A three-dimensional visualization emergency management system for a mining area

Abstract

The present application relates to the technical field of mine emergency management, and discloses a mine three-dimensional visual emergency management system, comprising a data acquisition module and a three-dimensional sand table generation module. The data acquisition module acquires dynamic data including geological structure (including static exploration and dynamic sensor data), disaster hidden danger and environmental monitoring; the three-dimensional sand table generation module processes data through a three-dimensional visual engine, wherein a spatial coordinate system one, missing data completion and abnormal value correction are completed by a data preprocessing unit, a scene rendering unit is subjected to texture mapping, spatial feature fusion, spatial coupling modeling and three-dimensional scene synthesis to generate a three-dimensional dynamic sand table. The system has function modules such as offline data loading, disaster diffusion simulation and escape route planning, and supports cross-platform rendering and user interactive operation. The present application realizes three-dimensional dynamic display and emergency analysis of mine multi-source data, and improves the accuracy and timeliness of disaster warning and emergency decision-making.

Description

Technical Field

[0001] This invention relates to the field of emergency management technology in mining areas, specifically a three-dimensional visualization emergency management system for mining areas. Background Technology

[0002] In mining operations, due to complex geological conditions and variable working environments, various safety risks and potential hazards remain key factors restricting the safe development of the industry. Traditional emergency management methods in mining areas mainly rely on two-dimensional maps and scattered monitoring data, which are difficult to intuitively and dynamically present the overall geological conditions, distribution of potential hazards, and environmental change trends of the mining area, resulting in a lack of comprehensiveness and timeliness in emergency decision-making.

[0003] From the perspective of geological data management, static and dynamic geological data are usually stored separately in different systems, lacking an effective fusion mechanism. Static geological data mainly comes from previous geological exploration platforms and can reflect the basic geological structure of the mining area, but it cannot be updated in real time. Dynamic geological data, on the other hand, is acquired through sensor networks deployed in the mining area and can monitor dynamic information such as stratum displacement and stress changes in real time. However, the two types of data differ in spatial coordinate systems, data formats, and accuracy, making it difficult for traditional methods to achieve accurate matching and comprehensive analysis, resulting in delayed geological risk early warning.

[0004] In terms of disaster hazard monitoring, existing systems process disaster hazard data in a relatively simple way, often only achieving threshold alarms for single-point data, and failing to deeply analyze the spatial coupling relationships between different types of disaster hazards. For example, there may be a complex correlation between fault activity and groundwater infiltration, but traditional technologies struggle to reveal this potential link through multi-source data fusion, making it impossible to predict the formation and development path of disaster chains in advance.

[0005] The processing of environmental monitoring data also faces challenges. Environmental monitoring in mining areas involves multi-dimensional data such as air quality, dust concentration, and harmful gas content. Traditional systems lack the precision to correct abnormal data and complete missing data, which may lead to distorted environmental risk assessments. In addition, various monitoring data lack intuitive visualization methods, making it difficult for managers to quickly grasp real-time changes in the mining area environment and affecting the speed of emergency response.

[0006] In terms of visualization technology, traditional two-dimensional visualization methods cannot realistically reproduce the three-dimensional spatial structure of mining areas, especially in terms of the layout of underground tunnels and the three-dimensional distribution of geological structures. For emergency management, the lack of support from a three-dimensional dynamic sand table makes it difficult to quickly plan escape routes and simulate the disaster spread process when a disaster occurs, resulting in low efficiency in emergency drills and actual rescue work.

[0007] Cross-platform compatibility issues also hinder the widespread application of emergency management systems in mining areas. Differences in the underlying graphics interfaces of different operating systems and hardware devices make it difficult for traditional visualization engines to achieve unified rendering across platforms. This results in inconsistent display effects and user experiences on different terminal devices, increasing the cost of system deployment and maintenance. Summary of the Invention

[0008] The purpose of this invention is to provide a three-dimensional visualization emergency management system for mining areas to solve the problems mentioned in the background art.

[0009] To achieve the above objectives, the present invention provides the following technical solution: a three-dimensional visualization emergency management system for mining areas, the system comprising:

[0010] The data acquisition module is used to acquire real-time dynamic data of the target mining area. The dynamic data includes a first spatial dataset corresponding to geological structure data, a second spatial dataset corresponding to disaster hazard data, and a third spatial dataset corresponding to environmental monitoring data. The geological structure data includes static geological data obtained from the geological exploration platform and dynamic geological data obtained through the sensor network deployed in the mining area.

[0011] The 3D sand table generation module is used to input the dynamic data into the 3D visualization engine for processing;

[0012] A 3D dynamic sand table of the target mining area is constructed based on the output of the 3D visualization engine. The 3D visualization engine includes a data preprocessing unit and a scene rendering unit. The data preprocessing unit is used to perform spatial transformation on the dynamic data, and the scene rendering unit performs dynamic rendering based on a preset 3D model template of the mining area and historical disaster data. The scene rendering unit includes a spatial mapping layer, a first rendering layer, a second rendering layer, and a display layer connected in sequence. The first rendering layer is used to perform texture mapping processing on multiple spatial datasets contained in the dynamic data to generate a spatial feature fusion model. The second rendering layer is used to model the spatial coupling relationship between the spatial feature fusion models corresponding to various types of dynamic data to generate spatially related topological data. The display layer is used to perform 3D scene synthesis based on the spatially related topological data and the spatial feature fusion model to generate a 3D dynamic sand table.

[0013] Preferably, the step of modeling the spatial coupling relationship between the spatial feature fusion models corresponding to various types of dynamic data to generate spatially correlated topological data includes:

[0014] An edge detection algorithm is used to identify abrupt boundary changes in the spatial feature fusion model, and the boundary distribution sequence corresponding to each type of dynamic data is determined based on the spatial location of each abrupt boundary change.

[0015] Calculate the sum of Manhattan distances between abrupt boundaries with the same spatial encoding in the boundary distribution sequences corresponding to any two types of dynamic data, and determine the spatial association topology data between the two types of dynamic data based on the sum of Manhattan distances; the calculation of the sum of Manhattan distances between abrupt boundaries with the same spatial encoding in the boundary distribution sequences corresponding to any two types of dynamic data includes:

[0016] When the number of mutation boundaries in the boundary distribution sequences corresponding to any two types of dynamic data is inconsistent, the sequence is filled by equidistant interpolation based on the spatial position of the last mutation boundary in the sequence with fewer mutation boundaries, and the sum of Manhattan distances between mutation boundaries with the same spatial encoding is calculated based on the filled sequence.

[0017] Preferably, the data preprocessing unit is specifically used for:

[0018] The first spatial dataset, the second spatial dataset, and the third spatial dataset are subjected to a unified projection transformation according to a preset spatial coordinate system, resulting in the transformed first spatial dataset, the transformed second spatial dataset, and the transformed third spatial dataset; and

[0019] The missing regions in the transformed first spatial dataset and the transformed second spatial dataset are filled in using a linear interpolation method, and the abnormal values ​​in the transformed third spatial dataset are corrected using a fixed threshold truncation method, so as to obtain the first target spatial dataset, the second target spatial dataset and the third target spatial dataset.

[0020] Preferably, the first target spatial dataset includes processed static geological data and processed dynamic geological data, and the data preprocessing unit is further used for:

[0021] Calculate the spatial overlap between the processed static geological data and the processed dynamic geological data within a historical time period;

[0022] Based on the spatial overlap and the geological parameters of the processed static geological data in the target area, predict the predicted geological parameters of the processed dynamic geological data in the target area.

[0023] A comprehensive geological dataset is generated based on the processed dynamic geological data and its predicted geological parameters, and the spatial data corresponding to the comprehensive geological dataset is used as the first target spatial dataset.

[0024] Preferably, the first rendering layer includes a texture mapping unit and a feature overlay unit; wherein,

[0025] The texture mapping unit is used to perform multi-resolution rasterization processing on each type of spatial dataset contained in the dynamic data, so as to extract the corresponding texture feature data from each type of spatial dataset.

[0026] The feature overlay unit is used to overlay the texture feature data extracted from each type of spatial dataset with the original spatial dataset in multiple layers to generate a spatial feature fusion model.

[0027] Preferably, the first rendering layer further includes a feature dimensionality reduction unit, which is used to perform principal component analysis dimensionality reduction processing on the spatial feature fusion model;

[0028] The process of modeling the spatial coupling relationship between the spatial feature fusion models corresponding to various types of dynamic data includes: performing spatial coupling analysis on the dimensionality-reduced spatial feature fusion models corresponding to various types of dynamic data to generate the spatially associated topological data.

[0029] Preferably, the display layer includes a visual interactive interface, which allows users to switch perspectives, overlay layers, and dynamically zoom the 3D dynamic sand table through operation commands.

[0030] The dynamic rendering generation based on the preset three-dimensional model template of the mining area and historical disaster data includes: updating the geometric mesh of the three-dimensional model template in real time through an incremental rendering algorithm to match the changes in the spatially associated topological data;

[0031] The process of synthesizing a 3D scene based on the spatially associated topological data and the spatial feature fusion model includes: rendering the spatially associated topological data in layers according to a preset rendering priority, and dynamically fusing it with the spatial feature fusion model.

[0032] Preferably, the three-dimensional dynamic sand table includes the following functional modules:

[0033] The offline data loading module is used to integrate offline terrain data, disaster hazard distribution maps, and underground tunnel model data of the mining area to ensure the complete display of the 3D scene in a network-free environment; the underground tunnel model includes 3D modeling of support structures, ventilation paths, and safety exits;

[0034] The disaster simulation module is used to dynamically render the disaster spread path based on real-time monitoring data and mark the risk level in the three-dimensional scene with color gradient.

[0035] The emergency passage planning module is used to automatically generate the optimal escape route by combining the alleyway model with the disaster spread path, and supports manual adjustment of path nodes.

[0036] Preferably, the three-dimensional dynamic sand table further includes:

[0037] The layer management module is used to classify and manage all spatial data within the sand table, and supports user-defined layer groups and display priorities; the layer groups include basic geographic layers, geological disaster layers, real-time monitoring layers, and emergency plan layers;

[0038] The element positioning module is used to quickly locate specific alleyways, equipment, or potential hazards based on keywords or codes entered by the user, and mark the target location in the 3D scene with a bright flashing effect;

[0039] The data export module is used to export user-annotated emergency response routes, hazard point distribution, and monitoring data into standard GIS format files.

[0040] Preferably, the 3D visualization engine is implemented based on a cross-platform rendering framework, which is compatible with the underlying hardware acceleration functions of different operating systems through an abstract graphics interface.

[0041] Compared with the prior art, the beneficial effects of the present invention are:

[0042] In terms of data acquisition and processing, the data acquisition module integrates three types of dynamic data: geological structure, disaster risks, and environmental monitoring. The geological structure data encompasses both static exploration data and dynamic sensor data, enabling multi-dimensional, all-time monitoring of the mining area's geological environment. The data preprocessing unit effectively addresses issues such as inconsistent spatial coordinate systems, missing data, and outliers in multi-source data through methods like unified projection transformation, linear interpolation completion, and fixed threshold truncation, ensuring the accuracy and reliability of the data input to the 3D visualization engine. Particularly for static and dynamic geological data, by calculating spatial overlap and generating a comprehensive geological dataset, the accuracy of geological parameter predictions is further improved, providing stronger data support for disaster early warning.

[0043] The 3D sand table generation module, through the innovative architecture of the 3D visualization engine, achieves efficient processing and realistic rendering of dynamic data. The spatialization of data preprocessing lays the foundation for subsequent rendering, while the multi-level processing mechanism of the scene rendering unit is highly innovative. The first rendering layer generates a spatial feature fusion model through texture mapping and feature overlay, achieving deep fusion of multiple spatial datasets. This allows geological structures, disaster risks, and environmental monitoring data to be intuitively presented in the 3D scene, revealing their individual characteristics and interrelationships. Principal component analysis in the feature dimensionality reduction unit reduces data dimensionality while ensuring data feature integrity, improving the efficiency of subsequent spatial coupling analysis. The second rendering layer accurately identifies spatial coupling relationships between different data through edge detection algorithms and Manhattan distance calculation, generating spatially correlated topological data. This reveals the potential connections between disaster risks and geological structures and environmental changes, providing a scientific basis for disaster prediction and emergency decision-making. The visualization interface of the display layer supports operations such as viewpoint switching, layer overlay, and dynamic scaling. Combined with incremental rendering algorithms and layered drawing technology, it achieves real-time updates and dynamic fusion of the 3D dynamic sand table, enabling managers to intuitively and flexibly grasp the real-time status of the mining area.

[0044] The functional modules of the 3D dynamic sand table are closely designed around emergency management needs, demonstrating significant practicality and innovation. The offline data loading module ensures the system can fully display the 3D scene even without a network connection. The detailed modeling of underground tunnels, including support structures, ventilation paths, and safety exits, provides reliable foundational data for emergency rescue. The disaster simulation module dynamically renders disaster spread paths and labels risk levels based on real-time monitoring data, helping to predict disaster development trends and providing a basis for developing rescue strategies. The emergency passage planning module automatically generates optimal escape routes by combining tunnel models and disaster spread paths, and supports manual adjustments, improving the efficiency and flexibility of emergency response. The layer management module enables classified management of spatial data and customizable display priorities, allowing users to quickly obtain key information according to different needs. The element positioning module's rapid positioning and highlighting functions help quickly locate hazard points and equipment positions, improving the targeting of emergency response. The data export module supports exporting emergency plan routes, hazard point distribution, and other data into standard GIS format files, facilitating data sharing and further analysis.

[0045] The system is based on a cross-platform rendering framework. By abstracting the graphics interface, it is compatible with the underlying hardware acceleration functions of different operating systems, ensuring the consistency of the display effect and operation experience of the 3D dynamic sandbox on different terminal devices, reducing the system deployment and maintenance costs, and improving the system's applicability and scalability. Attached Figure Description

[0046] Figure 1This is a schematic diagram illustrating the working principle of the three-dimensional visualization emergency management system for mining areas described in this invention.

[0047] Figure 2 A flowchart for generating spatially correlated topological data;

[0048] Figure 3 A flowchart illustrating the work of the data preprocessing unit;

[0049] Figure 4 This is a flowchart for compositing a 3D scene in the display layer. Detailed Implementation

[0050] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0051] Please see Figures 1-4 This invention relates to a three-dimensional visualization emergency management system for mining areas. The system includes a data acquisition module and a three-dimensional sand table generation module. The specific implementation steps are as follows:

[0052] The data acquisition module is used to acquire real-time dynamic data of the target mining area. This dynamic data includes a first spatial dataset corresponding to geological structure data, a second spatial dataset corresponding to hazard data, and a third spatial dataset corresponding to environmental monitoring data. The geological structure data includes static geological data obtained from the geological exploration platform, such as long-term stable geological information like stratigraphic distribution, rock type, and geological faults in the mining area; it also includes dynamic geological data acquired through a sensor network deployed in the mining area, such as real-time monitoring data on rock displacement, changes in ground stress, and fluctuations in groundwater levels. The second spatial dataset corresponding to hazard data mainly covers the location, scale, and stability indicators of potential hazard points such as landslides, collapses, and debris flows within the mining area. The third spatial dataset corresponding to environmental monitoring data includes real-time monitoring data of environmental parameters such as air quality (e.g., dust concentration, harmful gas content), water quality, and noise within the mining area.

[0053] The 3D sand table generation module is used to input the dynamic data into the 3D visualization engine for processing, and to construct a 3D dynamic sand table of the target mining area based on the output of the 3D visualization engine. The 3D visualization engine includes a data preprocessing unit and a scene rendering unit. The data preprocessing unit is used to spatialize the dynamic data, specifically converting dynamic data from different sources and in different formats into a unified spatial data format for subsequent processing and rendering. The scene rendering unit performs dynamic rendering based on a preset 3D model template of the mining area and historical disaster data. This scene rendering unit includes a spatial mapping layer, a first rendering layer, a second rendering layer, and a display layer connected in sequence. The first rendering layer performs texture mapping processing on multiple spatial datasets contained in the dynamic data to generate a spatial feature fusion model; the second rendering layer models the spatial coupling relationship between the spatial feature fusion models corresponding to various types of dynamic data to generate spatially related topological data; the display layer performs 3D scene synthesis based on the spatially related topological data and the spatial feature fusion model to generate a 3D dynamic sand table.

[0054] The present invention will be further described below with reference to Examples 1 to 5:

[0055] Example 1:

[0056] During the spatial transformation of dynamic data in the data preprocessing unit, two core processes need to be completed: unified projection transformation and data correction and completion. The following section provides a detailed explanation of this embodiment from the aspects of technical principles, operational steps, and implementation logic.

[0057] I. Technical Implementation of Unified Projection Conversion

[0058] The preset spatial coordinate system serves as the baseline framework for the entire data processing. Its selection must comprehensively consider the geographical location of the mining area, industry standards, and subsequent application requirements. For example, if the mining area is located within China and needs to be integrated with the national basic geographic information system, the National Geodetic Coordinate System 2000 is typically used. If the mining area is small or has special local modeling requirements, a custom local coordinate system can also be defined. The purpose of unified projection transformation is to convert spatial datasets from different sources and coordinate systems (i.e., the first spatial dataset, the second spatial dataset, and the third spatial dataset) into a standard format under the same coordinate system, eliminating problems such as data misalignment and inability to overlay data caused by coordinate system differences.

[0059] In practice, for the first spatial dataset (geological structure data), static geological data may come from borehole data and geological profiles from geological exploration platforms. These data may be acquired using different local coordinate systems or projection methods. Dynamic geological data is acquired by sensor networks deployed in the mining area, and the sensor coordinates are typically represented by the local coordinates of the equipment's installation location. For the second spatial dataset (hazard hazard data), it may be acquired through UAV aerial surveys, ground surveys, etc. Aerial survey data often uses a geographic coordinate system (latitude and longitude), while ground survey data may use a Cartesian coordinate system. The third spatial dataset (environmental monitoring data) comes from various environmental sensors, and its coordinate system may be related to the sensor's deployment location and acquisition method, such as a local coordinate system with the monitoring point as the origin.

[0060] The projection transformation process requires the use of coordinate transformation tools in a Geographic Information System (GIS). First, the original coordinate system parameters of each spatial dataset must be identified, including the datum, projection type, central meridian, and scale factor. For the first spatial dataset, if static geological data is based on the 1954 Beijing coordinate system and dynamic geological data is based on a custom local coordinate system, both must be converted to transition coordinate systems before being uniformly transformed to the preset target coordinate system. For the second spatial dataset, if the coordinates of hazard points obtained from aerial surveys are in the WGS84 geographic coordinate system, they must be converted to a Cartesian coordinate system using the Gauss-Kruger projection before being unified with the plane coordinate systems of other datasets. For the third spatial dataset, if environmental monitoring data is represented by relative coordinates with the sensor as the origin, the absolute coordinates of the sensor must be determined first (e.g., through GPS positioning or connection to the mining area control network), then the relative coordinates must be converted to absolute coordinates and incorporated into the preset spatial coordinate system.

[0061] During the conversion process, it is crucial to maintain data accuracy. For continuous data (such as rock displacement fields and geostress distribution), polynomial interpolation is used for coordinate transformation to ensure the continuity and smoothness of the converted data. For discrete data (such as coordinates of hazard points and monitoring point locations), coordinate transformation formulas are directly applied for precise calculations. After the conversion, three spatial datasets are obtained: the first, second, and third converted datasets. At this point, the three types of data are spatially consistent and can be used for subsequent overlay analysis and rendering.

[0062] II. Specific procedures for data correction and completion

[0063] (I) Linear interpolation to complete missing data

[0064] The converted first and second spatial datasets may contain missing data areas due to reasons such as sensor malfunction, data transmission interruption, and uneven distribution of geological exploration boreholes. For example, in geological structural data, if a certain area has missing stratigraphic interface data due to excessively large borehole spacing, or if a sensor fails to collect rock displacement data for a certain period due to equipment failure, data gaps will be formed.

[0065] Linear interpolation methods are based on the assumption of spatial continuity of data, meaning that attribute values ​​between adjacent known data points change linearly. The specific steps are as follows:

[0066] Determine the scope of missing regions: Identify missing regions in the first and second spatial datasets through data visualization or statistical analysis. These regions may appear as blank areas with no data values ​​on a two-dimensional plane or as data breaks in three-dimensional space.

[0067] Selecting adjacent known data points: For each missing data point, search for a certain number of known data points in its neighborhood. The neighborhood range can be determined based on the data distribution density; for example, select points within a 3×3 grid in densely populated areas and appropriately expand the search range in sparsely populated areas.

[0068] Establish a linear interpolation model: Assuming a linear relationship between the missing point and its adjacent known points on the spatial coordinate axes (e.g., X, Y, Z axes), fit the linear equation using the least squares method. Taking a two-dimensional plane as an example, let the coordinates of the missing point be (x0, y0), and the coordinates of the adjacent known points be (x...y0 ... i y i The corresponding attribute value is z. i Then the linear interpolation equation can be expressed as z = a·x + b·y + c. By solving the system of equations, the coefficients a, b, and c are obtained, and then the attribute value z0 of the missing point is calculated.

[0069] Complete the missing data point by point: Repeat the above steps for each data point within the missing area until all missing areas are filled. The completed data needs to be checked for reasonableness, such as comparing the trend of attribute value changes in adjacent areas, to ensure that the interpolation results conform to geological patterns or the distribution characteristics of potential disaster hazards, and to avoid abnormal abrupt changes.

[0070] Linear interpolation can effectively fill in the missing regions in the first and second spatial datasets, making the data spatially continuous and complete, and providing a reliable data foundation for subsequent spatial feature fusion and coupling relationship analysis.

[0071] (ii) Fixed threshold truncation to correct outlier values

[0072] The converted third-space dataset (environmental monitoring data) may contain anomalous values, such as dust concentration monitoring values ​​jumping beyond the range or harmful gas content data suddenly becoming negative. These anomalous values ​​may be caused by sensor malfunctions, electromagnetic interference, data transmission errors, etc. If they are not corrected, the distribution of environmental parameters in the 3D scene will be distorted, affecting the accuracy of emergency decision-making.

[0073] The core of the fixed threshold truncation method is to set a reasonable threshold range based on the physical meaning of environmental monitoring parameters and historical statistical data, and then forcibly correct abnormal values ​​exceeding this range to the threshold boundary value. The specific implementation steps are as follows:

[0074] Determine the threshold range: For each environmental monitoring parameter (such as dust concentration, CO concentration, O2 concentration, etc.), analyze its normal range of variation. For example, the normal range for dust concentration is typically 0-1000 mg / m³. 3 (The specific threshold is determined based on the mining environment). The safe threshold for CO concentration is generally 24 ppm (long-term exposure). Short-term high concentrations may reach several hundred ppm, but negative values ​​or values ​​exceeding the sensor's range will not occur (e.g., if the sensor's range is 0-1000 ppm, values ​​exceeding 1000 ppm are considered abnormal). The threshold setting can refer to industry standards, equipment manuals, and statistical analysis results of historical monitoring data. For example, the average value of historical data ± 3 times the standard deviation can be used as the threshold range, or the warning threshold recommended by the equipment manufacturer can be directly used as the upper limit.

[0075] Identifying outlier values: Iterate through each data point in the transformed third-space dataset and determine whether its value exceeds the threshold range. For example, if the CO concentration monitoring value at a certain moment is -5ppm or 1500ppm (assuming the sensor range is 0-1000ppm), it is determined to be an outlier value.

[0076] Truncation and correction of abnormal values: Values ​​exceeding the upper threshold are uniformly corrected to the upper threshold value; values ​​below the lower threshold (such as negative values) are uniformly corrected to the lower threshold value (usually 0). For example, a CO concentration of -5 ppm is corrected to 0 ppm, and 1500 ppm is corrected to 1000 ppm. The corrected data must be marked as abnormal for subsequent tracing and troubleshooting of sensor malfunctions.

[0077] By using a fixed threshold truncation method, abnormal interference in the third-space dataset can be effectively eliminated, ensuring the rationality and reliability of environmental monitoring data. This allows the environmental parameters displayed in the 3D dynamic sand table to match actual working conditions, providing accurate data support for mine environmental safety assessment and emergency response.

[0078] III. Processing Results and Data Output

[0079] After unified projection transformation, linear interpolation completion, and fixed threshold truncation correction, the first target space dataset, the second target space dataset, and the third target space dataset are obtained. Among them:

[0080] The first target spatial dataset contains processed static and dynamic geological data. The static geological data is aligned with the dynamic geological data in the same coordinate system after coordinate transformation. Missing geological interface data is filled in by linear interpolation. The data has good continuity and spatial consistency, and can fully display the geological structural features and dynamic changes of the mining area.

[0081] The second target spatial dataset: After the disaster hazard data has been standardized in coordinates and missing information has been filled in, the location, scale and other information of the hazard points are accurate and continuous, which is convenient for analyzing the spatial distribution pattern of disaster hazards and their relationship with geological structures.

[0082] The third target spatial dataset: the abnormal values ​​in the environmental monitoring data are corrected, the data distribution conforms to physical laws and actual environmental conditions, and can truly reflect the environmental quality of the mining area.

[0083] These three types of target spatial datasets serve as data inputs for the 3D visualization engine, providing high-quality data sources for subsequent texture mapping processing, spatial coupling relationship modeling, and 3D scene synthesis. This ensures that the 3D dynamic sand table can accurately and intuitively display the real-time dynamic information of the mining area, providing strong technical support for emergency management in the mining area.

[0084] Throughout the data preprocessing process, it is crucial to ensure the standardization and traceability of the processing flow. This includes recording parameter settings for projection transformation, the neighborhood range for linear interpolation, and the basis for threshold settings, for future reference during data verification or system maintenance. Furthermore, data visualization tools can be used to compare and analyze the data before and after processing, visually demonstrating the effectiveness of the preprocessing and ensuring the accuracy and validity of the data processing.

[0085] Example 2:

[0086] During the processing of the first rendering layer, the texture mapping unit and the feature overlay unit work together to convert dynamic data into a spatial feature fusion model with rich visual features. The following section elaborates on Example 2 from the aspects of data processing principles, technical implementation paths, and system interaction logic.

[0087] I. Technical Principles of Multi-Resolution Rasterization Processing

[0088] When performing multi-resolution rasterization on each type of spatial dataset, the texture mapping unit must adhere to the principle of hierarchical representation of spatial data. This principle is based on the characteristics of human visual perception: users have different needs for data detail at different observation scales. For example, at a macroscopic scale, users are more concerned with the overall geological structure outline and hazard distribution trends of the mining area; while at a microscopic scale, they need to view the details of the support structure and equipment layout of specific tunnels. Multi-resolution rasterization processing achieves differentiated representation of the same data at different scales by establishing a pyramid-shaped data structure.

[0089] For the first spatial dataset of geological structural data, multi-resolution rasterization first discretizes continuous geological bodies into regular grids. For example, a three-dimensional geological body is divided into a cubic grid with a side length of 10 meters, and each grid cell is assigned a corresponding geological attribute value (such as lithology, density, porosity, etc.). At the coarse resolution level, a larger grid size (e.g., 100 meters) is used to retain only the main geological interfaces and structural features; as the resolution increases, the grid size gradually decreases (e.g., 50 meters, 20 meters, 10 meters), gradually adding secondary geological interfaces and local structural details. The data at each level is generated through an aggregation algorithm, for example, defining the grid attribute values ​​of low-resolution levels as the statistical characteristics (such as mean, median, maximum, etc.) of all the high-resolution grid attribute values ​​they contain.

[0090] Multi-resolution processing of the second spatial dataset of disaster hazard data focuses on preserving spatial distribution characteristics. For areal hazard areas such as landslides and collapses, simplified polygonal boundaries are used at the low-resolution level to retain their approximate location and area; at the high-resolution level, the number of polygon vertices is increased, the boundary shape is refined, and internal structural features (such as sliding direction and crack distribution) are added. For point-like hazard points (such as monitoring equipment locations and hazardous chemical storage points), a single symbol is used at the low-resolution level, while at the high-resolution level, it is expanded into a composite symbol containing more attribute information (such as symbol size indicating risk level and color indicating hazard type).

[0091] Multi-resolution processing of third-space datasets for environmental monitoring data needs to consider both temporal and spatial dimensions. For continuously distributed environmental parameters such as air quality and water quality, spatial interpolation methods such as Kriging interpolation are used to convert discrete monitoring point data into a continuous raster surface. At low-resolution levels, the time sampling interval is longer (e.g., 1 hour) and the spatial raster size is larger (e.g., 100 meters), mainly reflecting the overall distribution trend of environmental parameters. At high-resolution levels, the time sampling interval is shorter (e.g., 10 minutes) and the spatial raster size is smaller (e.g., 20 meters), which can capture rapid changes and local anomalies in environmental parameters. Data at each level is generated through temporal aggregation and spatial downsampling algorithms to ensure data consistency between different resolution levels.

[0092] II. Implementation Path of Texture Feature Extraction

[0093] Texture feature extraction is a crucial step in transforming spatial datasets into visually perceptible textures. This process is based on feature extraction algorithms from computer vision, optimized for the specific characteristics of mining area data.

[0094] For geological structural data, the Gray-Level Co-occurrence Matrix (GLCM) is used to extract lithological distribution texture. The GLCM describes the spatial correlation of gray values ​​in an image. By calculating the joint probability distribution of gray values ​​at different directions and distances, statistical features of the texture (such as contrast, correlation, energy, and uniformity) are extracted. In geological data processing, geological attribute values ​​(such as density and acoustic velocity) are mapped to gray values ​​to generate pseudo-grayscale images, and then their GLCMs are calculated. For example, layered geological structures appear as a high-probability distribution along the diagonal in the GLCM, while fault structures appear as local abrupt changes in gray values. By analyzing these statistical features, texture patterns of different lithologies can be identified, such as the granular texture of sandstone and the layered texture of shale.

[0095] Texture feature extraction from disaster hazard data focuses on identifying spatial distribution patterns. For landslide hazard areas, Gabor filters are used to extract surface texture features. Gabor filters are bandpass filters capable of detecting local texture changes in images at different scales and directions. By adjusting filter parameters (such as wavelength, orientation, and bandwidth), information such as landslide boundary features, internal crack textures, and vegetation cover differences can be extracted. For example, the surface texture of active landslides typically exhibits an irregular blocky distribution, while stable areas have more continuous texture patterns. For collapse hazard points, Local Binary Pattern (LBP) is used to extract surface roughness features. LBP generates binary patterns by comparing the gray values ​​of the central pixel with those of its neighboring pixels, which are used to describe the microstructure of local textures. The surface of a collapse body typically has high roughness, resulting in increased diversity of LBP patterns.

[0096] Texture feature extraction from environmental monitoring data is based on spatiotemporal variation pattern analysis. For air quality data, wavelet transform is used to extract multi-scale features of its time series. Wavelet transform can decompose a signal into different frequency components, allowing analysis of signal variation characteristics at different time scales. For example, short-term fluctuations in pollutant concentrations can be represented by high-frequency components, while long-term trend changes are reflected by low-frequency components. Mapping wavelet coefficients at different scales to color or brightness values ​​can generate texture images reflecting spatiotemporal changes in air quality. For water quality data, principal component analysis (PCA) dimensionality reduction is used to extract comprehensive features from multiple monitoring indicators. PCA can project high-dimensional data into a low-dimensional space, preserving the main variation information of the data. Mapping the principal component scores obtained from PCA to texture parameters (such as color and transparency) can visually display the overall water quality status and its spatial distribution differences.

[0097] III. Logical Architecture of Multi-Layer Overlay Generation Fusion Model

[0098] When the feature overlay unit overlays the extracted texture feature data with the original spatial dataset in multiple layers, it uses layered rendering and transparency blending techniques to achieve seamless data fusion.

[0099] The entire overlay process is based on a layered structure design of the 3D scene. The bottom layer is the basic terrain layer, which consists of a digital elevation model (DEM) of the mining area and orthophotos, providing realistic surface morphology and texture. Above the basic terrain layer, geological structure layers, hazard hazard layers, and environmental monitoring layers are overlaid in sequence. Each layer of data is rendered according to its spatial location and attribute characteristics, and the visibility and blending degree between layers are controlled by adjusting the transparency settings.

[0100] The overlay of geological structural layers employs volumetric rendering technology. Multi-resolution rasterized geological data is combined with extracted lithological texture features, and a ray casting algorithm is used to calculate the absorption, scattering, and emission effects of light passing through the geological body, thus visualizing the internal structure of the geological body. For example, the texture features of different lithologies are modulated with color and transparency to clearly reveal geological interfaces and internal structures. For special structures such as faults and folds, their spatial morphology and distribution patterns are highlighted by enhancing the contrast and color saturation of their boundary textures.

[0101] The overlay of hazard hazard layers employs symbolization and buffer analysis techniques. Extracted hazard texture features are combined with spatial location information from the original hazard data to generate a symbol layer with visual warning effects. For example, for landslide hazard areas, a red semi-transparent polygon represents its impact range, filled with a landslide texture pattern; for collapse hazard points, symbols with warning signs are used, with symbol size and color dynamically adjusted according to the risk level. Simultaneously, different risk levels are generated based on buffer analysis, and a visual transition between risk levels is achieved through gradual changes in transparency.

[0102] The environmental monitoring layer overlay employs a combination of isosurface and volume rendering techniques. For continuously distributed environmental parameters (such as dust concentration and harmful gas content), isosurfaces are generated through interpolation, and the extracted spatiotemporal texture features are mapped onto these isosurfaces. Variations in color and transparency represent the magnitude and trend of parameter values. For discrete monitoring point data, volume rendering technology is used to generate a spatial distribution field of the monitoring data, correlating texture features with data values ​​to make the spatial distribution of environmental parameters more intuitive. For example, high-concentration areas are represented in red, and low-concentration areas in green; the density of the texture reflects the gradient of parameter value changes.

[0103] During the multi-layer overlay process, different types of data are organically integrated by adjusting parameters such as transparency, brightness, and contrast of each layer. For example, the transparency of the geological structure layer is set to 70%, making the terrain texture of the underlying layers visible and enhancing the realism of the scene; the transparency of the disaster hazard layer is set to 50%, highlighting the hazard areas without obscuring the geological and topographical information of the underlying layers; the transparency of the environmental monitoring layer is dynamically adjusted according to the reliability and importance of the data, ensuring that users can clearly perceive the distribution of environmental parameters. Through this multi-layer overlay method, the generated spatial feature fusion model can simultaneously display multiple types of dynamic data and their interrelationships, providing comprehensive and intuitive decision support for mine emergency management.

[0104] Example 3:

[0105] When modeling the spatial coupling relationship between the spatial feature fusion models corresponding to various dynamic data in the second rendering layer, it is necessary to achieve this through steps such as edge detection, distance calculation, and sequence filling. The following section elaborates on Example 3 from the aspects of technical principles, operation procedures, and data processing logic.

[0106] Edge detection algorithms are designed based on the physical characteristics of dynamic mining area data, employing different detection strategies for different data types. For geological structural data, abrupt boundaries typically correspond to discontinuities in geological structures such as rock strata interfaces and fault zones. When using the Canny edge detection algorithm, the spatial feature fusion model must first be Gaussian smoothed to reduce the impact of data noise. The size of the Gaussian smoothing kernel is determined based on the accuracy and noise level of the geological data, typically using a 3×3 or 5×5 convolution kernel. Next, the magnitude and direction of the image gradient are calculated, and edges are refined using non-maximum suppression techniques, retaining only points with local gradient maximums as candidate edge points. Finally, a dual-threshold method is used to determine the true edges; the low threshold is used to connect edge segments, and the high threshold is used to determine the starting point of the edge. This method effectively identifies abrupt boundaries in geological structures, forming a boundary distribution sequence.

[0107] The identification of abrupt boundary changes in disaster hazard data focuses on the boundary between hazard areas and normal areas. For example, the boundary of a landslide typically manifests as a sudden change in parameters such as terrain slope and surface displacement rate. For this type of data, an edge detection algorithm based on region growing is employed. First, known hazard points are marked as seed points in the spatial feature fusion model. Then, the region is gradually expanded according to a preset growth criterion (such as the attribute difference between adjacent pixels being less than a threshold) until the boundary condition is reached. The boundary condition can be that the attribute difference between adjacent regions exceeds a threshold, or that the boundary reaches the edge of the image. In this way, the boundaries of disaster hazard areas such as landslides and collapses can be accurately identified, forming a corresponding boundary distribution sequence.

[0108] Abrupt boundary identification in environmental monitoring data primarily targets areas of rapid change in parameters such as pollutant concentration, temperature, and humidity. For example, in mines, areas with large concentration gradients typically form around hazardous gas leaks. For this type of data, gradient-based edge detection algorithms, such as the Sobel or Prewitt operators, are employed. These operators detect areas of drastic grayscale value changes by calculating the gradients in the horizontal and vertical directions of the image. In environmental monitoring data processing, parameter values ​​are mapped to grayscale values, and then gradient operators are applied to calculate gradient amplitudes; areas with larger amplitudes are identified as abrupt boundary changes. By thresholding the gradient amplitude image, abrupt boundary changes in environmental parameters are extracted, forming a boundary distribution sequence.

[0109] When calculating the sum of Manhattan distances between abrupt boundary points with the same spatial code in the boundary distribution sequences corresponding to any two types of dynamic data, the boundary distribution sequences first need to be spatially encoded. Spatial encoding uses a quadtree or octree structure to divide the three-dimensional space into grid cells of different levels. Each grid cell is assigned a unique code; the code length represents the spatial resolution, and the code value represents the cell's position in space. For the boundary distribution sequences of geological structural data and hazard data, each abrupt boundary point is mapped to its corresponding grid cell to obtain its spatial code.

[0110] Before calculating the Manhattan distance, the boundary distribution sequences need to be aligned. Due to differences in data acquisition density and precision across different data types, the number of abrupt boundaries in their boundary distribution sequences may vary. To ensure the calculation of distances between abrupt boundaries with the same spatial encoding, the sequences need to be padded. Based on the spatial location of the last abrupt boundary among those with fewer abrupt boundaries, virtual abrupt boundaries are inserted around it at certain intervals. The insertion interval is determined based on the spatial distribution characteristics of the data and the required computational precision, typically chosen as half the average spacing of the original data points.

[0111] In the padded sequence, for each pair of mutation boundaries with the same spatial encoding, the Manhattan distance between them is calculated. In three-dimensional space, the Manhattan distance is defined as the sum of the absolute distances between two points on the X, Y, and Z coordinate axes, i.e., d = |x1-x2| + |y1-y2| + |z1-z2|. By iterating through all pairs of mutation boundaries with the same spatial encoding and summing their Manhattan distances, the Manhattan distance between any two classes of dynamic data is obtained. This sum of distances reflects the spatial similarity of the mutation boundaries between the two classes of data; the smaller the distance, the closer the mutation boundaries are spatially, and the stronger the spatial coupling between them.

[0112] In some cases, some mutation boundary points may not have corresponding points with the same spatial coding. For these points, they are matched with the nearest mutation boundary points with different spatial codings, and the Manhattan distance between them is calculated. To avoid the influence of such imprecise matching on the results, when summing the distances, the distance values ​​of such imprecise matches are multiplied by a weight coefficient less than 1. The weight coefficient is determined according to the degree of difference in spatial coding between the matched points; the greater the difference, the smaller the weight coefficient.

[0113] Based on the calculated sum of Manhattan distances, spatial association topology data between any two types of dynamic data is determined. This spatial association topology data is represented by a graph structure, where nodes represent different types of dynamic data, edges represent the spatial relationships between them, and the weight of each edge is the reciprocal of the sum of the Manhattan distances. In this way, spatial distance relationships are converted into association strength relationships; the smaller the distance, the stronger the association. This spatial association topology data provides crucial spatial association information for subsequent 3D scene synthesis, helping the system correctly handle the spatial relationships between different types of data during rendering, making the generated 3D dynamic sandbox more realistically reflect the actual situation of the mining area.

[0114] Example 4:

[0115] When compositing a 3D scene based on spatially correlated topological data and a spatial feature fusion model in the display layer, it is necessary to achieve the visual representation of the data through layered rendering, dynamic fusion, and interactive interface design. The following section elaborates on Example 4 from the aspects of scene construction principles, rendering technology implementation, and user interaction mechanisms.

[0116] The layered rendering process is based on the importance level classification of spatially correlated topological data, with different types of data assigned different rendering priorities. Geological structural data, as the base layer, carries the underlying structural information of the mining area and is set to the lowest rendering priority. Its rendering employs volumetric rendering technology, discretizing the 3D geological volume into regular meshes and calculating the absorption, scattering, and transmission effects of light passing through the geological volume using a ray casting algorithm. To enhance the display effect of geological interfaces, different color mapping schemes are applied to different lithological regions, such as yellow for sandstone, gray for shale, and black for coal seams. Simultaneously, texture details, such as bedding structures and fault striations, are added to the geological interfaces, and microscopic geological structures are simulated using normal perturbation technology.

[0117] Disaster hazard data, as crucial information for safety early warning, is given the highest rendering priority. For area hazard zones such as landslides and collapses, a semi-transparent polygon overlay texture pattern is used for rendering. The polygon boundaries are emphasized with bold red lines, and the interior is filled with irregular textures to indicate the activity level of the hazard. The texture pattern uses a perturbation pattern generated based on Perlin noise to simulate surface deformation and cracks. For point-like hazard points, warning symbols of different sizes and colors are designed according to the risk level. Low-risk points are represented by yellow triangles, medium-risk points by orange pentagrams, and high-risk points by red explosion symbols. The symbols are always facing the camera using Billboard technology to ensure clear visibility from any viewing angle.

[0118] The rendering priority of environmental monitoring data falls between that of geological structures and potential hazards. For continuously distributed environmental parameters (such as dust concentration and harmful gas content), a combination of isosurface and volume rendering is used. First, discrete monitoring point data is converted into a continuous scalar field using the Kriging interpolation algorithm, and then the MarchingCubes algorithm is used to extract isosurfaces. The color of the isosurfaces is mapped according to the parameter value, such as green for low concentration areas, yellow for medium concentration areas, and red for high concentration areas. To enhance the sense of spatial hierarchy, a transparency gradient is added to the isosurfaces, with increased transparency for isosurfaces further away. For discrete monitoring device locations, icons with dynamic particle effects are used to represent them, with the color and density of the particles reflecting the changing trends of real-time monitoring data.

[0119] During the dynamic fusion phase, the system monitors updates to spatially correlated topological data and spatial feature fusion models in real time. When new data arrives, its validity is first verified, checking its completeness and rationality. For updates to geological structural data, such as rock strata displacement monitoring data, the system employs a smooth transition algorithm to achieve seamless switching between old and new data. By calculating the interpolation function between the old and new data fields, the system gradually transitions from old data to new data over a certain time interval (e.g., 2 seconds) to avoid abrupt changes in the scene.

[0120] The update of disaster hazard data uses an alarm animation effect. When the stability index of a landslide hazard area changes, the boundary of the area flashes red, and the perturbation frequency of the internal texture pattern increases. The system dynamically adjusts the intensity of the alarm effect according to the change of hazard level, with higher flashing frequency and brightness in high-risk areas. For newly added hazard points, a growth animation is used, with the icon gradually enlarging from a small dot to the full size, accompanied by an outward ripple effect to attract user attention.

[0121] Environmental monitoring data updates utilize fluid simulation technology. When a hazardous gas leak is detected, the system renders the gas diffusion process in real-time within a 3D scene based on the leak source location and diffusion model. The diffusion effect employs a physics-based fluid simulation algorithm, considering factors such as gas density, wind speed, and terrain. Gas clouds are implemented using a semi-transparent particle system, with particle color and transparency dynamically adjusted according to gas concentration. Simultaneously, streamline effects are added to the gas diffusion path to visually demonstrate the gas flow direction.

[0122] The visual interactive interface is based on WebGL technology to achieve cross-platform compatibility, supporting multiple operation methods including mouse, keyboard, and touch devices. The perspective switching function offers several preset perspectives, such as a global overview, a walkthrough of the tunnel interior, and close-ups of potential hazards. Users can quickly switch perspectives by clicking the perspective buttons on the interface, freely rotate the scene by dragging the mouse, and zoom in and out using the scroll wheel. To ensure smooth operation, the system employs multi-threading technology, separating rendering calculations from user interaction processing to avoid interface lag.

[0123] The layer overlay function allows users to show or hide specific types of data as needed. The left side of the interface features a layer management panel listing all available layers, including basic terrain, geological structure, hazard hazard, and environmental monitoring layers. Each layer has a checkbox and an opacity slider; users can control the layer's display by checking the checkbox and adjust its opacity by dragging the slider. When there are many layers, the panel supports collapsing and expanding, allowing users to collapse unused layer groups to save interface space.

[0124] The dynamic zoom function employs a progressive loading strategy. As the user zooms in on the scene, the system automatically loads higher-resolution data. For example, in the global view, the system only loads low-resolution terrain models and approximate geological structure data; when the user zooms in to the tunnel level, the system asynchronously loads high-resolution geological data, tunnel support structure data, and equipment distribution data for that area. To avoid blank areas during loading, the system displays temporary placeholder graphics, such as low-precision mesh models or blurred textures.

[0125] During the interaction, the system provides a real-time feedback mechanism. When a user selects an object (such as a potential hazard point or monitoring equipment), the object is highlighted, and an information panel pops up displaying its detailed attributes. The information panel uses a card-style design, including basic information, historical data charts, and relevant multimedia materials. Users can perform further operations using buttons on the panel, such as viewing historical trends and retrieving on-site monitoring videos. To enhance the user experience, all interactive operations are equipped with appropriate animation effects, such as ripple effects when buttons are clicked and smooth transitions when the panel expands.

[0126] Example 5:

[0127] The integration and cross-platform implementation of functional modules in a 3D dynamic sand table requires collaboration of multiple technologies. The following section elaborates on Example 5 from the perspectives of module design principles, data processing flow, and system architecture.

[0128] The offline data loading module employs a hierarchical storage strategy, classifying and storing the mine's topographic data, hazard distribution maps, and underground tunnel model data into different file formats. Topographic data is stored in high-precision DEM (Digital Elevation Model) format as a regular grid of elevation values, supplemented by texture mapping technology to map satellite imagery or aerial photography onto the terrain surface, enhancing realism. Hazard distribution maps are stored in vector data format, with each hazard point or area containing detailed attribute information (such as type, level, and formation time), enabling rapid querying and location through spatial indexing technology. Underground tunnel model data uses the BIM (Building Information Modeling) standard format, including geometric information of the support structure, material properties, construction time, and other parameters. Ventilation paths and safety exit information are stored in a topological network format, recording the connection relationships between nodes (such as tunnel intersections and ventilation openings) and edges (tunnel segments).

[0129] In offline environments, the system caches offline data to local storage via a pre-loading mechanism. Upon startup, it first loads basic terrain data and main tunnel models, providing basic 3D scene browsing functionality. Users can select areas of interest using navigation controls on the interface, and the system dynamically loads detailed data for those areas based on the user's selection. To reduce data loading volume, spatial index structures such as quadtrees or octrees are used to organize the data into blocks, loading only data blocks within the currently visible area and a certain surrounding range. Simultaneously, the system supports data compression technology to reduce data storage volume and improve loading speed without affecting display accuracy.

[0130] The disaster simulation module constructs a disaster diffusion model based on fluid dynamics and discrete element theory. For gaseous disasters such as methane and dust, the Navier-Stokes equations are used to describe fluid motion, considering the influence of factors such as tunnel structure, ventilation conditions, and obstacles on airflow. The equations are discretized and solved using the finite difference method or the finite element method, and the computational domain is divided into a regular grid. Fluid state parameters (such as velocity, pressure, and concentration) are updated at each time step. To improve computational efficiency, GPU acceleration technology is used, distributing computational tasks across multiple GPU cores for parallel processing.

[0131] When rendering the disaster spread path, the system combines particle systems with volume rendering. For large-scale spread areas, volume rendering technology is used to generate a continuous concentration field visualization effect, with different concentration levels represented by color mapping. For localized high-concentration areas or leak sources, particle systems are used to simulate discrete gas cloud movement, with the particle size, color, and trajectory dynamically adjusted based on calculated fluid parameters. Risk level labeling uses a combination of color gradients and transparency variations. Semi-transparent risk area polygons are overlaid in the 3D scene, filled with a gradient color from green (low risk) to red (high risk), with transparency gradually decreasing with distance from the leak source to create a sense of depth.

[0132] The emergency passage planning module uses the A* algorithm as its core path search algorithm, combining the topology of the tunnel model and disaster spread path information to generate the optimal escape route. First, the underground tunnel network is abstracted as a graph structure, where nodes represent key locations such as tunnel intersections and safety exits, and edges represent tunnel segments. The weight of each edge is determined comprehensively based on factors such as tunnel length, passage difficulty, and risk level. During a disaster, the edge weights are dynamically adjusted according to the real-time disaster spread path, with higher-risk areas receiving greater weights.

[0133] The A* algorithm searches for the optimal path by maintaining an open list and a closed list. Starting from the starting point, it calculates the estimated cost (actual cost + heuristic function value) for each neighboring node, selects the node with the smallest estimated cost, adds it to the closed list, and adds its neighbors to the open list. This process is repeated until the destination is found or the open list is empty. In the tunnel network, the Manhattan distance heuristic function is used, which, considering the grid-like structure of the tunnels, can more accurately estimate the shortest path.

[0134] The system supports manual adjustment of path nodes. Users can add, delete, or move nodes on the automatically generated escape route, and the system will recalculate the path based on the new node positions. For ease of use, an intuitive drag-and-drop interface is provided; users can simply click and drag nodes on the path to change its direction. Simultaneously, the system will evaluate the feasibility of the new path in real time, ensuring that the adjusted path does not pass through high-risk areas or impassable obstacles.

[0135] The layer management module uses a tree structure to organize all spatial data within the sandbox. The root node represents the entire 3D scene, while child nodes are divided into major categories such as basic geographic layers, geological hazard layers, real-time monitoring layers, and emergency response plan layers. Each major category is further subdivided into specific layers; for example, the basic geographic layer includes topography, landforms, and water systems, while the geological hazard layer includes landslides, collapses, and debris flows. Users can expand or collapse layer groups using the tree control on the interface to quickly locate the layer they need to manipulate.

[0136] The custom layer grouping feature allows users to create personalized layer collections based on their work needs. Users can add different types of layers to custom groups by dragging and dropping, and set names and icons for the groups. Display priority is set through numerical sorting; each layer is assigned a unique integer value, with smaller values ​​indicating higher display priority. The system renders layers sequentially according to priority, ensuring that high-priority layers are not obscured by lower-priority layers.

[0137] The feature localization module utilizes spatial indexing technology to achieve rapid location functionality. A spatial index is established for each tunnel, equipment, and potential hazard point, recording its geometric location and attribute information. When a user enters keywords or codes, the system first performs a fuzzy match in the attribute database to filter out possible target features. Then, based on the spatial location of the target feature, the system adjusts the viewpoint of the 3D scene to center the target feature within the viewport.

[0138] The highlighting and flickering effect is achieved by modifying the rendering properties of the target feature. For point features, their display size is increased and their color is changed to a vibrant red, while periodic changes in transparency are added to create a flickering effect. For polygon features, a glowing outline is added to their boundaries, with the brightness and thickness of the outline changing periodically over time. To ensure that the highlighting effect does not affect the display of other features, depth testing and stencil buffering techniques are used to limit the highlighting effect to the geometric range of the target feature.

[0139] The data export module supports exporting files in various standard GIS formats. For emergency response routes, the system converts them to ESRIShapefile format, containing the route's geometric information (points, lines) and attribute information (route name, length, risk level, etc.). For hazard point distribution data, it exports to GeoJSON format, retaining detailed information such as coordinates, type, and level for each hazard point. Monitoring data is exported to CSV format, containing time-series data and spatial location information.

[0140] During the export process, the system performs format conversion and coordinate system unification on the data. Coordinates in the 3D scene are converted to the coordinate system required by the target GIS format, ensuring that the exported data can be correctly displayed and analyzed in other GIS software. Simultaneously, the data is compressed and optimized, removing redundant information and improving file transfer and processing efficiency. Users can select the export format, file path, and data range through the export dialog box on the interface. The system will process the data and generate the file in the background, providing progress feedback.

[0141] This cross-platform rendering framework is built on the WebGL 2.0 standard and achieves compatibility with the underlying hardware acceleration functions of different operating systems through an abstract graphics interface. The framework defines a unified API that encapsulates the differences in underlying graphics hardware, freeing upper-layer application code from concern itself with the specific graphics driver implementation. On Windows systems, the framework utilizes the hardware acceleration functions provided by DirectX 11 / 12; on Linux systems, it uses OpenGLES 3.0; and on macOS systems, it supports the Metal graphics API.

[0142] To ensure rendering performance, the framework employs a multi-threaded rendering architecture. The main thread handles user interaction and scene management, while the rendering thread executes the actual graphics rendering operations. Double buffering is used, rendering the scene in a background buffer and then exchanging it with the foreground buffer to avoid flickering during rendering. Simultaneously, texture compression and resource management mechanisms are implemented to optimize the storage and loading of textures, models, and other resources, reducing memory usage and loading time.

[0143] On mobile devices, the framework optimizes the interactive experience for touchscreens. It enables gesture controls such as touch drag, two-finger zoom, and pinch rotation, allowing users to easily browse and manipulate 3D scenes on small screens. Simultaneously, it employs adaptive rendering technology, dynamically adjusting rendering quality and level of detail based on device performance, ensuring smooth operation even on low-spec devices while maintaining visual quality. Through this cross-platform rendering framework, the system can provide a consistent user experience and high-performance 3D rendering effects across various devices, including desktops and mobile devices.

[0144] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0145] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A three-dimensional visualization emergency management system for mining areas, characterized in that, include: The data acquisition module is used to acquire real-time dynamic data of the target mining area. The dynamic data includes a first spatial dataset corresponding to geological structure data, a second spatial dataset corresponding to disaster hazard data, and a third spatial dataset corresponding to environmental monitoring data. The geological structure data includes static geological data obtained from the geological exploration platform and dynamic geological data obtained through the sensor network deployed in the mining area. The 3D sand table generation module is used to input the dynamic data into the 3D visualization engine for processing; A three-dimensional dynamic sand table of the target mining area is constructed based on the output of the three-dimensional visualization engine. The 3D visualization engine includes a data preprocessing unit and a scene rendering unit. The data preprocessing unit performs spatial transformation on the dynamic data, and the scene rendering unit dynamically renders the data based on a preset 3D model template of the mining area and historical disaster data. The scene rendering unit includes a spatial mapping layer, a first rendering layer, a second rendering layer, and a display layer connected in sequence. The first rendering layer performs texture mapping processing on multiple spatial datasets contained in the dynamic data to generate a spatial feature fusion model. The second rendering layer models the spatial coupling relationship between the spatial feature fusion models corresponding to various types of dynamic data to generate spatially related topological data. The display layer synthesizes a 3D scene based on the spatially related topological data and the spatial feature fusion model to generate a 3D dynamic sandbox. The process of modeling the spatial coupling relationship between spatial feature fusion models corresponding to various types of dynamic data to generate spatially correlated topological data includes: An edge detection algorithm is used to identify abrupt boundary changes in the spatial feature fusion model, and the boundary distribution sequence corresponding to each type of dynamic data is determined based on the spatial location of each abrupt boundary change. Calculate the sum of Manhattan distances between abrupt boundaries with the same spatial encoding in the boundary distribution sequences corresponding to any two types of dynamic data, and determine the spatial association topology data between the two types of dynamic data based on the sum of Manhattan distances; the calculation of the sum of Manhattan distances between abrupt boundaries with the same spatial encoding in the boundary distribution sequences corresponding to any two types of dynamic data includes: When the number of mutation boundaries in the boundary distribution sequences corresponding to any two types of dynamic data is inconsistent, the sequence is filled by equidistant interpolation based on the spatial position of the last mutation boundary in the sequence with fewer mutation boundaries, and the sum of Manhattan distances between mutation boundaries with the same spatial encoding is calculated based on the filled sequence.

2. The three-dimensional visualization emergency management system for mining areas as described in claim 1, characterized in that, The data preprocessing unit is specifically used for: The first spatial dataset, the second spatial dataset, and the third spatial dataset are uniformly projected and transformed according to a preset spatial coordinate system to obtain the transformed first spatial dataset, the transformed second spatial dataset, and the transformed third spatial dataset. as well as The missing regions in the transformed first spatial dataset and the transformed second spatial dataset are filled in using a linear interpolation method, and the abnormal values ​​in the transformed third spatial dataset are corrected using a fixed threshold truncation method, so as to obtain the first target spatial dataset, the second target spatial dataset, and the third target spatial dataset.

3. The three-dimensional visualization emergency management system for mining areas as described in claim 2, characterized in that, The first target spatial dataset contains processed static geological data and processed dynamic geological data. The data preprocessing unit is further used for: Calculate the spatial overlap between the processed static geological data and the processed dynamic geological data within a historical time period; Based on the spatial overlap and the geological parameters of the processed static geological data in the target area, predict the predicted geological parameters of the processed dynamic geological data in the target area. A comprehensive geological dataset is generated based on the processed dynamic geological data and its predicted geological parameters, and the spatial data corresponding to the comprehensive geological dataset is used as the first target spatial dataset.

4. The three-dimensional visualization emergency management system for mining areas as described in claim 1, characterized in that, The first rendering layer includes texture mapping units and feature overlay units; wherein, The texture mapping unit is used to perform multi-resolution rasterization processing on each type of spatial dataset contained in the dynamic data, so as to extract the corresponding texture feature data from each type of spatial dataset. The feature overlay unit is used to overlay the texture feature data extracted from each type of spatial dataset with the original spatial dataset in multiple layers to generate a spatial feature fusion model.

5. The three-dimensional visualization emergency management system for mining areas as described in claim 4, characterized in that, The first rendering layer further includes a feature dimensionality reduction unit, which is used to perform principal component analysis dimensionality reduction processing on the spatial feature fusion model; The process of modeling the spatial coupling relationship between the spatial feature fusion models corresponding to various types of dynamic data includes: performing spatial coupling analysis on the dimensionality-reduced spatial feature fusion models corresponding to various types of dynamic data to generate the spatially associated topological data.

6. The three-dimensional visualization emergency management system for mining areas as described in claim 1, characterized in that, The display layer includes a visual interactive interface, which allows users to switch perspectives, overlay layers, and dynamically zoom the 3D dynamic sand table through operation commands. The dynamic rendering generation based on the preset three-dimensional model template of the mining area and historical disaster data includes: updating the geometric mesh of the three-dimensional model template in real time through an incremental rendering algorithm to match the changes in the spatially associated topological data; The process of synthesizing a 3D scene based on the spatially associated topological data and the spatial feature fusion model includes: rendering the spatially associated topological data in layers according to a preset rendering priority, and dynamically fusing it with the spatial feature fusion model.

7. The three-dimensional visualization emergency management system for mining areas as described in claim 1, characterized in that, The three-dimensional dynamic sand table includes the following functional modules: The offline data loading module is used to integrate offline terrain data, disaster hazard distribution maps, and underground tunnel model data of the mining area to ensure the complete display of the 3D scene in a network-free environment; the underground tunnel model includes 3D modeling of support structures, ventilation paths, and safety exits; The disaster simulation module is used to dynamically render the disaster spread path based on real-time monitoring data and mark the risk level in the three-dimensional scene with color gradient. The emergency passage planning module is used to automatically generate the optimal escape route by combining the alleyway model with the disaster spread path, and supports manual adjustment of path nodes.

8. The three-dimensional visualization emergency management system for mining areas as described in claim 7, characterized in that, The three-dimensional dynamic sand table also includes: The layer management module is used to classify and manage all spatial data within the sand table, and supports user-defined layer groups and display priorities; the layer groups include basic geographic layers, geological disaster layers, real-time monitoring layers, and emergency plan layers; The element positioning module is used to quickly locate specific alleyways, equipment, or potential hazards based on keywords or codes entered by the user, and mark the target location in the 3D scene with a bright flashing effect; The data export module is used to export user-annotated emergency response routes, hazard point distribution, and monitoring data into standard GIS format files.

9. A three-dimensional visualization emergency management system for mining areas as described in claim 1, characterized in that, The 3D visualization engine is implemented based on a cross-platform rendering framework, which is compatible with the underlying hardware acceleration functions of different operating systems through an abstract graphics interface.

Images