the relationship between kinetic energy and potential energy in physical mechanics, and obtains the increment of potential energy by calculating the interaction between the landslide body and gravity during the sliding process. Then, according to the slope and the landslide volume, the entrainment force increment occurring during the landslide process is further calculated. The entrainment force increment is the friction and interaction force generated between the landslide body and the surrounding materials (such as mud, rocks, etc.) during the movement of the landslide body. Using the landslide dynamics model, combined with the actual terrain and soil mechanics parameters, the entrainment potential energy increment of each potential landslide area is calculated. Through the integration method, the incremental data of the landslide entrainment potential energy of the entire mining area are obtained. This data reflects the kinetic energy pressure that different areas are borne when a landslide occurs, and is of great significance for predicting landslide disasters in mining areas and their degree of hazard.

[0097] Step S23 includes the following steps:

[0098] Step S231: performing a vertical distribution density analysis of vegetation in the mining area on the global image of the open-pit non-metallic mine to obtain the vertical distribution density of vegetation in the mining area;

[0099] Step S232: Decomposing the degree of wind erosion and wear of rock and soil in the vertical direction of the vertical distribution density of vegetation in the mining area and the three-dimensional model of the open-pit non-metallic mine according to the weathering erosion mechanism data, and obtaining the data of wind erosion and wear of rock and soil in the vertical direction;

[0100] Step S233: Based on the rock and soil wind erosion and wear degree data, the open-pit non-metallic mine three-dimensional model is evaluated for the lateral support loss of the free surface to obtain the lateral support loss data of the free surface;

[0101] Step S234: deduce the vertical spatial distribution risk difference coefficient for the lateral support loss data of the air-facing surface to obtain the vertical spatial distribution risk difference coefficient;

[0102] Step S235: Estimate the probability of rock and soil landslide based on the rock and soil wind erosion and wear degree data, the lateral support loss data of the free surface, and the vertical spatial distribution risk difference coefficient to obtain rock and soil landslide probability data.

[0103] In an embodiment of the present invention, the vertical distribution density of vegetation in the mining area is analyzed for the global image of the open-pit non-metallic mine. First, the vegetation area of ​​the mining area is extracted from the global image of the open-pit non-metallic mine obtained by the drone, and the vegetation and non-vegetation areas are distinguished by image segmentation technology. The image is converted from the RGB color space to the HSV (hue, saturation, brightness) or CIE Lab color space by using the color space conversion technology, and the color characteristics of the vegetation are extracted from it, and then the vegetation pixels in the mining area are identified by the threshold segmentation method. In order to improve the segmentation accuracy, the K-means clustering algorithm can be used to perform multi-class segmentation on the image to obtain the vegetation area. Next, the morphological processing of the vegetation area, such as expansion and corrosion operations, is performed on the vegetation area by using morphological-based technology to remove noise and small non-vegetation areas. Then, the horizontal distribution density of vegetation in the mining area is obtained by calculating the pixel ratio of the vegetation area. In the vertical direction, the height change of the vegetation area is calculated in combination with the digital elevation model (DEM) of the mining area, and the vertical distribution density of vegetation in each height layer is obtained according to the distribution of vegetation area at different height levels. Finally, the vertical distribution density data of vegetation in the mining area is obtained, which reflects the distribution of vegetation at different height levels in the mining area. This data provides important environmental information for the subsequent analysis of wind erosion and wear. Based on the weathering erosion mechanism data, the vertical distribution density of vegetation in the mining area and the three-dimensional model of open-pit non-metallic mines are decomposed into the degree of wind erosion and wear of rock and soil in the vertical direction. First, according to the mountain crack trend data in step S21 and the weathering erosion mechanism data in step S22, the weathering degree of the rock and soil layer in the mining area is analyzed. On this basis, combined with wind data (such as wind speed, wind direction, wind duration, etc.), through the relationship between wind and rock and soil surface contact, the wind erosion model in fluid dynamics is used to estimate the degree of wind erosion and wear of the rock and soil surface. The model takes into account the changes in the slope, vegetation coverage, rock and soil strength and wind action in the mining area, and simulates the erosion effect of wind on rock and soil at different heights. The vertical distribution density of vegetation in the mining area is combined with wind erosion to analyze the influence of wind in areas with different vegetation coverage densities. In order to further improve the accuracy of the profile level, the wind action is combined with the slope information of the three-dimensional model of the mining area through the wear simulation method based on fractal theory, and the longitudinal distribution map of wind erosion is established in the three-dimensional model. Finally, the wind erosion wear degree data between each height level is obtained, which represents the wind wear of rock and soil at different vertical heights, and provides specific data support for the subsequent assessment of the loss of support on the free surface. Based on the wind erosion wear degree data of rock and soil, the lateral support loss of the free surface of the three-dimensional model of the open-pit non-metallic mine is evaluated. The lateral support of the free surface refers to the ability of the rock and soil layer to maintain its stability under the action of external forces, which is usually closely related to factors such as slope, wind erosion degree, and vegetation coverage. First, combining the three-dimensional model of the mining area and the wind erosion wear degree data, the free surface area with lateral support loss is selected.By calculating the slope change, crack direction and weathering erosion depth of the free surface of the mining area, the mechanical analysis method is used to calculate the stability of the rock and soil under different lateral support conditions. In particular, the friction angle and shear strength model in the landslide analysis are used in combination with the mechanical parameters of the rock and soil to evaluate the support loss. By comparing the thickness of the rock and soil layer, weathering depth and wind erosion wear degree in different free surface areas, it is determined which areas have a greater risk of lateral support loss. This process requires multiple calculations of the digital elevation model of the mining area to evaluate the support loss of each free surface under the current wind erosion conditions. Finally, the lateral support loss data of each free surface is obtained, including the specific risk area and loss degree, to provide quantitative data support for subsequent risk assessment. The vertical spatial distribution risk difference coefficient of the free surface lateral support loss data is deduced. First, based on the free surface lateral support loss data obtained in step S233, the risk distribution of each free surface area in the mining area is analyzed using a spatial statistical analysis method. The risk difference coefficient of each area is calculated to reflect the degree of risk difference in different areas. Specifically, spatial autocorrelation analysis methods, such as Moran's I index, are used to evaluate the spatial correlation of the support loss of the free surface. This method can determine which areas have a high clustering of support loss and identify potential high-risk areas. Then, based on the spatial distribution of free surface loss data, the risk assessment model is used to combine the terrain characteristics, weathering erosion degree and vegetation density data of the mining area to deduce the risk difference coefficient in the vertical space of the mining area. This coefficient reflects the difference in support loss between different height levels and further reveals the risk change trend of different vertical levels in the mining area. Finally, the vertical spatial distribution risk difference coefficient is obtained to provide spatial data support for the subsequent landslide probability estimation. The probability of rock and soil landslide is estimated based on the wind erosion wear degree data of the rock and soil, the lateral support loss data of the free surface and the vertical spatial distribution risk difference coefficient. First, the multi-factor landslide probability estimation model is used to calculate the landslide probability of each area, combining the wind erosion wear degree data, free surface support loss data and risk difference coefficient obtained in the previous steps. The model combines the rock and soil strength, weathering depth, slope, wind erosion and support loss in different areas of the mining area, and comprehensively evaluates each factor through weighted analysis. Specifically, each factor is assigned a different weight coefficient according to its influence on the occurrence of landslides, and the landslide probability value of each area is finally calculated by weighted sum. To ensure the accuracy of the calculation, the weighted average method is used to quantitatively analyze the landslide risk in each area, and the spatial interpolation method is combined to smooth the landslide probability data, so as to obtain the landslide probability data of each area in the mining area. These data provide a scientific basis for landslide warning and safety hazard identification in mining areas, and can effectively guide the safety management and monitoring of mines.

[0104] Step S24 includes the following steps:

[0105] Step S241: performing risk zoning processing on the rock and soil landslide probability data to obtain rock and soil landslide probability risk zoning data;

[0106] Step S242: Calculating the vertical drop between different risk areas based on the rock and soil landslide probability risk zoning data to obtain vertical drop data between different risk areas;

[0107] Step S243: Calculating the slope variance between different risk areas based on the rock and soil landslide probability risk zoning data to obtain the slope variance between different risk areas;

[0108] Step S244: simulating the kinetic energy of rock and soil carried by rainfall changes on the weathering erosion mechanism data according to the vertical drop data and the slope variance, and obtaining the kinetic energy of rock and soil carried by rainfall changes data;

[0109] Step S245: Calculate the incremental potential energy of the landslide entrainment force based on the kinetic energy data of rock and soil carried by rainfall changes to obtain the incremental potential energy data of the landslide entrainment force.

[0110] In an embodiment of the present invention, risk zoning is performed on the rock and soil landslide probability data. First, based on the rock and soil landslide probability data obtained in step S235, risk levels are divided for different areas in the mining area. This step first divides the mining area into multiple risk intervals based on the value of the landslide probability data. For example, the interval of the landslide probability can be set to five levels: 0-0.2, 0.2-0.4, 0.4-0.6, 0.6-0.8, and 0.8-1.0. Each level represents a different risk level, 0-0.2 is a low risk area, 0.2-0.4 is a medium-low risk area, 0.4-0.6 is a medium risk area, 0.6-0.8 is a medium-high risk area, and 0.8-1.0 is a high risk area. In specific implementation, the landslide probability value of each unit can be calculated by traversing the grid units in each mining area, and it can be classified according to the preset risk level threshold. In order to improve the accuracy of risk zoning, it is also possible to fine-tune the terrain undulation and other characteristic information. For example, even if the landslide probability is low in an area with a steeper slope, it can be divided into a higher risk area. In addition, the landslide probability data can be corrected by weighted average method according to environmental factors such as vegetation coverage in each area to ensure the accuracy and rationality of the zoning results. Finally, the rock and soil landslide probability risk zoning data is obtained to provide quantitative assessment and spatial distribution information of the risks in different areas of the mining area. The vertical drop between different risk areas is calculated based on the rock and soil landslide probability risk zoning data. First, according to the risk zoning data obtained in step S241, the boundary lines of each risk interval are extracted, and the boundaries of each area are analyzed. Then, the digital elevation model (DEM) of the mining area is used to calculate the vertical drop within the boundary of each risk area. The vertical drop refers to the elevation difference from the lowest point to the highest point of a risk area. This parameter can reflect the severity of the terrain undulation in the mining area. In the specific operation, the elevation data of each risk area is first extracted from the digital elevation model, and then the difference between the maximum height and the minimum height of all points in the area is calculated to obtain the vertical drop. In order to ensure the accuracy of the calculation results, the risk area can be gridded, the maximum and minimum values ​​of the elevation are calculated in each grid unit, and then the vertical drop of the entire area is calculated by the maximum and minimum values ​​of the entire area. This data helps to determine the degree of terrain change in different risk areas in the mining area, and further supports the assessment of landslide risks. The slope variance between different risk areas is calculated based on the rock and soil landslide probability risk zoning data. The slope variance reflects the changes in terrain undulations in a region, and can quantitatively analyze the stability of different risk areas in the mining area. First, based on the risk zoning data obtained in step S241, the boundaries of each area are extracted, and the slope in each area is calculated in combination with the digital elevation model (DEM). The calculation of the slope can be completed by calculating the rate of change of the elevation in the neighborhood. The specific method is to calculate the elevation difference between two adjacent points, and then divide it by the horizontal distance between them to obtain the slope value. Next, the slope variance in each risk area is calculated.Variance is the average value of the square of the deviation of a data set from the mean, which can reflect the degree of slope change in the area. By statistically analyzing the slope values ​​in each area, the mean of the slope is calculated, and then the variance between the slope and the mean of each point is calculated, and finally the slope variance of the area is obtained. Areas with larger variance values ​​indicate that the slope changes dramatically and there is a greater risk of landslides, while areas with smaller variance values ​​have relatively gentle terrain and lower landslide risks. The slope variance data finally obtained, combined with the rock and soil landslide probability risk zoning data, helps to make a more detailed assessment of the risks in the mining area. According to the vertical drop data and the slope variance, the weathering erosion mechanism data is simulated by the kinetic energy carried by the rock and soil under rainfall changes. First, the rainfall change simulation of the weathering erosion mechanism is performed for different risk areas in the mining area in combination with the vertical drop data in step S242 and the slope variance data in step S243. In this step, the rainfall erosion mechanics model is used to simulate the kinetic energy changes carried by the rock and soil under different rainfall conditions. The kinetic energy simulation of rainfall changes needs to consider the impact of factors such as precipitation, precipitation intensity and precipitation duration on rock and soil weathering. First, the scouring force generated by precipitation is calculated based on the rainfall intensity and slope variance data. The scouring force is related to the slope, precipitation intensity and rock and soil type. Especially in areas with larger slopes, the scouring force of precipitation is greater, which is easy to cause surface weathering of rock and soil. Then, according to the vertical drop data, the degree of erosion of the rock and soil layer by the fluid impact force generated by rainfall is simulated. During the rainfall process, the kinetic energy of the water flow can be calculated by the flow velocity and the density of the water body. Through the simulation of weathering and erosion mechanism, combined with factors such as the topography, precipitation intensity and slope of the mining area, the degree of influence of rainfall changes on the kinetic energy carried by rock and soil is obtained. This data can provide support for dynamic environmental change factors for subsequent landslide risk assessment. The landslide entrainment potential energy increment is calculated based on the kinetic energy data of rock and soil carried by rainfall changes. First, according to the kinetic energy data of rock and soil carried by rainfall changes obtained in step S244, the movement of rock and soil particles under different rainfall conditions is calculated. Specifically, under the action of rainfall, rock and soil will be pushed by the water flow and move, forming a landslide. According to the kinetic energy calculation formula, the kinetic energy of rock and soil is equal to the product of the mass of the object and the square of the velocity, divided by 2. Through a comprehensive analysis of factors such as precipitation, slope, and weathering erosion, the velocity change of rock and soil under different rainfall conditions is estimated, and the landslide entrainment force is further obtained. Entrainment force refers to the force that pushes rock and soil particles and slides together due to the action of water flow during the landslide process. According to the physical model of landslide, entrainment force is closely related to the potential energy increment of the landslide. Potential energy increment refers to the energy change caused by the change of position of an object during the sliding process. By calculating the potential energy increment of different risk areas during rainfall, the landslide entrainment force potential energy increment data is obtained. Finally, the obtained potential energy increment data can be used to predict landslide events in different areas of the mining area under specific rainfall conditions, providing a basis for mine safety management.

[0111] Step S244 includes the following steps:

[0112] Obtain historical rainfall data; analyze the spatial and temporal distribution of rainfall intensity in historical rainfall data to obtain spatial and temporal distribution data of rainfall intensity;

[0113] According to the vertical drop data and slope variance, the rock and soil looseness difference between different regions is analyzed for the weathering and erosion mechanism data, and the rock and soil looseness difference data between different regions are obtained;

[0114] According to the vertical drop data and slope variance, soil shear strength analysis is performed on the rock and soil looseness difference data to obtain soil shear strength data;

[0115] Based on the spatial and temporal distribution data of rainfall intensity, the rock looseness difference data and soil shear strength data between different regions were used to simulate and evaluate the erosion rock and soil carrying density, and the erosion rock and soil carrying density data were obtained;

[0116] Conduct vertical increment analysis on the scouring rock and soil carrying density data to obtain vertical increment data of carrying density;

[0117] The rainfall-variable rock-soil carrying kinetic energy data are obtained by simulating the rainfall-variable rock-soil carrying kinetic energy based on the vertical drop data and the slope variance and the vertical increment data of the carrying density.

[0118] In an embodiment of the present invention, first, historical rainfall data is obtained, and long-term rainfall data in and around the mining area is collected and sorted out by accessing meteorological stations, groundwater monitoring systems, and open meteorological data resource platforms around the mining area. These data include information such as the daily average value of rainfall, monthly total, and rainfall intensity. The data unit can be unified by using a unified precipitation standard (such as mm / hour), and precipitation data of different time scales can be standardized into a unified format. These historical rainfall data should cover different seasons and climate change cycles to ensure the extensiveness and representativeness of the collected data. This data serves as the basis for subsequent analysis, provides rainfall intensity change patterns and rainfall distribution information in the mining area, and helps analyze changes in rock and soil and landslide risks under different precipitation conditions. The historical rainfall data obtained is then analyzed for the spatiotemporal distribution of rainfall intensity. This process first requires classification and processing of rainfall in different time periods. The rainfall is divided according to different intensity intervals (such as light, moderate, heavy rainfall, etc.) to ensure that the data in each intensity interval is representative. Then, based on geographic spatial information, the rainfall data is spatially distributed. Spatial interpolation methods, such as inverse distance weighted (IDW) or Kriging interpolation, can be used to fill in the blank areas without observation data in space and generate a spatial distribution map of rainfall intensity covering the mining area and its surroundings. For temporal distribution, time series analysis methods, such as trend analysis and periodicity analysis, can be used to determine the variation pattern of rainfall intensity in different seasons, months and daytime periods. Through this analysis process, the spatiotemporal distribution data of rainfall intensity are obtained, which can help understand the impact pattern of precipitation on the mining area, especially in the landslide-prone areas of the mining area under extreme weather conditions. After obtaining the spatiotemporal distribution data of rainfall intensity, combined with the vertical drop data and slope variance data of the mining area, the difference analysis of the rock and soil looseness in different areas of the mining area is carried out. First, the vertical drop data reflects the elevation change in the area, and the slope variance reflects the severity of the terrain undulation. These terrain parameters directly affect the looseness of the rock and soil. The steeper the slope, the higher the possibility of rock and soil looseness. The rock and soil looseness of each area is analyzed by correlation analysis between terrain data and weathering and erosion mechanisms. Generally, the effect of weathering on rock looseness increases with the increase of rainfall. Therefore, rainfall intensity can be used as a correction factor, combined with vertical drop and slope variance, and weighted average or weighted regression methods can be used to calculate rock looseness. During the calculation process, areas with greater rock looseness usually have a higher risk of landslides. By analyzing the difference in rock looseness in different areas of the mining area, areas that are more affected by weathering erosion can be identified, further providing data support for subsequent shear strength and scour assessments. The analysis of soil shear strength is based on rock looseness difference data, vertical drop data, and slope variance data. This analysis mainly evaluates the differences in soil shear strength in various areas within the mining area. Shear strength is the ability of soil or rock to resist sliding or destruction under shear force.First, based on the difference data of rock and soil looseness, combined with the vertical drop and slope variance, the influence of different terrain characteristics on soil shear strength is analyzed. In areas with large slopes and high rock and soil looseness, the soil shear strength is usually low because the loose soil layer is easily affected by external forces. Therefore, it is necessary to combine soil type and rock and soil structure data, use empirical formulas or experimental data, and derive the soil shear strength under different terrain conditions. Commonly used soil shear strength calculation methods include the Mohr-Coulomb criterion, in which the shear strength τ is related to the normal force N and the friction angle. , the relationship between cohesion c is: τ=c+Ntan( ). Through these analyses, the soil shear strength data of different areas in the mining area are obtained, which provides basic data for the subsequent simulation of scouring rock and soil carrying density. After obtaining the soil shear strength data and the rock and soil looseness difference data, the scouring rock and soil carrying density simulation evaluation between different regions is carried out based on the spatiotemporal distribution data of rainfall intensity. The core of this step is to evaluate the soil carrying capacity and the possibility of landslide under rainfall. First, it is necessary to establish a dynamic soil erosion model with rainfall intensity as the dominant factor. When the rainfall intensity is high, rainwater flowing through the loose soil layer will carry away the surface soil particles and produce scouring. By adopting a tillage erosion model (such as the USLE model), combined with factors such as precipitation intensity, soil shear strength, and terrain slope, the rock and soil carrying density of each area is simulated. Specifically, the amount of soil particles carried away is related to factors such as rainfall, soil type, and slope, and the carrying density can be calculated by formula. The simulation results provide the density information of rock and soil carrying in different risk areas, which helps to determine the probability of landslides in each area after rainfall. After obtaining the scouring rock and soil carrying density data, vertical incremental analysis is performed. Vertical increment analysis aims to evaluate the change of soil or rock density at different height levels under the action of rainfall. First, through stratified analysis, the terrain in the mining area is divided into different height levels, and the rock density in each level is different. According to the influence of rainfall intensity and slope variance, the change between the rock density of each level and the previous level is calculated to obtain vertical increment data. Common methods include calculating the increment of rock density at each level and the density difference of the adjacent level. Through these increment data, it is possible to understand the impact of soil and rock layers at different height levels during the scouring process under different rainfall conditions, so as to more accurately predict the potential landslide risk in the mining area. Finally, based on the vertical increment data, combined with the vertical drop data and slope variance of the mining area, the kinetic energy of rock and soil carried by rainfall is simulated. The simulation of the kinetic energy of rock and soil carried after rainfall can be calculated by the kinetic energy formula, taking into account the water velocity and the density of soil particles after precipitation to simulate the movement of rock and soil. In this process, the kinetic energy of rock and soil is calculated as follows: kinetic energy = 1 / 2 times mass times velocity squared. By simulating the fluid flow and scouring caused by rainfall, the kinetic energy data of rock and soil in the mining area under rainfall conditions were obtained. These data not only reflect the erosion effect of precipitation on soil, but also provide quantitative support for subsequent landslide risk assessment.

[0119] Step S3 includes the following steps:

[0120] Step S31: normalizing the landslide entrainment potential energy increment data to obtain normalized landslide entrainment potential energy data;

[0121] Step S32: performing convolution calculation on the normalized data of the landslide entrainment force potential energy to obtain the convolution data of the landslide entrainment force potential energy;

[0122] Step S33: assigning a local risk level to the three-dimensional model of the open-pit non-metallic mine according to the landslide entrainment potential energy convolution data and the rock and soil landslide probability data to obtain the local risk level of the open-pit non-metallic mine.

[0123] As an example of the present invention, refer to Figure 3 As shown, in this example, step S3 includes:

[0124] Step S31: normalizing the landslide entrainment potential energy increment data to obtain normalized landslide entrainment potential energy data;

[0125] In the embodiment of the present invention, it is first necessary to normalize the landslide entrainment potential energy increment data calculated in the previous stage through Min-Max Normalization. The purpose of normalization is to standardize data of different magnitudes to a unified scale to avoid deviations in analysis results due to differences in data values. This operation can eliminate data range differences caused by factors such as terrain height differences and precipitation differences between different regions, and ensure that all data are within the same standard range (usually the interval [0, 1]), thereby making subsequent analysis more fair and effective.

[0126] Step S32: performing convolution calculation on the normalized data of the landslide entrainment force potential energy to obtain the convolution data of the landslide entrainment force potential energy;

[0127] In an embodiment of the present invention, after obtaining the normalized data of the potential energy of the landslide entrainment force, the data are then subjected to convolution calculation. The convolution operation can be used to extract local features in the data and highlight potential landslide risk points in different areas. The core idea of ​​the convolution calculation is to extract the data pattern in a specific area through sliding calculation of the convolution kernel and the data. The specific operation is to select a suitable convolution kernel (such as a matrix of 3×3 or 5×5 size) and perform weighted averaging on each normalized data of the potential energy of the landslide entrainment force. Assuming that the convolution kernel is K and the data is D, the convolution operation can be expressed as: C=DK; wherein C is the convolution result, representing the convolution data of the potential energy of the landslide entrainment force, D is the normalized data of the potential energy of the landslide entrainment force, and K is the convolution kernel. The convolution kernel highlights the high-risk areas by weighted processing of the potential energy data of the landslide entrainment force in adjacent areas. For example, for specific terrain or rainfall areas in mining areas, convolution kernels with different weights can be selected to adapt to different geographical features, thereby accurately extracting potential landslide risk points. After the convolution calculation, the convolution data of the landslide entrainment potential energy contains the local characteristics of the landslide entrainment potential energy under different terrain and rainfall conditions in the mining area. High-value areas represent areas of the mining area that are more at risk under certain conditions, while low-value areas represent areas with less risk.

[0128] Step S33: assigning a local risk level to the three-dimensional model of the open-pit non-metallic mine according to the landslide entrainment potential energy convolution data and the rock and soil landslide probability data to obtain the local risk level of the open-pit non-metallic mine.

[0129] In an embodiment of the present invention, after obtaining the landslide entrainment force potential energy convolution data, combined with the rock and soil landslide probability data, the local risk level is assigned to the open-pit non-metallic mine three-dimensional model. The core of this step is to fuse the calculated landslide entrainment force potential energy convolution data with the rock and soil landslide probability data, evaluate the landslide risk in each area, and finally assign a risk level to each local area in the mining area. First, it is necessary to classify different areas according to the rock and soil landslide probability data (including the landslide possibility and specific terrain characteristics of the rock and soil in the mining area), such as high-risk, medium-risk, and low-risk areas. Then, the landslide entrainment force potential energy convolution data and the rock and soil landslide probability data are fused by the weighted average method, and the formula is as follows: ;in, Indicates the local risk level, is the landslide entrainment potential energy convolution data, is the rock and soil landslide probability data, and They are the weight coefficients of the landslide entrainment potential energy convolution data and the rock and soil landslide probability data, respectively. By adjusting the weight coefficients, it is possible to better adapt to the landslide risk distribution under different terrain and rainfall conditions in the mining area. Then, based on the calculated local risk level data, different areas in the mining area are divided into different risk levels. Generally speaking, local risk levels can be divided into high risk, medium risk and low risk levels, and each level corresponds to a different probability of landslide occurrence. Mapping these data to the three-dimensional model of the open-pit non-metallic mine, each area in the mining area is assigned a clear risk level, thereby providing accurate safety hazard identification and early warning data for mine managers. By assigning risk levels to different local areas, key areas with landslide risks in the mining area can be effectively identified, and data support can be provided for subsequent protective measures.

[0130] Step S33 includes the following steps:

[0131] Step S331: recalibrate the spatial coordinates of the open-pit non-metallic mine three-dimensional model to obtain the spatial recalibrated coordinates of the open-pit non-metallic mine;

[0132] Step S332: performing coordinate risk linkage analysis on the spatial recalibration coordinates of the open-pit non-metallic mine according to the landslide entrainment potential energy convolution data and the rock and soil landslide probability data to obtain coordinate risk linkage data;

[0133] Step S333: Calculate a multivariate local risk index based on the landslide entrainment potential energy convolution data, the rock and soil landslide probability data, and the coordinate risk linkage data to obtain a multivariate local risk index;

[0134] Step S334: assigning a local risk level to the open-pit non-metallic mine three-dimensional model according to the multivariate local risk index to obtain the local risk level of the open-pit non-metallic mine.

[0135] In an embodiment of the present invention, it is first necessary to recalibrate the spatial coordinates of the three-dimensional model of the open-pit non-metallic mine. The recalibration of the coordinates of the three-dimensional model aims to align the spatial data of the mining area collected by the drone with the standard coordinate system, so as to ensure the accuracy of the position of each point in the model. To this end, the three-dimensional data collected by the drone are geometrically calibrated using the known reference points in the mining area (such as GPS calibration points or ground measured coordinate points). The specific operation is to select more than three ground control points, the positions of which have been accurately coordinates obtained by ground measurement tools (such as total stations or differential GPS). According to these control points, the three-dimensional point cloud data taken by the drone is translated, rotated and scaled using the least squares method. After the transformation, the coordinates of each point will be aligned with the actual geographical location, and the obtained spatial recalibration coordinates are the open-pit non-metallic mine. In this way, the accuracy of the three-dimensional data of the mining area in the spatial coordinate system is ensured, and the foundation is laid for subsequent analysis. After obtaining the spatial recalibration coordinates of the open-pit non-metallic mine, the coordinate risk linkage analysis is performed. This analysis evaluates the risk level corresponding to each coordinate point by combining the landslide entrainment potential energy convolution data with the rock and soil landslide probability data and the recalibrated spatial coordinates. First, the spatial coordinates need to be converted into a gridded grid so that each grid represents a spatial point. In each grid, the corresponding landslide entrainment potential energy convolution data and rock and soil landslide probability data are obtained. Then, these data are fused by the weighted average method to calculate the risk index of each grid. In this way, the risk value corresponding to each grid point is obtained, and then the coordinate risk linkage data is formed. The coordinate risk linkage data reflects the interaction between the landslide risk and the geological environment in different regions, and can provide a quantitative basis for subsequent local risk analysis. After obtaining the coordinate risk linkage data, a multivariate local risk index is calculated. This calculation evaluates the comprehensive risk of the local area of ​​the open-pit non-metallic mine by comprehensively considering the landslide entrainment potential energy convolution data, the rock and soil landslide probability data, and the coordinate risk linkage data. Specifically, the landslide entrainment potential energy convolution data, the rock and soil landslide probability data, and the coordinate risk linkage data of each region are first determined, and the multivariate local risk index of each region is calculated by combining these data by the weighted average method. These weight coefficients are adjusted according to the actual situation of the mining area and the degree of attention paid to different types of risks. The multivariate local risk index obtained by weighted synthesis reflects the comprehensive risk level of each area under multiple risk factors. This index provides a scientific basis for the subsequent identification of safety hazards in mining areas, and can clearly indicate which areas have a higher risk of landslides. Finally, based on the calculated multivariate local risk index, the local risk level is assigned to the three-dimensional model of the open-pit non-metallic mine. The specific operation is to divide each area in the three-dimensional model of the mining area into different risk levels according to the numerical value of the multivariate local risk index. First, all calculated multivariate local risk indices are divided into several levels (such as high risk, medium risk, and low risk) according to the set threshold.Then, in the 3D model, the areas are labeled with corresponding risk levels based on the multivariate local risk index corresponding to each area. High-risk areas indicate a high probability of landslides, while low-risk areas indicate a low probability of landslides. This process uses spatial data analysis tools to assign values ​​to the 3D model point by point, and finally generates a 3D mining area model showing each risk level, providing mine managers with an intuitive illustration of safety hazard identification.

[0136] Step S4 includes the following steps:

[0137] Step S41: constructing a mine safety hazard identification model based on the local risk level of the open-pit non-metallic mine to obtain a mine safety hazard identification model;

[0138] Step S42: performing deep optimization of the mine safety hazard identification model based on transfer learning to obtain a mine safety hazard identification optimization model;

[0139] Step S43: Send the mine safety hazard identification optimization model to the terminal to execute the mine safety hazard identification method.

[0140] In the embodiment of the present invention, it is first necessary to use the local risk level data of the open-pit non-metallic mine to construct a mine safety hazard identification model. The construction of the model is based on the geological characteristics of the mine, the landslide risk level and its distribution characteristics. First, the risk level data of each local area is extracted from the three-dimensional model of the mining area as the key variable of the model input. These risk levels are based on multiple data sources, such as landslide entrainment potential energy convolution data, rock and soil landslide probability data, coordinate risk linkage data and multivariate local risk index. Then, the local risk level data is classified by using traditional machine learning algorithms such as logistic regression method, decision tree algorithm or support vector machine (SVM), and the types of mine safety hazards under different risk levels are determined in combination with the geological environment characteristics of the mine. The core of this step is to identify which areas have the most serious safety hazards by analyzing the local risk levels of different areas, and to qualitatively classify the mine safety hazards. In the process of model construction, by selecting variables related to the risk level and inputting them into the model for training, a recognition model for mine safety hazards is finally constructed. The model can judge the potential safety hazards in the mining area based on the input local risk level data. The constructed mine safety hazard identification model is deeply optimized through transfer learning technology. Transfer learning is to optimize the model training process by transferring knowledge from one field to another related field. In this step, it is first necessary to prepare existing models similar to the mine safety hazard identification task. Usually, these models come from mine safety hazard identification data in other similar fields (for example, existing landslide identification models, engineering safety hazard identification models, etc.). Through transfer learning, borrowing the knowledge of these existing models can effectively improve the training efficiency and accuracy of the new model. In specific implementation, the pre-trained parameters of the existing model are first used as the initial model parameters and migrated to the new mine safety hazard identification task. Then, fine-tuning is performed using local mine data, that is, adjusting some layers or weights in the network to ensure that the optimized model can adapt to the characteristics of the mine safety hazard identification task. This process involves fine-tuning the neural network model, convolutional neural network (CNN) or other deep learning models, using the local risk level data of the mine, landslide probability data, etc. as training data, and optimizing the parameters of the model through multiple iterations. The optimization effect of transfer learning is reflected in improving the convergence speed and accuracy of the model, reducing overfitting, and improving the stability and reliability of the recognition results. In step S43, the optimized mine safety hazard identification optimization model will be sent to the terminal device to execute the mine safety hazard identification method. The terminal device generally refers to the drone, monitoring equipment, mobile terminal or computer at the mine site. First, the optimized model is deployed to the terminal device via a wireless network. After the terminal device receives the optimization model, it can perform the safety hazard identification task based on the environmental data collected in real time in the mining area.During the specific operation, the terminal device inputs the real-time received mining data (such as image data collected by drones, geographic information data, etc.) into the deployed recognition model. The model determines the type of safety hazards in the mining area based on the input data and outputs the recognition results. This process can be supported by cloud computing platforms or edge computing devices, making the data transmission and processing process efficient and real-time. During the execution process, mine managers can view the identification results of mine safety hazards in real time through terminal devices and take corresponding preventive or emergency measures. In addition, the terminal device can also upload data to the central database regularly to ensure the continuous monitoring and management of safety hazards. Through this process, the automated identification of mine safety hazards based on the optimization model is realized, providing an efficient and real-time monitoring method for mine safety.

[0141] The present invention also provides a mine safety hazard identification system based on a drone, which is used to execute the mine safety hazard identification method based on a drone as described above. The mine safety hazard identification system based on a drone includes:

[0142] The three-dimensional model building module is used to collect global images of the open-pit non-metallic mine through the airborne camera of the drone to obtain a global image of the open-pit non-metallic mine; and to build a three-dimensional model of the open-pit non-metallic mine based on the global image of the open-pit non-metallic mine to obtain a three-dimensional model of the open-pit non-metallic mine;

[0143] The entrainment potential energy increment calculation module is used to restore the weathering and erosion mechanism of the open-pit non-metallic mine three-dimensional model to obtain weathering and erosion mechanism data; estimate the probability of rock and soil landslides on the open-pit non-metallic mine three-dimensional model based on the weathering and erosion mechanism data and the global image of the open-pit non-metallic mine to obtain rock and soil landslide probability data; calculate the landslide entrainment potential energy increment on the rock and soil landslide probability data to obtain the landslide entrainment potential energy increment data;

[0144] The local risk level assignment module is used to assign a local risk level to the three-dimensional model of the open-pit non-metallic mine according to the rock and soil landslide probability data and the landslide entrainment potential energy increment data, so as to obtain the local risk level of the open-pit non-metallic mine;

[0145] The hidden danger identification model construction module is used to construct a mine safety hidden danger identification model based on the local risk level of the open-pit non-metallic mine, and obtain a mine safety hidden danger identification optimization model; the mine safety hidden danger identification optimization model is sent to the terminal to execute the mine safety hidden danger identification method.

[0146] Therefore, the embodiments should be regarded as illustrative and non-restrictive from all points, and the scope of the present invention is limited by the appended claims rather than the above description, and it is therefore intended that all changes falling within the meaning and range of equivalent elements of the application documents are included in the present invention.

[0147] The above description is only a specific embodiment of the present invention, so that those skilled in the art can understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but should conform to the widest scope consistent with the principles and novel features invented herein.

Claims

1. A method for identifying safety hazards in mines based on drones, characterized in that: The following steps are involved: Step S1: using an airborne camera of a drone to collect a global image of the open-pit non-metallic mine to obtain a global image of the open-pit non-metallic mine; constructing a three-dimensional model of the open-pit non-metallic mine based on the global image of the open-pit non-metallic mine to obtain a three-dimensional model of the open-pit non-metallic mine; Step S2: performing weathering erosion mechanism restoration on the three-dimensional model of the open-pit non-metallic mine to obtain weathering erosion mechanism data; performing rock and soil landslide probability estimation on the three-dimensional model of the open-pit non-metallic mine according to the weathering erosion mechanism data and the global image of the open-pit non-metallic mine to obtain rock and soil landslide probability data; The landslide potential energy increment is calculated for the rock and soil landslide probability data to obtain the landslide potential energy increment data; Step S2 includes: Step S21: Analyze the mountain crack trend of the open-pit non-metallic mine three-dimensional model to obtain mountain crack trend data; Step S22: based on the mountain crack trend data, the weathering and erosion mechanism of the open-pit non-metallic mine three-dimensional model is restored to obtain weathering and erosion mechanism data; Step S23: estimating the probability of rock and soil landslide on the three-dimensional model of the open-pit non-metallic mine according to the weathering erosion mechanism data and the global image of the open-pit non-metallic mine to obtain rock and soil landslide probability data; wherein step S23 includes: Step S231: performing a vertical distribution density analysis of vegetation in the mining area on the global image of the open-pit non-metallic mine to obtain the vertical distribution density of vegetation in the mining area; Step S232: Decomposing the degree of wind erosion and wear of rock and soil in the vertical direction of the vertical distribution density of vegetation in the mining area and the three-dimensional model of the open-pit non-metallic mine according to the weathering erosion mechanism data, and obtaining the data of wind erosion and wear of rock and soil in the vertical direction; Step S233: Based on the rock and soil wind erosion and wear degree data, the open-pit non-metallic mine three-dimensional model is evaluated for the lateral support loss of the free surface to obtain the lateral support loss data of the free surface; Step S234: deduce the vertical spatial distribution risk difference coefficient for the lateral support loss data of the air-facing surface to obtain the vertical spatial distribution risk difference coefficient; Step S235: estimating the probability of rock and soil landslide according to the rock and soil wind erosion wear degree data, the lateral support loss data of the free surface, and the vertical spatial distribution risk difference coefficient to obtain rock and soil landslide probability data; Step S24: calculating the landslide entrainment potential energy increment on the rock and soil landslide probability data to obtain the landslide entrainment potential energy increment data; Step S3: assigning a local risk level to the three-dimensional model of the open-pit non-metallic mine according to the rock and soil landslide probability data and the landslide entrainment potential energy increment data to obtain the local risk level of the open-pit non-metallic mine; Step S4: construct a mine safety hazard identification model based on the local risk level of the open-pit non-metallic mine to obtain a mine safety hazard identification optimization model; send the mine safety hazard identification optimization model to the terminal to execute the mine safety hazard identification method.

2. The method for identifying safety hazards in mines based on drones according to claim 1 is characterized in that: Step S1 includes the following steps: Step S11: using the airborne camera of the UAV to collect a global image of the open-pit non-metallic mine to obtain a global image of the open-pit non-metallic mine; Step S12: performing color distribution ratio harmonization on the global image of the open-pit non-metallic mine to obtain a global color harmonized image of the open-pit non-metallic mine; Step S13: marking the morphological points of the global color-harmonized image of the open-pit non-metallic mine to obtain the morphological point marking data of the open-pit non-metallic mine; Step S14: constructing a three-dimensional model of the open-pit non-metallic mine according to the global color-harmonizing image of the open-pit non-metallic mine and the morphological point marking data of the open-pit non-metallic mine to obtain a three-dimensional model of the open-pit non-metallic mine.

3. The method for identifying safety hazards in mines based on drones according to claim 1 is characterized in that: Step S24 includes the following steps: Step S241: performing risk zoning processing on the rock and soil landslide probability data to obtain rock and soil landslide probability risk zoning data; Step S242: Calculating the vertical drop between different risk areas based on the rock and soil landslide probability risk zoning data to obtain vertical drop data between different risk areas; Step S243: Calculating the slope variance between different risk areas based on the rock and soil landslide probability risk zoning data to obtain the slope variance between different risk areas; Step S244: simulating the kinetic energy of rock and soil carried by rainfall changes on the weathering erosion mechanism data according to the vertical drop data and the slope variance, and obtaining the kinetic energy of rock and soil carried by rainfall changes data; Step S245: Calculate the incremental potential energy of the landslide entrainment force based on the kinetic energy data of rock and soil carried by rainfall changes to obtain the incremental potential energy data of the landslide entrainment force.

4. The method for identifying safety hazards in mines based on drones according to claim 3 is characterized in that: Step S244 includes the following steps: Obtain historical rainfall data; analyze the spatial and temporal distribution of rainfall intensity in historical rainfall data to obtain spatial and temporal distribution data of rainfall intensity; According to the vertical drop data and slope variance, the rock and soil looseness difference between different regions is analyzed for the weathering and erosion mechanism data, and the rock and soil looseness difference data between different regions are obtained; According to the vertical drop data and slope variance, soil shear strength analysis is performed on the rock and soil looseness difference data to obtain soil shear strength data; Based on the spatial and temporal distribution data of rainfall intensity, the rock looseness difference data and soil shear strength data between different regions were used to simulate and evaluate the erosion rock and soil carrying density, and the erosion rock and soil carrying density data were obtained; Conduct vertical increment analysis on the scouring rock and soil carrying density data to obtain vertical increment data of carrying density; The rainfall-variable rock-soil carrying kinetic energy data are obtained by simulating the rainfall-variable rock-soil carrying kinetic energy based on the vertical drop data and the slope variance and the vertical increment data of the carrying density.

5. The method for identifying safety hazards in mines based on drones according to claim 1 is characterized in that: Step S3 includes the following steps: Step S31: normalizing the landslide entrainment potential energy increment data to obtain normalized landslide entrainment potential energy data; Step S32: performing convolution calculation on the normalized data of the landslide entrainment force potential energy to obtain the convolution data of the landslide entrainment force potential energy; Step S33: assigning a local risk level to the three-dimensional model of the open-pit non-metallic mine according to the landslide entrainment potential energy convolution data and the rock and soil landslide probability data to obtain the local risk level of the open-pit non-metallic mine.

6. The method for identifying safety hazards in mines based on drones according to claim 5 is characterized in that: Step S33 includes the following steps: Step S331: recalibrate the spatial coordinates of the open-pit non-metallic mine three-dimensional model to obtain the spatial recalibrated coordinates of the open-pit non-metallic mine; Step S332: performing coordinate risk linkage analysis on the spatial recalibration coordinates of the open-pit non-metallic mine according to the landslide entrainment potential energy convolution data and the rock and soil landslide probability data to obtain coordinate risk linkage data; Step S333: Calculate a multivariate local risk index based on the landslide entrainment potential energy convolution data, the rock and soil landslide probability data, and the coordinate risk linkage data to obtain a multivariate local risk index; Step S334: assigning a local risk level to the open-pit non-metallic mine three-dimensional model according to the multivariate local risk index to obtain the local risk level of the open-pit non-metallic mine.

7. The method for identifying safety hazards in mines based on drones according to claim 1, characterized in that: Step S4 includes the following steps: Step S41: constructing a mine safety hazard identification model based on the local risk level of the open-pit non-metallic mine to obtain a mine safety hazard identification model; Step S42: performing deep optimization of the mine safety hazard identification model based on transfer learning to obtain a mine safety hazard identification optimization model; Step S43: Send the mine safety hazard identification optimization model to the terminal to execute the mine safety hazard identification method.

8. A mine safety hazard identification system based on drones, characterized in that: Used to execute the mine safety hazard identification method based on drone as claimed in claim 1, the mine safety hazard identification system based on drone includes: The three-dimensional model building module is used to collect global images of the open-pit non-metallic mine through the airborne camera of the drone to obtain a global image of the open-pit non-metallic mine; and to build a three-dimensional model of the open-pit non-metallic mine based on the global image of the open-pit non-metallic mine to obtain a three-dimensional model of the open-pit non-metallic mine; The entrainment potential energy increment calculation module is used to restore the weathering and erosion mechanism of the open-pit non-metallic mine three-dimensional model to obtain weathering and erosion mechanism data; estimate the probability of rock and soil landslides on the open-pit non-metallic mine three-dimensional model based on the weathering and erosion mechanism data and the global image of the open-pit non-metallic mine to obtain rock and soil landslide probability data; calculate the landslide entrainment potential energy increment on the rock and soil landslide probability data to obtain the landslide entrainment potential energy increment data; The local risk level assignment module is used to assign a local risk level to the three-dimensional model of the open-pit non-metallic mine according to the rock and soil landslide probability data and the landslide entrainment potential energy increment data, so as to obtain the local risk level of the open-pit non-metallic mine; The hidden danger identification model construction module is used to construct a mine safety hidden danger identification model based on the local risk level of the open-pit non-metallic mine, and obtain a mine safety hidden danger identification optimization model; the mine safety hidden danger identification optimization model is sent to the terminal to execute the mine safety hidden danger identification method.

Images