Mine hidden disaster risk assessment method and system based on space-air-ground integrated perception

Abstract

This invention relates to the field of risk assessment technology, specifically to a method and system for assessing hidden disaster risks in mines based on integrated air-space-ground sensing. The method includes the following steps: collecting data from satellite remote sensing, synthetic aperture radar, UAV aerial surveys, underground exploration, slope monitoring, and dam monitoring; completing multi-scene registration within a unified time window; forming fused features through noise reduction, anomaly removal, and spatial alignment; and performing three-dimensional reconstruction by combining engineering survey, mining layout, and tailings operation information, extracting displacement rate, settlement, stress concentration factor, seepage pressure gradient, microseismic energy density, and slope stability coefficient. Based on fuzzy comprehensive evaluation and changes in adjacent time windows, risk classification, migration discrimination, and early warning assessment are completed. In this invention, the hidden disaster state of underground mines, open-pit mines, and tailings dams is collaboratively characterized through multi-source observation, enhancing the continuity of risk identification, the accuracy of classification, and the ability to track evolution, providing a more stable basis for early warning and response.

Description

Technical Field

[0001] This invention relates to the field of risk assessment technology, and in particular to a method and system for assessing hidden disaster risks in mines based on integrated air-space-ground sensing. Background Technology

[0002] The field of risk assessment technology involves a technical system for identifying, analyzing, and classifying potential hazards in various engineering systems and natural environments. It mainly covers core aspects such as data acquisition and monitoring deployment, hazard type classification, correlation analysis of influencing factors, construction of risk indicator systems, and risk level determination. This field typically involves comprehensive analysis of spatial environment, operating conditions, equipment status, and historical disaster records to form risk identification and evaluation methods for complex scenarios. In mining scenarios, it also includes the systematic integration and analysis of multi-source information such as geological structure, mining activities, hydrological conditions, and surface changes.

[0003] Traditional methods for assessing hidden disaster risks in mines refer to methods for identifying and assessing disaster-causing factors that are not easily observed directly during the mining process. These methods mainly target technical issues such as hidden faults, water accumulation in goaf areas, gas-rich areas, and abnormal stress in surrounding rock. Typically, this involves obtaining lithological and water-bearing information by setting up boreholes underground, recording gas concentration changes by installing gas detectors in the roadways, marking fault distribution based on geological exploration maps, and periodically measuring surface displacement data using surface subsidence observation points. The above data are then compared and judged according to established risk classification standards to complete the risk assessment of hidden disaster-causing factors.

[0004] Traditional assessment methods rely on discrete operations such as borehole exploration, gas detection, map annotation, and regular observation of surface subsidence. The various information sources are scattered and have inconsistent time scales. Data from underground, open-pit, and tailings dam scenarios are difficult to corroborate each other, resulting in blurred spatial correspondences, difficulty in continuously depicting dynamic changes, and the potential for omissions in identifying complex disaster-causing factors and deviations in risk level determination. It is also difficult to grasp risk migration trends in a timely manner, and there is a lack of unified judgment criteria for the relationship between hidden structures and disasters. Consequently, the assessment results do not provide sufficient support for early warning and response. Summary of the Invention

[0005] To address the technical problems existing in the prior art, embodiments of the present invention provide a method for assessing hidden disaster risks in mines based on integrated air-space-ground sensing, comprising the following steps: To achieve the above objectives, the present invention adopts the following technical solution: a method for assessing hidden disaster risks in mines based on integrated air-space-ground sensing, comprising the following steps: S1: Collect satellite remote sensing data, synthetic aperture radar data, UAV aerial survey data, underground mine underground detection data, open-pit mine slope monitoring data, and tailings dam monitoring data. Divide the scene sampling units according to the underground mine roadway zone, open-pit mine slope zone, and tailings dam zone, and register the data according to a unified time window to obtain the scene observation sequence. S2: Call the scene observation sequence to collect deformation data, electromagnetic response data, surrounding rock stress data, seepage pressure data, microseismic data and video frame data. After wavelet transform noise decomposition, screening by three times the standard deviation criterion and coordinate reference space alignment, a fused feature body is obtained. S3: Based on the fused feature body, mine engineering survey data, mining layout data and tailings operation data are collected, three-dimensional reconstruction is performed and displacement rate, settlement, stress concentration factor, seepage pressure gradient, microseismic energy density and slope stability factor are calculated to obtain the disaster-causing parameter set; S4: Evaluate the disaster-causing parameter set based on the partition membership function, aggregate the membership values ​​of each partition, and divide the risk levels to obtain a risk level map; S5: Based on the risk level map, risk zoning is performed, and the direction of risk migration is determined by combining displacement changes, stress changes, seepage changes and microseismic changes within adjacent time windows. The magnitude of risk level changes is calculated, the risk warning level is assessed, and risk warning is implemented to obtain the mine disaster risk assessment results.

[0006] As a further aspect of the present invention, the specific steps of S1 are as follows: S101: Collect satellite remote sensing data, synthetic aperture radar data, UAV aerial survey data, downhole detection data, slope monitoring data and dam monitoring data; detect spatial resolution, timestamps and coordinate identifiers; remove invalid records based on missing data markers and anomaly codes; and generate observation datasets. S102: Based on the observation dataset, extract the boundaries of underground mine roadways, open-pit mine slopes, and tailings dams, determine the spatial adjacency relationships and monitoring coverage, establish the mapping relationship between roadway sampling units, slope sampling units, and dam sampling units, and obtain the scene sampling unit set; S103: Based on the scene sampling unit set associated with the observation dataset, verify the correspondence of timestamps for each type of data, perform time slicing according to a unified time window, and perform spatial registration according to coordinate identifiers to obtain the scene observation sequence.

[0007] As a further aspect of the present invention, the specific steps of S2 are as follows: S201: Collect deformation data, electromagnetic response data, surrounding rock stress data, seepage pressure data, microseismic data and video frame data corresponding to the scene observation sequence, detect the consistency of timestamps and the integrity of coordinate labels, reorganize multi-source data frames according to sampling interval and dimensional attributes, and generate feature data clusters; S202: Decompose each channel fluctuation component according to the feature data cluster, calculate the mean and standard deviation of each sampling point, remove deviation values ​​according to the three-times-standard-deviation criterion, retain the amplitude interval and time sequence correspondence, and obtain the purification feature group; S203: For the purification feature group, convert the longitude, latitude and elevation benchmarks, verify the correspondence between the dimensions of displacement, stress, seepage pressure and microseismic energy, establish spatial index association based on unified coordinate reference, and obtain the fused feature body.

[0008] As a further aspect of the present invention, the specific steps of S3 are as follows: S301: Collect the mine engineering survey data, mining layout data and tailings operation data corresponding to the fused feature body, check the consistency of coordinate benchmark, elevation benchmark and time mark, associate the goaf boundary, slope structure surface and dam seepage channel according to the spatial adjacency relationship, and establish a reconstruction constraint set; S302: Based on the reconstructed constraint set, splice the boundary points of the goaf, the slope structural surface points and the seepage channel points of the dam body, verify the correspondence between the depth value and the high-order sequence, and reorganize the three-dimensional unit according to the connection order and topological closure relationship to obtain the disaster space volume; S303: For the disaster-affected spatial body associated fusion feature body, calculate the displacement rate by the ratio of displacement change to time interval, calculate the settlement by the elevation change, determine the stress concentration factor based on the ratio of peak stress to average stress, determine the seepage pressure gradient based on the ratio of seepage pressure difference to path length, determine the microseismic energy density based on the ratio of microseismic released energy to volume, determine the slope stability coefficient based on the ratio of sliding force to anti-sliding force, and generate a disaster-causing parameter set.

[0009] As a further aspect of the present invention, the process of verifying the consistency of coordinate reference, elevation reference and time mark specifically involves comparing the coordinate origin, coordinate axis direction and elevation reference surface of the mine engineering survey data, mining layout data and tailings operation data according to the spatial location identifier corresponding to the fusion feature body, comparing the time mark according to the unified sampling time, and retaining data records that have the same coordinate origin, the same coordinate axis direction, the same elevation reference surface and the time mark falling in the same sampling period.

[0010] As a further aspect of the present invention, the specific steps of S4 are as follows: S401: Call the disaster-causing parameter set, collect parameter values ​​for underground mine zones, open-pit mine zones, and tailings dam zones, verify the dimensional correspondence between displacement rate, settlement, stress concentration factor, seepage pressure gradient, microseismic energy density, and slope stability coefficient, establish a parameter mapping table based on the zone number, and generate a zone parameter table; S402: Based on the partition parameter table, retrieve the value range of the partition membership function corresponding to each parameter, calculate the membership value of displacement rate, settlement, stress concentration factor, seepage gradient, microseismic energy density and slope stability coefficient, aggregate the membership values ​​of each partition, and obtain the risk assessment value; S403: For the risk assessment value, compare it with the preset grading threshold range, determine the risk level of each scene sampling unit, associate the risk level with the spatial location of the scene sampling unit, establish a risk distribution mapping, and obtain a risk level map.

[0011] As a further aspect of the present invention, the process of retrieving the value range of the membership function corresponding to each parameter is specifically as follows: for underground mine zones, open-pit mine zones, and tailings dam zones, the upper and lower boundary values ​​of the corresponding parameter columns in each zone parameter table are extracted respectively; the interval endpoints are divided according to the sorting position of each parameter column within the same zone; and the adjacent interval endpoints are connected to form the value range of the membership function.

[0012] As a further aspect of the present invention, the specific steps of S5 are as follows: S501: Based on the risk level map, collect the grid coordinates and risk level values ​​of the three-dimensional reconstruction area, verify the correspondence between the grid coordinates and the area boundary, mark the risk points of the underground mine area, open-pit mine area and tailings dam area according to the spatial adjacency relationship, and establish a risk zoning map; S502: Based on the risk zoning map, call the scene observation sequence, calculate the difference in displacement change, stress change, seepage change, and microseismic change between adjacent time windows, compare the distribution of each difference in adjacent grid directions, determine the risk migration direction, and obtain the migration criterion set; S503: For the disaster-causing parameter set associated with the migration criterion set, calculate the risk level change range, compare the risk level change range with the early warning classification threshold, determine the risk early warning level, establish the mapping between the early warning level and the partition location, and generate the mine disaster risk assessment result.

[0013] As a further aspect of the present invention, the process of verifying the correspondence between the grid coordinates and the region boundary specifically involves comparing the grid coordinates with the set of boundary points of the three-dimensional reconstructed region through point-by-point projection, and retaining the grid coordinates that fall within the closed range of the three-dimensional reconstructed region boundary in the risk zoning map. The process of determining the direction of risk migration specifically involves comparing the distribution of differences in adjacent grid directions and determining the direction of risk migration based on the synchronous increase or decrease relationship between the differences in displacement, stress, seepage, and microseismic changes on both sides of the adjacent grid boundary.

[0014] A mine hidden disaster risk assessment system based on integrated air-space-ground sensing, the system comprising: The multi-source sensing and collaboration module collects satellite remote sensing data, synthetic aperture radar data, UAV aerial survey data, underground mine underground detection data, open-pit mine slope monitoring data, and tailings dam monitoring data. It divides the scene sampling units according to the underground mine roadway zone, open-pit mine slope zone, and tailings dam zone, and registers the data according to a unified time window to obtain the scene observation sequence. The heterogeneous data processing module calls the scene observation sequence to collect deformation data, electromagnetic response data, surrounding rock stress data, seepage pressure data, microseismic data and video frame data. After wavelet transform noise decomposition, three-times-standard-deviation criterion screening and coordinate reference space alignment, a fused feature body is obtained. The disaster mechanism interpretation module, based on the fused feature body, collects mine engineering survey data, mining layout data and tailings operation data, performs three-dimensional reconstruction and calculates displacement rate, settlement, stress concentration factor, seepage pressure gradient, microseismic energy density and slope stability factor to obtain a disaster-causing parameter set. The partition identification and decision-making module evaluates the disaster-causing parameter set based on the partition membership function, aggregates the membership values ​​of each partition, and divides the risk levels to obtain a risk level map. The dynamic early warning and linkage module performs risk zoning mapping based on the risk level map, combines displacement changes, stress changes, seepage changes and microseismic changes within adjacent time windows to determine the risk migration direction, calculates the risk level change amplitude, assesses the risk early warning level, implements risk early warning, and obtains the mine disaster risk assessment results.

[0015] Compared with the prior art, the advantages and positive effects of the present invention are as follows: In this invention, information from satellite remote sensing, radar aerial surveys, underground exploration, slope monitoring, and dam monitoring is registered according to a unified time window. Combined with noise reduction, anomaly removal, and spatial alignment processing, comparable fusion features are formed, enabling a unified expression of deformation, stress, seepage, and microseismic changes under different scenarios. By combining survey data and operational conditions, three-dimensional reconstruction and disaster-causing parameter extraction are completed. Risk assessment and migration discrimination are carried out based on zoning affiliation, strengthening the continuity and pertinence of hidden disaster-causing factor identification, and improving the reliability and timeliness of risk classification, evolution tracking, and early warning and disposal. This provides a stable basis for cross-scenario risk linkage assessment and continuous updating. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is a schematic diagram of the steps of the present invention; Figure 2 This is a detailed schematic diagram of S1 of the present invention; Figure 3 This is a detailed schematic diagram of S2 of the present invention; Figure 4 This is a detailed schematic diagram of S3 of the present invention; Figure 5 This is a detailed schematic diagram of S4 of the present invention; Figure 6 This is a detailed schematic diagram of S5 of the present invention; Figure 7 This is a system module diagram of the present invention. Detailed Implementation

[0018] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0019] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0020] Please see Figure 1 This invention provides a method for assessing hidden disaster risks in mines based on integrated air-space-ground sensing, comprising the following steps: S1: Collect satellite remote sensing data, synthetic aperture radar data, UAV aerial survey data, underground mine underground detection data, open-pit mine slope monitoring data, and tailings dam monitoring data. Divide the scene sampling units according to the underground mine roadway zoning, open-pit mine slope zoning, and tailings dam zoning, and register the data in each scene sampling unit according to a unified time window to obtain the scene observation sequence. S2: Call the scene observation sequence to collect deformation data, electromagnetic response data, surrounding rock stress data, seepage pressure data, microseismic data and video frame data. Input the scene observation sequence into wavelet transform for noise decomposition. Filter outlier measurements according to the three-standard-deviation criterion. Spatially align the multi-source data according to the World Geodetic System coordinate benchmark to obtain the fused feature body. S3: Based on the fusion feature body, mine engineering survey data, mining layout data and tailings operation data are collected. The three-dimensional reconstruction of the boundary of the underground mine goaf, the slope structure of the open mine and the seepage channel of the tailings dam is carried out. The displacement rate, settlement, stress concentration factor, seepage pressure gradient, microseismic energy density and slope stability factor are calculated according to the fusion feature body to obtain the disaster-causing parameter set. S4: Call the disaster-causing parameter set to collect parameter values ​​of underground mine zone, open-pit mine zone and tailings dam zone. Based on the zone membership function, perform fuzzy comprehensive evaluation of displacement rate, settlement, stress concentration factor, seepage pressure gradient, microseismic energy density and slope stability factor. Calculate the risk assessment value of each scene sampling unit, and classify the risk level according to the risk assessment value to obtain the risk level map. S5: Based on the risk level map, risk zoning mapping is performed on the 3D reconstructed area. The direction of risk migration is determined according to the displacement, stress, seepage and microseismic changes in adjacent time windows of the scene observation sequence. The magnitude of risk level change is calculated in combination with the disaster-causing parameter set, the risk warning level is assessed, the risk warning is implemented, and the mine disaster risk assessment result is obtained.

[0021] Please see Figure 2 The specific steps of S1 are as follows: S101: Collect satellite remote sensing data, synthetic aperture radar data, UAV aerial survey data, downhole detection data, slope monitoring data and dam monitoring data; detect spatial resolution, timestamps and coordinate identifiers; remove invalid records based on missing data markers and anomaly codes; and generate observation datasets. The Gaofen-6 satellite remote sensing interface and the DJI Phantom drone aerial survey system were used to acquire raw remote sensing images and aerial survey pictures covering three major scenarios: underground mines, open-pit mines, and tailings ponds, through a multi-threaded data acquisition strategy. The images were divided into grids according to a satellite spatial resolution of 1 meter and a drone spatial resolution of 0.1 meters, and the timestamp information and WGS84 coordinate system latitude and longitude identifiers were extracted from the image files. Simultaneously, YCS400 transient electromagnetic detectors and surrounding rock stress monitoring equipment deployed in the underground mines were used to acquire electromagnetic response sequences at depths exceeding 100 meters and stress signals in the range of 0 to 50 MPa. The Huace T30 global navigation satellite system single-point monitoring equipment in the open-pit mines was used to acquire slope displacement change values ​​with a positioning accuracy of 3 millimeters, and microseismic monitoring equipment in the tailings ponds was used to acquire microseismic amplitude waveforms with frequencies ranging from 1 to 1000 Hz. The spatial resolution identifier, timestamp identifier, and three-dimensional coordinate identifier attached to the above multi-source data were extracted and serialized in chronological order to form the original multimodal feature sequence. The system scans data status bits in the multimodal feature sequence line by line using regular expression matching rules. It reads the pre-defined missing data markers from the device communication protocol; when a data status bit matches a missing data marker exactly, the data is considered a missing record. The system extracts the context anomaly code for this data; if the anomaly code value is greater than 0 and corresponds to a power outage, sensor malfunction, or channel congestion status code, the data line containing both the anomaly code and the missing data marker is directly deleted from the original sequence. An average interpolation strategy is used to fill in occasional packet loss due to network latency, obtaining all remaining valid data entries. These cleaned valid multimodal data are then classified and reorganized according to their sensor source labels. Satellite remote sensing data, synthetic aperture radar data, UAV aerial survey data, downhole detection data, slope monitoring data, and dam monitoring data are concatenated column-wise to generate a multi-dimensional observation dataset. The advantage of this operational logic is that, through missing data marker comparison and anomaly code filtering, the signal-to-noise ratio of the observation dataset is guaranteed from the ground up. In a real-world data collection instance, a total of 15,000 original observation entries were collected for a certain period of time. The scan revealed 125 entries with a missing measurement marker of 1 and an anomaly code of 3. These were removed according to the elimination logic, leaving 14,875 valid data entries to form the observation dataset.

[0022] S102: Based on the observation dataset, extract the boundaries of underground mine roadways, open-pit mine slopes, and tailings dams, determine the spatial adjacency relationships and monitoring coverage, establish the mapping relationship between roadway sampling units, slope sampling units, and dam sampling units, and obtain the scene sampling unit set; Based on the spatial coordinate distribution in the observation dataset, the accompanying latitude, longitude, and elevation 3D coordinate values ​​are extracted, and high-density point cloud regions are identified using a density clustering algorithm. In the clustering operation, a neighborhood radius of 5 meters and a minimum number of points are set to 10. The envelope of the clustered point cloud is smoothed, and 3D contour lines representing the boundaries of underground mine roadways, open-pit mine slopes, and tailings dams are extracted. Spatial adjacency and monitoring coverage are determined by calculating the shortest Euclidean distance between different 3D contour lines. The shortest Euclidean distance = the square of the difference in the node's x-coordinate + the square of the difference in the node's y-coordinate + the square of the difference in the node's elevation, and finally, the square root of the sum is taken. It is then determined whether the shortest Euclidean distance is less than a preset adjacency determination distance benchmark. The adjacency determination distance benchmark is set to 15 meters. If the shortest Euclidean distance is less than 15 meters, the two are determined to have a physical spatial adjacency relationship, and the area is marked as an overlapping monitoring coverage area. Along the extraction lines of the roadway boundary, slope boundary, and dam boundary, the sample is divided at fixed steps of 20 meters to generate multiple independent geometric patch units. The underground mine roadway is divided into multiple independent roadway sampling units, the open-pit slope into multiple independent slope sampling units, and the tailings dam into multiple independent dam sampling units. The centroid coordinates within each unit are extracted, and distance matching is performed between these centroid coordinates and the coordinates of the equipment deployment locations in the observation data. A unique mapping relationship is established between the sensing data channel closest to the centroid coordinates and its corresponding sampling unit. Through this mapping relationship, multi-source sensing data is accurately allocated to each specific geospatial area, forming a scene sampling unit set. In a practical calculation example, in calculating the shortest Euclidean distance between a roadway contour node and a nearby slope contour node, the difference in the horizontal coordinate is 8 meters, the difference in the vertical coordinate is 6 meters, and the difference in elevation is 2 meters. Substituting these values ​​into the calculation, the shortest Euclidean distance = (8² + 6² + 2²) / (square root), resulting in 10.19 meters. Since 10.19 meters is less than the baseline value of 15 meters, it is determined that the two units are adjacent. The area adjacent to the roadway and the slope was divided and numbered, and a total of 200 roadway sampling units, 150 slope sampling units, and 100 dam sampling units were established, forming a complete scene sampling unit set.

[0023] S103: Based on the scene sampling unit set associated with the observation dataset, verify the correspondence of timestamps for each type of data, perform time slicing according to a unified time window, and perform spatial registration according to coordinate identifiers to obtain the scene observation sequence; The process iterates through each observation dataset record associated with the scene sampling unit set, extracting the absolute timestamp string from the record header. Asynchronous timestamps collected by different sensing devices are converted to a unified Coordinated Universal Time (UTC) format, and the correspondence between timestamps of each data type is verified. A unified time window with a width of 30 minutes is set, mapping UTC-formatted timestamps to the corresponding time window sequence. If data from multiple sensors with different sampling frequencies fall within the same 30-minute time window, high-frequency data is downsampled using an arithmetic mean, and low-frequency data is upsampled using linear interpolation. Time slicing is performed based on the unified time window, ensuring that each time slice contains a set of merged multi-dimensional sensing features. Local coordinate identifiers are extracted from the data within each slice, and spatial registration is performed according to the national geodetic coordinate system transformation matrix. The independent coordinate system references of different sensors are uniformly transformed to a preset global coordinate reference frame using translation vectors and rotation matrices, eliminating spatial position offsets caused by different references. Through time-dimensional slice alignment and spatial-dimensional coordinate registration operations, a tensor structure containing multiple time steps and multiple spatial dimensions is constructed to obtain the scene observation sequence. The tensor structure's dimensions are set as the number of time windows multiplied by the number of scene sampling units multiplied by the number of sensing feature types. The advantage of this operational logic is that it overcomes the problem of spatiotemporal misalignment of heterogeneous data through spatiotemporal dimension registration. In a specific implementation example, the underground stress sensor samples once per minute, and UAV data is collected once per day. A 30-minute time window slice is used, averaging 30 values ​​from the underground stress sensor to obtain the representative value for that time window. The UAV data is then linearly interpolated into this 30-minute window based on values ​​from the preceding and following two days. After coordinate transformation, a scene observation sequence with dimensions of 48 time windows multiplied by 450 sampling units multiplied by 10 sensing features is generated.

[0024] Please see Figure 3 The specific steps of S2 are as follows: S201: Collect deformation data, electromagnetic response data, surrounding rock stress data, seepage pressure data, microseismic data and video frame data corresponding to the scene observation sequence, detect the consistency of timestamps and the integrity of coordinate labels, reorganize multi-source data frames according to sampling interval and dimensional attributes, and generate feature data clusters; The system utilizes perception channels from different business types within the scene observation sequence to collect data on roadway surrounding rock stress and electromagnetic response transmitted from sensors in underground mines, slope deformation and high-definition video frames from radar and single-point positioning devices in open-pit mines, and seepage pressure and microseismic data from online monitoring of tailings dams. A second round of in-depth testing is performed on the timestamp consistency and coordinate label integrity of these six categories of business data. The millisecond-level deviation between the timestamp encapsulated in the packet header and the timestamp verified in the packet tail is compared. If the deviation exceeds 500 milliseconds, the timestamp consistency is deemed compromised and marked as an unreliable time cluster. The system also checks whether the values ​​of the latitude, longitude, and elevation coordinate dimensions are empty or all zeros. If empty values ​​exist, the coordinate label integrity is deemed missing. After removing these unreliable and missing-labeled data, the theoretical sampling interval and corresponding physical dimensional attributes for each data category are extracted. The dimensions of the roadway surrounding rock stress data are megapascals (MPa), slope deformation data are millimeters (mm), seepage pressure data are kilopascals (kPa), microseismic data are joules (J), and video frame data are pixel grayscale values. Multi-source data frames are reconstructed based on sampling intervals and dimensional attributes, combining values ​​of different dimensions at the same sampling time into a multi-feature vector. During the reconstruction process, channels with significantly different dimensions are normalized by dividing the original dimensional values ​​by the historical maximum value of that type of data, mapping them to a dimensionless range of 0 to 1, ensuring that features of different dimensions within the combined multi-source data frames are on the same numerical order. The multi-source data frames, after dimensional unification and time alignment, are packaged sequentially to generate feature data clusters. In a practical example, the surrounding rock stress data collected at a certain time slice was 35 MPa, and the slope deformation data was 12 mm. The mine's historical maximum stress was 50 MPa, and the historical maximum deformation was 50 mm. The normalized stress is calculated as 35 divided by 50 = 0.7, and the normalized deformation is calculated as 12 divided by 50 = 0.24. Combining 0.7 and 0.24 into the same data frame creates a numerically balanced feature data cluster sequence.

[0025] S202: Decompose the fluctuation component of each channel according to the feature data cluster, calculate the mean and standard deviation of each sampling point, remove the deviation values ​​according to the three-times-standard-deviation criterion, retain the amplitude interval and time series correspondence, and obtain the purification feature group; The one-dimensional time-series signal sequence from the feature data cluster is imported, and the wavelet transform algorithm is used to decompose the fluctuation components of each channel. The Daubechies wavelet is selected as the mother wavelet basis function to divide the time series of the feature data cluster into low-frequency approximation components and high-frequency detail components. Hard-threshold denoising is performed on the decomposed high-frequency detail components, retaining the low-frequency approximation components as the fluctuation components reflecting the true trend. For the reconstructed time-series signal after denoising, the local mean and local standard deviation of each sampling point within a sliding window are calculated. The sliding window size is set to 60 consecutive sampling points. The local mean is equal to the sum of the values ​​of all sampling points within the sliding window divided by the total number of sampling points, and the local standard deviation is equal to the square root of the sum of the squares of the differences between the sampling point values ​​and the local mean divided by the total number of sampling points. The actual measured value of each current sampling point is compared with the local mean. When the absolute value of this difference is greater than the product of the local standard deviation multiplied by 3, the deviation value is removed according to the three-standard-deviation criterion. The advantage of this operational logic is that it automatically identifies singular isolated points caused by sudden environmental noise through statistical principles and separates them from the fluctuation components. After stripping away the outlier measurements, the amplitude interval boundaries of the retained stable values ​​are extracted, while maintaining their timestamp correspondence in the original sequence, resulting in a cleaned feature set. For example, in a window containing 60 stress sampling points, the calculated local mean is 20 MPa, and the local standard deviation is 1.5 MPa. According to the three-standard-deviation criterion, the upper limit for removal is calculated as 20 + 3 * 1.5 = 24.5 MPa, and the lower limit for removal is calculated as 20 - 3 * 1.5 = 15.5 MPa. If an isolated data point with a measurement of 26 MPa appears in the window, since 26 is greater than 24.5, this outlier measurement is directly removed. After this operation, the sequence containing abnormal peaks is transformed into a smooth and reliable cleaned feature set.

[0026] S203: For the purification feature group, convert longitude, latitude and elevation benchmarks, verify the correspondence between displacement, stress, seepage pressure and microseismic energy dimensions, establish spatial index association based on unified coordinate reference, and obtain fused feature body; The system retrieves latitude, longitude, degree, minute, and second coordinates and geodetic elevation data stored in the purification feature group, and performs benchmark conversion on the different coordinate system data generated by different sensors due to factory default settings. It reads the 3D translation parameters, 3D rotation parameters, and scale factor from the coordinate transformation parameter file, calculates the converted target system benchmark coordinates according to the seven-parameter coordinate transformation model, and uniformly converts all longitude, latitude, and elevation to the WGS84 national standard geodetic coordinate system benchmark. The system verifies the correspondence between the dimensions of displacement, stress, seepage pressure, and microseismic energy, ensuring that the data units for displacement channels are all millimeters, stress channels are all megapascals, and seepage pressure channels are all kilopascals. Based on the unified coordinate reference, a 3D spatial R-tree index structure is constructed, with each spatial 3D coordinate as an index node. The corresponding purification features such as displacement, stress, seepage, and microseismic activity are bound to the data domain of this spatial index node, establishing spatial index associations. Through this R-tree-based spatial index tree structure, all multimodal features within a specific spatial bounding box can be quickly retrieved, thereby obtaining a fused feature body with strong spatial correlation attributes. The advantage of this process is that it avoids the computational overhead of extensive full-table traversal during subsequent spatial analysis. In the example demonstration, the local coordinates of a specific purification feature point are extracted, with 3D translation parameters of 10.5 meters, -5.2 meters, and 3.1 meters. Applying translation operations, the original coordinates are adjusted by adding 10.5 meters to the horizontal axis, subtracting 5.2 meters from the vertical axis, and adding 3.1 meters to the elevation. Combining rotation and scale parameters, the standard WGS84 latitude and longitude coordinates are finally calculated. This coordinate point is then entered into an R-tree structure and bound to the corresponding 5 MPa stress feature and 2 mm displacement feature, forming a fused feature body that can be directly used for subsequent 3D modeling and risk calculation.

[0027] Please see Figure 4 The specific steps of S3 are as follows: S301: Collect and integrate the mine engineering survey data, mining layout data and tailings operation data corresponding to the feature body, check the consistency of coordinate benchmark, elevation benchmark and time mark, associate the goaf boundary, slope structure surface and dam seepage channel according to the spatial adjacency relationship, and establish the reconstruction constraint set; The process of verifying the consistency of coordinate benchmarks, elevation benchmarks and time markers is as follows: according to the spatial location identifiers corresponding to the fusion feature body, the coordinate origin, coordinate axis direction and elevation reference surface of the mine engineering exploration data, mining layout data and tailings operation data are compared; the time markers are compared according to the unified sampling time; and data records with consistent coordinate origins, consistent coordinate axis directions, consistent elevation reference surfaces and time markers falling in the same sampling period are retained. Collect and integrate mine exploration borehole data, mining progress layout data, and tailings dam daily operation log data that completely correspond to the area where the fused feature is located from the database server. Read the coordinate origin definition, Z-axis elevation reference surface, and log entry timestamp from the header of this background data, and check the consistency of coordinate datum, elevation datum, and timestamp. Based on the spatial location identifier corresponding to the fused feature, compare the coordinate origin, coordinate axis direction, and elevation reference surface corresponding to the mine exploration data, mining layout data, and tailings operation data. Calculate the Euclidean distance between the origin of the exploration data and the origin of the fused feature datum. If the distance is 0 and the coordinate axis rotation is completely consistent, the coordinate datum is considered consistent. Compare the absolute elevation values ​​of the elevation reference surface; if the difference is less than 0.01 meters, the elevation datum is considered consistent. Compare the timestamps according to a unified sampling time, requiring a timestamp difference of less than 1 hour. Retain data records with consistent coordinate origins, coordinate axis directions, elevation reference surfaces, and timestamps falling within the same sampling period. Based on the preserved high-quality engineering data, the topological mesh files were read, and a reconstruction constraint set for constraining spatial mesh evolution was established by associating the geometric boundaries of the goaf, the slope rock mass structure, and the internal seepage channels of the dam, according to the spatial coordinate overlap and patch intersection determination algorithm. In the actual verification example, the elevation reference surface of the engineering survey data was set as the Yellow Sea elevation datum, with a datum value of 0 meters. The elevation datum of the fused feature body was also the Yellow Sea elevation, with a difference of 0 meters, meeting the elevation consistency requirement. The time stamp of the mining layout data was 10:00 AM on April 13, 2026, and the time window of the fused feature body was from 10:00 AM to 10:30 AM on the same day, with the time stamp successfully falling within the same time period. Finally, this batch of engineering survey and working condition data that met all datum verification requirements was aggregated to generate a reconstruction constraint set containing geometric constraints.

[0028] S302: Based on the reconstructed constraint set, splice the boundary points of the goaf, the structural surface points of the slope, and the seepage channel points of the dam body, verify the correspondence between the depth value and the high-order sequence, and reorganize the three-dimensional unit according to the connection order and topological closure relationship to obtain the disaster space volume; The point cloud data file within the reconstructed constraint set is retrieved, and the Deloitte triangulation algorithm is used to mesh the discrete point set distributed in 3D space, stitching together the boundary points of the goaf, slope structural surface points, and dam seepage channel points. During the triangulation construction process, each generated 3D node is checked to verify the correspondence between depth values ​​and elevation sequences. If the elevation value of a goaf floor node is found to be abnormally greater than the elevation value of the roof node directly above it, it indicates a depth-elevation inversion error, and the inverted node coordinates are automatically reversed and corrected. Normal vectors are generated according to the connection sequence in a fixed direction, and it is checked whether the geometry formed by the triangular facets satisfies the Eulerian topological closure relation, i.e., whether the number of vertices minus the number of edges plus the number of faces equals 2. For unclosed regions with topological cracks, a hole-filling algorithm is used to generate new facets for stitching, reassembling them into seamless solid 3D units. All 3D units were merged and rendered according to the scenes of underground mining areas, open-pit slopes, and tailings dams to obtain a catastrophe spatial volume that accurately reflects the outline of underground tunnels, the orientation of slope joints and fissures, and the distribution of hydraulic channels in the dam. In the spatial volume reconstruction calculation, for a specific local goaf boundary point set, a total of 5000 discrete points were read. After Delaunay subdivision, 9996 triangular faces and 14994 edges were generated. Substituting into the topological verification formula, 5000 minus 14994 plus 9996 equals 2, perfectly satisfying the Eulerian closure relation. This indicates that the surface of the reconstructed goaf model is closed and there are no geometric topological loopholes. The resulting 3D geological model of the underground mine achieved a high accuracy of ±10 cm, laying the spatial geometric model foundation for the next step of extracting disaster-causing parameters.

[0029] S303: For the fusion of the characteristics of the disaster space body, calculate the displacement rate by the ratio of displacement change to time interval, calculate the settlement by the elevation change, determine the stress concentration factor based on the ratio of peak stress to average stress, determine the seepage pressure gradient based on the ratio of seepage pressure difference to path length, determine the microseismic energy density based on the ratio of microseismic released energy to volume, determine the slope stability coefficient based on the ratio of sliding force to anti-sliding force, and generate a disaster-causing parameter set; The process of determining the stress concentration factor based on the ratio of peak stress to average stress is as follows: extract the peak stress and the average stress in the monitoring interval corresponding to the disaster space body, and set the stress concentration factor according to the correspondence between the peak stress and the average stress. The process of determining the slope stability coefficient based on the ratio of sliding force to anti-sliding force is as follows: extract the sliding force and anti-sliding force in the monitoring interval of the slope corresponding to the disaster space body, and set the slope stability coefficient according to the correspondence between sliding force and anti-sliding force. Load the catastrophic space volume into the 3D visualization engine, and extract various monitoring values ​​for key indicator calculations based on the fusion feature volume attribute values ​​associated with the internal mapping of the catastrophic space volume. Extract the current displacement and the previous displacement of the slope monitoring points, and calculate the displacement change. Obtain the time interval between these two moments, and calculate the displacement rate = displacement change divided by the time interval. Extract the current absolute elevation and the initial calibration elevation of the dam monitoring points, and calculate the settlement = initial calibration elevation minus the current absolute elevation. Extract the maximum measured peak stress in the monitoring interval of the underground tunnel corresponding to the catastrophic space volume, and the arithmetic mean stress of all stress measuring points in the same interval. According to the correspondence between the peak stress and the mean stress, calculate the stress concentration factor = peak stress divided by the mean stress. Extract the seepage pressure values ​​at both ends of the seepage channel, and calculate the seepage pressure difference. Obtain the geometric curve length of the seepage channel in the catastrophic space volume as the path length, and calculate the seepage pressure gradient = seepage pressure difference divided by the path length. The total energy released from the seismic source and the geometric volume of the affected area around the seismic source are extracted from the microseismic event monitoring data. The microseismic energy density is calculated as the microseismic released energy divided by the volume. The sliding force along the sliding surface, generated by the self-weight of the soil and rock and pore water pressure, and the anti-sliding force generated by the cohesion and internal friction angle of the soil and rock within the slope monitoring section corresponding to the catastrophic spatial volume, are extracted. Based on the correspondence between sliding force and anti-sliding force, the slope stability coefficient is calculated as the anti-sliding force divided by the sliding force. These calculations are then summarized to form a set of disaster-causing parameters. The advantage of this operational logic is that it transforms simple physical monitoring values ​​into assessment indicators with clear mechanical catastrophic significance. In the calculation example, the slope displacement change was extracted as 5 mm, with a time interval of 1 day. The displacement rate was calculated as 5 mm / day (5 / 1). The peak stress in the underground tunnel was extracted as 45 MPa, and the average stress was 15 MPa. The stress concentration factor was calculated as 3 (45 / 15). The slope resistance was extracted as 200 kN, and the sliding force as 160 kN. The slope stability factor was calculated as 1.25 (200 / 160). These specific parameters, including 5 mm / day, stress concentration factor 3, and stability factor 1.25, were combined and output as a complete set of disaster-causing parameters.

[0030] Table 1. Calculation of characteristic parameters for slope monitoring points Open-air Zone 1 5 mm 1 day 5 millimeters per day 200 kN 160 kN 1.25 Open-air Zone 2 8 mm 1 day 8 millimeters per day 180 kN 180 kN 1.00 Open-air Zone 3 2 mm 1 day 2 millimeters per day 220 kN 150 kN 1.47 Table 1 lists the monitoring characteristics and disaster-causing parameter calculation results of different slope zones in open-pit mines, and shows the specific estimated data of displacement rate and slope stability coefficient.

[0031] Please see Figure 5 The specific steps of S4 are as follows: S401: Call the disaster-causing parameter set, collect parameter values ​​for underground mine zones, open-pit mine zones, and tailings dam zones, verify the dimensional correspondence between displacement rate, settlement, stress concentration factor, seepage pressure gradient, microseismic energy density, and slope stability coefficient, establish a parameter mapping table based on the zone number, and generate a zone parameter table. The calculated disaster-causing parameter set is retrieved from the memory stack. Based on the scene sampling unit set, each parameter is broken down and unpacked, and detailed parameter values ​​are collected for each of the underground mine, open-pit mine, and tailings dam zones. The data structure dictionary of these parameter values ​​is traversed, and the dimension declaration fields are validated a second time. This validates whether the displacement rate is in millimeters per time unit, the settlement is in millimeters, the stress concentration factor is a dimensionless pure number, the seepage pressure gradient is in kilopascals per meter, the microseismic energy density is in joules per cubic meter, and the slope stability coefficient is a dimensionless pure number. If any dimension declaration fields are missing or abnormal, a conversion script is initiated to correct and complete them. After ensuring complete dimension consistency, the unique identification code set for each geographic zone in the database is read. Using the zone number as the primary key, a data association relationship for the parameter mapping table is established. Each disaster-causing parameter, after dimension validation, is inserted into the corresponding column of the mapping table, generating a zone parameter table containing complete feature columns. In a practical example, the parameter dictionary of the first zone of the underground mine was extracted. Verification revealed that the displacement rate unit was recorded as 0.5 cm / day. A conversion script was triggered to multiply 0.5 cm / day by 10, converting it to 5 mm / day to ensure absolute consistency of the dimensional standard. Next, the identification code "Underground Main Mining Area 001" for the first zone of the underground mine was read. Using this code as the header index, the displacement rate of 5 mm / day, stress concentration factor of 3, and microseismic energy density of 150 joules per cubic meter were sequentially stored into the corresponding parameter cells, constructing a well-organized and rigorous zone parameter table.

[0032] S402: Based on the zoning parameter table, retrieve the value range of the zoning membership function corresponding to each parameter, calculate the membership values ​​of displacement rate, settlement, stress concentration factor, seepage gradient, microseismic energy density, and slope stability coefficient, aggregate the membership values ​​of each zoning, and obtain the risk assessment value; The process of retrieving the value range of the membership function corresponding to each parameter is as follows: for underground mines, open-pit mines, and tailings dams, extract the upper and lower boundary values ​​of the corresponding parameter columns in each partition parameter table, namely displacement rate, settlement, stress concentration factor, seepage pressure gradient, microseismic energy density, and slope stability factor; divide the interval endpoints according to the sorting position of each parameter column in the same partition; and connect the adjacent interval endpoints to form the value range of the membership function of the partition. The parameter columns stored in the zoning parameter table are retrieved, and the upper and lower boundary values ​​of displacement rate, settlement, stress concentration factor, seepage pressure gradient, microseismic energy density, and slope stability coefficient are extracted for the underground mine, open-pit mine, and tailings dam zones, respectively. The specific measured values ​​of each parameter column within the same zone are sorted from low to high, and the interval endpoints are defined based on the quartile sorting position of the data. For example, the minimum endpoint, lower quartile, median, upper quartile, and maximum endpoint are used as dividing markers, and adjacent interval endpoints are connected to form the interval of the zoning membership function. For the membership function, a parabolic fuzzy distribution function is used to calculate the membership value corresponding to the specific measured value within its respective interval, calculating the membership values ​​for displacement rate, settlement, stress concentration factor, seepage pressure gradient, microseismic energy density, and slope stability coefficient. Dynamic weighting coefficients are pre-assigned to each parameter. Each calculated membership value is multiplied by its corresponding parameter's weighting coefficient. All products are then summed to aggregate the membership values ​​for each partition, yielding a risk assessment value reflecting the overall hazard level. In the calculation example, the upper and lower boundary values ​​of the slope stability coefficient sequence are extracted, with a minimum endpoint of 0.9 and a maximum endpoint of 1.5. The current slope stability coefficient measurement for a certain partition is 1.25, falling within this range, and the calculated membership value is 0.6. Simultaneously, the displacement rate measurement is extracted as 5 mm / day, and its membership value is calculated as 0.4. The weight of the slope stability coefficient is set to 0.7, and the displacement rate weight is set to 0.3. The products are summed, and the aggregated risk assessment value is calculated as: 0.6 x 0.7 + 0.4 x 0.3 = 0.4² + 0.1² = 0.54. This result, 0.54, is the final aggregated risk assessment value for this partition, providing a quantitative benchmark for subsequent determination of specific risk levels.

[0033] S403: For the risk assessment value, compare it with the preset grading threshold range, determine the risk level of each scene sampling unit, associate the risk level with the spatial location of the scene sampling unit, establish a risk distribution mapping, and obtain a risk level map; The process of determining the risk level of each scene sampling unit is as follows: the risk assessment value is arranged in the risk assessment value sequence corresponding to each scene sampling unit, and the risk level is determined based on the risk assessment value falling into the risk assessment value sequence.

[0034] A risk classification configuration array, defined by safety management standards, is pre-loaded into memory. For the risk assessment values ​​obtained in the previous step, a preset classification threshold range is compared. The classification threshold range boundaries are set according to the overall order of the risk assessment values ​​in the risk assessment value sequence corresponding to each scene sampling unit, and in conjunction with the statistical probability of a normal distribution. The classification threshold range boundaries correspond one-to-one with the quartile interval breakpoints of the risk assessment value sequence. The first threshold boundary is set to 0.3, the second threshold boundary to 0.6, and the third threshold boundary to 0.8. The risk level is determined based on the classification threshold range into which the risk assessment value falls: if the risk assessment value is less than 0.3, it is classified as a general risk; if it is between 0.3 and 0.6, it is classified as a relatively large risk; if it is between 0.6 and 0.8, it is classified as a major risk; and if it is greater than 0.8, it is classified as an extremely major risk. After determining the risk level of each scene sampling unit, the grid code within the geographic information system is extracted based on the spatial coordinates of the scene sampling unit. The risk level label is matched and associated with this spatial location code to establish a risk distribution mapping matrix. Using a 3D visualization engine, different risk levels are rendered onto the surface of the scene's 3D model according to a preset color spectrum, obtaining a risk level map that presents a color gradient. In the judgment instance, the previously calculated risk assessment value of 0.54 is retrieved and compared with the grading threshold boundary range. Because 0.54 is greater than 0.3 and less than 0.6, it is determined to fall into the larger risk range. The "larger risk" label is assigned to this open slope zone, and a yellow warning color code is written into the established risk distribution mapping relationship. After judging and rendering each unit, a risk level map that intuitively reflects the distribution of hidden disaster hazards is generated in the panoramic 3D view, achieving the phased judgment goal of risk assessment.

[0035] Please see Figure 6 The specific steps of S5 are as follows: S501: Based on the risk level map, collect the grid coordinates and risk level values ​​of the three-dimensional reconstruction area, verify the correspondence between the grid coordinates and the area boundary, mark the risk points of the underground mine area, open-pit mine area and tailings dam area according to the spatial adjacency relationship, and establish a risk zoning map. The process of verifying the correspondence between grid coordinates and region boundaries is as follows: the grid coordinates are compared point by point with the boundary point set of the 3D reconstructed region, and the grid coordinates that fall within the closed range of the 3D reconstructed region boundary are retained in the risk zoning map. The rendered risk level map is invoked, and the coordinate array data of all vertices and meshes contained in the 3D reconstructed region are extracted. Simultaneously, the quantified risk level values ​​bound to each mesh in the map are captured. The mesh coordinate point set is cross-validated with the closed boundary line of the 3D reconstructed region, and a point-by-point projection comparison is performed using a ray projection algorithm. A horizontal ray is drawn from the centroid coordinates of each mesh, and the number of intersection points between this ray and the outer polygon boundary of the 3D reconstructed region is calculated. If the number of intersection points is odd, the mesh coordinates are determined to fall within the closed boundary of the 3D reconstructed region; if the number of intersection points is even or zero, they are determined to fall outside the region, and the outer coordinates are discarded. The mesh coordinates falling within the closed boundary of the 3D reconstructed region and their bound risk level values ​​are retained. Based on the spatial adjacency map, the validated mesh coordinates are traversed to find their respective macro-scene labels, accurately pinpointing the specific anomaly points within underground mining areas, the potential slope slip surface points in open-pit mining areas, and the seepage and piping risk points in tailings dam areas. These risk points, each with clearly defined coordinates and classification labels, are aggregated and projected onto a global two-dimensional engineering plane coordinate system to generate a risk zoning map that eliminates spatial projection redundancy and provides precise location. In the verification calculation example, the centroid coordinates of a certain grid are extracted, and a ray is drawn. Calculations show that the ray intersects the three-dimensional closed boundary line of the open-pit mine at only one point. Since 1 is an odd number, the algorithm confirms that the grid is effectively within the closed area of ​​the open-pit mine. The "significant risk" attribute and its absolute coordinates bound to this grid are mapped onto the global plane, ultimately generating an open-pit risk zoning map containing the alarm point.

[0036] Table 2. Statistics on Intersection Verification of Risk Grid Projection Points Unit A underground mines 1 Effective retention Major risks Unit B Open-pit mine 2 Remove coordinates General risks Unit C Tailings pond 3 Effective retention Significant risk As shown in Table 2, the calculation process of verifying the boundary enclosure relationship of several three-dimensional mesh elements by ray projection method is statistically presented, clarifying the direct impact of different numbers of intersection points on the determination of coordinate validity.

[0037] S502: Based on the risk zoning map, call the scene observation sequence, calculate the difference in displacement, stress, seepage, and microseismic changes between adjacent time windows, compare the distribution of each difference in adjacent grid directions, determine the risk migration direction, and obtain the migration criterion set; The process of determining the direction of risk migration is as follows: compare the distribution of the difference in adjacent grid directions, and determine the direction of risk migration according to the synchronous increase and decrease relationship of the difference in displacement, stress, seepage and microseismic changes on both sides of the boundary of adjacent grids in adjacent time windows; Import the generated risk zoning map and associated spatial coordinates, and retrieve previously stored scene observation sequences from the historical database for multiple consecutive periods. Extract the original parameter arrays of the two most recent adjacent time windows along the time axis, and calculate the difference between the historical values ​​and the current values ​​of the corresponding grid nodes. Subtract the displacement of the previous time window from the displacement of the current time window to calculate the displacement change difference between adjacent time windows; subtract the stress of the previous time window from the stress of the current time window to calculate the stress change difference; similarly, calculate the seepage change difference and the microseismic energy change difference. Traverse the grid topology matrix and compare the numerical distribution gradient of each difference in the four directions of east, south, west, and north of adjacent grids. Check the synchronous increase and decrease relationship of the differences on both sides of the boundary of adjacent grids. If the displacement change difference gradually increases in the eastward grid and the stress change difference also gradually decreases and shifts synchronously in the eastward grid, it is determined that stress release leads to an increase in displacement, thus determining that the risk evolution is moving eastward, and determining the risk migration direction. Package and encapsulate the grid coordinates, difference data, and determined evolution vector angles to obtain the migration criterion set. The advantage of this logical process lies in introducing spatial difference operations to capture the dynamic characteristics of the spatiotemporal evolution during the incubation period of disasters. In the simulation exercise, the current time window stress of the central grid of the roof of an underground goaf is extracted as 30 MPa, and the stress of the previous time window is 35 MPa. The stress change difference is calculated as 30 minus 35 = -5 MPa. At the same time, the current stress of its southward adjacent grid is extracted as 40 MPa, and the stress of the previous time window is 35 MPa. The stress change difference is calculated as 40 minus 35 = 5 MPa. By comparison, the stress of the central grid decreases while the stress of the southward grid increases, and the synchronous increase and decrease relationship shows a southward stress concentration trend. Therefore, the algorithm calculates that the risk is shifting southward and generates a migration criterion set for the southward expansion of the roof collapse risk in this area.

[0038] S503: For the disaster-causing parameter set associated with the migration criterion set, calculate the risk level change range, compare the risk level change range with the early warning classification threshold, determine the risk early warning level, establish the mapping between the early warning level and the zoning location, and generate the mine disaster risk assessment result. The process of determining the risk warning level is as follows: the warning level threshold is set in layers according to the cumulative distribution range of risk level changes in each partition, and the risk warning level is determined according to the correspondence between the risk level change range and the warning level threshold range. The system integrates dynamic evolution vectors from the migration criterion set and static assessment data from the disaster parameter set in the memory processor. It extracts the risk assessment value of the grid from the previous period and the risk assessment value from the current period, calculating the absolute difference as the risk level change magnitude. It reads the preset disaster emergency response grading standard and compares the risk level change magnitude with the warning grading threshold. The warning grading threshold is set hierarchically based on the percentile of the historical risk level change magnitude in each partition's cumulative distribution interval. The value corresponding to the 80th percentile distribution position is set as the general warning grading threshold, and the value corresponding to the 90th percentile distribution position is set as the severe warning grading threshold. It determines whether the currently calculated risk level change magnitude exceeds the severe warning grading threshold. If it does, the risk warning level is determined to be a red emergency warning; if it does not exceed the severe threshold but exceeds the general warning grading threshold, the risk warning level is determined to be an orange warning; otherwise, it is determined to be a blue safe state. After generating warning level labels, it queries and establishes a spatial mapping table between warning levels and partition location codes. All warning judgment conclusions, migration direction trends, and underlying risk grading data are summarized and exported into a final digital report format, generating mine disaster risk assessment results to guide on-site safety management. In a practical calculation example, the current period's risk assessment value is read as 0.54, and the previous period's assessment value is 0.44. The risk level change is calculated as 0.54 minus 0.44 = 0.10. The database is queried to determine the severe warning threshold as 0.15 and the general warning threshold as 0.08. By comparison, the calculated result 0.10 is greater than 0.08 and less than 0.15, therefore, the area is determined to trigger an orange warning response level. Subsequently, the fields containing "orange warning," the change range "0.10," and the warning location are integrated and packaged to output a complete full-chain mine disaster risk assessment result.

[0039] Please see Figure 7 A mine hidden disaster risk assessment system based on integrated air-space-ground perception includes: The multi-source sensing and collaboration module collects satellite remote sensing data, synthetic aperture radar data, UAV aerial survey data, underground mine underground detection data, open-pit mine slope monitoring data, and tailings dam monitoring data. It divides the scene sampling units according to the underground mine roadway zone, open-pit mine slope zone, and tailings dam zone, and registers the data according to a unified time window to obtain the scene observation sequence. The heterogeneous data integration module calls the scene observation sequence to collect deformation data, electromagnetic response data, surrounding rock stress data, seepage pressure data, microseismic data and video frame data. After wavelet transform noise decomposition, screening by three times the standard deviation criterion and alignment with coordinate reference space, a fused feature body is obtained. The disaster mechanism interpretation module, based on the fusion feature body, collects mine engineering survey data, mining layout data and tailings operation data, performs three-dimensional reconstruction and calculates displacement rate, settlement, stress concentration factor, seepage pressure gradient, microseismic energy density and slope stability factor to obtain a disaster-causing parameter set. The partition identification and decision-making module evaluates the disaster-causing parameter set based on the partition membership function, aggregates the membership values ​​of each partition, and classifies the risk levels to obtain a risk level map. The dynamic early warning and linkage module maps risk zones based on the risk level map, combines displacement changes, stress changes, seepage changes and microseismic changes within adjacent time windows to determine the direction of risk migration, calculates the magnitude of risk level changes, assesses the risk warning level, implements risk warning, and obtains the mine disaster risk assessment results.

[0040] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of protection of the described technical solutions.

Claims

1. A method for assessing hidden disaster risks in mines based on integrated air-space-ground sensing, characterized in that, Includes the following steps: S1: Collect satellite remote sensing data, synthetic aperture radar data, UAV aerial survey data, underground mine underground detection data, open-pit mine slope monitoring data, and tailings dam monitoring data. Divide the scene sampling units according to the underground mine roadway zone, open-pit mine slope zone, and tailings dam zone, and register the data according to a unified time window to obtain the scene observation sequence. S2: Call the scene observation sequence to collect deformation data, electromagnetic response data, surrounding rock stress data, seepage pressure data, microseismic data and video frame data. After wavelet transform noise decomposition, screening by three times the standard deviation criterion and coordinate reference space alignment, a fused feature body is obtained. S3: Based on the fused feature body, mine engineering survey data, mining layout data and tailings operation data are collected, three-dimensional reconstruction is performed and displacement rate, settlement, stress concentration factor, seepage pressure gradient, microseismic energy density and slope stability factor are calculated to obtain the disaster-causing parameter set; The specific steps for S3 are as follows: S301: Collect the mine engineering survey data, mining layout data and tailings operation data corresponding to the fused feature body, check the consistency of coordinate benchmark, elevation benchmark and time mark, associate the goaf boundary, slope structure surface and dam seepage channel according to the spatial adjacency relationship, and establish a reconstruction constraint set; S302: Based on the reconstructed constraint set, splice the boundary points of the goaf, the slope structural surface points and the seepage channel points of the dam body, verify the correspondence between the depth value and the high-order sequence, and reorganize the three-dimensional unit according to the connection order and topological closure relationship to obtain the disaster space volume; S303: For the disaster-affected spatial body associated fusion feature body, calculate the displacement rate by the ratio of displacement change to time interval, calculate the settlement by the elevation change, determine the stress concentration factor based on the ratio of peak stress to average stress, determine the seepage pressure gradient based on the ratio of seepage pressure difference to path length, determine the microseismic energy density based on the ratio of microseismic released energy to volume, determine the slope stability coefficient based on the ratio of sliding force to anti-sliding force, and generate a disaster-causing parameter set; S4: Evaluate the disaster-causing parameter set based on the partition membership function, aggregate the membership values ​​of each partition, and divide the risk levels to obtain a risk level map; S5: Based on the risk level map, risk zoning is performed, and the direction of risk migration is determined by combining displacement changes, stress changes, seepage changes and microseismic changes within adjacent time windows. The magnitude of risk level changes is calculated, the risk warning level is assessed, and risk warning is implemented to obtain the mine disaster risk assessment results.

2. The method for assessing hidden disaster risks in mines based on integrated air-space-ground sensing according to claim 1, characterized in that, The specific steps of S1 are as follows: S101: Collect satellite remote sensing data, synthetic aperture radar data, UAV aerial survey data, downhole detection data, slope monitoring data and dam monitoring data; detect spatial resolution, timestamps and coordinate identifiers; remove invalid records based on missing data markers and anomaly codes; and generate observation datasets. S102: Based on the observation dataset, extract the boundaries of underground mine roadways, open-pit mine slopes, and tailings dams, determine the spatial adjacency relationships and monitoring coverage, establish the mapping relationship between roadway sampling units, slope sampling units, and dam sampling units, and obtain the scene sampling unit set; S103: Based on the scene sampling unit set associated with the observation dataset, verify the correspondence of timestamps for each type of data, perform time slicing according to a unified time window, and perform spatial registration according to coordinate identifiers to obtain the scene observation sequence.

3. The method for assessing hidden disaster risks in mines based on integrated air-space-ground sensing as described in claim 1, characterized in that, The specific steps of S2 are as follows: S201: Collect deformation data, electromagnetic response data, surrounding rock stress data, seepage pressure data, microseismic data and video frame data corresponding to the scene observation sequence, detect the consistency of timestamps and the integrity of coordinate labels, reorganize multi-source data frames according to sampling interval and dimensional attributes, and generate feature data clusters; S202: Decompose each channel fluctuation component according to the feature data cluster, calculate the mean and standard deviation of each sampling point, remove deviation values ​​according to the three-times-standard-deviation criterion, retain the amplitude interval and time sequence correspondence, and obtain the purification feature group; S203: For the purification feature group, convert the longitude, latitude and elevation benchmarks, verify the correspondence between the dimensions of displacement, stress, seepage pressure and microseismic energy, establish spatial index association based on unified coordinate reference, and obtain the fused feature body.

4. The method for assessing hidden disaster risks in mines based on integrated air-space-ground sensing according to claim 1, characterized in that, The process of verifying the consistency of coordinate reference, elevation reference and time mark is as follows: according to the spatial location identifier corresponding to the fusion feature body, the coordinate origin, coordinate axis direction and elevation reference surface of the mine engineering exploration data, mining layout data and tailings operation data are compared; the time mark is compared according to the unified sampling time; and data records with consistent coordinate origin, consistent coordinate axis direction, consistent elevation reference surface and time mark falling in the same sampling period are retained.

5. The method for assessing hidden disaster risks in mines based on integrated air-space-ground sensing according to claim 1, characterized in that, The specific steps of S4 are as follows: S401: Call the disaster-causing parameter set, collect parameter values ​​for underground mine zones, open-pit mine zones, and tailings dam zones, verify the dimensional correspondence between displacement rate, settlement, stress concentration factor, seepage pressure gradient, microseismic energy density, and slope stability coefficient, establish a parameter mapping table based on the zone number, and generate a zone parameter table; S402: Based on the partition parameter table, retrieve the value range of the partition membership function corresponding to each parameter, calculate the membership value of displacement rate, settlement, stress concentration factor, seepage gradient, microseismic energy density and slope stability coefficient, aggregate the membership values ​​of each partition, and obtain the risk assessment value; S403: For the risk assessment value, compare it with the preset grading threshold range, determine the risk level of each scene sampling unit, associate the risk level with the spatial location of the scene sampling unit, establish a risk distribution mapping, and obtain a risk level map.

6. The method for assessing hidden disaster risks in mines based on integrated air-space-ground sensing according to claim 5, characterized in that, The process of retrieving the value range of the membership function corresponding to each parameter is as follows: for underground mines, open-pit mines, and tailings dams, extract the upper and lower boundary values ​​of the corresponding parameter columns in each partition parameter table, namely displacement rate, settlement, stress concentration factor, seepage pressure gradient, microseismic energy density, and slope stability factor. Divide the interval endpoints according to the sorting position of each parameter column in the same partition, and connect the adjacent interval endpoints to form the value range of the partition membership function.

7. The method for assessing hidden disaster risks in mines based on integrated air-space-ground sensing according to claim 1, characterized in that, The specific steps of S5 are as follows: S501: Based on the risk level map, collect the grid coordinates and risk level values ​​of the three-dimensional reconstruction area, verify the correspondence between the grid coordinates and the area boundary, mark the risk points of the underground mine area, open-pit mine area and tailings dam area according to the spatial adjacency relationship, and establish a risk zoning map; S502: Based on the risk zoning map, call the scene observation sequence, calculate the difference in displacement change, stress change, seepage change, and microseismic change between adjacent time windows, compare the distribution of each difference in adjacent grid directions, determine the risk migration direction, and obtain the migration criterion set; S503: For the disaster-causing parameter set associated with the migration criterion set, calculate the risk level change range, compare the risk level change range with the early warning classification threshold, determine the risk early warning level, establish the mapping between the early warning level and the partition location, and generate the mine disaster risk assessment result.

8. The method for assessing hidden disaster risks in mines based on integrated air-space-ground sensing according to claim 7, characterized in that, The process of verifying the correspondence between grid coordinates and region boundaries specifically involves comparing the grid coordinates with the set of boundary points of the 3D reconstructed region through point-by-point projection, and retaining the grid coordinates that fall within the closed range of the 3D reconstructed region boundary in the risk zoning map. The process of determining the direction of risk migration specifically involves comparing the distribution of differences in adjacent grid directions and determining the direction of risk migration based on the synchronous increase or decrease relationship between the differences in displacement, stress, seepage, and microseismic changes on both sides of the adjacent grid boundary.

9. A mine hidden disaster risk assessment system based on integrated air-space-ground sensing, characterized in that: The system is used to implement the mine hidden disaster risk assessment method based on integrated air-space-ground sensing as described in any one of claims 1-8, and the system includes: The multi-source sensing and collaboration module collects satellite remote sensing data, synthetic aperture radar data, UAV aerial survey data, underground mine underground detection data, open-pit mine slope monitoring data, and tailings dam monitoring data. It divides the scene sampling units according to the underground mine roadway zone, open-pit mine slope zone, and tailings dam zone, and registers the data according to a unified time window to obtain the scene observation sequence. The heterogeneous data processing module calls the scene observation sequence to collect deformation data, electromagnetic response data, surrounding rock stress data, seepage pressure data, microseismic data and video frame data. After wavelet transform noise decomposition, three-times-standard-deviation criterion screening and coordinate reference space alignment, a fused feature body is obtained. The disaster mechanism interpretation module, based on the fused feature body, collects mine engineering survey data, mining layout data and tailings operation data, performs three-dimensional reconstruction and calculates displacement rate, settlement, stress concentration factor, seepage pressure gradient, microseismic energy density and slope stability factor to obtain a disaster-causing parameter set. The partition identification and decision-making module evaluates the disaster-causing parameter set based on the partition membership function, aggregates the membership values ​​of each partition, and divides the risk levels to obtain a risk level map. The dynamic early warning and linkage module performs risk zoning mapping based on the risk level map, combines displacement changes, stress changes, seepage changes and microseismic changes within adjacent time windows to determine the risk migration direction, calculates the risk level change amplitude, assesses the risk early warning level, implements risk early warning, and obtains the mine disaster risk assessment results.

Images