A geological disaster chain propagation path identification method based on space-time correlation

Abstract

The application relates to a geological disaster chain propagation path identification method based on space-time correlation. The method comprises the following steps: acquiring geological disaster multi-source monitoring data and extracting space-time features to obtain geological disaster space-time feature data; extracting disaster correlation relationship of the space-time feature data to obtain disaster chain correlation relationship data; combining the correlation relationship data and the space-time feature data to construct a geological disaster chain knowledge graph, wherein the knowledge graph comprises disaster entity nodes, relationship edges and space-time attribute annotations; calculating the time interval, the spatial distance and the attribute similarity according to the space-time attribute annotations of the node pairs, and introducing a Gaussian attenuation function to quantify the space-time correlation strength; and finally taking the correlation strength threshold as a constraint, traversing the nodes and edges along the relationship edge direction, and extracting the disaster chain propagation path. The method can improve the accuracy of geological disaster chain propagation path identification, and reduce misjudgment and omission.

Description

Technical Field

[0001] This invention belongs to the field of geological disaster detection, and in particular relates to a method for identifying the propagation path of geological disaster chains based on spatiotemporal correlation. Background Technology

[0002] With the development of geological disaster monitoring technology, multi-source sensor collaborative monitoring technology has emerged. This technology integrates data from various monitoring devices such as displacement sensors, rain gauges, and seismic detectors to achieve multi-dimensional perception of geological disasters such as landslides, collapses, and debris flows, providing a data foundation for disaster early warning. This leads to the current methods of geological disaster chain analysis.

[0003] In traditional techniques, geological disaster chain analysis mainly relies on manual interpretation of monitoring data, identifying the triggering relationships between disaster events through expert experience, and combining simple time-series statistical methods to determine the disaster propagation trend, or using a single monitoring indicator threshold determination method for disaster correlation analysis.

[0004] However, current methods for analyzing geological disaster chains have the following technical problems: Existing methods struggle to effectively handle the spatiotemporal coupling relationships of multi-source heterogeneous monitoring data, and cannot accurately quantify the degree of correlation between different disaster events in the temporal and spatial dimensions, resulting in a lack of objective spatiotemporal consistency constraints in the identification of disaster chain propagation paths; simultaneously, existing technologies lack systematic modeling methods for complex causal relationships between disaster entities, making it difficult to automatically extract and characterize the directed propagation structure between nodes in the disaster chain, thus limiting the efficiency and accuracy of disaster chain propagation path identification; furthermore, existing methods fail to fully consider the dynamic decay characteristics of spatiotemporal correlation strength during the search for disaster chain propagation paths, easily leading to misjudgments or omissions of propagation paths, and failing to meet the needs for refined identification and early warning of geological disaster chains. Summary of the Invention

[0005] Therefore, it is necessary to provide a method for identifying the propagation path of geological disaster chains based on spatiotemporal correlation to address the aforementioned technical problems.

[0006] Firstly, this application provides a method for identifying the propagation path of geological disaster chains based on spatiotemporal correlation, including:

[0007] S1. Obtain multi-source monitoring data of geological disasters, extract spatiotemporal features from the multi-source monitoring data of geological disasters, and obtain spatiotemporal feature data of geological disasters;

[0008] S2. Extract disaster correlation data from the spatiotemporal characteristic data of geological disasters to obtain disaster chain correlation data;

[0009] S3. Construct a knowledge graph from the disaster chain relationship data and the spatiotemporal feature data of geological disasters to obtain geological disaster chain knowledge graph data; the geological disaster chain knowledge graph data includes disaster entity nodes, relationship edges and spatiotemporal attribute annotations;

[0010] S4. Based on the spatiotemporal attribute annotations of disaster entity nodes in the geological disaster chain knowledge graph data, calculate the spatiotemporal correlation strength between each disaster entity node pair to obtain the spatiotemporal correlation strength data between each disaster entity node.

[0011] S5. Starting from the disaster entity nodes in the geological disaster chain knowledge graph data, based on the spatiotemporal correlation strength data, traverse the correlation edges that satisfy the correlation strength threshold and the disaster entity nodes connected to them along the direction of the correlation edges, extract the geological disaster chain propagation path composed of disaster entity nodes and correlation edges that satisfy the conditions, and obtain the geological disaster chain propagation path identification result.

[0012] In one embodiment, the multi-source monitoring data for geological hazards includes multiple monitoring data points, and S1 includes:

[0013] S11. Time stamp alignment and spatial coordinate unification are performed on multi-source geological disaster monitoring data to obtain standardized geological disaster monitoring data;

[0014] S12. Extract time-series features from standardized geological disaster monitoring data to obtain geological disaster time-series feature data;

[0015] S13. Extract spatial distribution features from standardized geological disaster monitoring data to obtain spatial distribution feature data of geological disasters;

[0016] S14. Perform spatiotemporal feature fusion on the temporal feature data and spatial distribution feature data of geological disasters to obtain spatiotemporal feature data of geological disasters.

[0017] In one embodiment, S2 includes:

[0018] S21. Based on the spatiotemporal characteristics data of geological disasters, calculate the spatiotemporal distance between any two monitoring data points, and generate spatiotemporal distance matrix data based on the spatiotemporal distance;

[0019] S22. Based on the spatiotemporal distance matrix data, cluster analysis is performed on the spatiotemporal characteristic data of geological disasters to obtain disaster event clustering results data; the disaster event clustering results data includes multiple clusters;

[0020] S23. Based on the disaster event clustering results data, determine the causal direction of the spatiotemporal characteristic data of geological disasters in each cluster to obtain disaster causal direction data;

[0021] S24. Based on the disaster event clustering results data and disaster causal direction data, construct the disaster correlation topology structure to obtain disaster chain correlation data.

[0022] In one embodiment, S3 includes:

[0023] S31. Based on the disaster event clustering results data, perform feature aggregation on the spatiotemporal characteristic data of geological disasters within each cluster to obtain the attribute data of disaster entity nodes;

[0024] S32. Based on disaster entity node attribute data and disaster causal direction data, the directed edges in the disaster chain relationship data are labeled with relationship types to obtain disaster relationship data; wherein, the disaster relationship data is labeled with relationship types.

[0025] S33. Based on the spatiotemporal characteristic data of geological disasters, the spatiotemporal attributes of each disaster entity node are calculated to obtain the spatiotemporal attribute data of the disaster entity node; among which, the spatiotemporal attribute data of the disaster entity node includes time center attribute and spatial center attribute;

[0026] S34. Based on disaster entity node attribute data, disaster correlation data, and disaster entity node spatiotemporal attribute data, a geological disaster chain knowledge graph data is constructed.

[0027] In one embodiment, S4 includes:

[0028] S41. Based on the time center attribute, calculate the time interval between any two disaster entity nodes to obtain disaster entity node time interval data;

[0029] S42. Based on the spatial center attribute, calculate the spatial distance between any two disaster entity nodes to obtain the spatial distance data of disaster entity nodes;

[0030] S43. Based on the attribute data of disaster entity nodes, calculate the attribute similarity between any two disaster entity nodes to obtain disaster entity node attribute similarity data;

[0031] S44. Based on the time interval data, spatial distance data, and attribute similarity data of disaster entity nodes, the spatiotemporal association strength of the relationship edges in the disaster chain association data is calculated to obtain the spatiotemporal association strength data between each disaster entity node; wherein, the expression for the spatiotemporal association strength is:

[0032] ;

[0033] in, Indicates the first The first disaster entity node and the first The spatiotemporal correlation strength between individual disaster entity nodes This represents the preset time decay scale parameter. This indicates the preset spatial attenuation scale parameter. Indicates the first The first disaster entity node and the first Attribute similarity between disaster entity nodes Indicates the first The first disaster entity node and the first The time interval between disaster entity nodes Indicates the first The first disaster entity node and the first Spatial distance between disaster entity nodes.

[0034] In one embodiment, S5 includes:

[0035] S51. Based on the spatiotemporal correlation strength data and correlation strength threshold, threshold filtering is performed on each relation edge in the disaster chain correlation data to obtain candidate propagation edge data;

[0036] S52. Based on the candidate propagation edge data, perform in-degree statistics on the disaster entity nodes in the geological disaster chain knowledge graph data, and select disaster entity nodes with zero in-degree as the starting point of the propagation path to obtain the propagation path starting point set data.

[0037] S53. Based on the propagation path starting point set data and candidate propagation edge data, starting from each propagation path starting point in the propagation path starting point set data, traverse layer by layer along the direction of the directed edge in the candidate propagation edge data, record the sequence of disaster entity nodes and relation edges passed during the traversal, and terminate the traversal when the out-degree of the current disaster entity node is zero or the spatiotemporal correlation strength of all downstream nodes corresponding to the out-degree is less than the correlation strength threshold, and obtain the candidate propagation path data.

[0038] S54. Based on the candidate propagation path data, calculate the cumulative path strength of each candidate propagation path to obtain the candidate propagation path strength data; wherein, the expression for the cumulative path strength is:

[0039] ;

[0040] In the formula, Indicates the first The cumulative strength of each candidate propagation path. Indicates the first The set of all relation edges contained in the candidate propagation path, Indicates the first From the candidate propagation path, from the first The disaster entity node points to the first The relationship edges between disaster entity nodes Indicates the first The first disaster entity node and the first The spatiotemporal correlation strength between individual disaster entity nodes;

[0041] S55. Based on the candidate propagation path intensity data, the candidate propagation paths are ranked to obtain the geological disaster chain propagation path identification results.

[0042] Secondly, this application also provides a geological disaster chain propagation path identification system based on spatiotemporal correlation, including:

[0043] The multi-source monitoring data acquisition and spatiotemporal feature extraction module is used to acquire multi-source monitoring data of geological disasters, extract spatiotemporal features from the multi-source monitoring data of geological disasters, and obtain spatiotemporal feature data of geological disasters.

[0044] The disaster correlation extraction module is used to extract disaster correlations from the spatiotemporal feature data of geological disasters to obtain disaster chain correlation data.

[0045] The geological hazard chain knowledge graph construction module is used to construct a knowledge graph from hazard chain relationship data and geological hazard spatiotemporal feature data to obtain geological hazard chain knowledge graph data; the geological hazard chain knowledge graph data includes hazard entity nodes, relationship edges, and spatiotemporal attribute annotations;

[0046] The spatiotemporal correlation strength calculation module is used to calculate the spatiotemporal correlation strength between each disaster entity node pair based on the spatiotemporal attribute annotations of disaster entity nodes in the geological disaster chain knowledge graph data, and obtain the spatiotemporal correlation strength data between each disaster entity node.

[0047] The geological disaster chain propagation path identification module is used to start from the disaster entity nodes in the geological disaster chain knowledge graph data, and based on the spatiotemporal correlation strength data, traverse the relation edges that meet the correlation strength threshold and the disaster entity nodes connected to them along the direction of the relation edges, extract the geological disaster chain propagation path composed of disaster entity nodes and relation edges that meet the conditions, and obtain the geological disaster chain propagation path identification result.

[0048] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the method described in the first aspect.

[0049] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method described in the first aspect.

[0050] The aforementioned method for identifying the propagation path of geological disaster chains based on spatiotemporal correlation extracts spatiotemporal features from multi-source monitoring data of geological disasters to obtain unified spatiotemporal feature data. Based on this, it extracts the correlation between disaster events and constructs a knowledge graph containing disaster entity nodes, relation edges, and spatiotemporal attribute annotations. Then, it calculates the spatiotemporal correlation strength based on the Gaussian decay of the fusion time interval, the Gaussian decay of the spatial distance, and attribute similarity according to the spatiotemporal attribute annotations between nodes. Finally, based on this strength, it traverses nodes and edges that meet the threshold along the relation edges to extract the propagation path. This method can solve the problem of existing technologies struggling to handle the spatiotemporal coupling relationship of multi-source heterogeneous data, achieving accurate quantification of the spatiotemporal correlation degree of disaster events and systematic modeling of causal relationships. It can improve the accuracy and automation level of disaster chain propagation path identification, reduce subjective errors in manual interpretation, and reduce path misjudgment or omissions caused by weak correlations, thereby meeting the needs of refined identification and early warning of geological disaster chains. Attached Figure Description

[0051] To more clearly illustrate the technical solutions in the embodiments or related technologies of this application, the accompanying drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0052] Figure 1 This is a flowchart illustrating a method for identifying the propagation path of a geological disaster chain based on spatiotemporal correlation in one embodiment.

[0053] Figure 2 This is a schematic diagram of the structure of a geological disaster chain propagation path identification system based on spatiotemporal correlation in one embodiment;

[0054] Figure 3 This is a schematic diagram of a computer device in one embodiment. Detailed Implementation

[0055] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0056] In one embodiment, reference Figure 1The document presents a flowchart illustrating a method for identifying the propagation path of a geological disaster chain based on spatiotemporal correlation, as provided in this application. This embodiment uses the application of this method to an identification terminal (hereinafter referred to as the terminal) as an example for illustration. It is understood that this method can also be applied to a server, and further to a system including both a terminal and a server, and is implemented through interaction between the terminal and the server. In this embodiment, the method includes the following steps:

[0057] S1. Obtain multi-source monitoring data of geological disasters, extract spatiotemporal features from the multi-source monitoring data of geological disasters, and obtain spatiotemporal feature data of geological disasters.

[0058] For example, the identification terminal first acquires multi-source monitoring data of geological hazards from multiple data sources. This multi-source monitoring data includes surface displacement sequence data collected by displacement sensors, rainfall intensity time series data recorded by rain gauges, seismic ground motion signal data captured by seismic detectors, and surface three-dimensional deformation data acquired by GPS receivers. Different sensors have different sampling frequencies, data formats, and spatial reference frames. The identification terminal performs spatiotemporal feature extraction on all acquired monitoring data. This operation aims to extract numerical features that characterize the dynamic evolution of geological hazards from the raw observations. Spatiotemporal feature extraction includes parameters such as the trend change rate, fluctuation amplitude, and peak occurrence time in the time dimension, and parameters such as gradient direction, variogram value, and local density in the spatial dimension. The output of spatiotemporal feature extraction is spatiotemporal feature data of geological hazards, which is organized in the form of feature vectors. Each feature vector corresponds to a comprehensive description of a monitoring data point in both time and space dimensions. This spatiotemporal feature data of geological hazards is subsequently used for the identification of hazard correlations. When extracting spatiotemporal features, the identification terminal can employ a combination of sliding window statistics and spatial differencing. For example, the observation sequence of monitoring data points within a time window can be... Temporal characteristics The calculation expression is as follows ,in, This is the series mean, expressed in millimeters squared or meters per second squared. This represents the number of data points within the window. Spatial characteristics. The expression can be ,in, For spatial gradient operators, The spatial distribution function of the monitoring data points is expressed in millimeters per meter or meters per second squared per meter.

[0059] S2. Extract disaster correlation data from the spatiotemporal characteristic data of geological disasters to obtain disaster chain correlation data.

[0060] For example, after obtaining spatiotemporal feature data of geological disasters, the identification terminal can extract disaster correlation data from the spatiotemporal feature data to obtain disaster chain correlation data. The purpose of disaster correlation extraction is to determine whether the abnormal changes corresponding to different monitoring data points originate from the same disaster event, and whether there is a triggering and being triggered causal relationship between different disaster events. The identification terminal can use a clustering algorithm based on spatiotemporal distance metrics to group spatiotemporally adjacent and feature-similar data points into the same disaster event, and then use temporal order and spatial propagation direction to determine the causal direction between events. Disaster chain correlation data can describe the connection structure between disaster events in the form of a directed graph. The nodes in the disaster chain correlation data represent discretized disaster events, and the directed edges in the disaster chain correlation data represent the triggering relationship from one disaster event to another. For example, the first The feature vector of each monitoring data point can be , No. The feature vector of each monitoring data point can be Then the first The monitoring data point and the first Spatiotemporal distance between monitoring data points The expression can be ,in, and The first The and the first The timestamps for each monitoring data point are in seconds. This is the time normalization factor, in seconds. and These are spatial coordinate vectors, with units in meters. This is the spatial normalization factor, in meters. This represents the Euclidean norm. The identification terminal can determine this based on... The data points are clustered based on their numerical values. After clustering, the data points within each cluster are arranged in chronological order. The clusters with earlier times are labeled as causal events, and the clusters with later times are labeled as result events, thus obtaining disaster chain correlation data.

[0061] S3. Construct a knowledge graph of the disaster chain relationship data and the spatiotemporal characteristic data of geological disasters to obtain geological disaster chain knowledge graph data.

[0062] For example, the identification terminal can take disaster chain correlation data and geological disaster spatiotemporal feature data as input, perform a knowledge graph construction operation, and obtain geological disaster chain knowledge graph data. The purpose of knowledge graph construction is to organize discrete disaster events, the relationships between events, and the spatiotemporal attributes of the events themselves into a graph-structured knowledge representation. Geological disaster chain knowledge graph data can include disaster entity nodes, relationship edges, and spatiotemporal attribute annotations. Disaster entity nodes can correspond to each disaster event, and each node has a unique identifier. Relationship edges are used to connect two disaster entity nodes and carry directional information, with the direction pointing from the cause event to the result event. Spatiotemporal attribute annotations are attached to disaster entity nodes, and these annotations can be used to record extended attributes such as the time center value, spatial center coordinates, duration, and impact range of the disaster event. The knowledge graph construction process includes three sub-operations: entity alignment, relationship fusion, and attribute filling. Specifically, during the construction process, the identification terminal can merge redundant connections in the disaster chain correlation data and disambiguate the same entity from different monitoring data sources. For example, the first... The attribute set of each disaster entity node is Time center value in spatiotemporal attribute annotation The expression can be ,in, For belonging to the first The number of monitoring data points for each disaster entity node. For the first The first disaster entity node Timestamps for each monitoring data point, in seconds. Spatial center coordinates. The expression can be ,in, For the first The first disaster entity node The spatial coordinate vectors of each monitoring data point are expressed in meters. The identification terminal can integrate disaster entity node attribute data, relationship type annotations, and spatiotemporal attribute annotations to generate geological disaster chain knowledge graph data that conforms to the requirements of graph database format.

[0063] S4. Based on the spatiotemporal attribute annotations of disaster entity nodes in the geological disaster chain knowledge graph data, calculate the spatiotemporal correlation strength between each disaster entity node pair to obtain the spatiotemporal correlation strength data between each disaster entity node.

[0064] For example, the identification terminal can read the spatiotemporal attribute annotations of each disaster entity node from the geological disaster chain knowledge graph data, and calculate the spatiotemporal association strength for any two disaster entity node pairs connected by a relational edge in the graph, thus obtaining the spatiotemporal association strength data between each disaster entity node. Spatiotemporal association strength is a comprehensive metric that can be used to comprehensively measure the temporal proximity, spatial proximity, and similarity of attribute features between node pairs. The closer two disaster entity nodes are in time, space, and attribute features, the higher their spatiotemporal association strength value. Spatiotemporal association strength also considers the attenuation characteristics during the propagation of the geological disaster chain; that is, as the time interval or spatial distance increases, the causal association between disaster events decreases exponentially. The calculation of spatiotemporal association strength requires obtaining the time interval, spatial distance, and attribute similarity of the node pairs. The time interval can be obtained by subtracting the time center values ​​in the spatiotemporal attribute annotations of the two nodes; the spatial distance can be obtained by the Euclidean distance of the spatial center coordinates; and the attribute similarity can be obtained by comparing the cosine similarity or Pearson correlation coefficient between the attribute sets of the two nodes. The identification terminal assigns the calculated spatiotemporal correlation strength to the corresponding relation edge, so that each relation edge carries a weight value that reflects the degree of causal correlation.

[0065] S5. Starting from the disaster entity nodes in the geological disaster chain knowledge graph data, based on the spatiotemporal correlation strength data, traverse the correlation edges that satisfy the correlation strength threshold and the disaster entity nodes connected to them along the direction of the correlation edges, extract the geological disaster chain propagation path composed of disaster entity nodes and correlation edges that satisfy the conditions, and obtain the geological disaster chain propagation path identification result.

[0066] For example, the identification terminal can combine geological disaster chain knowledge graph data and spatiotemporal correlation strength data to perform path search and extraction operations, obtaining the identification results of geological disaster chain propagation paths. The correlation strength threshold is a preset value, which can be used to filter out edges with significant causal relationships. The identification terminal can retain edges with spatiotemporal correlation strength greater than or equal to the correlation strength threshold and delete edges with correlation strength below the threshold. The identification terminal can search for all disaster entity nodes with an in-degree of zero in the geological disaster chain knowledge graph data. An in-degree of zero indicates that no other disaster entity node points to that node, and that node is a possible disaster chain initiation event. Starting from each node with an in-degree of zero, the identification terminal can perform a depth-first or breadth-first traversal along the retained relation edges. During the traversal, the sequence of disaster entity nodes traversed and the relation edges connecting these nodes are recorded. When a node has an out-degree of zero (i.e., no edge pointing to other nodes) or the spatiotemporal correlation strength of all downstream nodes pointed to by that node is less than the correlation strength threshold, the traversal terminates at that node. Each acyclic path from the starting node to the ending node constitutes a candidate propagation path. The identification terminal can calculate the cumulative path strength of each candidate propagation path. The cumulative path strength is equal to the product of the spatiotemporal correlation strengths of all relation edges along the path. The cumulative path strength reflects the overall reliability of the entire disaster chain propagation. The identification terminal can sort all candidate propagation paths in descending order of cumulative path strength and output the sorted list as the geological disaster chain propagation path identification result. (Path Cumulative Strength) The expression can be ,in, For the first The first disaster entity node and the first The spatiotemporal correlation strength between individual disaster entity nodes For the first The set of all relation edges on a candidate propagation path. The greater the cumulative strength of the path, the higher the probability that the path is a propagation path of an actual geological disaster chain.

[0067] In the aforementioned method for identifying geological disaster chain propagation paths based on spatiotemporal correlation, multi-source heterogeneous monitoring data is converted into unified spatiotemporal feature data through spatiotemporal feature extraction. This allows for the extraction of correlations between disaster events and the construction of a knowledge graph containing disaster entity nodes, relation edges, and spatiotemporal attribute annotations. Based on this, a spatiotemporal correlation strength calculation method integrating time interval, spatial distance, and attribute similarity, and incorporating an exponential decay function, is used to accurately quantify the causal correlation between node pairs. This fully considers the dynamic decay characteristics of spatiotemporal correlation strength during disaster chain propagation. Finally, effective relation edges are selected based on a preset correlation strength threshold, and a traversal search is performed along the relation edge pointing from the starting node with zero in-degree to extract candidate propagation paths, which are then sorted by cumulative path strength. This method systematically models the complex causal relationships between disaster entities, extracts directed propagation structures, improves the objectivity and accuracy of geological disaster chain propagation path identification, effectively reduces subjective misjudgments due to reliance on human experience, and avoids omissions of propagation paths caused by ignoring spatiotemporal decay characteristics.

[0068] In an optional embodiment, the multi-source monitoring data for geological hazards includes multiple monitoring data points, and S1 includes:

[0069] S11. Time stamp alignment and spatial coordinate unification are performed on multi-source geological disaster monitoring data to obtain standardized geological disaster monitoring data.

[0070] Optionally, after acquiring multi-source geological hazard monitoring data, the identification terminal first performs data standardization preprocessing. Displacement sensor data in multi-source geological hazard monitoring typically uses the device's local time, rain gauge data uses Coordinated Universal Time (UTC), and seismic detector data uses GPS timing; the time bases of different sensors differ. The identification terminal converts the timestamps of all monitoring data to a unified time base, such as Coordinated Universal Time (UTC). Specifically, the identification terminal can perform linear interpolation resampling on the time series of each sensor, aligning the time coordinates of all data points to the same equally spaced time axis. Different sensors also use different spatial coordinate reference systems; displacement sensors often use local engineering coordinate systems, rain gauges use geographic latitude and longitude coordinates, and seismic detectors use station center coordinates. The identification terminal can transform the spatial coordinates of all monitoring data to the same geodetic coordinate reference system, such as the Universal Transverse Mercator Projection System (UTM). Timestamp alignment can be performed using nearest neighbor matching or linear interpolation, while spatial coordinate unification can be achieved using a seven-parameter Bursa-Taylor coordinate transformation model. The data obtained after timestamp alignment and spatial coordinate unification is called standardized geological hazard monitoring data. Each monitoring data point in the standardized geological hazard monitoring data carries a unified format timestamp and spatial coordinates. Specifically, the original timestamp alignment... Each sensor at time The collected values ​​are The converted unified time tag is The expression for linear interpolation can be: ,in, and The units depend on the sensor type (displacement is in millimeters, rainfall intensity is in millimeters per hour, and ground motion velocity is in meters per second). and The unit is seconds. The expression for unified spatial coordinate transformation can be: ,in, These are translation parameters, in meters. As a scale factor, Let be a rotation matrix. These are the original coordinates, in meters. These are the transformed coordinates, in meters.

[0071] S12. Extract time-series features from standardized geological disaster monitoring data to obtain geological disaster time-series feature data.

[0072] Optionally, the identification terminal can extract statistical and dynamic features along the time dimension from standardized geological disaster monitoring data to obtain geological disaster time-series feature data. The purpose of time-series feature extraction is to transform the original time series into feature parameters that can reflect the evolution law of geological disasters. For each monitoring data point's neighborhood window on the time axis, the identification terminal can calculate statistical quantities such as mean, variance, skewness, kurtosis, autocorrelation coefficient, trend slope, and abrupt change intensity within that window. The mean can be used to reflect the average level of the data, the variance can be used to reflect the fluctuation range of the data, the skewness can be used to reflect the asymmetry of the data distribution, the kurtosis can be used to reflect the tail thickness of the data distribution, the autocorrelation coefficient can be used to reflect the time dependence of the data, the trend slope can be used to reflect the monotonic rate of change of the data, and the abrupt change intensity can be used to reflect the magnitude of the jump in the data within a short period of time. The identification terminal can use a sliding window method to calculate the above features point by point along the entire time axis, and the window length can be set according to the typical evolution cycle of geological disasters. For example, in the standardized geological disaster monitoring data, the first... The observed values ​​of each monitoring data point are The corresponding time is Take the window length as There are [number] data points, and the index range of the data points within the window is [number]. arrive The slope of the trend within that window The trend slope can be obtained through least squares fitting, and the expression for the trend slope can be: ,in, This represents the average time within the window, in seconds. The mean of the observations within the window, in units of and . Same (e.g., displacement in millimeters, rainfall intensity in millimeters per hour, earthquake velocity in meters per second). The unit is the observation unit divided by seconds; all time series features are combined into a multi-dimensional feature vector, namely geological disaster time series feature data.

[0073] S13. Extract spatial distribution features from standardized geological disaster monitoring data to obtain spatial distribution feature data of geological disasters.

[0074] Optionally, the identification terminal can extract spatial distribution and structural features from standardized geological hazard monitoring data to obtain spatial distribution feature data of geological hazards. The purpose of spatial distribution feature extraction is to characterize the spatial variation pattern of geological hazard monitoring data. For each monitoring data point, the identification terminal considers the numerical differences and geometric relationships of other data points in its surrounding neighborhood. Spatial distribution features may include local variogram values, gradient vectors, curvature, local Moran's index, and anisotropy index. Local variograms can be used to reflect the characteristic that the correlation of data in space decays with distance; gradient vectors can be used to characterize the direction of maximum change of data in space; curvature can be used to reflect the degree of curvature of the spatial surface; local Moran's index can be used to measure the strength of spatial autocorrelation; and anisotropy index can be used to describe the spatial variation differences in different directions. The identification terminal can use the semi-variogram analysis method in Kriging interpolation to calculate spatial features. For example, the spatial coordinates of the monitoring data points are... All units are meters, and the observed values ​​are... Then any two points and The semivariogram values ​​between The calculation formula is: The unit is the square of the observed unit. For a point Its local gradient vector The two components are obtained by finite difference approximation, and the expression is: The unit is per meter of observation. The unit is per meter of observation. The magnitude of the gradient vector. and direction angle Together they constitute part of the spatial distribution characteristics; all spatial characteristics constitute the spatial distribution characteristic data of geological disasters.

[0075] S14. Perform spatiotemporal feature fusion on the temporal feature data and spatial distribution feature data of geological disasters to obtain spatiotemporal feature data of geological disasters.

[0076] Optionally, the identification terminal fuses temporal feature data and spatial distribution feature data of geological hazards to generate spatiotemporal feature data of geological hazards. The purpose of spatiotemporal feature fusion is to integrate information from the time and spatial dimensions into a unified feature space, so that subsequent hazard correlation extraction can utilize complementary information from both time and space dimensions simultaneously. The fusion operation can employ feature concatenation, weighted summation, or a dynamic fusion method based on an attention mechanism. Feature concatenation is the most direct fusion method, which involves sequentially concatenating the temporal feature vector and spatial feature vector of each monitoring data point into a longer vector. The weighted summation method assigns different weight coefficients to the two feature vectors, and the weight coefficients can be dynamically adjusted according to the type of geological hazard. For example, for landslide hazards, the gradient direction in the spatial features has a higher weight; for debris flow hazards, the abrupt change intensity in the temporal features has a higher weight. The fusion method based on an attention mechanism can adaptively determine the contribution of each feature to the current hazard state by calculating the interaction correlation between temporal and spatial features. After fusion, the identification terminal normalizes the resulting feature vector to eliminate differences in dimensions and orders of magnitude between different features. For example, the first... The geological disaster time-series feature vector of each monitoring data point is: , dimension The dimensionless component has a mean of 0 and a variance of 1; the spatial distribution feature vector of geological hazards is... , dimension Similarly, dimensionless transformation is performed. When using feature stitching, the feature vectors in the spatiotemporal feature data of geological disasters... The expression can be ,in, This represents a vector concatenation operation. The dimension is When using weighted summation, ,in and Let be the weight coefficient, and satisfy... The fused feature vector is the spatiotemporal feature data of geological disasters.

[0077] In an optional embodiment, S2 includes:

[0078] S21. Based on the spatiotemporal characteristic data of geological disasters, calculate the spatiotemporal distance between any two monitoring data points, and generate spatiotemporal distance matrix data based on the spatiotemporal distance.

[0079] Optionally, the identification terminal can use spatiotemporal feature data of geological disasters as input to calculate the spatiotemporal distance between any two monitoring data points in the spatiotemporal feature data of geological disasters, and generate spatiotemporal distance matrix data based on the spatiotemporal distance. Spatiotemporal distance is a comprehensive metric that can be used to characterize the interval between monitoring data points on the time axis, the straight-line distance in space, and the feature differences in the feature space. The larger the time interval, the greater the spatial distance, and the more significant the feature differences between two data points, the larger their spatiotemporal distance value. The spatiotemporal distance matrix is ​​a symmetric square matrix, where the rows and columns of the matrix correspond to the numbers of the monitoring data points, and each element in the matrix represents the spatiotemporal distance value between the corresponding two data points. Before calculating the spatiotemporal distance, the identification terminal needs to normalize the time coordinates, spatial coordinates, and feature vectors respectively to avoid the unbalanced influence of different physical dimensions on the distance calculation. Time coordinate normalization uses the standard deviation of the time coordinates of all monitoring data points as the denominator, spatial coordinate normalization uses the standard deviation of the spatial coordinates of all monitoring data points as the denominator, and feature vector normalization uses the standard deviation of each feature dimension as the denominator. For example, there are a total of The monitoring data point, the first The time coordinates of each monitoring data point are The unit is seconds, and the spatial coordinate vector is... The unit is meters, and the feature vector is... ;No. The and the first Spatiotemporal distance between monitoring data points The standardized expression can be ,in, The unit is seconds. For all The mean, in seconds; The unit is meters. For all The mean of the eigenvectors, in meters; the Euclidean distance of the eigenvectors. Dimensionless. Spatiotemporal distance matrix data is a... The matrix is ​​denoted as ,in, ,and .

[0080] S22. Based on the spatiotemporal distance matrix data, cluster analysis is performed on the spatiotemporal characteristic data of geological disasters to obtain clustering results of disaster events.

[0081] Optionally, the identification terminal can perform clustering analysis on all monitoring data points in the spatiotemporal feature data of geological disasters based on the spatiotemporal distance matrix data to obtain disaster event clustering results. The purpose of clustering analysis is to group data points that are close in spatiotemporal distance and have similar characteristics into the same disaster event, thereby discretizing continuous monitoring data points into independent disaster event entities. The identification terminal can use a density-based spatial clustering algorithm for clustering analysis. The core parameters of the density-based spatial clustering algorithm can include the neighborhood radius and the minimum number of points. For each monitoring data point, the number of data points in the neighborhood centered at that point and within a spatiotemporal distance not exceeding the neighborhood radius is calculated. If the number of data points is greater than or equal to the minimum number of points, the point is marked as a core point. All core points and their density-reachable points constitute a cluster. Data points that do not belong to any cluster are marked as noise points. For example, the distance threshold in the spatiotemporal distance matrix data is... The minimum number of points is Dimensionless. For the first... Each monitoring data point, - Neighborhood is defined as .like Then point The core point. Two core points. and The condition for density to be directly reached is Density connectivity can be obtained through transitive closure. Each density-connected maximal set forms a cluster. After clustering, the... The set of monitoring data point indices contained in each cluster is denoted as The disaster event clustering results data can include a list of all clusters and information on the monitoring data points contained in each cluster. Each cluster corresponds to a disaster event to be identified.

[0082] S23. Based on the clustering results of disaster events, determine the causal direction of the spatiotemporal characteristics of geological disasters within each cluster to obtain disaster causal direction data.

[0083] Optionally, the identification terminal analyzes the temporal sequence and spatial propagation pattern of monitoring data points within each cluster of disaster event clustering results data to determine whether a causal triggering relationship and causal direction exist between different clusters. The rule for determining causal direction can be: if a triggering relationship exists between two disaster events, the causal event must be earlier than the result event in time, and the spatial location of the causal event must conform to the propagation pattern of geological disasters (e.g., landslides propagate from the top to the bottom of the slope, and debris flows propagate from upstream to downstream). The identification terminal can calculate the time center value of each cluster, i.e., the average timestamp of all monitoring data points within the cluster. The identification terminal can sort all clusters in ascending order of time center value. For any two clusters... and ,like The time center value is less than The time center value is initially considered to be... may be The reason is that the identification terminal can further calculate the vector between the spatial centers of two clusters and check whether the direction of this vector is consistent with the dominant direction in the geological topography (such as slope direction, gully direction). The identification terminal can also quantify the causal strength between two clusters using statistical methods such as transfer entropy or Granger causality tests. For example, clusters... and clusters The time center values ​​are respectively and The unit is seconds, and the coordinates of the spatial center are respectively and The unit is meters. From arrive Confidence of causal orientation The expression can be ,in, This is the time sensitivity coefficient, expressed in negative first power of seconds. The direction sensitivity coefficient, From point to The angle between the vector and the geologically dominant direction, expressed in radians. When When the value exceeds a preset causal threshold, the identification terminal determines that there is a causal relationship. arrive The causal relationship. Disaster causal direction data can include all cluster pairs determined to have a causal relationship and their directions.

[0084] S24. Based on the disaster event clustering results data and disaster causal direction data, construct the disaster correlation topology structure to obtain disaster chain correlation data.

[0085] Optionally, the identification terminal can define each cluster in the disaster event clustering results data as a disaster event node, and add directed edges between nodes based on the causal relationships in the disaster causal direction data, thereby constructing a disaster correlation topology. This topology can be a directed graph, where nodes represent disaster events and directed edges represent triggering relationships from one disaster event to another. The disaster correlation topology may contain multiple connected components, each corresponding to one or more interconnected disaster chains. During the construction process, the identification terminal needs to perform a consistency check on the causal direction data. If a circular causal relationship (i.e., ...) is found... arrive and arrive If both edges exist (i.e., both have strong causal relationships), then the edge with the stronger causal relationship is retained, and the edge with the weaker causal relationship is deleted. The recognition terminal can assign a weight to each directed edge; the weight value can be used to calculate the causal direction confidence. Alternatively, the strength of the spatiotemporal correlation can be determined. For example, suppose the total number of disaster event nodes is... Disaster causal direction data includes A directed edge. Disaster chain correlation data can be represented using an adjacency matrix. It means that, among them This indicates the existence of slave nodes. To the node The directed edge, This indicates that no directed edges exist. To facilitate subsequent knowledge graph construction, the recognition terminal can also store disaster chain relationship data in an edge list format, with each row containing... ,in, As an identifier for the nodes of a causal disaster event, For identifying the nodes of the resulting disaster event, The confidence level for causal direction can be temporarily set to 1. Disaster chain correlation data can be used to characterize the propagation topology between geological disaster events.

[0086] In an optional embodiment, S3 includes:

[0087] S31. Based on the disaster event clustering results data, feature aggregation is performed on the spatiotemporal characteristic data of geological disasters within each cluster to obtain the attribute data of disaster entity nodes.

[0088] Optionally, the identification terminal can read each cluster in the disaster event clustering result data, and perform an aggregation operation on the spatiotemporal characteristic data of geological disasters of all monitoring data points within that cluster to generate attribute data of the disaster entity nodes corresponding to that cluster. The aggregation operation may include calculating statistics such as the mean, median, maximum, minimum, and standard deviation for each feature dimension. The mean can be used to reflect the average state of the disaster event, the maximum value can be used to reflect the extreme intensity of the disaster event, and the standard deviation can be used to reflect the degree of unevenness within the disaster event. The identification terminal can also extract specific attributes based on knowledge in the field of geological disasters, such as the sliding direction of a landslide, the deposition area of ​​a debris flow, or the volume of falling rocks in a collapse. For example, the first... Each disaster entity node contains There are [number] monitoring data points, and the spatiotemporal feature vector of geological disaster for each monitoring data point can be represented as: , mean vector The expression can be Maximum vector Each component is ,in, For the first The monitoring data point of the first Each feature component. Standard deviation vector. The expression can be In addition to statistical features, the identification terminal can also calculate physical properties based on the raw sensor data. For example, the sliding direction angle... The expression can be ,in, and The first Each monitoring data point corresponds to the northward and eastward components of the displacement, in meters. The unit is radians. All aggregated features and physical attributes together constitute the disaster entity node attribute data.

[0089] S32. Based on the disaster entity node attribute data and disaster causal direction data, the directed edges in the disaster chain relationship data are labeled with relationship types to obtain disaster relationship data.

[0090] Optionally, the identification terminal can combine disaster entity node attribute data and disaster causal direction data to label the relationship type of each directed edge in the disaster chain correlation data, thus obtaining disaster correlation data. The purpose of relationship type labeling is to clarify the specific triggering mechanism between two disaster entity nodes, such as "landslide triggering debris flow," "collapse blocking river channel forming a barrier lake," or "rainfall inducing landslide," etc. Specifically, the labeling process can adopt a combination of rule reasoning and machine learning. The identification terminal can determine the basic category of the node based on the disaster type label in the disaster entity node attribute data. The disaster type can be automatically identified through feature patterns within clusters; for example, nodes with dominant sliding displacement are labeled as landslides, nodes with high flow velocities are labeled as debris flows, and nodes with significant vertical displacement are labeled as collapses. For each directed edge, the identification terminal can check the matching pattern between the attributes of the cause node and the attributes of the result node. The rule base can include triggering conditions for various typical relationship types. For example, if the hazard type of the cause node is landslide and the hazard type of the result node is debris flow, and the spatial location of the cause node is in the upper part of the slope while the spatial location of the result node is in the gully, then the directed edge is labeled as "landslide transforming into debris flow". If a predefined rule cannot be matched, the identification terminal can use a graph neural network-based relationship classification model. This model takes the feature vectors from the attribute data of the hazard entity nodes as input and outputs the probability distribution of the relationship type. For example, the cause node... The attribute vector is Result node The attribute vector is relation type posterior probability The expression can be ,in, and This is the weight matrix. and For bias vectors, This indicates concatenating two attribute vectors. For activation function, This is a normalized exponential function. The identification terminal can select the relationship type with the highest probability as the labeling result to obtain disaster correlation data. The disaster correlation data can include the original disaster chain correlation data and newly added relationship type labels.

[0091] S33. Based on the spatiotemporal characteristic data of geological disasters, the spatiotemporal attributes of each disaster entity node are calculated to obtain the spatiotemporal attribute data of the disaster entity node.

[0092] Optionally, the identification terminal calculates the spatiotemporal attributes of each disaster entity node based on the original time and spatial coordinates in the geological disaster spatiotemporal feature data. Spatiotemporal attributes can include time center attributes, time range attributes, spatial center attributes, and spatial range attributes. The time center attribute can be used to characterize the typical moment of the disaster event, and the time range attribute can be used to characterize the duration of the disaster event. The spatial center attribute characterizes the core location of the disaster event, and the spatial range attribute can be used to characterize the size of the affected area of ​​the disaster event. The identification terminal can calculate these attributes independently for each disaster entity node, without relying on other nodes. The time center attribute can be the arithmetic mean of the timestamps of all monitoring data points within a cluster, and the time range attribute can be the standard deviation or range of the timestamps. The spatial center attribute can be the arithmetic mean of the spatial coordinates of all monitoring data points within a cluster, and the spatial range attribute can be the standard deviation or convex hull area of ​​the spatial coordinates. The spatiotemporal attribute data of the disaster entity node can include all attributes organized into structured records, with one record corresponding to each disaster entity node.

[0093] S34. Based on disaster entity node attribute data, disaster correlation data, and disaster entity node spatiotemporal attribute data, a geological disaster chain knowledge graph data is constructed.

[0094] Optionally, the identification terminal can integrate disaster entity node attribute data, disaster association data, and disaster entity node spatiotemporal attribute data into a unified graph data structure, which is the geological disaster chain knowledge graph data. The construction process can include three sub-operations: node creation, edge creation, and attribute attachment. The node creation process is as follows: the identification terminal generates a unique identifier for each disaster entity node and stores all attribute fields (such as disaster type, sliding direction, maximum displacement, etc.) from the disaster entity node attribute data as the node's attribute list. The edge creation process is as follows: the identification terminal traverses each directed edge in the disaster association data, establishes connections between nodes based on the source node identifier and target node identifier of the directed edge, and stores the relationship type label as the edge's attribute. The attribute attachment process is as follows: the identification terminal attaches the time center attribute, time range attribute, spatial center attribute, and spatial range attribute from the disaster entity node's spatiotemporal attribute data to the corresponding node as part of the node's spatiotemporal attribute annotation. The geological disaster chain knowledge graph data can be stored using a resource description framework format or as a set of nodes and edges in a graph database. In the resource description framework, each disaster entity node corresponds to a Uniform Resource Identifier (URI), each relation edge corresponds to a predicate, and node attributes and edge attributes correspond to literals. The identification terminal can also perform consistency checks on the knowledge graph, checking for isolated nodes or duplicate edges, and filling in missing attributes with default values. After construction, each disaster entity node in the geological disaster chain knowledge graph data includes a complete attribute description and spatiotemporal annotation, and each relation edge includes direction information and type annotation.

[0095] In an optional embodiment, S4 includes:

[0096] S41. Based on the time center attribute, calculate the time interval between any two disaster entity nodes to obtain the disaster entity node time interval data.

[0097] Optionally, the identification terminal reads the time center attribute of each disaster entity node from the geological disaster chain knowledge graph data, and calculates the absolute difference of the time center attributes of two nodes for all node pairs connected by relational edges, obtaining the disaster entity node time interval data. The time interval reflects the proximity of two disaster events in terms of occurrence time. The smaller the time interval, the more likely the two events are causally related in time; the larger the time interval, the lower the probability of causal relationship. The identification terminal only calculates the time interval of node pairs with existing directed edges in the geological disaster chain relationship data, and does not perform full calculation on any two nodes, in order to reduce computational overhead. For example, the first... The time center attribute of each disaster entity node is The unit is seconds, the first The time center attribute of each disaster entity node is The unit is seconds, and the time interval is... The expression can be The unit is seconds. If there is a directed edge between two nodes and the direction is from... point to Then it is generally expected The time interval is still calculated using absolute values. The identification terminal stores the calculated time interval values ​​as disaster entity node time interval data. The disaster entity node time interval data can be in sparse matrix form, where non-zero elements correspond to node pairs with related edges. For node pairs without related edges, no time interval is calculated.

[0098] S42. Based on the spatial center attribute, calculate the spatial distance between any two disaster entity nodes to obtain the spatial distance data of disaster entity nodes.

[0099] Optionally, the identification terminal can read the spatial center attribute of each disaster entity node from the geological disaster chain knowledge graph data, and calculate the Euclidean distance between the spatial center attributes of two nodes for all node pairs connected by related edges, thus obtaining the spatial distance data of the disaster entity nodes. Spatial distance can be used to reflect the proximity of two disaster events in space. The smaller the spatial distance, the more likely the two events are to have a propagation relationship in space; the larger the spatial distance, the lower the probability of propagation. The identification terminal can calculate only the spatial distance between node pairs with directed edges. For example, the first... The spatial center attribute of each disaster entity node is The unit is meters, the first The spatial center attribute of each disaster entity node is If the unit is meters, then the spatial distance The expression can be The unit is meters. In identifying the propagation path of geological disaster chains, spatial distance is considered not only as a straight-line distance in a plane, but also, for areas with significant topographic relief, a corrected Euclidean distance that takes into account elevation differences can be used for the identification terminal. ,in, The elevation is expressed in meters. The identification terminal stores the calculated spatial distance values ​​as disaster entity node spatial distance data, which can be stored in a sparse storage format.

[0100] S43. Based on the disaster entity node attribute data, calculate the attribute similarity between any two disaster entity nodes to obtain disaster entity node attribute similarity data.

[0101] Optionally, the identification terminal can read the attribute data of each disaster entity node from the geological disaster chain knowledge graph data, and calculate the attribute similarity between two nodes for all node pairs connected by relational edges. Attribute similarity can be used to measure the degree of similarity between two disaster events in their intrinsic features, such as disaster type, sliding direction, displacement amplitude, and energy release. Similar disaster events are more likely to have causal relationships because events on the same disaster chain often have similar dynamic characteristics. The identification terminal can use cosine similarity or weighted Euclidean similarity for calculation. Cosine similarity is suitable for scenarios where the dimensions of the attribute vectors are consistent and direction-sensitive, while weighted Euclidean similarity is suitable for scenarios where the differences in the importance of each attribute need to be considered. For example, the first... The attribute vector of each disaster entity node is , No. The attribute vector of each disaster entity node is Both vectors have dimensions. cosine similarity The expression can be ,in, This represents the dot product operation of vectors. The Euclidean norm of a vector. The value of is between 0 and 1. Weighted Euclidean similarity. The expression can be ,in, For the first The weight coefficients of each attribute dimension. For the first The first node attribute vector There are two similarity components. The identification terminal can choose one of the two similarity methods or a combination of both. The higher the attribute similarity, the stronger the spatiotemporal correlation.

[0102] S44. Based on the disaster entity node time interval data, disaster entity node spatial distance data, and disaster entity node attribute similarity data, calculate the spatiotemporal association strength of the relationship edges in the disaster chain association data to obtain the spatiotemporal association strength data between each disaster entity node.

[0103] Optionally, the identification terminal takes disaster entity node time interval data, disaster entity node spatial distance data, and disaster entity node attribute similarity data as input, and calculates the spatiotemporal correlation strength for each relation edge in the geological disaster chain correlation data. The expression for the spatiotemporal correlation strength can be: ,in, Indicates the first The first disaster entity node and the first The spatiotemporal correlation strength between individual disaster entity nodes ranges from 0 to 1. Indicates the first The first disaster entity node and the first The similarity of attributes between disaster entity nodes. Indicates the first The first disaster entity node and the first The time interval between disaster entity nodes is in seconds. This represents the preset time decay scale parameter, in seconds, used to control the degree to which time intervals affect the correlation strength. The larger the value, the slower the time decay, allowing for longer time intervals to maintain a high correlation strength; The smaller the value, the faster the time decays, and only events with very close time intervals can achieve a high correlation strength. Indicates the first The first disaster entity node and the first The spatial distance between disaster entity nodes is expressed in meters. This represents the preset spatial attenuation scale parameter, in meters, used to control the influence of spatial distance on the correlation strength. (Exponential term) and All are Gaussian decay functions; the function value is 1 when the independent variable is zero, and monotonically decreases to 0 as the independent variable increases. The identification terminal calculates the spatiotemporal correlation strength between each node pair according to the above formula, and then processes the calculated results... The data is stored as spatiotemporal correlation strength data between various disaster entity nodes. The spatiotemporal correlation strength data and the relationship edges correspond one-to-one, providing edge weights for subsequent propagation path extraction.

[0104] In an optional embodiment, S5 includes:

[0105] S51. Based on the spatiotemporal correlation strength data and correlation strength threshold, threshold filtering is performed on each relation edge in the disaster chain correlation data to obtain candidate propagation edge data.

[0106] Optionally, the identification terminal reads the spatiotemporal correlation strength data between each disaster entity node, as well as a pre-set correlation strength threshold. The correlation strength threshold ranges from [0,1] and is used to filter out weak spatiotemporal correlation edges. Only correlation edges with a spatiotemporal correlation strength greater than or equal to the threshold are considered reliable propagation connections and can participate in the subsequent construction of the disaster chain propagation path; correlation edges below the threshold are considered noise or weak correlations and are excluded. The identification terminal traverses each correlation edge in the disaster chain correlation data, obtains the corresponding spatiotemporal correlation strength value, and compares this value with the correlation strength threshold. If the spatiotemporal correlation strength is greater than or equal to the correlation strength threshold, the correlation edge is retained; if it is less than the correlation strength threshold, the correlation edge is deleted from the disaster chain correlation data. The threshold filtering operation can significantly reduce invalid branches in the propagation path search space and avoid including weak or false correlations in the propagation path. For example, the correlation strength threshold is... Its value can be adjusted according to the conservatism of the geological hazard chain identification: higher values... A higher value will produce a shorter and more reliable propagation path, but may miss some true propagation relationships; a lower value... Higher values ​​result in longer propagation paths but may introduce false alarms. The identification terminal combines the relationship edges retained after threshold filtering with their corresponding disaster entity nodes into a new subgraph. The edges in this subgraph are candidate propagation edges, and the set of all candidate propagation edges is the candidate propagation edge data. Edges with a spatiotemporal correlation strength exactly equal to the threshold are retained by the identification terminal.

[0107] S52. Based on the candidate propagation edge data, perform in-degree statistics on the disaster entity nodes in the geological disaster chain knowledge graph data, and select disaster entity nodes with zero in-degree as the starting point of the propagation path to obtain the propagation path starting point set data.

[0108] Optionally, the identification terminal can calculate the in-degree of all disaster entity nodes in the geological disaster chain knowledge graph data based on the candidate propagation edge data. In-degree refers to the number of candidate propagation edges pointing to a given node. A disaster entity node with an in-degree of zero indicates that no other disaster entity node points to it through candidate propagation edges; this node is not triggered by other disaster events and is therefore a potential disaster chain initiation event. The identification terminal traverses all disaster entity nodes and counts the number of times each node appears as a target node in the candidate propagation edge data; this number is the in-degree. For nodes with an in-degree of zero, the identification terminal marks them as the starting point of a propagation path. All disaster entity nodes marked as starting points constitute the propagation path starting point set data. If multiple nodes have an in-degree of zero, each node serves as the starting point for searching an independent propagation path. If a node with an in-degree of zero has multiple outgoing edges, multiple branch paths can be generated from that node. If there are no nodes with an in-degree of zero in the candidate propagation edge data, it means that all nodes are pointed to by at least one other node. In this case, there may be a loop structure or all nodes forming one or more closed loops. The identification terminal will forcibly select the source node connected by the edge with the highest spatiotemporal correlation strength as the virtual starting point. In normal geological disaster chain scenarios, it is common for the starting point of the propagation path to have an in-degree of zero, such as in the initial landslide or initial rainfall event. For example, let the first... The in-degree of each disaster entity node is Its calculation formula is ,in, The set of all disaster entity nodes. For the set of directed edges in the candidate propagation edge data, This is an indicator function; it takes the value 1 when the condition within the parentheses is true, and 0 otherwise. The set of propagation path starting points can be denoted as... .

[0109] S53. Based on the propagation path starting point set data and candidate propagation edge data, starting from each propagation path starting point in the propagation path starting point set data, traverse layer by layer along the direction of the directed edges in the candidate propagation edge data, record the sequence of disaster entity nodes and relation edges passed during the traversal, and terminate the traversal when the out-degree of the current disaster entity node is zero or the spatiotemporal correlation strength of all downstream nodes corresponding to the out-degree is less than the correlation strength threshold, thus obtaining the candidate propagation path data.

[0110] Optionally, the identification terminal uses each propagation path origin in the propagation path origin set as an initial node and performs a depth-first or breadth-first traversal on the directed graph defined by the candidate propagation edge data. The traversal process strictly follows the direction of the directed edges, that is, only movement from the cause node to the result node is allowed, and reverse movement is not permitted. For the disaster entity node currently being visited, the identification terminal first obtains all outgoing edges of that node in the candidate propagation edge data, that is, edges pointing from that node to other nodes. If the number of outgoing edges is zero, then the node is a leaf node, the traversal terminates at this node, and the complete path from the origin to the current node is recorded. If the number of outgoing edges is greater than zero, the identification terminal sequentially visits the downstream nodes along each outgoing edge, and records the current node, the edge followed, and the downstream node in the path sequence. After visiting the downstream node, the outgoing edges of the downstream node are recursively traversed. The termination condition can be: the outgoing degree of the current node is zero, that is, there are no candidate propagation edges pointing to any other node. During traversal, the identification terminal needs to avoid repeatedly visiting the same node to prevent loops. However, since the candidate propagation edge data comes from directed acyclic graphs or near-acyclic graphs, the probability of loops is low. If a loop is detected, the identification terminal will terminate the current branch when visiting a visited node. After traversal, all acyclic paths from the starting point to a leaf node are recorded as candidate propagation paths. Each candidate propagation path data includes a sequence of disaster entity nodes on the path (arranged in the direction of propagation) and a sequence of relational edges connecting these nodes. The identification terminal summarizes the paths obtained from traversing all starting points to obtain the candidate propagation path data.

[0111] S54. Based on the candidate propagation path data, calculate the cumulative path strength of each candidate propagation path to obtain the candidate propagation path strength data.

[0112] Optionally, the identification terminal calculates the cumulative intensity of each candidate propagation path in the candidate propagation path data. The cumulative intensity of a path can be used to measure the overall reliability of the entire geological hazard chain from its starting point to its ending point. Since the spatiotemporal correlation strength of each relation edge reflects the reliability of that segment of propagation, the reliability of the entire path can be defined as the product of the spatiotemporal correlation strengths of all relation edges on the path. This product reflects the vulnerability of the propagation path: as long as there is a segment with low correlation strength on the path, the cumulative intensity of the entire path will decrease significantly; only when all segments have high correlation strength can the cumulative intensity remain at a high level. For example, the first... The sequence of disaster entity nodes included in the candidate propagation path is as follows ,in, The number of nodes on the path is given by the set of corresponding relation edges. . No. Path cumulative strength of candidate propagation paths The expression can be ,in, Indicates the first The cumulative strength of each candidate propagation path. Indicates the first The set of all relation edges contained in the candidate propagation path. Indicates the first From the candidate propagation path, from the first The disaster entity node points to the first The relationship edges between disaster entity nodes Indicates the first The first disaster entity node and the first The spatiotemporal correlation strength between individual disaster entity nodes. The estimated joint probability of the entire disaster chain propagation path under the assumption that each segment propagates independently; the larger the product value, the higher the probability that the path is a real geological disaster chain. Because... All values ​​are between 0 and 1. The cumulative intensity of a path decreases exponentially with increasing path length. Therefore, longer paths naturally have lower cumulative intensity. The identification terminal calculates the cumulative intensity of each candidate propagation path and associates the calculation results with the corresponding path identifier to obtain candidate propagation path intensity data. For isolated nodes with a path length of zero (i.e., the starting point is also the ending point), their cumulative intensity is set to 1. However, such paths are generally not considered valid disaster chain propagation paths.

[0113] S55. Based on the candidate propagation path intensity data, the candidate propagation paths are ranked to obtain the geological disaster chain propagation path identification results.

[0114] Optionally, the identification terminal sorts all candidate propagation paths in descending order based on the cumulative intensity values ​​of the paths in the candidate propagation path intensity data. The purpose of this sorting is to prioritize the most likely geological hazard chain propagation paths, allowing users or downstream early warning systems to focus on the most reliable hazard chains. After sorting, the identification terminal outputs a list of sorted paths as the geological hazard chain propagation path identification results. Each identification result includes the following information: the propagation path number, the identifiers of the hazard entity nodes arranged sequentially on the path and their spatiotemporal attributes, the edge type of the relationship between adjacent nodes on the path, and the value of the cumulative intensity of the path. If multiple candidate propagation paths share the same prefix, the identification terminal can choose to retain the complete path or truncate it to a certain length according to the cumulative intensity threshold. In practical applications, the identification terminal can present the sorting results visually, such as drawing the spatiotemporal trajectory of the path on a geographic information system map, or exporting it in tabular form. The output format can be structured text, such as JavaScript object notation or comma-separated value format, facilitating integration with other systems. The identification terminal can also mark the top few paths with the highest cumulative intensity as primary hazard chain propagation paths, and the remaining paths as secondary paths. Let the set of candidate propagation paths be... The corresponding path cumulative intensity is Sorting operations find a permutation. Make The sorted list of paths can be represented as follows: The terminal ultimately outputs this sorted list as the result of geological hazard chain propagation path identification; the result of geological hazard chain propagation path identification can be used for geological hazard early warning decision-making, disaster evolution trend analysis, and disaster prevention and mitigation resource deployment.

[0115] In the aforementioned method for identifying the propagation path of geological disaster chains based on spatiotemporal correlation, spatiotemporal features are extracted from multi-source monitoring data to obtain unified spatiotemporal feature data. Then, based on this data, disaster correlation relationships are extracted, constructing a knowledge graph containing disaster entity nodes, relation edges, and spatiotemporal attribute annotations. On this basis, the fusion time interval, spatial distance, and attribute similarity between each node pair are calculated, and a Gaussian decay function is introduced to assess the spatiotemporal correlation strength, thereby accurately quantifying the closeness of causal relationships between disaster events. Finally, starting from the nodes in the knowledge graph, a directed traversal search is performed along the relation edges and based on the correlation strength threshold to extract the complete disaster chain propagation path. This method enables automatic identification and quantitative analysis of geological disaster chain propagation paths, improving the objectivity and accuracy of path identification, effectively reducing the subjective bias of manual interpretation and the misjudgment and omission of weakly correlated paths in traditional methods, and enabling a refined reconstruction of the disaster evolution process.

[0116] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0117] Based on the same inventive concept, this application also provides a system for identifying the propagation path of a geological disaster chain based on spatiotemporal correlation, used to implement the aforementioned method for identifying the propagation path of a geological disaster chain based on spatiotemporal correlation. The solution provided by this system is similar to the solution described in the above method. Therefore, the specific limitations of one or more embodiments of the system for identifying the propagation path of a geological disaster chain based on spatiotemporal correlation provided below can be found in the limitations of the method for identifying the propagation path of a geological disaster chain based on spatiotemporal correlation described above, and will not be repeated here.

[0118] In one exemplary embodiment, such as Figure 2 As shown, a schematic diagram of a geological disaster chain propagation path identification system 10 based on spatiotemporal correlation is provided, including:

[0119] The multi-source monitoring data acquisition and spatiotemporal feature extraction module 11 is used to acquire multi-source monitoring data of geological disasters, extract spatiotemporal features from the multi-source monitoring data of geological disasters, and obtain spatiotemporal feature data of geological disasters.

[0120] The disaster correlation extraction module 12 is used to extract disaster correlations from the spatiotemporal feature data of geological disasters to obtain disaster chain correlation data.

[0121] The geological hazard chain knowledge graph construction module 13 is used to construct a knowledge graph from the disaster chain relationship data and the spatiotemporal feature data of geological hazards, so as to obtain geological hazard chain knowledge graph data; wherein, the geological hazard chain knowledge graph data includes disaster entity nodes, relationship edges and spatiotemporal attribute annotations;

[0122] The spatiotemporal correlation strength calculation module 14 is used to calculate the spatiotemporal correlation strength between each disaster entity node pair based on the spatiotemporal attribute annotation of the disaster entity nodes in the geological disaster chain knowledge graph data, and obtain the spatiotemporal correlation strength data between each disaster entity node.

[0123] The geological disaster chain propagation path identification module 15 is used to start from the disaster entity nodes in the geological disaster chain knowledge graph data, and based on the spatiotemporal correlation strength data, traverse the relation edges that satisfy the correlation strength threshold and the disaster entity nodes connected to them along the direction of the relation edges, extract the geological disaster chain propagation path composed of disaster entity nodes and relation edges that satisfy the conditions, and obtain the geological disaster chain propagation path identification result.

[0124] In one embodiment, such as Figure 3 A computer device 300 is provided, comprising:

[0125] At least one processor 301, and at least one memory 302 communicatively connected to said processor 301; said memory stores application code executable by said processor, said application code being executed by said processor to enable said processor to perform the steps of a method for identifying the propagation path of a geological disaster chain based on spatiotemporal correlation as described above;

[0126] The computer device may also include: sensor 303;

[0127] The processor 301, memory 302, and sensor 303 can be connected via bus 304 or other means. The figure shows an example of connection via bus 304. Figure 3 The character is represented by a single thick line, but this does not mean that there is only one bus or a type of bus.

[0128] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps in the above method embodiments.

[0129] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The device embodiments described above are merely illustrative. The components described as separate parts may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this disclosure according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0130] The above-described embodiments are merely illustrative of several implementation methods of the embodiments of this application, and their descriptions are relatively specific and detailed. However, they should not be construed as limiting the scope of the patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the embodiments of this application, and these modifications and improvements all fall within the protection scope of the embodiments of this application.

Claims

1. A method for identifying the propagation path of geological disaster chains based on spatiotemporal correlation, characterized in that, The method includes: S1. Obtain multi-source monitoring data of geological disasters, and extract spatiotemporal features from the multi-source monitoring data of geological disasters to obtain spatiotemporal feature data of geological disasters; S2. Extract disaster correlation data from the spatiotemporal feature data of the geological disasters to obtain disaster chain correlation data; S3. Construct a knowledge graph from the disaster chain relationship data and the spatiotemporal feature data of geological disasters to obtain geological disaster chain knowledge graph data; wherein, the geological disaster chain knowledge graph data includes disaster entity nodes, relationship edges and spatiotemporal attribute annotations; S4. Based on the spatiotemporal attribute annotations of the disaster entity nodes in the geological disaster chain knowledge graph data, calculate the spatiotemporal correlation strength between each pair of disaster entity nodes to obtain the spatiotemporal correlation strength data between each pair of disaster entity nodes; S5. Starting from the disaster entity node in the geological disaster chain knowledge graph data, based on the spatiotemporal correlation strength data, traverse the correlation edge that satisfies the correlation strength threshold and the disaster entity node connected to it along the direction of the correlation edge, extract the geological disaster chain propagation path composed of the disaster entity node that satisfies the condition and the correlation edge, and obtain the geological disaster chain propagation path identification result.

2. The method according to claim 1, characterized in that, The multi-source monitoring data for geological disasters includes multiple monitoring data points, and S1 includes: S11. The multi-source geological disaster monitoring data is time-stamp aligned and spatial coordinates are unified to obtain standardized geological disaster monitoring data; S12. Extract time-series features from the standardized geological disaster monitoring data to obtain geological disaster time-series feature data; S13. Extract spatial distribution features from the standardized geological disaster monitoring data to obtain spatial distribution feature data of geological disasters; S14. Perform spatiotemporal feature fusion on the temporal feature data and the spatial distribution feature data of the geological disaster to obtain the spatiotemporal feature data of the geological disaster.

3. The method according to claim 2, characterized in that, S2 includes: S21. Based on the spatiotemporal characteristic data of the geological disaster, calculate the spatiotemporal distance between any two monitoring data points, and generate spatiotemporal distance matrix data based on the spatiotemporal distance; S22. Based on the spatiotemporal distance matrix data, perform cluster analysis on the spatiotemporal characteristic data of geological disasters to obtain disaster event clustering result data; the disaster event clustering result data includes multiple clusters; S23. Based on the disaster event clustering results data, determine the causal direction of the spatiotemporal characteristic data of the geological disasters in each cluster to obtain disaster causal direction data; S24. Based on the disaster event clustering results data and the disaster causal direction data, construct the disaster correlation topology structure to obtain the disaster chain correlation data.

4. The method according to claim 3, characterized in that, S3 includes: S31. Based on the disaster event clustering result data, perform feature aggregation on the spatiotemporal feature data of the geological disasters within each cluster to obtain disaster entity node attribute data; S32. Based on the disaster entity node attribute data and the disaster causal direction data, the directed edges in the disaster chain association data are labeled with relationship types to obtain disaster association data; wherein, the disaster association data is labeled with relationship types. S33. Based on the spatiotemporal characteristic data of the geological disaster, the spatiotemporal attributes of each disaster entity node are calculated to obtain the spatiotemporal attribute data of the disaster entity node; wherein, the spatiotemporal attribute data of the disaster entity node includes time center attribute and spatial center attribute; S34. Based on the disaster entity node attribute data, the disaster association data, and the disaster entity node spatiotemporal attribute data, the geological disaster chain knowledge graph data is constructed.

5. The method according to claim 4, characterized in that, S4 includes: S41. Based on the time center attribute, calculate the time interval between any two disaster entity nodes to obtain disaster entity node time interval data; S42. Based on the spatial center attribute, calculate the spatial distance between any two disaster entity nodes to obtain disaster entity node spatial distance data; S43. Based on the disaster entity node attribute data, calculate the attribute similarity between any two disaster entity nodes to obtain disaster entity node attribute similarity data; S44. Based on the disaster entity node time interval data, the disaster entity node spatial distance data, and the disaster entity node attribute similarity data, calculate the spatiotemporal association strength of the relationship edges in the disaster chain association data to obtain the spatiotemporal association strength data between each disaster entity node; wherein, the expression for the spatiotemporal association strength is: ； in, Indicates the first The first disaster entity node and the first The spatiotemporal correlation strength between individual disaster entity nodes This indicates the preset time decay scale parameter. This indicates the preset spatial attenuation scale parameter. Indicates the first The first disaster entity node and the first Attribute similarity between disaster entity nodes Indicates the first The first disaster entity node and the first The time interval between disaster entity nodes Indicates the first The first disaster entity node and the first Spatial distance between disaster entity nodes.

6. The method according to claim 1, characterized in that, S5 includes: S51. Based on the spatiotemporal correlation strength data and the correlation strength threshold, perform threshold filtering on each of the relation edges in the disaster chain correlation data to obtain candidate propagation edge data; S52. Based on the candidate propagation edge data, perform in-degree statistics on the disaster entity nodes in the geological disaster chain knowledge graph data, and select disaster entity nodes with zero in-degree as the starting point of the propagation path to obtain the propagation path starting point set data. S53. Based on the propagation path starting point set data and the candidate propagation edge data, starting from each of the propagation path starting point sets in the propagation path starting point set data, traverse layer by layer along the direction of the directed edges in the candidate propagation edge data, record the sequence of disaster entity nodes and the relation edges passed during the traversal, and terminate the traversal when the out-degree of the current disaster entity node is zero or the spatiotemporal correlation strength of all downstream nodes corresponding to the out-degree is less than the correlation strength threshold, thereby obtaining the candidate propagation path data. S54. Based on the candidate propagation path data, calculate the cumulative path intensity of each candidate propagation path to obtain candidate propagation path intensity data; wherein, the expression for the cumulative path intensity is: ； In the formula, Indicates the first The cumulative strength of each candidate propagation path. Indicates the first The set of all relation edges contained in the candidate propagation path. Indicates the first From the candidate propagation path, from the first The disaster entity node points to the first The relationship edges between disaster entity nodes Indicates the first The first disaster entity node and the first The spatiotemporal correlation strength between individual disaster entity nodes; S55. Based on the candidate propagation path intensity data, sort the candidate propagation paths to obtain the geological disaster chain propagation path identification results.

7. A geological disaster chain propagation path identification system based on spatiotemporal correlation, characterized in that, The system includes: The multi-source monitoring data acquisition and spatiotemporal feature extraction module is used to acquire multi-source monitoring data of geological disasters and extract spatiotemporal features from the multi-source monitoring data of geological disasters to obtain spatiotemporal feature data of geological disasters. The disaster correlation extraction module is used to extract disaster correlations from the spatiotemporal feature data of geological disasters to obtain disaster chain correlation data. A geological hazard chain knowledge graph construction module is used to construct a knowledge graph from the hazard chain relationship data and the spatiotemporal feature data of geological hazards, thereby obtaining geological hazard chain knowledge graph data; wherein, the geological hazard chain knowledge graph data includes hazard entity nodes, relationship edges, and spatiotemporal attribute annotations; The spatiotemporal correlation strength calculation module is used to calculate the spatiotemporal correlation strength between each pair of disaster entity nodes based on the spatiotemporal attribute annotations of the disaster entity nodes in the geological disaster chain knowledge graph data, and to obtain the spatiotemporal correlation strength data between each pair of disaster entity nodes. The geological disaster chain propagation path identification module is used to start from the disaster entity node in the geological disaster chain knowledge graph data, and based on the spatiotemporal correlation strength data, traverse the correlation edge that satisfies the correlation strength threshold and the disaster entity node connected to it along the direction of the correlation edge, extract the geological disaster chain propagation path composed of the disaster entity node that meets the condition and the correlation edge, and obtain the geological disaster chain propagation path identification result.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the method of any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the method of any one of claims 1 to 6.

Images