Geological disaster intelligent monitoring and early warning method and system based on Beidou

Abstract

The invention relates to the technical field of geological disaster monitoring and early warning, and discloses a Beidou-based geological disaster intelligent monitoring and early warning method and system. Beidou high-precision monitoring equipment is deployed by selecting a geological disaster prone area, earth surface displacement, settlement and inclination deformation data are collected in real time, and a multi-modal database is constructed in combination with environmental parameters. And performing alignment and noise correction on the spatio-temporal data by adopting Kalman filtering and a weighted evidence theory, extracting short-term and long-term deformation characteristics by utilizing a DBSCAN spatial clustering algorithm, and realizing multi-scale abnormal change pattern recognition in combination with a GeoHash grid index. Dimensional differences are eliminated through Z-score standardization processing, a geological stability index and change rate model is established, a causal reasoning framework is further constructed based on a Bayesian network, and a risk prediction model is trained in combination with a space-time neural network. The system can dynamically adjust a monitoring period threshold value and automatically trigger graded early warning, and supports hidden danger rectification whole-process tracing and multi-level gridding management. According to the scheme, the limitation of traditional single-source monitoring is broken through, the full-chain prevention and control of geological disasters from deformation feature extraction, causal relationship modeling to dynamic risk prediction is realized, and the early warning timeliness and accuracy are remarkably improved.

Description

Technical Field

[0001] The present invention relates to the field of geological disaster monitoring, and in particular to a method and system for intelligent geological disaster monitoring and early warning based on the Beidou satellite navigation system, which is suitable for real-time monitoring and early warning of geological disasters such as landslides, debris flows, and ground subsidence. Background Art

[0002] Geological disasters (such as landslides, debris flows, and ground subsidence) are characterized by sudden and destructive nature. Traditional geological disaster monitoring methods suffer from low accuracy, poor real-time performance, and limited coverage, making them inadequate for modern geological disaster early warning. With the maturity of China's Beidou satellite navigation system, its high-precision positioning capabilities provide new technical support for geological disaster monitoring.

[0003] The BeiDou Navigation Satellite System, a global satellite navigation system independently developed by China, offers high-precision, all-weather, and all-region positioning capabilities, providing a reliable technical means for geological disaster monitoring and early warning. While some technologies currently utilize the BeiDou system for geological disaster monitoring, the following challenges remain: 1. Insufficient multi-source data fusion capabilities prevent the comprehensive reflection of the dynamic changes in geological hazards. 2. The system's intelligence level is low, requiring improved warning accuracy and timeliness. 3. Existing systems lack a unified monitoring network, hindering large-scale application and deployment. Therefore, a geological disaster monitoring system with high-precision positioning capabilities is urgently needed. Summary of the Invention

[0004] In view of this, it is necessary to provide a Beidou-based intelligent monitoring and early warning method and system for geological disasters. Through Beidou's high-precision positioning, multi-source sensor fusion, artificial intelligence analysis and real-time data transmission technology, all-weather, high-precision, intelligent monitoring and early warning of geological disasters can be achieved, the accuracy and timeliness of geological disaster early warnings can be improved, and strong technical support can be provided for disaster prevention and mitigation.

[0005] In a first aspect, an embodiment of the present application provides a BeiDou-based intelligent monitoring and early warning method for geological disasters, which is applied to a three-level architecture of cloud-edge-drone. The method includes:

[0006] S1: Select areas prone to geological disasters as monitoring points and deploy Beidou high-precision monitoring equipment to record the initial geological parameters and Beidou equipment information of the monitoring points;

[0007] S2: Use BeiDou monitoring equipment to collect real-time surface displacement, settlement, and tilt deformation data at monitoring points, and divide the data into multiple monitoring periods according to time series;

[0008] S3: Use multi-source data fusion algorithms to process BeiDou monitoring data and identify surface deformation characteristics and abnormal change patterns in each monitoring period;

[0009] S4: Statistical analysis is performed on the processing results of S3, and the monitoring data of the same period are normalized to obtain the expected value of the deformation characteristics of each monitoring point in different periods;

[0010] S5: Based on the deformation geological stability index, the geological stability change rate between cycles is calculated;

[0011] S6: Using the geological environment information and geological stability change rate of the monitoring points, a geological disaster causal reasoning network framework is constructed based on the Bayesian network for each period;

[0012] S7: Based on the causal reasoning network framework, the directionality of the causal relationship is determined through the conditional independence test algorithm to establish a complete geological disaster cause network;

[0013] S8: Construct and train a geological disaster early warning model based on a spatiotemporal neural network, wherein the model input is the key factors directly affecting geological stability selected from the genetic network, and the output is the geological disaster risk level of different periods;

[0014] S9: Based on the trained geological disaster early warning model, the geological disaster risks in different periods in the future are predicted in real time, and the early warning mechanism is automatically triggered when the risk exceeds the threshold.

[0015] Optionally, in an implementation of the first aspect of the present invention, the step S1 of selecting areas prone to geological disasters as monitoring points, deploying Beidou high-precision monitoring equipment, and recording initial geological parameters and Beidou equipment information at the monitoring points includes:

[0016] Through field exploration and geological surveys, areas prone to geological disasters are identified, and monitoring points with a wide field of view and a ground elevation angle of more than 15 degrees are selected;

[0017] Enter the project's potential risk point equipment list and obtain the equipment identification information by scanning the equipment's QR code;

[0018] Enter the hidden danger description, responsible personnel, and rectification requirements associated with the equipment and save them to the system database;

[0019] Select Manually add in the point list and enter the point name, type, and risk level information;

[0020] Get real-time geographic coordinates by activating the GPS module on your phone, and automatically mark the coordinates on the map interface using the positioning button;

[0021] Combined with the map selection function, coordinate accuracy is reconfirmed to complete geographic information binding; all equipment and point information are linked to a unified management platform to support progress tracking and status updates of the hidden danger rectification process;

[0022] The system automatically generates an electronic ledger with time and space tags, including equipment QR code scanning records, GPS coordinate data and operation timestamps;

[0023] The latitude and longitude coordinates of the monitoring points are encoded into string identifiers using the GeoHash algorithm, and these encoded string identifiers are sharded and stored according to the preset prefix length;

[0024] The GeoHash prefix length is dynamically adjusted based on the spatial density of monitoring points. Specifically, when the spatial density of monitoring points exceeds a preset high-density threshold, the GeoHash prefix length is increased to subdivide the grid. When the spatial density of monitoring points falls below a preset low-density threshold, the GeoHash prefix length is reduced to merge grids. A multi-level grid index is generated based on GeoHash prefixes of different lengths, and each monitoring point is associated with multiple levels of GeoHash codes to enable cross-level queries.

[0025] Optionally, in an implementation of the first aspect of the present invention, S2: collecting surface displacement, settlement, and tilt deformation characteristic data of monitoring points in real time using Beidou monitoring equipment, and dividing the data into multiple monitoring periods in time series, includes:

[0026] Preprocessing of the collected multimodal data: Integrating Beidou monitoring data, deformation data, and environmental parameters on the device side, using Kalman filtering and weighted evidence theory for spatiotemporal alignment and noise correction, and using LMS front-end solving technology to reduce atmospheric errors and filter multipath errors on the raw data, and outputting corrected deformation observation values;

[0027] Dynamic monitoring cycle division and data encapsulation: Continuously collected deformation data are divided into multi-level monitoring cycles according to time series. Each cycle data contains displacement increment, settlement accumulation, inclination change rate and environmental parameters. The cycle length is dynamically adjusted according to the disaster risk level: high-risk conditions trigger high-frequency monitoring in seconds; stable conditions switch to low-frequency mode.

[0028] Multi-level GeoHash grid index construction: The GeoHash algorithm is applied to the latitude and longitude coordinates of monitoring points, with an initial prefix length of 6 bits. The prefix length is dynamically adjusted based on spatial density: when the number of monitoring points in a single grid is greater than 50, the prefix length is increased to 7 bits; when the number of monitoring points in a single grid is ≤10, the prefix length is shortened to 5 bits. Multi-level codes are generated for each monitoring point, supporting cross-level data association queries.

[0029] Optionally, in an implementation of the first aspect of the present invention, S3: processing Beidou monitoring data using a multi-source data fusion algorithm to identify surface deformation characteristics and abnormal change patterns in each monitoring period includes:

[0030] Based on the fused data, DBSCAN spatial clustering is used to extract surface deformation features and detect abnormal change patterns;

[0031] Among them, the deformation characteristics include: short-term characteristics and long-term characteristics. The short-term characteristics include single-cycle displacement acceleration and sudden change in tilt direction; the long-term characteristics include cumulative settlement and trend offset vector; the abnormal change pattern is a local deformation intensive area identified by DBSCAN clustering.

[0032] Optionally, in an implementation of the first aspect of the present invention, the step S4 of performing statistical analysis on the processing results of S3, normalizing the monitoring data of the same period, and obtaining expected values ​​of deformation characteristics of each monitoring point in different periods includes:

[0033] S4.1: The deformation characteristics and anomaly detection results output by S3 multi-source data fusion are divided into time series according to the monitoring period;

[0034] S4.2: Select the cycle length based on the monitoring target, covering the complete evolution of the deformation characteristics, and use the sliding window method to handle non-stationary data;

[0035] S4.3: Check the integrity of the data, detect the sampling rate of each cycle, and use Kalman filtering or linear interpolation to fill in missing values;

[0036] S4.4: Use the Z-score normalization method to normalize the deformation variables of each monitoring point within the same period to eliminate dimensional differences. The calculation formula is:

[0037]

[0038] Among them, μ cycle is the single-period data mean, σ cycle is the standard deviation, IQR is the interquartile range, is the abnormal suppression coefficient, stable area: Abnormally high incidence areas: X i 、X norm They are the original deformation feature data and the normalized deformation feature data respectively;

[0039] S4.5: Perform Min-Max normalization on meteorological factors and map them to the interval [0,1] to eliminate environmental interference;

[0040] S4.6: Group the normalized data by monitoring point and period, and calculate the mean within each group:

[0041]

[0042] Among them, Ei,j represents the expected value of the deformation feature of the i-th monitoring point at the j-th period t, k i represents the BeiDou device confidence weight of the i-th monitoring point, N represents the number of data collections in the cycle, and the output is an expected value matrix, where the matrix rows represent monitoring points, the matrix columns represent cycles, and the matrix values ​​represent expected values;

[0043] S4.7: The mean μ and variance σ of each cycle must be calculated independently for normalization. For areas with sparse monitoring points, Kriging spatial interpolation is used to supplement the data and then normalize. The output expectation value matrix is ​​directly input into the geological stability index calculation in S5. The value range must be ensured to be between [-3, 3]. If it exceeds the range, the original data must be reviewed.

[0044] Optionally, in an implementation of the first aspect of the present invention, the step S5 of calculating the geological stability index of the monitoring point based on the expected value of the deformation characteristic and calculating the geological stability change rate during the period includes:

[0045] Based on the expected value of deformation characteristics after S4 normalization, a multi-factor weighted dynamic evaluation model is used to calculate the stability index (GSI), and its core formula is:

[0046]

[0047] Among them, F short,i,t represents the short-term characteristics of the t-th period of the i-th monitoring point, F long,i,t represents the long-term characteristics of the t-th period of the i-th monitoring point, w i represents the weight of the i-th monitoring point, α and β represent the short-term and long-term feature contribution coefficients respectively, max(·) represents the historical maximum deformation feature cumulative value, which is used for normalization, n represents the number of data monitoring points in the period, and T represents the number of periods;

[0048] The sliding window difference method is used to quantify the dynamic evolution of stability and calculate the stability change rate between cycles. The formula is:

[0049]

[0050] k represents the number of sliding window cycles, and Δt represents the duration of a single cycle.

[0051] Optionally, in an implementation of the first aspect of the present invention, S6: using the geological environment information and geological stability change rate of the monitoring points to construct a geological disaster causal reasoning network framework based on a Bayesian network for each period, includes:

[0052] Construct the Bayesian network model structure and define the node set V = {GSI t ,ΔGSI t,X,Y}, where X represents the geological environment information vector X=[x1,x2,...,x5] of the monitoring point, where x1,x2,...,x5∈R represents geological environment parameters, including lithology coefficient, slope, annual rainfall, groundwater level change rate, and vegetation coverage rate, and Y∈{0,1} represents the disaster type, where 0 is stable state and 1 is disaster occurrence;

[0053] The causal edge set E is established through expert knowledge and historical data analysis, including:

[0054] x i →ΔGSI t ,

[0055] GSI t →Y,

[0056] x i →Y;

[0057] The maximum likelihood estimation method is used to learn the Bayesian network parameter set θ={θ1,θ2,...θ k}, where θ k Conditional probability distribution between corresponding nodes:

[0058] For continuous variables GSI t and ΔGSI t , modeled using Gaussian distribution:

[0059] Among them, μ t =ξΔGSI t +ψx1+ζx2,

[0060] P(ΔGSI t |pa(ΔGSI t ))=N(μ_Δ,σ_Δ 2 ), μ_Δ=δx3+εx4, ξ, ψ, ζ, δ, ε are regression

[0061] coefficient,

[0062] For discrete variables Y, Bernoulli distribution is used:

[0063] w is the weight vector;

[0064] Given the current period observation data e={GSI t ,x1,x2,x3,x4,x5], calculate the posterior probability of disaster occurrence through variable elimination algorithm:

[0065]

[0066] When P(Y=1|e)≥τ, a disaster warning is triggered, where τ is the threshold, τ∈[0.5,0.9].

[0067] Optionally, in an implementation of the first aspect of the present invention, S8: constructing and training a geological hazard early warning model based on a spatiotemporal neural network, wherein the model input is key factors directly affecting geological stability selected from the genetic network, and the output is geological hazard risk levels in different periods, including:

[0068] Constructing a spatiotemporal neural network model: Building a spatiotemporal neural network consisting of a temporal feature extraction layer, a spatial feature extraction layer, and a spatiotemporal fusion layer. The temporal feature extraction layer uses a long short-term memory (LSTM) network to extract the sequential dependencies of key factors over time.

[0069] The spatial feature extraction layer uses a convolutional neural network (CNN) to extract the spatial correlation between monitoring points;

[0070] The spatiotemporal fusion layer adopts an attention mechanism to fuse temporal features and spatial features and output spatiotemporal joint features;

[0071] Training the early warning model: collecting historical monitoring data, including key factor inputs for short-term, medium-term, and long-term periods and corresponding geological hazard risk level labels of low risk, medium risk, and high risk; inputting the historical monitoring data into the spatiotemporal neural network, using the cross-entropy loss function to calculate the error between the predicted risk level and the true label, and adjusting the model parameters using the Adam optimizer until the model converges to obtain a trained geological hazard early warning model;

[0072] Output risk level: The early warning model takes key factors collected in real time as input and outputs geological disaster risk levels in different periods.

[0073] In a second aspect, an embodiment of the present application provides a Beidou-based intelligent monitoring and early warning system for geological disasters, which is applied to the Beidou-based intelligent monitoring and early warning method for geological disasters as described in the first aspect. The system includes:

[0074] The monitoring point deployment module is used to select areas prone to geological disasters as monitoring points, deploy Beidou high-precision monitoring equipment, and record the initial geological parameters and Beidou equipment information of the monitoring points;

[0075] The data acquisition module uses Beidou monitoring equipment to collect real-time surface displacement, settlement, and tilt deformation data of monitoring points, and divides the data into multiple monitoring periods according to time series;

[0076] Multi-source data fusion module, which uses a multi-source data fusion algorithm to process Beidou monitoring data and identify surface deformation characteristics and abnormal change patterns in each monitoring cycle;

[0077] The statistical analysis module performs statistical analysis on the processing results of S3, normalizes the monitoring data of the same period, and obtains the expected value of the deformation characteristics of each monitoring point in different periods;

[0078] Stability analysis module, based on deformation geological stability index, and calculates the geological stability change rate between cycles;

[0079] The causal reasoning network construction module uses the geological environment information and geological stability change rate of the monitoring points to construct a geological disaster causal reasoning network framework based on the Bayesian network for each cycle;

[0080] The causal relationship orientation module, based on the causal reasoning network framework, determines the directionality of the causal relationship through the conditional independence test algorithm and establishes a complete geological disaster cause network;

[0081] An early warning model training module is used to construct and train a geological disaster early warning model based on a spatiotemporal neural network. The model input is the key factors directly affecting geological stability selected from the genetic network, and the output is the geological disaster risk level of different periods;

[0082] The early warning trigger module, based on the trained geological disaster early warning model, predicts the geological disaster risks in different periods in the future in real time, and automatically triggers the early warning mechanism when the risk exceeds the threshold.

[0083] In a third aspect, an embodiment of the present application provides an electronic device, including:

[0084] processor;

[0085] a memory for storing processor-executable instructions;

[0086] Among them, the processor is configured to implement the Beidou-based intelligent monitoring and early warning method for geological disasters as described in the first aspect when executing the instructions.

[0087] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, which stores a program, and the program instructs a device to execute the Beidou-based intelligent monitoring and early warning method for geological disasters as described in the first aspect.

[0088] The present invention discloses a multi-source data fusion monitoring and intelligent early warning method based on the Beidou satellite system. By selecting areas prone to geological disasters and deploying Beidou high-precision monitoring equipment, surface displacement, settlement and tilt deformation data are collected in real time, and a multimodal database is constructed in combination with environmental parameters. Kalman filtering and weighted evidence theory are used to align and correct noise in spatiotemporal data, and the DBSCAN spatial clustering algorithm is used to extract short-term and long-term deformation features. The GeoHash grid index is combined to realize multi-scale abnormal change pattern recognition. The dimensional differences are eliminated through Z-score standardization, and a geological stability index and change rate model are established. A causal reasoning framework is further constructed based on the Bayesian network, and a risk prediction model is trained in combination with the spatiotemporal neural network. The system can dynamically adjust the monitoring cycle threshold and automatically trigger graded early warnings, supporting full-process traceability of hidden danger rectification and multi-level grid management. This solution breaks through the limitations of traditional single-source monitoring, realizes the full-chain prevention and control of geological disasters from deformation feature extraction, causal relationship modeling to dynamic risk prediction, and significantly improves the timeliness and accuracy of early warnings.

[0089] Beneficial effects:

[0090] (1) High-precision monitoring and real-time warning: By combining Beidou high-precision positioning equipment with multi-source data fusion algorithms, millimeter-level deformation monitoring can be achieved, significantly improving data accuracy; dynamic adjustment of the monitoring cycle ensures real-time response in high-risk areas.

[0091] (2) Intelligent cause analysis: A causal reasoning framework for geological disasters is constructed based on the Bayesian network, and the driving factors of disasters are identified through conditional independence tests to enhance the scientific nature and interpretability of early warnings.

[0092] (3) Adaptive spatial management: The GeoHash algorithm is used to dynamically shard and store monitoring point data, automatically adjust the grid accuracy according to the spatial density, optimize storage efficiency, and support cross-level fast queries.

[0093] (4) Multi-dimensional risk assessment: Spatiotemporal features are extracted through spatiotemporal neural networks, combined with the normalized deformation expectation value and stability change, to output multi-period risk levels and reduce the false alarm rate.

[0094] (5) System scalability and practicality: It supports the automatic generation of electronic records of hidden danger points and multi-terminal early warning linkage. It is suitable for various disaster scenarios such as landslides and subsidence, and can be integrated into existing geological monitoring platforms. BRIEF DESCRIPTION OF THE DRAWINGS

[0095] Figure 1 A flowchart of a BeiDou-based intelligent monitoring and early warning method for geological disasters provided in one embodiment of the present application.

[0096] Figure 2This is an architecture diagram of a BeiDou-based intelligent geological disaster monitoring and early warning system provided in one embodiment of the present application.

[0097] Figure 3 A schematic diagram of an electronic device provided in one embodiment of the present application. DETAILED DESCRIPTION

[0098] The technical solutions in the embodiments of the present application will be described clearly and completely below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments.

[0099] It should be noted that, in the embodiments of the present application, "at least one" refers to one or more, and "more" refers to two or more. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art in the art to which this application relates. The terms used in the specification of this application are only for the purpose of describing specific embodiments and are not intended to limit this application.

[0100] It should be noted that, in the embodiments of the present application, words such as "first" and "second" are only used for the purpose of distinguishing descriptions, and cannot be understood as indicating or implying relative importance, nor can they be understood as indicating or implying an order. Features defined as "first" and "second" may explicitly or implicitly include one or more of the features. In the description of the embodiments of the present application, words such as "exemplary" or "for example" are used to indicate examples, illustrations or explanations. Any embodiment or design scheme described as "exemplary" or "for example" in the embodiments of the present application should not be interpreted as being more preferred or more advantageous than other embodiments or design schemes. Specifically, the use of words such as "exemplary" or "for example" is intended to present related concepts in a concrete way.

[0101] Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0102] Example 1

[0103] Figure 1 This is a flowchart of a method for intelligent monitoring and early warning of geological disasters based on Beidou provided in one embodiment of this application. Figure 1 As shown, a BeiDou-based intelligent geological disaster monitoring and early warning method includes:

[0104] S1: Select areas prone to geological disasters as monitoring points, deploy Beidou high-precision monitoring equipment, and record the initial geological parameters and Beidou equipment information of the monitoring points.

[0105] Specifically, the S1: selects areas prone to geological disasters as monitoring points, and deploys Beidou high-precision monitoring equipment to record the initial geological parameters and Beidou equipment information of the monitoring points, including: determining areas prone to geological disasters through field surveys and geological surveys, and selecting points with a wide field of view and a ground elevation angle of more than 15 degrees as monitoring points; entering the project hidden danger point equipment list, and obtaining equipment identification information by scanning the equipment QR code; entering the hidden danger description, responsible personnel, and rectification requirements associated with the equipment, and saving them to the system database; selecting manual addition in the point list, and entering the point name, type, and risk level information; obtaining real-time geographic coordinates by activating the mobile phone GPS module, and automatically marking the coordinate position through the positioning button on the map interface; combining the map selection function to confirm the coordinate accuracy for a second time to complete the geographic information binding; all equipment and point information are associated with the unified management platform, supporting hidden dangers The system automatically generates a spatiotemporally tagged electronic ledger containing equipment QR code scan records, GPS coordinate data, and operation timestamps. The system uses the GeoHash algorithm to encode the latitude and longitude coordinates of monitoring points into string identifiers, which are then sharded and stored according to a preset prefix length. The GeoHash prefix length is dynamically adjusted based on the spatial density of the monitoring points. Specifically, when the spatial density of the monitoring points exceeds the preset high-density threshold, the GeoHash prefix length is increased to subdivide the grid. When the spatial density of the monitoring points is lower than the preset low-density threshold, the GeoHash prefix length is reduced to merge the grids. A multi-level grid index is generated based on GeoHash prefixes of different lengths, with each monitoring point simultaneously associated with multiple levels of GeoHash codes to enable cross-level query capabilities.

[0106] Collect geological data on the study area, including geological structure, stratigraphic lithology, etc. For example, in mountainous areas, areas with developed folds and faults may be more prone to geological disasters such as landslides and collapses. Organize historical geological disaster records to understand the location, scale, and frequency of previous disasters. For example, mudslides often occur in certain areas, and these areas require special attention. Analyze topographic data. For example, areas with steep slopes and large elevation differences are more likely to experience geological disasters. Conduct field surveys of the possible areas initially screened to determine their geological conditions, topographic characteristics, and surrounding environment. For example, observe whether there are cracks in the mountain or whether there is loose rock and soil on the slope. Based on data analysis and on-site survey results, representative areas are selected as monitoring points, taking into account the type of geological disaster, likelihood of occurrence, and degree of harm. Generally speaking, 3-5 monitoring points can be selected for each area prone to geological disasters.

[0107] Select appropriate Beidou equipment based on monitoring needs and the type of geological disaster. For example, for landslide monitoring, choose Beidou high-precision displacement monitoring equipment; for ground subsidence monitoring, choose Beidou equipment with elevation monitoring capabilities. Ensure that the equipment has high accuracy, good stability, and strong anti-interference capabilities to adapt to complex field environments.

[0108] Determine the parameters that need to be recorded for different types of geological hazards. For landslide monitoring, primarily record parameters such as displacement, displacement rate, and tilt angle; for ground subsidence monitoring, record parameters such as elevation change and subsidence rate. Environmental parameters such as rainfall, temperature, and humidity can also be recorded, as these factors may influence the occurrence and development of geological hazards. Recorded parameter information is promptly stored in the database of the data receiving center, establishing a comprehensive data management system. Regularly back up data to prevent data loss. Furthermore, data is collated and analyzed to promptly identify potential geological hazard risks.

[0109] GeoHash is an algorithm that encodes longitude and latitude into strings. It divides the Earth's surface into grids, each of which corresponds to a unique GeoHash code. Longer codes correspond to smaller grids and higher precision. For example, a short GeoHash prefix indicates a larger area, while a long prefix indicates a smaller area.

[0110] First, obtain the latitude and longitude information of the monitoring point and then use the GeoHash algorithm to encode it into a string. The encoded GeoHash string is then sharded and stored according to different prefix lengths. For example, prefix lengths of 2, 4, or 6 can be selected, with different prefix lengths corresponding to different area sizes. This allows for quick location of relevant areas based on different query requirements. The prefix length of the shards is dynamically adjusted based on the density of monitoring points. In areas with high point density, longer prefixes can be used to improve storage accuracy; in areas with low point density, shorter prefixes can be used to reduce storage overhead.

[0111] S2: The surface displacement, settlement, and tilt deformation data of the monitoring points are collected in real time through Beidou monitoring equipment, and the data are divided into multiple monitoring periods according to the time series.

[0112] Specifically, the S2: collects the surface displacement, settlement, and tilt deformation characteristic data of the monitoring points in real time through Beidou monitoring equipment, and divides the data into multiple monitoring cycles according to the time series, including: preprocessing the collected multimodal data: integrating Beidou monitoring data, deformation data and environmental parameters on the device side, using Kalman filtering and weighted evidence theory to perform spatiotemporal alignment and noise correction, using LMS front-end solution technology to weaken the atmospheric error and multipath error filtering of the original data, and outputting the corrected deformation observation value.

[0113] Continuously collected deformation data is divided into multiple monitoring cycles based on time series. Each cycle contains displacement increments, accumulated settlement, inclination change rates, and environmental parameters. The cycle duration is dynamically adjusted based on the disaster risk level: in high-risk situations, the system triggers high-frequency monitoring in seconds; in stable situations, it switches to low-frequency mode to improve monitoring efficiency and resource utilization.

[0114] Dynamic monitoring cycle division and data encapsulation: The continuously collected deformation data are divided into multi-level monitoring cycles according to the time series. Each cycle data contains displacement increment, settlement accumulation, inclination change rate and environmental parameters; the cycle duration is dynamically adjusted according to the disaster risk level: high-risk status triggers high-frequency monitoring in seconds; stable status switches to low-frequency mode.

[0115] Multi-level GeoHash grid index construction: The GeoHash algorithm is applied to the latitude and longitude coordinates of monitoring points, with an initial prefix length of 6 bits. The prefix length is dynamically adjusted based on spatial density: when the number of monitoring points in a single grid is greater than 50, the prefix length is increased to 7 bits; when the number of monitoring points in a single grid is ≤10, the prefix length is shortened to 5 bits. Multi-level codes are generated for each monitoring point, supporting cross-level data association queries, thereby achieving efficient spatial data management and query.

[0116] Step S2 achieves high-precision, timely monitoring of surface deformation by integrating Beidou monitoring equipment with multi-source data, combining spatiotemporal filtering with dynamic monitoring periodization. GeoHash grid indexing technology improves the efficiency and flexibility of data management. The combined application of these technologies provides strong support for geological disaster early warning and infrastructure safety monitoring.

[0117] S3: Use multi-source data fusion algorithm to process Beidou monitoring data and identify surface deformation characteristics and abnormal change patterns in each monitoring period.

[0118] Specifically, S3: using a multi-source data fusion algorithm to process Beidou monitoring data to identify surface deformation characteristics and abnormal change patterns in each monitoring period, including:

[0119] Based on the fused data, DBSCAN spatial clustering is used to extract surface deformation features and detect abnormal change patterns;

[0120] Among them, the deformation characteristics include: short-term characteristics and long-term characteristics. The short-term characteristics include single-cycle displacement acceleration and sudden change in tilt direction; the long-term characteristics include cumulative settlement and trend offset vector; the abnormal change pattern is a local deformation intensive area identified by DBSCAN clustering.

[0121] Specifically, single-cycle displacement acceleration uses multi-temporal InSAR technology (such as SBAS-InSAR) to monitor instantaneous dynamic changes in the surface. The difference in displacement rates between adjacent observation periods is calculated to reflect sudden deformation (such as landslide precursors). Sudden changes in tilt direction are based on radar line-of-sight (LOS) deformation data, combined with a terrain slope model, to identify deviations in local deformation direction from the principal stress direction of geological structures.

[0122] Cumulative subsidence can be calculated by fitting time series to multi-source InSAR data (e.g., Sentinel-1 and ALOSPALSAR) to calculate the total deformation over a specific time period. Trend-based offset vectors can be extracted using linear regression or Kalman filtering to quantify the overall movement trend of the region (e.g., fault creep).

[0123] The DBSCAN clustering algorithm uses InSAR-inverted deformation point clouds (including latitude and longitude, deformation rate, and acceleration) as input. Parameter settings include adjusting the neighborhood radius (eps) and minimum sample size (min_samples) based on regional geological conditions. The algorithm outputs concentrated areas of local deformation (such as landslide sliding surfaces and goaf collapse clusters) as high-risk early warning targets. This method effectively extracts key information from complex monitoring data, helping researchers and engineers better understand the dynamics of surface deformation and promptly identify potential geological hazards.

[0124] S4: Perform statistical analysis on the processing results of S3, perform normalization on the monitoring data of the same period, and obtain the expected value of the deformation characteristics of each monitoring point in different periods.

[0125] Extract all relevant monitoring data from the S3 processing system to ensure data integrity and accuracy. This data may contain measurements from different monitoring points at different times. Perform a preliminary data cleansing to remove obvious errors and outliers, such as those caused by instrument failure or data transmission issues.

[0126] Calculate common statistical indicators such as mean, median, standard deviation, minimum, and maximum. These indicators can help understand the central tendency and dispersion of the data. Draw visual charts such as histograms and boxplots to intuitively display the distribution of the data. By observing the charts, you can determine whether the data has characteristics such as skewness and outliers. Analyze the correlation between different monitoring points and different monitoring indicators. You can use methods such as the Pearson correlation coefficient to measure the linear correlation between variables. Identifying variables with strong correlations will help understand the inherent connections between monitoring data and provide a basis for subsequent data analysis and modeling.

[0127] Specifically, the step S4: performing statistical analysis on the processing results of step S3, normalizing the monitoring data of the same period, and obtaining the expected value of the deformation characteristics of each monitoring point in different periods, includes:

[0128] S4.1: Divide the deformation features and anomaly detection results from the multi-source data fusion output in S3 into time series by monitoring period. Based on the monitoring time span and actual needs, divide the monitoring data into different periods, such as daily, monthly, and annual periods. For each monitoring point, extract normalized monitoring data for each period.

[0129] S4.2: Select the cycle length based on the monitoring target, which should cover the complete evolution process of the deformation characteristics and process non-stationary data using the sliding window method.

[0130] S4.3: Check the data integrity, detect the sampling rate of each cycle, and use Kalman filtering or linear interpolation to fill in missing values.

[0131] S4.4: Use the Z-score normalization method to normalize the deformation variables of each monitoring point within the same period to eliminate dimensional differences. The calculation formula is:

[0132]

[0133] Among them, μ cycle is the single-period data mean, σ cycle is the standard deviation, IQR is the interquartile range, is the abnormal suppression coefficient, stable area: Abnormally high incidence areas: X i 、X norm They are the original deformation feature data and the normalized deformation feature data respectively.

[0134] S4.5: Perform Min-Max normalization on meteorological factors, mapping them to the [0, 1] range to eliminate environmental interference. Because the dimensions and value ranges of different monitoring indicators can vary significantly, normalization can unify the data into a specific range, eliminating the impact of dimension and making different indicators comparable. Common normalization methods include Min-Max normalization and Z-score normalization.

[0135] S4.6: Group the normalized data by monitoring point and period, and calculate the mean within each group:

[0136]

[0137] Among them, E i,j represents the expected value of the deformation feature of the i-th monitoring point at the j-th period t, k irepresents the Beidou device confidence weight for the i-th monitoring point, N represents the number of data collections within that cycle, and the output is an expected value matrix, where the rows represent monitoring points, the columns represent cycles, and the values ​​represent expected values. The expected value can be approximated by calculating the mean of the data within each cycle. For each monitoring point in each cycle, the mean of the normalized data is calculated. These means are the expected values ​​of the deformation characteristics of each monitoring point in different cycles, reflecting the average deformation of the monitoring point over different time periods.

[0138] S4.7: The mean μ and variance σ of each cycle must be calculated independently for normalization. For areas with sparse monitoring points, Kriging spatial interpolation is used to supplement the data and then normalize. The output expectation value matrix is ​​directly input into the geological stability index calculation in S5. The value range must be ensured to be between [-3, 3]. If it exceeds the range, the original data must be reviewed.

[0139] Analyze the expected deformation characteristics of each monitoring point over different periods and compare the differences between different monitoring points and between different periods. Based on these expected values, a deformation model can be established to predict future deformation trends, providing a scientific basis for relevant decision-making, such as safety assessment of engineering structures and early warning of geological disasters.

[0140] S5: Based on the deformation geological stability index, the geological stability change rate between cycles is calculated.

[0141] Specifically, the step S5: calculating the geological stability index of the monitoring point based on the expected value of the deformation characteristic, and calculating the geological stability change rate during the period, includes:

[0142] Based on the expected value of deformation characteristics after S4 normalization, a multi-factor weighted dynamic evaluation model is used to calculate the stability index (GSI), and its core formula is:

[0143]

[0144] Among them, F short,i,t represents the short-term characteristics of the t-th period of the i-th monitoring point, F long,i,t represents the long-term characteristics of the t-th period of the i-th monitoring point, w i represents the weight of the i-th monitoring point, α and β represent the short-term and long-term characteristic contribution coefficients, respectively, max(·) represents the historical maximum deformation characteristic cumulative value, which is used for normalization, n represents the number of data monitoring points in the cycle, and T represents the number of cycles. The sliding window difference method is used to quantify the dynamic evolution of stability and calculate the stability change rate between cycles. The formula is:

[0145]

[0146] k represents the number of sliding window cycles, and Δt represents the duration of a single cycle.

[0147] S6: Using the geological environment information and geological stability change rate of the monitoring points, a geological disaster causal reasoning network framework is constructed based on the Bayesian network for each cycle.

[0148] Specifically, the S6: using the geological environment information and geological stability change rate of the monitoring points, constructing a geological disaster causal reasoning network framework based on the Bayesian network for each cycle, including:

[0149] Construct the Bayesian network model structure and define the node set V = {GSI t ,ΔGSI t ,X,Y}, where X represents the geological environment information vector X=[x1,x2,...,x5] of the monitoring point, where x1,x2,...,x5∈R represents geological environment parameters, including lithology coefficient, slope, annual rainfall, groundwater level change rate, and vegetation coverage rate, and Y∈{0,1} represents the disaster type, where 0 is stable state and 1 is disaster occurrence;

[0150] Through expert knowledge and historical data analysis, a causal edge set E is established, including: x i →ΔGSI t , GSI t →Y,x i →Y; use the maximum likelihood estimation method to learn the Bayesian network parameter set θ={θ1,θ2,...θ k}, where θ k Conditional probability distribution between corresponding nodes: for continuous variables GSI t and ΔGSI t , modeled using Gaussian distribution:

[0151] Among them, μ t =ξΔGSI t +ψx1+ζx2,

[0152] P(ΔGSI t |pa(ΔGSI t ))=N(μ_Δ,σ_Δ 2 ), μ_Δ=δx3+εx4, ξ, ψ, ζ, δ, ε are regression

[0153] coefficient,

[0154] For discrete variables Y, Bernoulli distribution is used:

[0155] w is the weight vector;

[0156] Given the current period observation data e={GSI t,x1,x2,x3,x4,x5], calculate the posterior probability of disaster occurrence through variable elimination algorithm:

[0157]

[0158] When P(Y=1|e)≥τ, a disaster warning is triggered, where τ is the threshold, τ∈[0.5,0.9].

[0159] S7: Based on the causal reasoning network framework, the directionality of the causal relationship is determined through the conditional independence test algorithm to establish a complete geological disaster cause network.

[0160] Specifically, the causal reasoning network framework is constructed using the PCMCI algorithm, which specifically includes:

[0161] For static geological environment data (such as lithology coefficient, slope, and vegetation coverage), the PC algorithm is used to construct a causal network in two stages:

[0162] (1) Undirected graph skeleton generation: remove irrelevant edges through conditional independence tests (such as partial correlation coefficients); (2) Edge orientation: use V-structure (such as "rainfall → soil moisture → vegetation coverage") and propagation rules to determine the causal direction.

[0163] For time series geological environmental data (such as annual rainfall, groundwater level change rate, and slope displacement), the PCMCI algorithm is used to process high-dimensional autocorrelated data, and dynamic causal relationships are modeled through lagged time dependencies (such as the impact of current rainfall on landslide risk in the next three days).

[0164] The conditional independence test algorithm adopts different methods for the linear / nonlinear relationship of geological environment parameters:

[0165] For linear relationship parameters (such as slope and rock and soil stability, vegetation coverage and soil moisture), partial correlation coefficient combined with Fisher Z transformation was used for testing, and the formula is:

[0166]

[0167] Among them, r xy|s is the partial correlation coefficient between variables X and Y under the condition set S, and n is the sample size;

[0168] For nonlinear relationship parameters (such as earthquake intensity and disaster scale, engineering excavation depth and ground subsidence), the kernel-based conditional independence test (Kernel CI) is used to convert the data into a high-dimensional space through kernel function mapping, and calculate the conditional mutual information to judge independence.

[0169] The node set includes geological environment information vectors and disaster type variables, which are specifically defined as:

[0170] The geological environment information vector X = [x1, x2, ..., x5], where: x1 is the lithology coefficient (such as hard rock / soft rock / loose deposit, coded as 1 / 2 / 3); x2 is the slope (continuous variable, unit: degree); x3 is the annual rainfall (continuous variable, unit: mm); x4 is the groundwater level change rate (continuous variable, unit: m / year); x5 is the vegetation coverage rate (continuous variable, unit: %); the disaster type variable Y is a discrete binary variable, Y = 0 indicates that the monitoring point is in a stable state, and Y = 1 indicates that a disaster has occurred (such as landslide, mud-rock flow).

[0171] The causal relationship edge set is established by fusing expert knowledge with historical data. The specific steps are as follows:

[0172] Step 1: Expert knowledge initialization: Based on geological theory (such as "rainfall → soil saturation → shear strength reduction → landslide"), construct preliminary directed edges, for example:

[0173] x3 (annual rainfall) → x4 (groundwater level change rate) → Y (disaster type);

[0174] x2(slope)→Y(disaster type).

[0175] Step 2: Historical data correction: Based on historical disaster records (such as rainfall, water levels and disaster events at monitoring points in the past 10 years), edges are screened through conditional independence tests.

[0176] If x i If x1 (lithology coefficient) is independent of Y under the condition set S (e.g., x1 (lithology coefficient) is independent of Y under the condition x2 (slope)), remove the edge x1→Y.

[0177] If x j →x k → If Y is a V-structure (e.g., x3 → x4 → x5), the structure is retained and oriented.

[0178] The learning of the Bayesian network parameter set Θ adopts the maximum likelihood estimation method (MLE), and models the continuous variables and discrete variables separately: for continuous geological environment parameters (such as x3 annual rainfall and x4 groundwater level change rate), the Gaussian distribution is used to model the conditional probability distribution:

[0179]

[0180] Among them, P(x i ) is x i The parent node set, β0, β j is the regression coefficient, σ 2 is the error variance.

[0181] For discrete disaster type variables (Y), the Bernoulli distribution is used to model the conditional probability distribution:

[0182]

[0183] Where σ(·) is the Sigmoid function, w0, w k is the weight vector, pa(Y) is the parent node set of Y, such as x2, x3, and x4.

[0184] The calculation of the posterior probability of disaster occurrence P(Y=1|x) adopts the variable elimination algorithm, and the specific steps are as follows:

[0185] Step 1: Input the observation data of the current period: X = [x1, x2, ..., x5] (e.g., the lithology coefficient of a monitoring point = 2, the slope = 35 degrees, the annual rainfall = 1200 mm, etc.).

[0186] Step 2: Construct a factor graph: Convert the Bayesian network into a factor graph, where each node corresponds to a factor (conditional probability distribution).

[0187] Step 3: Eliminate irrelevant variables: Eliminate irrelevant variables according to the node order (such as x1 → x2 → x3 → x4 → x5) and calculate the joint probability distribution:

[0188]

[0189] Step 4: Marginalize to get the posterior probability:

[0190]

[0191] Step 5: Warning triggering: When P(Y=1|X)≥τ, the disaster warning is triggered, where τ is the threshold, τ∈[0.5,0.9], such as 0.8.

[0192] S8: Construct and train a geological disaster early warning model based on a spatiotemporal neural network, wherein the model input is the key factors directly affecting geological stability selected from the causal network, and the output is the geological disaster risk level in different periods.

[0193] Specifically, S8: constructing and training a geological disaster early warning model based on a spatiotemporal neural network, wherein the model input is the key factors directly affecting geological stability selected from the genetic network, and the output is the geological disaster risk level of different periods, including:

[0194] Construct a spatiotemporal neural network model: build a spatiotemporal neural network including a time feature extraction layer, a spatial feature extraction layer and a spatiotemporal fusion layer, wherein: the time feature extraction layer adopts a long short-term memory network LSTM to extract the serial dependency of key factors changing over time.

[0195] The spatial feature extraction layer uses a convolutional neural network (CNN) to extract the spatial correlation between monitoring points.

[0196] The spatiotemporal fusion layer adopts an attention mechanism to fuse temporal features with spatial features and output spatiotemporal joint features.

[0197] Training the early warning model: collecting historical monitoring data, which includes key factor inputs of short-term, medium-term and long-term periods and corresponding geological disaster risk level labels of low risk, medium risk and high risk; inputting the historical monitoring data into the spatiotemporal neural network, using the cross entropy loss function to calculate the error between the predicted risk level and the true label, and adjusting the model parameters through the Adam optimizer until the model converges to obtain a trained geological disaster early warning model.

[0198] Output risk level: The early warning model takes key factors collected in real time as input and outputs geological disaster risk levels in different periods.

[0199] Select the appropriate neural network type: Considering the spatiotemporal characteristics of geological hazard data, a spatiotemporal neural network can be selected, such as a long short-term memory network (LSTM), a gated recurrent unit (GRU), or a spatiotemporal convolutional neural network (STCNN). LSTM and GRU can effectively handle long-term dependencies in time series data, while STCNN can capture both spatial and temporal features. Construct the model structure: Design the model's input layer, hidden layer, and output layer. The number of neurons in the input layer is equal to the number of key factors in the causal network, and the number of neurons in the output layer is equal to the number of risk levels in different periods. In the hidden layer, a multi-layer neural network structure can be used to increase the complexity and expressiveness of the model.

[0200] Divide the preprocessed dataset into a training set, a validation set, and a test set according to a specific ratio, for example, 70% for training, 15% for validation, and 15% for testing. Select an appropriate loss function, such as the cross-entropy loss function, to measure the difference between the model's predictions and the true labels. Use optimization algorithms such as stochastic gradient descent (SGD) and adaptive moment estimation (Adam) to optimize the model parameters. Input the training set data into the model for training, and continuously adjust the model parameters to gradually reduce the loss function. During training, use the validation set to evaluate the model to prevent overfitting. Select appropriate evaluation metrics, such as precision, recall, and F1 score, to assess model performance. Based on the evaluation results, optimize the model. Model hyperparameters, such as the learning rate, number of hidden layer neurons, and number of training epochs, can be adjusted. Regularization methods, such as L1 and L2 regularization, can also be used to reduce overfitting.

[0201] S9: Based on the trained geological disaster early warning model, the geological disaster risks in different periods in the future are predicted in real time, and the early warning mechanism is automatically triggered when the risk exceeds the threshold.

[0202] Specifically, the key geological environmental factors extracted from the causal network (such as real-time rainfall, groundwater level change rate, slope displacement, vegetation coverage, lithology coefficient, etc., are all "factors directly affecting geological stability" determined in the S6-S7 causal network); it includes three layers: time feature extraction (LSTM), spatial feature extraction (CNN), and spatiotemporal fusion (attention mechanism), which can process spatiotemporal heterogeneous data (such as time series data of different monitoring points and spatial location associations); multi-period risk levels (short-term: next 1-3 days; medium-term: next 1-2 weeks; long-term: next 1-3 months), and each period corresponds to a discrete label (or probability value) of "low risk / medium risk / high risk".

[0203] Example 2

[0204] like Figure 2 As shown, the present application provides an architecture diagram of a Beidou-based intelligent monitoring and early warning system for geological disasters, which is applied to the Beidou-based intelligent monitoring and early warning system for geological disasters as described in Example 1, including a monitoring point deployment module 11, a data acquisition module 12, a multi-source data fusion module 13, a statistical analysis module 14, a stability analysis module 15, a causal reasoning network construction module 16, a causal relationship orientation module 17, an early warning model training module 18, and an early warning trigger module 19.

[0205] The monitoring point deployment module 11 is used to select areas prone to geological disasters as monitoring points, deploy Beidou high-precision monitoring equipment, and record the initial geological parameters and Beidou equipment information of the monitoring points;

[0206] The data acquisition module 12 collects the surface displacement, settlement, and tilt deformation data of the monitoring points in real time through Beidou monitoring equipment, and divides the data into multiple monitoring periods according to the time series;

[0207] The multi-source data fusion module 13 processes BeiDou monitoring data using a multi-source data fusion algorithm to identify surface deformation characteristics and abnormal change patterns in each monitoring period;

[0208] The statistical analysis module 14 performs statistical analysis on the processing results of S3, normalizes the monitoring data of the same period, and obtains the expected value of the deformation characteristics of each monitoring point in different periods;

[0209] The stability analysis module 15 is based on the deformation geological stability index and calculates the geological stability change rate between cycles;

[0210] The causal reasoning network construction module 16 uses the geological environment information and geological stability change rate of the monitoring points to construct a geological disaster causal reasoning network framework based on the Bayesian network for each cycle;

[0211] Causal relationship orientation module 17, based on the causal reasoning network framework, determines the directionality of causal relationships through the conditional independence test algorithm and establishes a complete geological disaster cause network;

[0212] An early warning model training module 18 is used to construct and train a geological disaster early warning model based on a spatiotemporal neural network. The model input is the key factors directly affecting geological stability selected from the genetic network, and the output is the geological disaster risk level of different periods.

[0213] The early warning trigger module 19 predicts the geological disaster risks in different periods in the future in real time based on the trained geological disaster early warning model, and automatically triggers the early warning mechanism when the risk exceeds the threshold.

[0214] Figure 3 This is an electronic device provided by an embodiment of the present application. Figure 3 As shown, the electronic device includes at least the following parts: a processor 101 and a memory 100 , a communication interface 103 , and a bus 102 .

[0215] In an embodiment of the present application, the memory 100 is used to store instructions executable by the processor 101, and the processor 101 is configured to implement the method of the first aspect when executing the instructions.

[0216] In an embodiment of the present application, a computer-readable storage medium includes instructions, and the instructions instruct a device to execute the method of the first aspect. For example, the instructions instruct the device to execute Figure 1 The method is shown in the process steps.

[0217] The program running in the electronic device involved in one embodiment of the present application can be a program that controls a central processing unit (CPU) and the like to realize the functions of the above-mentioned embodiment involved in one embodiment of the present invention (a program that enables a computer to function). Then, the information processed by these devices is temporarily stored in a random access memory (RAM) while being processed, and then stored in various ROMs such as read-only memory (Flash ROM) and hard disk drive (HDD), and is read, modified, and written by the CPU as needed.

[0218] It should be noted that a portion of the electronic device of the above embodiment may also be implemented by a computer. In this case, a program for implementing the control function may be recorded on a computer-readable recording medium, and the program recorded on the recording medium may be read into a computer and executed.

[0219] It should be noted that the "computer" mentioned here refers to a computer built into an electronic device, employing hardware including an operating system (OS) and peripheral devices. Furthermore, "computer-readable recording medium" refers to removable media such as floppy disks, magneto-optical disks, ROMs, and CD-ROMs, as well as storage devices such as hard disks built into computers.

[0220] Furthermore, "computer-readable recording media" may include: media that dynamically store programs for a short period of time, such as communication lines when transmitting programs via networks such as the Internet or communication lines such as telephone lines; and media that store programs for a fixed period of time, such as volatile memory within computers acting as servers or clients in this context. Furthermore, the aforementioned program may be a program for implementing a portion of the aforementioned functions, or a program that can achieve the aforementioned functions by combining with a program already stored in a computer.

[0221] Furthermore, the electronic device in the above-described embodiments can also be implemented as a collection (device group) consisting of multiple devices. Each device constituting the device group may have a portion or all of the functions or functional blocks of the electronic device in the above-described embodiments. A device group only needs to have all the functions or functional blocks of the electronic device.

[0222] Those skilled in the art should recognize that the above embodiments are merely intended to illustrate the present application and are not intended to limit the present application. As long as they are within the spirit of the present application, appropriate changes and modifications to the above embodiments are within the scope of protection claimed in the present application.

Claims

1. A Beidou-based intelligent monitoring and early warning method for geological disasters, characterized in that: The method comprises: S1: Select areas prone to geological disasters as monitoring points and deploy Beidou high-precision monitoring equipment to record the initial geological parameters and Beidou equipment information of the monitoring points; S2: Use BeiDou monitoring equipment to collect real-time surface displacement, settlement, and tilt deformation data at monitoring points, and divide the data into multiple monitoring periods according to time series; S3: Use multi-source data fusion algorithms to process BeiDou monitoring data and identify surface deformation characteristics and abnormal change patterns in each monitoring cycle; S4: Statistical analysis is performed on the processing results of S3, and the monitoring data of the same period are normalized to obtain the expected value of the deformation characteristics of each monitoring point in different periods; S5: Based on the deformation geological stability index, the geological stability change rate between cycles is calculated; S6: Using the geological environment information and geological stability change rate of the monitoring points, a geological disaster causal reasoning network framework is constructed based on the Bayesian network for each period; S7: Based on the causal reasoning network framework, the directionality of the causal relationship is determined through the conditional independence test algorithm to establish a complete geological disaster cause network; S8: Construct and train a geological disaster early warning model based on a spatiotemporal neural network, wherein the model input is the key factors directly affecting geological stability selected from the genetic network, and the output is the geological disaster risk level of different periods; S9: Based on the trained geological disaster early warning model, the geological disaster risks in different periods in the future are predicted in real time, and the early warning mechanism is automatically triggered when the risk exceeds the threshold.

2. The Beidou-based intelligent geological disaster monitoring and early warning method according to claim 1 is characterized in that: S1: Selecting areas prone to geological disasters as monitoring points, deploying Beidou high-precision monitoring equipment, and recording the initial geological parameters and Beidou equipment information of the monitoring points, including: Through field exploration and geological surveys, areas prone to geological disasters are identified, and monitoring points with a wide field of view and a ground elevation angle of more than 15 degrees are selected; Enter the project's potential risk point equipment list and obtain the equipment identification information by scanning the equipment's QR code; Enter the hidden danger description, responsible personnel, and rectification requirements associated with the equipment and save them to the system database; Select Manually add in the point list and enter the point name, type, and risk level information; Get real-time geographic coordinates by activating the GPS module on your phone, and automatically mark the coordinates on the map interface using the positioning button; Combined with the map selection function, coordinate accuracy is reconfirmed to complete geographic information binding; all equipment and point information are linked to a unified management platform to support progress tracking and status updates of the hidden danger rectification process; The system automatically generates an electronic ledger with time and space tags, including equipment QR code scanning records, GPS coordinate data and operation timestamps; The latitude and longitude coordinates of the monitoring points are encoded into string identifiers using the GeoHash algorithm, and these encoded string identifiers are sharded and stored according to the preset prefix length; The GeoHash prefix length is dynamically adjusted based on the spatial density of monitoring points. Specifically, when the spatial density of monitoring points exceeds a preset high-density threshold, the GeoHash prefix length is increased to subdivide the grid. When the spatial density of monitoring points falls below a preset low-density threshold, the GeoHash prefix length is reduced to merge grids. A multi-level grid index is generated based on GeoHash prefixes of different lengths, and each monitoring point is associated with multiple levels of GeoHash codes to enable cross-level queries.

3. The method for intelligent monitoring and early warning of geological disasters based on Beidou according to claim 2 is characterized in that: S2: Using BeiDou monitoring equipment to collect real-time surface displacement, settlement, and tilt deformation characteristic data of monitoring points, and dividing the data into multiple monitoring periods according to time series, including: Preprocessing of the collected multimodal data: Integrating Beidou monitoring data, deformation data, and environmental parameters on the device side, using Kalman filtering and weighted evidence theory for spatiotemporal alignment and noise correction, and using LMS front-end solving technology to reduce atmospheric errors and filter multipath errors on the raw data, and outputting corrected deformation observation values; Dynamic monitoring cycle division and data encapsulation: Continuously collected deformation data are divided into multi-level monitoring cycles according to time series. Each cycle data contains displacement increment, settlement accumulation, inclination change rate and environmental parameters. The cycle length is dynamically adjusted according to the disaster risk level: high-risk conditions trigger high-frequency monitoring in seconds; stable conditions switch to low-frequency mode. Multi-level GeoHash grid index construction: The GeoHash algorithm is applied to the latitude and longitude coordinates of monitoring points, with an initial prefix length of 6 bits. The prefix length is dynamically adjusted based on spatial density: when the number of monitoring points in a single grid is greater than 50, the prefix length is increased to 7 bits; when the number of monitoring points in a single grid is ≤10, the prefix length is shortened to 5 bits. Multi-level codes are generated for each monitoring point, supporting cross-level data association queries.

4. The method for intelligent monitoring and early warning of geological disasters based on Beidou according to claim 2 is characterized in that: S3: Using a multi-source data fusion algorithm to process Beidou monitoring data to identify surface deformation characteristics and abnormal change patterns in each monitoring period, including: Based on the fused data, DBSCAN spatial clustering is used to extract surface deformation features and detect abnormal change patterns; Among them, the deformation characteristics include: short-term characteristics and long-term characteristics. The short-term characteristics include single-cycle displacement acceleration and sudden change in tilt direction; the long-term characteristics include cumulative settlement and trend offset vector; the abnormal change pattern is a local deformation intensive area identified by DBSCAN clustering.

5. The method for intelligent monitoring and early warning of geological disasters based on Beidou according to claim 4 is characterized in that: S4: Statistically analyzing the processing results of S3, normalizing the monitoring data of the same period, and obtaining the expected value of the deformation characteristics of each monitoring point in different periods, including: S4.1: The deformation characteristics and anomaly detection results output by S3 multi-source data fusion are divided into time series according to the monitoring period; S4.2: Select the cycle length based on the monitoring target, covering the complete evolution of the deformation characteristics, and use the sliding window method to handle non-stationary data; S4.3: Check the integrity of the data, detect the sampling rate of each cycle, and use Kalman filtering or linear interpolation to fill in missing values; S4.4: Use the Z-score normalization method to normalize the deformation variables of each monitoring point within the same period to eliminate dimensional differences. The calculation formula is: Among them, μ cycle is the single-period data mean, σ cycle is the standard deviation, IQR is the interquartile range, is the abnormal suppression coefficient, stable area: Abnormally high incidence areas: X i 、X norm They are the original deformation feature data and the normalized deformation feature data respectively; S4.5: Perform Min-Max normalization on meteorological factors and map them to the interval [0,1] to eliminate environmental interference; S4.6: Group the normalized data by monitoring point and period, and calculate the mean within each group: Among them, E i,j represents the expected value of the deformation feature of the i-th monitoring point at the j-th period t, k i represents the BeiDou device confidence weight of the i-th monitoring point, N represents the number of data collections in the cycle, and the output is an expected value matrix, where the matrix rows represent monitoring points, the matrix columns represent cycles, and the matrix values ​​represent expected values; S4.7: The mean μ and variance σ of each cycle must be calculated independently for normalization. For areas with sparse monitoring points, Kriging spatial interpolation is used to supplement the data and then normalize. The output expectation value matrix is ​​directly input into the geological stability index calculation in S5. The value range must be ensured to be between [-3, 3]. If it exceeds the range, the original data must be reviewed.

6. The Beidou-based intelligent geological disaster monitoring and early warning method according to claim 5 is characterized in that: S5: calculating the geological stability index of the monitoring point based on the expected value of the deformation characteristic, and calculating the geological stability change rate during the cycle, including: Based on the expected value of deformation characteristics after S4 normalization, a multi-factor weighted dynamic evaluation model is used to calculate the stability index (GSI), and its core formula is: Among them, F short,i,t represents the short-term characteristics of the t-th period of the i-th monitoring point, F long,i,t represents the long-term characteristics of the t-th period of the i-th monitoring point, w i represents the weight of the i-th monitoring point, α and β represent the short-term and long-term feature contribution coefficients respectively, max(·) represents the historical maximum deformation feature cumulative value, which is used for normalization, n represents the number of data monitoring points in the period, and T represents the number of periods; The sliding window difference method is used to quantify the dynamic evolution of stability and calculate the stability change rate between cycles. The formula is: k represents the number of sliding window cycles, and Δt represents the duration of a single cycle.

7. The Beidou-based intelligent geological disaster monitoring and early warning method according to claim 6 is characterized in that: S6: Using the geological environment information and geological stability change rate of the monitoring points, a geological disaster causal reasoning network framework is constructed based on the Bayesian network for each cycle, including: Construct the Bayesian network model structure and define the node set V = {GSI t ,ΔGSI t ,X,Y}, where X represents the geological environment information vector X=[x1,x2,...,x5] of the monitoring point, where x1,x2,...,x5∈R represents geological environment parameters, including lithology coefficient, slope, annual rainfall, groundwater level change rate, and vegetation coverage rate, and Y∈{0,1} represents the disaster type, where 0 is stable state and 1 is disaster occurrence; The causal edge set E is established through expert knowledge and historical data analysis, including: x i →ΔGSI t , GSI t →Y, x i →Y; The maximum likelihood estimation method is used to learn the Bayesian network parameter set θ={θ1,θ2,...θ k }, where θ k Conditional probability distribution between corresponding nodes: For continuous variables GSI t and ΔGSI t , modeled using Gaussian distribution: Among them, m t =ξΔGSI t +ψx1+ζx2, P(ΔGSI t |pa(ΔGSI t ))=N(μ_Δ,σ_Δ) 2 ),μ_Δ=δx3+εx4, ξ, ψ, ζ, δ, ε are regression coefficient, For discrete variables Y, Bernoulli distribution is used: w is the weight vector; Given the current period observation data e={GSI t ,x1,x2,x3,x4,x5], calculate the posterior probability of disaster occurrence through variable elimination algorithm: When P(Y=1|e)≥τ, a disaster warning is triggered, where τ is the threshold, τ∈[0.5,0.9].

8. The Beidou-based intelligent geological disaster monitoring and early warning method according to claim 7 is characterized in that: in, S8: Constructing and training a geological disaster early warning model based on a spatiotemporal neural network, wherein the model input is the key factors directly affecting geological stability selected from the genetic network, and the output is the geological disaster risk level of different periods, including: Constructing a spatiotemporal neural network model: Building a spatiotemporal neural network consisting of a temporal feature extraction layer, a spatial feature extraction layer, and a spatiotemporal fusion layer. The temporal feature extraction layer uses a long short-term memory (LSTM) network to extract the sequential dependencies of key factors over time. The spatial feature extraction layer uses a convolutional neural network (CNN) to extract the spatial correlation between monitoring points; The spatiotemporal fusion layer adopts an attention mechanism to fuse temporal features and spatial features and output spatiotemporal joint features; Training the early warning model: collecting historical monitoring data, including key factor inputs for short-term, medium-term, and long-term periods and corresponding geological hazard risk level labels of low risk, medium risk, and high risk; inputting the historical monitoring data into the spatiotemporal neural network, using the cross-entropy loss function to calculate the error between the predicted risk level and the true label, and adjusting the model parameters using the Adam optimizer until the model converges to obtain a trained geological hazard early warning model; Output risk level: The early warning model takes key factors collected in real time as input and outputs geological disaster risk levels in different periods.

9. A BeiDou-based intelligent geological disaster monitoring and early warning system, applied to the BeiDou-based geological disaster intelligent monitoring and early warning method according to any one of claims 1 to 8, characterized in that: The system comprises: The monitoring point deployment module is used to select areas prone to geological disasters as monitoring points, deploy Beidou high-precision monitoring equipment, and record the initial geological parameters and Beidou equipment information of the monitoring points; The data acquisition module uses Beidou monitoring equipment to collect real-time surface displacement, settlement, and tilt deformation data of monitoring points, and divides the data into multiple monitoring periods according to time series; Multi-source data fusion module, which uses a multi-source data fusion algorithm to process Beidou monitoring data and identify surface deformation characteristics and abnormal change patterns in each monitoring cycle; The statistical analysis module performs statistical analysis on the processing results of S3, normalizes the monitoring data of the same period, and obtains the expected value of the deformation characteristics of each monitoring point in different periods; Stability analysis module, based on deformation geological stability index, and calculates the geological stability change rate between cycles; The causal reasoning network construction module uses the geological environment information and geological stability change rate of the monitoring points to construct a geological disaster causal reasoning network framework based on the Bayesian network for each cycle; The causal relationship orientation module, based on the causal reasoning network framework, determines the directionality of the causal relationship through the conditional independence test algorithm and establishes a complete geological disaster cause network; An early warning model training module is used to construct and train a geological disaster early warning model based on a spatiotemporal neural network. The model input is the key factors directly affecting geological stability selected from the genetic network, and the output is the geological disaster risk level of different periods; The early warning trigger module, based on the trained geological disaster early warning model, predicts the geological disaster risks in different periods in the future in real time, and automatically triggers the early warning mechanism when the risk exceeds the threshold.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a program, and the program instructs the device to execute the Beidou-based intelligent monitoring and early warning method for geological disasters as described in any one of claims 1 to 8.

Images