Strip mine air-space-ground slope integrated supervision system and method

Abstract

The invention relates to a strip mine air-space-ground slope integrated supervision system and method, through designing a unique algorithm model and an innovative data processing flow, in satellite remote sensing image semantic segmentation, an algorithm combining a U-Net network structure and an attention mechanism is improved, and compared with a traditional image segmentation method, the method has the advantages that the algorithm is more accurate, and the efficiency is higher. The slope boundary extraction precision is improved by more than 20%, and the processing speed is higher; in unmanned aerial vehicle L i DAR point cloud data processing, based on a morphological filtering and cloth simulation filtering combined method, compared with a traditional experience evaluation method, the classification accuracy is improved by 35% or above, and the change trend of the slope state can be predicted 1-2 days in advance.

Description

Technical Field

[0001] The present invention relates to the technical field of open-pit mining, and in particular to an open-pit mine air-ground-slope integrated monitoring system and method. Background Art

[0002] During open-pit mining, the stability of the slope is directly related to the production safety and economic benefits of the mine. With the continuous expansion of open-pit mining and the gradual increase in mining depth, the slope faces increasingly complex geological conditions and engineering environments, such as complex geological structures, variable mechanical properties of rock and soil, the impact of natural disasters such as rainfall and earthquakes, and stress redistribution caused by mining activities, which significantly increases the risk of slope instability accidents.

[0003] Traditional open-pit mine slope supervision methods mainly rely on on-site manual monitoring and simple empirical judgment. Monitoring personnel regularly arrange measurement points on the slope, use total stations, levels and other instruments to measure the displacement, settlement and other parameters of the slope, and then evaluate the stability of the slope based on limited monitoring data and empirical formulas. This method has many limitations: first, manual monitoring is inefficient and the monitoring cycle is long. It is difficult to achieve real-time and continuous monitoring of the slope state, and it is impossible to capture the small deformation and mutation information of the slope in time. Often, when problems are discovered, the best prevention and control time has been missed; second, the amount of data obtained by traditional monitoring methods is limited, and it is mainly concentrated on local displacement and geometric morphology information. It is difficult to fully reflect the stress and strain distribution inside the slope and the changes in the physical and mechanical properties of the rock and soil, and the prediction accuracy of slope instability is low; third, the experience-based stability assessment method lacks scientificity and objectivity, cannot adapt to the changes in slopes under different geological conditions and mining processes, and is prone to misjudgment and omission.

[0004] In recent years, with the rapid development of aerospace technology, satellite remote sensing technology, Internet of Things technology and computer technology, some new technical means have begun to be applied to the field of open-pit mine slope monitoring, but there are still many problems. For example, although satellite remote sensing technology can obtain large-area slope image information, it is limited by resolution and revisit period, and its monitoring ability for local details and short-term changes of slopes is insufficient; although drone monitoring can achieve high-resolution image acquisition, the data processing and analysis methods are relatively simple, mainly relying on manual visual interpretation, and it is difficult to automatically extract effective slope deformation features and conduct quantitative analysis; in addition, most of the existing slope monitoring systems lack the ability to deeply integrate and intelligently analyze multi-source data, and cannot give full play to the advantages of different monitoring methods in the air, ground and space, and it is difficult to establish an accurate slope stability prediction model and realize refined supervision decisions. Therefore, there is an urgent need for an innovative open-pit mine air-ground-ground integrated slope supervision system and method to improve the level and effect of open-pit mine slope supervision. Summary of the invention

[0005] The present invention provides an open-pit mine air-ground-slope integrated supervision system and method, which effectively solves the problems in the prior art by designing a unique algorithm model and an innovative data processing flow.

[0006] An integrated open-pit mine air-ground slope supervision system includes: a satellite remote sensing monitoring subsystem, which is used to use multiple high-resolution optical satellites and synthetic aperture radar (SAR) satellites to perform regular large-scale monitoring of the open-pit mine slope, wherein the optical satellite can obtain high-resolution visible light and infrared images of the slope, and the SAR satellite can obtain radar interferograms of the slope; a satellite data receiving station, which is used to receive data sent back by the satellite and perform preprocessing, including operations such as data format conversion, geometric correction, and radiation correction, and is also used to process optical satellite images using an image semantic segmentation algorithm based on deep learning, and to process SAR satellite interferograms using multi-baseline interferometry technology and a phase unwrapping algorithm;

[0007] The UAV low-altitude monitoring subsystem is used to equip multiple UAVs with high-precision optical cameras, LiDAR, thermal imagers and other equipment to conduct low-altitude fine monitoring of open-pit mine slopes. The data collected by the UAVs is transmitted to the ground control center in real time through the wireless transmission module for data processing;

[0008] The ground sensor monitoring subsystem is used to deploy various types of sensors at key locations on the slopes of open-pit mines, including high-precision global positioning system (GPS) receivers, total stations, inclination sensors, strain sensors, and groundwater level sensors. The data collected by the sensors are transmitted to the data processing center for processing via a wired or wireless data transmission network.

[0009] When the deep learning-based image semantic segmentation algorithm processes optical satellite images, an improved U-Net network structure is adopted. The network structure adds an attention mechanism module between the encoding and decoding paths, which can automatically focus on the key features of the slope area, accurately segment the slope from the surrounding environment, and extract information such as the slope boundary and slope surface morphology.

[0010] In the UAV low-altitude monitoring subsystem, data processing of optical camera images includes stitching and orthorectification processing to generate a complete orthophoto map of the slope; filtering of LiDAR point cloud data adopts a method based on a combination of morphological filtering and cloth simulation filtering to extract the point cloud features of the slope, and by comparing point cloud data of different periods, based on feature matching and least squares adjustment principles, calculate the local deformation of the slope; temperature anomaly detection algorithm is applied to thermal imaging data, based on statistical analysis and threshold setting, to identify temperature anomaly areas, and combined with geological data of the slope and other monitoring data, analyze the causes and potential risks of temperature anomaly.

[0011] In the ground sensor monitoring subsystem, when the data processing center processes the data collected by the sensors, a slope stability assessment algorithm based on multi-sensor data fusion is adopted. The algorithm first standardizes the data of different types of sensors, then uses the principal component analysis (PCA) method to reduce the dimension of the data and extract the main feature information. Then, a slope stability classification model based on a support vector machine (SVM) is constructed, and the feature data after dimension reduction is used as input and the stability state of the slope is used as output. The model is trained through a large amount of historical data, and the Kalman filter algorithm is used to filter and predict the sensor data in real time.

[0012] On the other hand, the present invention also provides an open-pit mine air-ground slope integrated supervision method, including a satellite remote sensing data processing and analysis step, in which:

[0013] After the satellite remote sensing monitoring subsystem acquires the optical satellite images and SAR satellite interferograms, it performs preprocessing operations, converts the data into a standard format, and performs geometric and radiation corrections. It then uses a deep learning-based image semantic segmentation algorithm to process the optical satellite images. After image semantic segmentation, it extracts information such as the slope boundary and slope morphology, and performs quantitative analysis on this information. For the SAR satellite interferograms, it uses multi-baseline interferometry technology and phase unwrapping algorithm to calculate the three-dimensional displacement information of the slope, and uses time series analysis methods to analyze the changing trend of the slope displacement.

[0014] In the process of processing optical satellite images by the deep learning-based image semantic segmentation algorithm, the improved U-Net network structure gradually extracts the feature information of the image through the convolution layer and the pooling layer during the encoding process, and restores the resolution of the image by using the deconvolution layer and the upsampling layer during the decoding process. At the same time, the attention mechanism module between the encoding and decoding paths enhances the attention and extraction of the slope area features according to the importance weights of different feature channels.

[0015] It also includes the step of drone monitoring data processing and analysis, in which:

[0016] After the optical camera images, LiDAR point cloud data and thermal imaging data collected by the UAV low-altitude monitoring subsystem are transmitted back to the ground control center, the optical camera images are first stitched and orthorectified; the LiDAR point cloud data are filtered, and the point cloud content of the slope is extracted by combining morphological filtering and cloth simulation filtering; then, by comparing the point cloud data of different periods, the local deformation of the slope is calculated based on feature matching and least squares adjustment principles; for thermal imaging data, the temperature anomaly detection algorithm is used to identify the temperature anomaly area based on statistical analysis and threshold setting, and the causes and potential risks of the temperature anomaly are analyzed in combination with the historical data of the slope and other monitoring data.

[0017] It also includes a ground sensor data processing and analysis step, in which:

[0018] After the data collected by the ground sensor monitoring subsystem, such as GPS receivers, total stations, inclination sensors, strain sensors, and groundwater level sensors, are transmitted to the data processing center, the data are first standardized; the data are reduced in dimension using the principal component analysis (PCA) method to extract the main feature information; a slope stability classification model based on support vector machine (SVM) is constructed, with the reduced feature data as input and the slope stability status as output, and the model is trained using a large amount of historical data; and the Kalman filter algorithm is used to perform real-time filtering and prediction on the sensor data.

[0019] The algorithm innovations based on deep learning and advanced measurement technology include: in the semantic segmentation of satellite remote sensing images, an algorithm that combines an improved U-Net network structure with an attention mechanism improves the accuracy of slope boundary extraction by more than 20% and has a faster processing speed than traditional image segmentation methods; in the processing of UAV Li DAR point cloud data, a method based on a combination of morphological filtering and cloth simulation filtering improves the integrity of feature points by about 30% and reduces the error of deformation monitoring by more than 15% when extracting slope point cloud features, compared with traditional filtering methods; in the processing of ground sensor data, a slope stability classification model based on principal component analysis and support vector machine and the application of the Kalman filtering algorithm improves the classification accuracy by more than 35% compared with traditional empirical evaluation methods, and can predict the changing trend of slope status 1-2 days in advance.

[0020] The intelligent slope stability prediction and supervision decision-making mechanism includes: establishing a slope stability prediction model based on machine learning through deep fusion and analysis of multi-source data from the air, ground and space, which can accurately judge the stability state of the slope based on real-time monitoring data and predict its future change trend; establishing an intelligent supervision decision-making mechanism based on the slope stability prediction results. When it is predicted that the slope is in a potentially unstable state, the system can automatically generate early warning information and provide targeted prevention and control measures based on the specific conditions of the slope. At the same time, it can monitor and evaluate the effectiveness of the prevention and control measures in real time, and adjust the supervision strategy in time according to the evaluation results.

[0021] Beneficial technical effects of the invention:

[0022] First, an innovative air-ground-ground multi-source data fusion monitoring system has been built. Satellite remote sensing, drone low-altitude monitoring and ground sensor monitoring are organically combined to form an all-round monitoring network. Satellite remote sensing provides large-area periodic data, drones obtain high-resolution local information, and ground sensors monitor changes in physical parameters of key parts. The three complement each other, overcoming the limitations of traditional single monitoring methods and greatly improving the comprehensiveness and accuracy of monitoring data.

[0023] Second, high-precision data processing algorithms were designed based on the data characteristics of different monitoring methods. In satellite remote sensing, image semantic segmentation algorithms based on deep learning and multi-baseline interferometry technology have improved the accuracy of image and interferogram analysis; in drone monitoring, the relevant processing algorithms for optical camera images, LiDAR point cloud data, and thermal imaging data have achieved effective data utilization and in-depth analysis; ground sensor data processing uses multi-sensor data fusion evaluation algorithms and Kalman filtering algorithms to improve processing efficiency and stability prediction accuracy.

[0024] Third, algorithm innovation based on deep learning and advanced measurement technology has been achieved. For example, the satellite remote sensing image segmentation algorithm has improved the accuracy of boundary extraction and processing speed; the UAV LiDAR point cloud data processing method has improved the accuracy of feature extraction and deformation monitoring; the relevant models and algorithms for ground sensor data processing have improved the scientificity of stability assessment and the timeliness of prediction.

[0025] Fourth, an intelligent slope stability prediction and supervision decision-making mechanism has been created. Through data fusion analysis, a prediction model is established to accurately judge stability and predict trends based on multiple factors. It can also automatically issue warnings based on prediction results, provide prevention and control suggestions, and monitor and evaluate the effects of measures in real time, thus achieving closed-loop management and dynamic optimization. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 : Flow chart of the working principle of the present invention. DETAILED DESCRIPTION

[0027] Example 1

[0028] 1. System Architecture

[0029] 1. Satellite remote sensing monitoring subsystem:

[0030] 1. Use multiple high-resolution optical satellites and synthetic aperture radar (SAR) satellites to conduct regular large-scale monitoring of open-pit mine slopes. Optical satellites can obtain high-resolution visible light and infrared images of slopes with a resolution better than 0.5 meters, and can clearly identify information such as slope topography, vegetation coverage, and surface cracks; SAR satellites are not restricted by weather and lighting conditions. By transmitting and receiving microwave signals, they can obtain radar interference patterns of slopes, which can be used to monitor small displacement changes of slopes with a monitoring accuracy of up to millimeter level.

[0031] 2. The satellite data receiving station receives the data transmitted back by the satellite and performs preprocessing, including data format conversion, geometric correction, radiation correction and other operations, to convert the raw data into standard image data for subsequent analysis. Then, the optical satellite image is processed using an image semantic segmentation algorithm based on deep learning. The algorithm constructs an improved U-Net network structure and adds an attention mechanism module between the encoding and decoding paths. It can automatically focus on the key features of the slope area, accurately segment the slope from the surrounding environment, and extract information such as the slope boundary and slope surface morphology. For SAR satellite interferograms, multi-baseline interferometry technology and phase unwrapping algorithm are used to accurately calculate the three-dimensional displacement information of the slope, and the changing trend of the slope displacement is monitored through time series analysis methods. 2. UAV low-altitude monitoring subsystem:

[0032] 1. Equipped with multiple drones, equipped with high-precision optical cameras, LiDAR, thermal imagers and other equipment, to conduct low-altitude and detailed monitoring of open-pit mine slopes. The drones fly over the mining area regularly according to the preset flight path and mission planning, and the flight altitude can be adjusted between 50-500 meters according to actual needs. The optical camera can obtain ultra-high-resolution images of the slope with a resolution better than 0.1 meters, which is used to observe the rock texture, joints and cracks and local deformation characteristics on the slope surface in detail; LiDAR can emit laser pulses and measure the time from the laser emission to the return, so as to obtain the three-dimensional point cloud data of the slope. The point cloud density can reach more than 100 points per square meter, which can accurately reflect the topographic undulations of the slope and the spatial position of objects; the thermal imager can detect the temperature distribution on the slope surface, and by analyzing the temperature abnormality area, potential geological disasters can be found, such as changes in groundwater levels and temperature increases caused by stress concentration inside the rock mass.

[0033] 2. The data collected by the drone is transmitted back to the ground control center in real time through the wireless transmission module, and the data is processed in the ground control center. First, the optical camera images are spliced ​​and orthorectified to generate a complete orthophoto map of the slope. Then, a slope deformation monitoring algorithm based on point cloud features is used. The algorithm first filters the LiDAR point cloud data, removes noise points and ground points, and extracts the point cloud features of the slope. Then, by comparing the point cloud data of different periods, based on feature matching and least squares adjustment principles, the local deformation of the slope is calculated, including displacement, settlement, tilt and other information. For thermal imaging data, a temperature anomaly detection algorithm is used to identify temperature anomaly areas based on statistical analysis and threshold setting, and the causes and potential risks of temperature anomalies are analyzed in combination with the geological data of the slope and other monitoring data.

[0034] 3. Ground sensor monitoring subsystem:

[0035] 1. Various types of sensors are deployed at key locations of open-pit mine slopes, such as the top, shoulder, foot and potential sliding surfaces, including high-precision global positioning system (GPS) receivers, total stations, inclination sensors, strain sensors and groundwater level sensors. GPS receivers and total stations are used to measure the three-dimensional coordinates and displacement changes of slope monitoring points in real time, with a measurement accuracy of up to millimeter level; inclination sensors can measure the inclination angle changes of slopes with an accuracy better than 0.01°; strain sensors are buried inside the rock and soil to monitor the strain state of the rock and soil and reflect the stress changes inside the slope; groundwater level sensors are used to monitor the depth and change trend of groundwater levels, providing a basis for analyzing the impact of the hydrogeological conditions of the slope on stability.

[0036] 2. The data collected by the sensor is transmitted to the data processing center through a wired or wireless data transmission network, and the data processing center collects, stores and processes the data in real time. A slope stability assessment algorithm based on multi-sensor data fusion is adopted. The algorithm first standardizes the data of different types of sensors, and then uses the principal component analysis (PCA) method to reduce the dimension of the data and extract the main feature information. Then, a slope stability classification model based on support vector machine (SVM) is constructed, and the reduced-dimensional feature data is used as input, and the stability state of the slope (stable, potentially unstable, unstable) is used as output. The model is trained with a large amount of historical data, so that the model can accurately judge the stability of the slope based on the real-time data of the sensor. At the same time, the Kalman filter algorithm is used to filter and predict the sensor data in real time to improve the accuracy and reliability of the data and predict the changing trend of the slope state in advance.

[0037] 2. Methodology

[0038] 1. Satellite remote sensing data processing and analysis steps:

[0039] 1. After the satellite remote sensing monitoring subsystem acquires the optical satellite image and SAR satellite interferogram, it performs preprocessing operations, converts the data into a standard format, and performs geometric correction and radiation correction. Then, the optical satellite image is processed using the image semantic segmentation algorithm based on deep learning. In the encoding process, the improved U-Net network structure gradually extracts the feature information of the image through the convolution layer and the pooling layer. In the decoding process, the resolution of the image is restored by the deconvolution layer and the upsampling layer. At the same time, the attention mechanism module between the encoding and decoding paths enhances the attention and extraction of the slope area features according to the importance weights of different feature channels. For example, when processing a slope image covered with vegetation, the attention mechanism will automatically increase the feature weight of the transition area between the vegetation and the slope boundary, so as to more accurately segment the slope area. After the image semantic segmentation, the slope boundary, slope morphology and other information are extracted, and the information is quantitatively analyzed, such as calculating the slope gradient, slope aspect, slope area and other parameters.

[0040] 2. For SAR satellite interferograms, use multi-baseline interferometry technology and phase unwrapping algorithm. Multi-baseline interferometry technology improves the accuracy and reliability of interferometry by jointly processing multiple SAR images. The phase unwrapping algorithm solves the problem of phase entanglement in the interferogram, obtains accurate phase difference information, and then calculates the three-dimensional displacement information of the slope. Through the time series analysis method, the displacement data of different periods are compared and fitted, and displacement change models such as linear regression models and polynomial models are established to analyze the change trend of slope displacement and determine whether there is an accelerated growth trend of displacement, so as to predict the possibility of slope instability.

[0041] 2. UAV monitoring data processing and analysis steps:

[0042] 1. After the optical camera images, LiDAR point cloud data and thermal imaging data collected by the drone low-altitude monitoring subsystem are transmitted back to the ground control center, the optical camera images are first stitched and orthorectified. The stitching algorithm stitches multiple images into a complete orthophoto map based on image feature matching and geometric transformation. The orthorectification corrects the image into a vertically projected map image based on the drone's flight attitude, camera parameters and terrain information for accurate measurement and analysis.

[0043] 2. Filter the LiDAR point cloud data using a method based on a combination of morphological filtering and cloth simulation filtering. Morphological filtering first dilates and erodes the point cloud to remove some small noise points and isolated points; cloth simulation filtering simulates the point cloud as a piece of cloth that naturally droops under the action of gravity and fits the ground, thereby separating ground points and non-ground points and extracting the point cloud features of the slope. Then, by comparing point cloud data from different periods, the local deformation of the slope is calculated based on feature matching and least squares adjustment principles. For example, for the monitoring point on the top of the slope, by matching the same-name points in the point cloud data from different periods, the change in its three-dimensional coordinates is calculated to obtain information such as displacement, settlement and tilt, and a spatial distribution histogram of the deformation is drawn to intuitively display the deformation of the slope.

[0044] 3. For thermal imaging data, use the temperature anomaly detection algorithm. First, perform radiation calibration and temperature correction on the thermal imaging data to convert the grayscale values ​​in the image into actual temperature values. Then, based on the statistical analysis method, calculate the statistical parameters such as the mean and standard deviation of the temperature data in the image, set a reasonable temperature threshold, and determine the area above the threshold as a temperature anomaly area. Combined with the geological data of the slope, such as stratum lithology, geological structure, etc., and other monitoring data, such as groundwater level data, analyze the causes of the formation of temperature anomaly areas, determine whether there is abnormal thermal conduction of the rock mass due to rising groundwater levels, frictional heating caused by internal stress concentration in the rock mass, etc., and evaluate its potential impact on slope stability.

[0045] 3. Ground sensor data processing and analysis steps:

[0046] 1. After the data collected by the ground sensor monitoring subsystem, such as GPS receivers, total stations, inclination sensors, strain sensors, and groundwater level sensors, are transmitted to the data processing center, the data is first standardized to convert the data of different types of sensors into a unified numerical range and data format. For example, the displacement data is normalized according to its range, and the strain data is converted into a dimensionless relative strain value.

[0047] 2. Use the principal component analysis (PCA) method to reduce the dimensionality of the data. PCA calculates the covariance matrix of the data, determines its eigenvalues ​​and eigenvectors, selects several principal components with larger eigenvalues, projects the original high-dimensional data into a low-dimensional space, and extracts the main characteristic information, such as integrating multiple sensor data into several key eigenvectors that reflect the overall deformation trend, stress state, and hydrological conditions of the slope.

[0048] 3. Construct a slope stability classification model based on support vector machine (SVM). The reduced feature data is used as input, the slope stability state is used as output, and the model is trained through a large amount of historical data. During the training process, the cross-validation method is used to select the kernel function of SVM (such as radial basis function) and optimize the model parameters (such as penalty factor and kernel function parameters) so that the model can accurately classify the stability of the slope. At the same time, the Kalman filter algorithm is used to filter and predict the sensor data in real time. The Kalman filter algorithm continuously updates the system's state estimate and covariance matrix based on the sensor's measurement value and the system's state equation, improves the accuracy and reliability of the data, and predicts the change trend of the slope state in advance based on the current state estimate and the change law of historical data, such as predicting the range of slope displacement changes and possible changes in stability state in the future.

[0049] 3. Innovation

[0050] 1. Innovative air-ground multi-source data fusion monitoring system:

[0051] 1. An all-round monitoring system combining satellite remote sensing, drone low-altitude monitoring and ground sensor monitoring has been built, realizing multi-scale and multi-dimensional monitoring of open-pit mine slopes from macro to micro, from surface to interior. Satellite remote sensing provides large-area, periodic monitoring data, drone low-altitude monitoring obtains high-resolution, refined local information, and ground sensors monitor the changes in physical parameters of key parts of the slope in real time. The three monitoring methods complement each other, effectively solving the limitations of traditional single monitoring methods and improving the comprehensiveness and accuracy of monitoring data.

[0052] 2. According to the data characteristics of different monitoring methods, corresponding high-precision data processing algorithms are designed. In satellite remote sensing data processing, the image semantic segmentation algorithm based on deep learning and multi-baseline interferometry technology improve the analysis accuracy of optical satellite images and SAR satellite interferograms; in UAV monitoring data processing, the stitching and orthorectification of optical camera images, the filtering and deformation monitoring algorithm of LiDAR point cloud data, and the temperature anomaly detection algorithm of thermal imaging data realize the effective use and in-depth analysis of UAV collected data; in ground sensor data processing, the slope stability assessment algorithm and Kalman filter algorithm based on multi-sensor data fusion improve the processing efficiency of sensor data and the accuracy of stability prediction.

[0053] 2. Algorithm innovation based on deep learning and advanced measurement technology:

[0054] 1. In the semantic segmentation of satellite remote sensing images, the proposed algorithm combining the improved U-Net network structure with the attention mechanism can automatically focus on the key features of the slope area, improving the accuracy and efficiency of image segmentation. Compared with traditional image segmentation methods, the accuracy of slope boundary extraction is improved by more than 20%, and the processing speed is faster, which can meet the needs of rapid analysis of large-scale satellite image data.

[0055] 2. In the processing of UAV LiDAR point cloud data, the method based on the combination of morphological filtering and cloth simulation filtering effectively removes the noise and ground points in the point cloud data, improves the accuracy of point cloud feature extraction and slope deformation monitoring. Through comparative experiments with traditional filtering methods, this method improves the integrity of feature points by about 30% when extracting slope point cloud features, and reduces the error of deformation monitoring by more than 15%.

[0056] 3. In the ground sensor data processing, the slope stability classification model based on principal component analysis and support vector machine and the application of Kalman filter algorithm have improved the scientific nature of slope stability assessment and the timeliness of prediction. Compared with the traditional empirical evaluation method, the classification accuracy of this model has increased by more than 35%, and it can predict the change trend of slope status 1-2 days in advance, providing more time for disaster prevention.

[0057] 3. Intelligent slope stability prediction and supervision decision-making mechanism:

[0058] 1. Through the deep integration and analysis of multi-source data from the air, ground and space, a slope stability prediction model based on machine learning was established, which can accurately judge the stability state of the slope based on real-time monitoring data and predict its future change trend. This prediction model not only takes into account traditional factors such as the geometric shape and displacement change of the slope, but also integrates multiple factors such as the physical and mechanical properties of the rock and soil, hydrogeological conditions, and temperature field distribution, thus improving the accuracy and reliability of the prediction.

[0059] 2. Based on the slope stability prediction results, an intelligent supervision decision-making mechanism has been established. When the slope is predicted to be in a potentially unstable state, the system can automatically generate early warning information and provide targeted prevention and control measures based on the specific conditions of the slope, such as adjusting the mining operation plan, strengthening slope support, and implementing drainage projects. At the same time, the system can also monitor and evaluate the effects of prevention and control measures in real time, and adjust the supervision strategy in a timely manner based on the evaluation results, thus realizing closed-loop management and dynamic optimization of slope supervision.

[0060] Example 2

[0061] (I) Air-space-ground data acquisition module

[0062] 1. Satellite remote sensing data collection

[0063] Satellite images of the open-pit mine area are obtained regularly using high-resolution optical satellites and synthetic aperture radar (SAR) satellites. Optical satellite images can be used to obtain surface information, while SAR images can penetrate clouds to achieve all-weather monitoring and obtain information such as topographic changes and surface deformation of the slope.

[0064] 2. Aerial data collection

[0065] Aerial surveying is performed by using drones equipped with a variety of sensors (such as high-resolution cameras and lidar). The camera can take high-definition images of the slope to identify cracks, landslides and other signs of geological disasters; the lidar can accurately obtain three-dimensional terrain data of the slope, with measurement accuracy up to centimeters.

[0066] 3. Ground data collection

[0067] Ground sensors, including displacement sensors, stress sensors, pore water pressure sensors, etc., are deployed at key locations on the slope to collect real-time data on displacement, stress, pore water pressure, etc. inside and on the surface of the slope to reflect changes in the mechanical state of the slope.

[0068] (II) Data transmission and preprocessing module

[0069] 1. Data transmission network

[0070] Establish a reliable data transmission network, including satellite communications, wireless communications (such as 4G / 5G networks for drone data transmission) and wired communications (for ground sensor data transmission), to ensure that data collected from the air, ground and space can be transmitted to the data processing center in a timely and accurate manner.

[0071] 2. Data Preprocessing

[0072] The collected data is preprocessed, including data cleaning (removing abnormal data caused by weather, equipment failure, etc.), data alignment (unifying data from different sources into the same coordinate system) and data normalization (standardizing data of different magnitudes) to provide high-quality data for subsequent analysis and supervision.

[0073] (III) Slope stability analysis and assessment module

[0074] 1. Multi-source data fusion and feature extraction

[0075] The satellite remote sensing, aerial survey and ground sensor data are integrated, and the complementary information between different data is mined using data fusion algorithms. Features related to slope stability are extracted from the integrated data, such as terrain change features, crack development features, mechanical parameter change features, etc.

[0076] 2. Stability analysis driven by new algorithm model

[0077] The following three unreported algorithm models are used to analyze and evaluate the slope stability, comprehensively considering the influence of geological conditions, mining activities, meteorological factors, etc. on slope stability, and generating slope stability assessment results.

[0078] (IV) Early warning and decision support module

[0079] 1. Early warning mechanism

[0080] A multi-level early warning mechanism is established based on the slope stability assessment results. When the slope stability index exceeds the set threshold, a warning message of the corresponding level is issued, including SMS, email, sound and light alarm, etc., to notify mine managers and relevant operators.

[0081] 2. Decision support

[0082] Provide decision-making support for slope management and safe production in mines, and propose reasonable response measures based on early warning information and slope stability analysis results, such as adjusting mining plans, strengthening slope support, and evacuating personnel.

[0083] 4. Solution

[0084] 1. Slope stability assessment model based on quantum evolution neural network

[0085] 1. Model Principle

[0086] The quantum evolutionary neural network combines the advantages of quantum computing and evolutionary neural networks. The quantum computing part uses the superposition and entanglement characteristics of quantum bits to increase the search and representation capabilities of the network, enabling the neural network to handle more complex nonlinear relationships. The evolutionary algorithm is used to optimize the structure and parameters of the neural network, and to find the optimal network model by simulating the biological evolution process, selection, crossover and mutation operations. In the slope stability assessment, this model can comprehensively consider the impact of multiple factors on slope stability.

[0087] 2. Algorithm Implementation

[0088] First, the slope data collected from the air and ground are encoded as the input of the quantum evolutionary neural network. The parameters of the quantum evolutionary neural network are initialized, including the initial state of the quantum bit, the network structure (number of layers, number of neurons, etc.) and the parameters of the evolutionary algorithm (population size, crossover probability, mutation probability, etc.). During the training process, the network structure and parameters are optimized through the evolutionary algorithm, and the state of the quantum bit and the network connection weight are adjusted according to the input data and the known slope stability state (training samples). After multiple iterations, a quantum evolutionary neural network model that can accurately evaluate the slope stability is obtained.

[0089] 3. Advantages

[0090] The model has strong nonlinear processing and global search capabilities, and can effectively mine the complex relationship between multi-source data from the air, space, and ground and slope stability. Compared with traditional neural networks, quantum evolution neural networks converge faster, have better adaptability to complex geological conditions and multi-factor coupling problems, and can improve the accuracy of slope stability assessment.

[0091] (II) Slope stability analysis model integrating graph convolutional neural network and fuzzy cognitive map

[0092] 1. Model Principle

[0093] Graph convolutional neural networks (GCNNs) are good at processing data with graph structures. In open-pit mine slope supervision, the slope is regarded as a graph, where nodes represent different monitoring points (including satellite remote sensing monitoring points, aerial measurement points, and ground sensor locations), and edges represent the spatial relationship or physical connection between monitoring points. GCNN can learn the feature representation of nodes to extract the overall characteristics of the slope. Fuzzy cognitive maps (FCMs) are used to deal with uncertainty and causal relationships in slope stability analysis, and to make inferences and decisions by establishing fuzzy relationships between nodes (such as geological factors, mining factors, monitoring data, etc.).

[0094] 2. Algorithm Implementation

[0095] Construct the graph structure of the slope and assign the data collected from the air and ground to the corresponding nodes. Use GCNN to perform convolution operations on the graph structure to extract the features of the nodes and the global features of the graph. Input the features extracted by GCNN into FCM to determine the node status and causal weight of FCM. Through the iterative calculation of FCM, update the node status according to the changes in the input data, and finally output the evaluation results of the slope stability. During the training process, use known slope stability and instability cases to train the model and adjust the parameters of GCNN and FCM.

[0096] 3. Advantages

[0097] This fusion model can make full use of the spatial feature extraction capability of GCNN and the uncertainty processing capability of FCM to better analyze the stability of the slope. Especially for open-pit mine slopes with complex spatial structures and multiple uncertain factors, this model can more accurately reflect the actual state of the slope and provide strong support for supervision.

[0098] 3. Slope deformation prediction and early warning model based on spatiotemporal attention mechanism and long short-term memory network

[0099] 1. Model Principle

[0100] The spatiotemporal attention mechanism can automatically learn the importance of spatiotemporal features in the data. In the open-pit mine slope monitoring data, data at different time and spatial locations have different effects on slope deformation. The spatiotemporal attention mechanism can focus on key spatiotemporal information. The long short-term memory network (LSTM) is good at processing data with time series characteristics, such as the displacement and stress of the slope that change over time. Combining the two can accurately predict the deformation trend of the slope and achieve early warning.

[0101] 2. Algorithm Implementation

[0102] The slope data collected from the air and ground are organized according to the time and space dimensions to form a spatiotemporal data sequence. In the model, the spatiotemporal attention mechanism is first used to process the data sequence, calculate the attention weight of each time step and spatial position, and highlight the key information. The weighted data is then input into the LSTM network, which learns the time series law of slope deformation based on historical data and predicts future deformation trends. According to the prediction results and the set warning threshold, the slope deformation warning information is issued. During the training process, the historical slope deformation data is used to train the model and adjust the parameters of the spatiotemporal attention mechanism and LSTM.

[0103] 3. Advantages

[0104] The model can effectively capture the spatiotemporal characteristics of slope monitoring data and improve the accuracy of deformation prediction. By focusing on key spatiotemporal information, the interference of irrelevant data is reduced, enabling the LSTM network to better learn deformation laws. Compared with traditional prediction models, the model based on the spatiotemporal attention mechanism and LSTM can issue early warning information more timely and accurately, providing protection for the safety supervision of open-pit mine slopes.

Claims

1. An integrated monitoring system for open-pit mines, characterized in that: include: A satellite remote sensing monitoring subsystem is used to perform regular large-scale monitoring of open-pit mine slopes using multiple high-resolution optical satellites and synthetic aperture radar (SAR) satellites. The optical satellites can obtain high-resolution visible light and infrared images of the slopes, and the SAR satellites can obtain radar interferograms of the slopes. A satellite data receiving station is used to receive data sent back by the satellites and perform preprocessing, including data format conversion, geometric correction, and radiation correction operations. It is also used to process optical satellite images using an image semantic segmentation algorithm based on deep learning, and to process SAR satellite interferograms using multi-baseline interferometry technology and phase unwrapping algorithms. The UAV low-altitude monitoring subsystem is used to equip multiple UAVs with high-precision optical cameras, LiDAR, thermal imagers and other equipment to conduct low-altitude fine monitoring of open-pit mine slopes. The data collected by the UAVs is transmitted to the ground control center in real time through the wireless transmission module for data processing; The ground sensor monitoring subsystem is used to deploy various types of sensors at key locations on the slopes of open-pit mines, including high-precision global positioning system (GPS) receivers, total stations, inclination sensors, strain sensors, and groundwater level sensors. The data collected by the sensors are transmitted to the data processing center for processing via a wired or wireless data transmission network.

2. The open-pit mine air-ground-slope integrated monitoring system according to claim 1 is characterized in that: When the deep learning-based image semantic segmentation algorithm processes optical satellite images, an improved U-Net network structure is adopted. The network structure adds an attention mechanism module between the encoding and decoding paths, which can automatically focus on the key features of the slope area, accurately segment the slope from the surrounding environment, and extract information such as the slope boundary and slope surface morphology.

3. The open-pit mine air-ground-slope integrated monitoring system according to claim 1 is characterized in that: In the UAV low-altitude monitoring subsystem, data processing of optical camera images includes stitching and orthorectification processing to generate a complete orthophoto map of the slope; filtering processing of Li DAR point cloud data adopts a method based on a combination of morphological filtering and cloth simulation filtering to extract the point cloud features of the slope, and by comparing point cloud data of different periods, based on feature matching and least squares adjustment principles, calculate the local deformation of the slope; using a temperature anomaly detection algorithm for thermal imaging data, based on statistical analysis and threshold setting, identify temperature anomaly areas, and combine the geological data of the slope and other monitoring data to analyze the causes and potential risks of temperature anomalies.

4. The open-pit mine air-ground-slope integrated monitoring system according to claim 1 is characterized in that: In the ground sensor monitoring subsystem, when the data processing center processes the data collected by the sensors, a slope stability assessment algorithm based on multi-sensor data fusion is adopted. The algorithm first standardizes the data of different types of sensors, then uses the principal component analysis (PCA) method to reduce the dimension of the data and extract the main feature information. Then, a slope stability classification model based on a support vector machine (SVM) is constructed, and the feature data after dimension reduction is used as input and the stability state of the slope is used as output. The model is trained through a large amount of historical data, and the Kalman filter algorithm is used to filter and predict the sensor data in real time.

5. An integrated monitoring method for open-pit mines, characterized in that: The method includes the steps of processing and analyzing satellite remote sensing data, in which: after the satellite remote sensing monitoring subsystem obtains the optical satellite image and the SAR satellite interferogram, it performs a preprocessing operation to convert the data into a standard format and perform geometric correction and radiation correction; then, the optical satellite image is processed using an image semantic segmentation algorithm based on deep learning, and after the image semantic segmentation, the boundary and slope morphology information of the slope are extracted, and the information is quantitatively analyzed; for the SAR satellite interferogram, the multi-baseline interferometric measurement technology and the phase unwrapping algorithm are used to calculate the three-dimensional displacement information of the slope, and the changing trend of the slope displacement is analyzed through the time series analysis method.

6. The open-pit mine air-ground-slope integrated supervision method according to claim 5 is characterized in that: In the process of processing optical satellite images by the deep learning-based image semantic segmentation algorithm, the improved U-Net network structure gradually extracts the feature information of the image through the convolution layer and the pooling layer during the encoding process, and restores the resolution of the image by using the deconvolution layer and the upsampling layer during the decoding process. At the same time, the attention mechanism module between the encoding and decoding paths enhances the attention and extraction of the slope area features according to the importance weights of different feature channels.

7. The open-pit mine air-ground-slope integrated supervision method according to claim 5 is characterized in that: It also includes the step of processing and analyzing the UAV monitoring data, in which: after the optical camera images, LiDAR point cloud data and thermal imaging data collected by the UAV low-altitude monitoring subsystem are transmitted back to the ground control center, the optical camera images are first stitched and orthorectified; the LiDAR point cloud data are filtered, and the point cloud content of the slope is extracted by a method based on a combination of morphological filtering and cloth simulation filtering; then, by comparing the point cloud data of different periods, the local deformation of the slope is calculated based on feature matching and least squares adjustment principles; for thermal imaging data, a temperature anomaly detection algorithm is used to identify temperature anomaly areas based on statistical analysis and threshold setting, and the causes and potential risks of temperature anomalies are analyzed in combination with the historical data of the slope and other monitoring data.

8. The open-pit mine air-ground-slope integrated supervision method according to claim 5 is characterized in that: It also includes a ground sensor data processing and analysis step, in which: after the data of the GPS receiver, total station, inclination sensor, strain sensor and groundwater level sensor collected by the ground sensor monitoring subsystem is transmitted to the data processing center, the data is first standardized; the data is reduced in dimension using the principal component analysis (PCA) method to extract the main feature information; a slope stability classification model based on a support vector machine (SVM) is constructed, the feature data after dimensionality reduction is used as input, the stability state of the slope is used as output, and the model is trained using a large amount of historical data; and the Kalman filter algorithm is used to filter and predict the sensor data in real time.

Images