Geographical event knowledge graph alignment method, device, equipment and medium

Abstract

The present application belongs to the technical field of geographic information, and provides a geographic event knowledge graph alignment method, device, equipment and medium, wherein the method comprises: constructing a plurality of scale spatial neighborhood windows, and acquiring multi-source heterogeneous data of the spatial neighborhood windows; dividing the spatial neighborhood windows into spatial semantic units; aggregating and classifying multi-source attribute data of the spatial semantic units to obtain attribute vectors and structured labels of the spatial semantic units, and then constructing a multi-scale scene semantic structure graph; encoding the multi-scale scene semantic structure graph, performing multi-layer aggregation and updating on the multi-scale scene semantic structure graph through the encoding result to obtain scene semantic feature representation; and performing similarity clustering on the scene semantic feature representation to obtain a knowledge graph alignment result. The present application has the beneficial effect of improving the accuracy of recognition of geographic disasters.

Description

Technical Field

[0001] This invention relates to the field of geographic information technology, and in particular to a method, apparatus, device, and medium for aligning geographic event knowledge graphs. Background Technology

[0002] In existing technologies, knowledge graph entity alignment has become one of the key technologies for solving the problem of heterogeneous semantic fusion. Its main goal is to identify semantically equivalent or highly related entity pairs in two or more graphs to achieve cross-source data integration and enhanced reasoning. With the continuous growth of multi-source heterogeneous data, the importance of entity alignment is becoming increasingly prominent, especially in geospatial and disaster scenarios, where accurate fusion of cross-source information is the foundation for complex relationship modeling and knowledge discovery.

[0003] However, existing methods still face significant challenges when applying knowledge graph alignment techniques to geographic event scenarios. Many knowledge graph embedding models, such as SE-KGE, directly encode the geographic coordinates or boundary information of entities into the embedding space to improve spatial reasoning capabilities. However, such methods rely solely on the geometric relationships between point coordinates and do not characterize the complex spatial environmental features of the landslide, such as slope, land use distribution, fault structure, and vegetation cover. Therefore, while coordinates can express location, they cannot reflect the semantic contextual information of the event.

[0004] Current research attempts to align entities with spatiotemporal characteristics in knowledge graphs. However, existing technologies have several drawbacks: they lack the ability to structurally model geographic semantic scenes; they cannot identify spatially dispersed but semantically similar landslide event pairs; and they lack the ability to transition from spatial alignment to cognitive modeling. Summary of the Invention

[0005] Aimed at at least in solving one of the technical problems existing in the prior art, the present invention provides a method, apparatus, device and medium for aligning geographic event knowledge graphs, thereby improving the accuracy of geographic disaster identification.

[0006] One aspect of the present invention provides a method for aligning geographical event knowledge graphs, comprising:

[0007] Based on the target geographic event, construct spatial neighborhood windows at multiple scales and acquire multi-source heterogeneous data from the spatial neighborhood windows;

[0008] Based on the digital elevation data of the spatial neighborhood window, the spatial neighborhood window is divided into spatial semantic units;

[0009] The multi-source attribute data of spatial semantic units are aggregated and classified to obtain the attribute vectors and structured labels of spatial semantic units;

[0010] Using the center point of the target geographic event as the anchor node, a multi-scale scene semantic structure diagram is determined based on the anchor node, the spatial semantic units adjacent to the anchor node, and the attribute vectors of the spatial semantic units.

[0011] The multi-scale scene semantic structure graph is encoded, and the encoding result is used to perform multi-level aggregation and update of the multi-scale scene semantic structure graph to obtain the scene semantic feature representation.

[0012] Similarity clustering is performed on the semantic feature representation of the scene to obtain the knowledge graph alignment result.

[0013] According to the aforementioned geographic event knowledge graph alignment method, multiple scale spatial neighborhood windows are constructed based on the target geographic event, and multi-source heterogeneous data of the spatial neighborhood windows are acquired, including:

[0014] Multiple spatial neighborhood windows of different scales are constructed with the target geographic event as the center point;

[0015] Multi-source heterogeneous data for each scale of spatial neighborhood window are collected and integrated, including monitoring data and event factor data.

[0016] According to the aforementioned geographic event knowledge graph alignment method, the spatial neighborhood window is divided into spatial semantic units based on the digital elevation data of the spatial neighborhood window, including:

[0017] Based on digital elevation data, ArcGIS is used to divide the spatial neighborhood window into multiple spatial semantic units, and the boundaries between adjacent spatial semantic units are defined by watersheds or terrain features.

[0018] According to the aforementioned geographic event knowledge graph alignment method, multi-source attribute data of spatial semantic units are aggregated and classified to obtain attribute vectors and structured labels for spatial semantic units, including:

[0019] The multi-source attribute data of each spatial semantic unit are aggregated to obtain the attribute vector of the spatial semantic unit;

[0020] Based on the attribute vectors of spatial semantic units, semantic classification is performed using event expertise and preset threshold rules to obtain structured labels for spatial semantic units.

[0021] According to the aforementioned geographic event knowledge graph alignment method, where the center point of the target geographic event is used as the anchor node, a multi-scale scene semantic structure graph is determined based on the anchor node, the spatial semantic units adjacent to the anchor node, and the attribute vectors of the spatial semantic units, including:

[0022] The spatial semantic unit where the center point of the target geographic event is located is used as the anchor node;

[0023] Using the spatial semantic units adjacent to the anchor node as unit nodes, the topological edges between the anchor node and the unit node are determined. The types of topological edges include grid adjacency, slope aspect continuity, and structural unit connectivity.

[0024] The multi-scale scene semantic structure graph is determined based on the attribute vectors of anchor nodes, unit nodes, topological edges, and spatial semantic units. for:

[0025] ;

[0026] in, It is a set of nodes consisting of anchor nodes and element nodes. For the set of topological edges, It is a set of attribute vectors of spatial semantic units.

[0027] According to the aforementioned geographic event knowledge graph alignment method, the multi-scale scene semantic structure graph is encoded, and the encoding result is used to perform multi-level aggregation and updating of the multi-scale scene semantic structure graph to obtain a scene semantic feature representation, including:

[0028] Based on the attribute types of anchor nodes and unit nodes in the multi-scale scene semantic structure graph, encoding processing is performed to obtain the initial node representation, where the encoding includes:

[0029] The numerical attributes of anchor nodes and unit nodes are normalized to obtain numerical vectors.

[0030] The structured labels of anchor nodes and unit nodes are converted into semantic embedding vectors using a BERT pre-trained model;

[0031] The initial node representation is obtained by concatenating the numerical vector and the semantic embedding vector. ,in It is a numerical vector. It is a semantic embedding vector;

[0032] Based on the initial node representations, the multi-scale scene semantic structure graph is aggregated and updated in multiple layers to obtain the final node representations. The formula for multi-level aggregation and updating of the semantic structure graph of a multi-scale scene is as follows:

[0033] ;

[0034] in, For node vectors, To control the weights of a node's own features and its neighbors' features, Representative node The initial feature vector, Represents a node The set of neighboring nodes.

[0035] According to the aforementioned geographic event knowledge graph alignment method, similarity clustering is performed on the semantic feature representation of the scene to obtain the knowledge graph alignment result, including:

[0036] The final representation of the node Perform weighted average aggregation to obtain graph-level vectors. for:

[0037] ;

[0038] in, As weight, The total number of node vectors;

[0039] The semantic structure graph of a multi-scale scene is identified by a distance threshold and density clustering method to obtain spatially semantically similar events or event clusters, and the knowledge graph alignment results are obtained.

[0040] Obtain event nodes of the same type of scenario from events or event clusters with similar spatial semantics, and then obtain an aligned list of event nodes.

[0041] Another aspect of the present invention provides a geographic event knowledge graph alignment device, comprising:

[0042] The first module is used to construct spatial neighborhood windows of multiple scales based on the target geographic event and to acquire multi-source heterogeneous data of the spatial neighborhood windows;

[0043] The second module is used to divide the spatial neighborhood window into spatial semantic units based on the digital elevation data of the spatial neighborhood window.

[0044] The third module is used to aggregate and classify the multi-source attribute data of spatial semantic units to obtain the attribute vectors and structured labels of spatial semantic units.

[0045] The fourth module is used to determine a multi-scale scene semantic structure diagram based on the center point of the target geographic event as the anchor node, the spatial semantic units adjacent to the anchor node, and the attribute vectors of the spatial semantic units.

[0046] The fifth module is used to encode the multi-scale scene semantic structure graph, and to perform multi-level aggregation and update of the multi-scale scene semantic structure graph through the encoding results to obtain the scene semantic feature representation.

[0047] The sixth module is used to perform similarity clustering on the semantic feature representation of the scene to obtain the knowledge graph alignment result.

[0048] Another aspect of the present invention provides an electronic device, including a processor and a memory;

[0049] The memory is used to store programs;

[0050] The processor executes the program to implement the method as described above.

[0051] This invention also discloses a computer program product or computer program, which includes computer instructions stored in a computer-readable storage medium. A processor of a computer device can read the computer instructions from the computer-readable storage medium and execute the computer instructions, causing the computer device to perform the methods described above.

[0052] The beneficial effects of this invention are as follows: By constructing a graph structure based on spatial semantic units and introducing topological relationships such as adjacency and connectivity, the spatial organization structure and geographic semantics of the environment in which an event is located can be effectively expressed; by setting neighborhood windows at multiple spatial scales, environmental structure information can be extracted at different spatial perception levels, solving the problem of incomplete or oversimplified semantic expression at fixed scales and improving the ability to characterize the heterogeneity of geographic event environments; by obtaining scene structure vectors through graph structure embedding and performing matching and clustering based on structural similarity, the spatial proximity assumption can be broken, supporting the need for cross-regional and non-local event alignment; the spatial semantic structure graph construction and alignment framework does not depend on specific event types and can be extended to scene recognition and merging tasks of various geographic phenomena, showing high application and promotion potential. Attached Figure Description

[0053] Figure 1 This is a schematic diagram of the geographic event knowledge graph alignment process according to an embodiment of the present invention;

[0054] Figure 2 This is a schematic diagram of the spatial neighborhood window construction process according to an embodiment of the present invention;

[0055] Figure 3 This is a schematic diagram of the aggregation and classification process of multi-source attribute data according to an embodiment of the present invention;

[0056] Figure 4 This is a schematic diagram illustrating the construction process of a multi-scale scene semantic structure graph according to an embodiment of the present invention;

[0057] Figure 5 This is a schematic diagram of the scene semantic feature generation process according to an embodiment of the present invention;

[0058] Figure 6 This is a schematic diagram of the scene semantic feature generation process according to an embodiment of the present invention;

[0059] Figure 7This is a schematic diagram of the knowledge graph alignment process according to an embodiment of the present invention;

[0060] Figure 8 This is a flowchart of the knowledge graph alignment method for hidden danger events and landslide events according to an embodiment of the present invention;

[0061] Figure 9 This is a schematic diagram of a geographic event knowledge graph alignment device according to an embodiment of the present invention. Detailed Implementation

[0062] The embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings. Throughout the description, the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions. In the following description, suffixes such as "module," "part," or "unit" used to denote elements are used only for the purpose of illustrative purposes and have no specific meaning in themselves. Therefore, "module," "part," or "unit" can be used interchangeably. Terms such as "first," "second," etc., are used only to distinguish technical features and should not be construed as indicating or implying relative importance, or implicitly indicating the number of indicated technical features, or implicitly indicating the sequential relationship of the indicated technical features. In the following description, the consecutive reference numerals for method steps are for ease of review and understanding. Adjusting the implementation order of steps, in conjunction with the overall technical solution of the present invention and the logical relationship between the various steps, will not affect the technical effect achieved by the technical solution of the present invention. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0063] refer to Figure 1 , Figure 1 This is a schematic flowchart of the geographic event knowledge graph alignment method according to an embodiment of the present invention, which includes, but is not limited to, steps S100~S600:

[0064] S100 constructs spatial neighborhood windows at multiple scales based on the target geographic event and acquires multi-source heterogeneous data from these spatial neighborhood windows.

[0065] In some embodiments, reference Figure 2 The schematic diagram of the spatial neighborhood window construction process shown includes, but is not limited to, steps S110 to S120:

[0066] S110 constructs multiple spatial neighborhood windows of different scales, with the target geographic event as the center point;

[0067] S120 collects and integrates multi-source heterogeneous data for each scale's spatial neighborhood window, including monitoring data and event factor data.

[0068] In some embodiments, taking a landslide event as an example, multiple spatial neighborhood windows of different scales are constructed centered on each landslide disaster point or landslide hazard point. For example, concentric circle regions with radii of 60m, 180m, 600m, and 1km are used as spatial neighborhood windows, which correspond to the spatial environmental impact range of small, medium, large, and giant landslides, respectively, to obtain the environmental context of the event at different scales.

[0069] In some embodiments, multi-source heterogeneous geographic data are collected and integrated within each spatial neighborhood window, including monitoring data such as rainfall and displacement, and event factor data such as geomorphic zoning, earthquakes, DEM (digital elevation data), and land use types.

[0070] It should be noted that the event factor data is used to characterize data that cause geographical events (such as landslides and flash floods).

[0071] S200 divides the spatial neighborhood window into spatial semantic units based on the digital elevation data of the spatial neighborhood window.

[0072] In some embodiments, based on digital elevation data (DEM), ArcGIS is used to divide the spatial neighborhood window into multiple spatial semantic units, and the boundaries between adjacent spatial semantic units are defined by watersheds or terrain features.

[0073] For example, taking a landslide event as an example, the DEM data within the neighborhood window is segmented into slope units using the terrain analysis tools of ArcGIS to obtain multiple natural slope units covering the neighborhood. These natural slope units are spatial semantic units that represent landforms.

[0074] S300 aggregates and classifies the multi-source attribute data of spatial semantic units to obtain the attribute vectors and structured labels of spatial semantic units.

[0075] In some embodiments, reference Figure 3 The diagram shown illustrates the aggregation and classification process for multi-source attribute data, including but not limited to steps S310-S320:

[0076] S310, aggregate the multi-source attribute data of each spatial semantic unit to obtain the attribute vector of the spatial semantic unit;

[0077] S320: Based on the attribute vector of the spatial semantic unit, semantic classification is performed using event expertise and preset threshold rules to obtain the structured label of the spatial semantic unit.

[0078] In some implementations, multi-source attribute data is aggregated for each slope unit (spatial semantic unit). The multi-source attribute data includes average slope, average aspect, mean NDVI, land use type ratio, fault density, geological unit type, etc. Finally, an attribute vector of the slope unit (spatial semantic unit) is formed, and the attribute vector of the slope unit (spatial semantic unit) serves as the node feature of subsequent nodes (spatial semantic units).

[0079] In some embodiments, landslide disaster expertise and existing threshold rules are used to semantically classify slope units based on their attributes to obtain structured labels for slope units (spatial semantic units). These structured labels include descriptions such as steep slope woodland, gentle slope farmland, fault edge zones, and collapse gullies. The structured labels enhance the semantic expressive power of subsequent nodes.

[0080] S400 uses the center point of the target geographic event as the anchor node, and determines the multi-scale scene semantic structure diagram based on the anchor node, the spatial semantic units adjacent to the anchor node, and the attribute vectors of the spatial semantic units.

[0081] In some embodiments, reference Figure 4 The diagram shown illustrates the construction process of the multi-scale scene semantic structure graph, which includes, but is not limited to, steps S410~S430:

[0082] S410 uses the spatial semantic unit where the center point of the target geographic event is located as the anchor node;

[0083] S420, using the spatial semantic units adjacent to the anchor node as unit nodes, determines the topological edge between the anchor node and the unit node, where the types of topological edges include grid adjacency, slope aspect continuity and structural unit connectivity.

[0084] S430 determines the multi-scale scene semantic structure graph based on the attribute vectors of anchor nodes, unit nodes, topological edges, and spatial semantic units.

[0085] In some embodiments, multi-scale scene semantic structure graph for:

[0086] ;

[0087] in, It is a set of nodes consisting of anchor nodes and element nodes. For the set of topological edges, It is a set of attribute vectors of spatial semantic units.

[0088] It should be noted that the attribute vector of the spatial semantic unit is represented by the node features in the node set, which represents the multi-source attribute data after clustering as shown in the above embodiment.

[0089] S500 encodes the multi-scale scene semantic structure graph, and uses the encoding results to perform multi-layer aggregation and update of the multi-scale scene semantic structure graph to obtain the scene semantic feature representation.

[0090] In some embodiments, reference Figure 5 The schematic diagram of the scene semantic feature generation process shown includes, but is not limited to, steps S510-S520:

[0091] S510, based on the attribute types of anchor nodes and unit nodes in the multi-scale scene semantic structure graph, perform encoding processing to obtain the initial representation of the nodes;

[0092] In some embodiments, the encoding includes: normalizing the numerical attributes of anchor nodes and unit nodes to obtain numerical vectors; converting the structured labels of anchor nodes and unit nodes into semantic embedding vectors using a BERT pre-trained model; and concatenating the numerical vectors and semantic embedding vectors to obtain the initial node representation. ,in It is a numerical vector. This is a semantic embedding vector.

[0093] S520: Based on the initial node representation, multi-level aggregation and updating of the multi-scale scene semantic structure graph are performed to obtain the final node representation;

[0094] In some embodiments, a node is ultimately represented as The formula for multi-level aggregation and updating of the semantic structure graph of a multi-scale scene is as follows:

[0095] ;

[0096] in, For node vectors, To control the weights of a node's own features and those of its neighbors; Representative node The initial feature vector, the numerical attributes of the nodes, and the semantic label information; Represents a node The set of neighboring nodes, that is, the set of nodes with which the node is located. A set of nodes connected by topological edges in a multi-scale scene semantic structure graph.

[0097] In some embodiments, the scene semantic feature vector embedding is performed by the above formula, the scene semantic structure graph is aggregated and updated in multiple layers, node attributes and topology information are fused, and a fixed-dimensional embedding vector of the event graph as a whole is generated, which accurately represents the geographic scene semantic features of physical events.

[0098] S600 performs similarity clustering on the semantic feature representation of the scene to obtain the knowledge graph alignment result.

[0099] In some embodiments, reference Figure 6 The diagram showing the image recognition and alignment process includes, but is not limited to, steps S610 to S630:

[0100] S610, perform weighted average aggregation on the final node representations to obtain graph-level vectors;

[0101] S620 uses a distance threshold and density clustering method to identify multi-scale scene semantic structure graphs, obtain spatially semantically similar events or event clusters, and obtain knowledge graph alignment results;

[0102] S630: Obtain event nodes of the same type of scenario from events or event clusters with similar spatial semantics, and then obtain an aligned list of event nodes.

[0103] Among them, graph-level vectors for:

[0104] ;

[0105] in, As weight, This represents the total number of node vectors.

[0106] It should be noted that the graph-level vector implemented in this application is used to integrate the scattered node features in the graph into a unified vector representation.

[0107] In some embodiments, the similarity of the embedding vectors of all landslide disaster event maps and landslide hazard event maps is calculated, and spatially semantically similar event pairs or event clusters are identified based on the distance threshold and density clustering (DBSCAN) method, thereby achieving knowledge graph alignment between different data sources and different types of landslide events.

[0108] In some embodiments, based on the event nodes corresponding to the vectors in the cluster, event nodes belonging to the same type of scenario are extracted, and an aligned list of event node IDs is output.

[0109] In some implementations, refer to Figure 7 and Figure 8 ,in Figure 7 This is a schematic diagram of a geographic event knowledge graph alignment model. Figure 8 A flowchart illustrating the method for aligning knowledge graphs of potential hazard events and landslide events, combined with... Figure 7 and Figure 8 The process is as follows:

[0110] (1) Obtain data on potential hazards and landslide events.

[0111] (2) Referring to the definition of landslide event scale, the event neighborhood is divided into concentric circle areas of 60m, 180m, 600m and 1km.

[0112] (3) Use DEM data to construct slope units and aggregate multi-factor attributes as spatial semantic entities (spatial semantic units) to carry events, that is, obtain landslide spatial attribute data and event spatiotemporal location.

[0113] (4) Based on monitoring data such as rainfall and displacement, and (event) factor data such as geomorphic zoning, earthquake, DEM, and land use type, slope units are constructed using DEM data and multi-factor attributes are aggregated as spatial semantic entities carrying events; then, spatial overlay analysis is used to construct basic spatial topological relationships between entities, hydrological connections are modeled based on DEM, and other semantic relationships are modeled based on prior mechanism knowledge.

[0114] (5) Construct an event subgraph in accordance with the above embodiments. It should be noted that the event subgraph is a multi-scale scene semantic structure graph constructed at different scales.

[0115] (6) For each subgraph, embedding calculation of entity and event subgraphs, vector clustering and event subgraph identification of similar scenarios are performed to obtain the aligned hidden danger-landslide event map.

[0116] In some embodiments, examples of the output results are shown in Table 1.

[0117] Table 1. Examples of some output results

[0118]

[0119] In Table 1, the knowledge graph alignment of event cluster IDs C-001 and C-002 under different event and scene entity types shows a high degree of similarity, proving that the technical solution of the present invention achieves a good geographic event mining effect.

[0120] Figure 9 This is a schematic diagram of a geographic event knowledge graph alignment device according to an embodiment of the present invention. The device includes a first module 910, a second module 920, a third module 930, a fourth module 940, a fifth module 950, and a sixth module 960.

[0121] The system comprises six modules: First, it constructs spatial neighborhood windows at multiple scales based on the target geographic event and acquires multi-source heterogeneous data for these windows. Second, it divides the spatial neighborhood windows into spatial semantic units based on their digital elevation data. Third, it aggregates and classifies the multi-source attribute data of these spatial semantic units to obtain their attribute vectors and structured labels. Fourth, it uses the center point of the target geographic event as an anchor node and determines a multi-scale scene semantic structure graph based on the anchor node, the adjacent spatial semantic units, and their attribute vectors. Fifth, it encodes the multi-scale scene semantic structure graph and performs multi-level aggregation and updates based on the encoding results to obtain a scene semantic feature representation. Sixth, it performs similarity clustering on the scene semantic feature representation to obtain knowledge graph alignment results.

[0122] For example, with the cooperation of the first, second, third, fourth, fifth, and sixth modules in the device, the embodiment device can implement any of the aforementioned geographic event knowledge graph alignment methods, namely, constructing spatial neighborhood windows at multiple scales based on the target geographic event and acquiring multi-source heterogeneous data of the spatial neighborhood windows; dividing the spatial neighborhood windows into spatial semantic units based on the digital elevation data of the spatial neighborhood windows; aggregating and classifying the multi-source attribute data of the spatial semantic units to obtain the attribute vectors and structured labels of the spatial semantic units; using the center point of the target geographic event as the anchor node, determining a multi-scale scene semantic structure graph based on the anchor node, the spatial semantic units adjacent to the anchor node, and the attribute vectors of the spatial semantic units; encoding the multi-scale scene semantic structure graph, and performing multi-layer aggregation and updating of the multi-scale scene semantic structure graph through the encoding results to obtain a scene semantic feature representation; and performing similarity clustering on the scene semantic feature representation to obtain the knowledge graph alignment result. The beneficial effects of this invention are as follows: By constructing a graph structure based on spatial semantic units and introducing topological relationships such as adjacency and connectivity, the spatial organization structure and geographic semantics of the environment in which an event is located can be effectively expressed; by setting neighborhood windows at multiple spatial scales, environmental structure information can be extracted at different spatial perception levels, solving the problem of incomplete or oversimplified semantic expression at fixed scales and improving the ability to characterize the heterogeneity of geographic event environments; by obtaining scene structure vectors through graph structure embedding and performing matching and clustering based on structural similarity, the spatial proximity assumption can be broken, supporting the need for cross-regional and non-local event alignment; the spatial semantic structure graph construction and alignment framework does not depend on specific event types and can be extended to scene recognition and merging tasks of various geographic phenomena, showing high application and promotion potential.

[0123] This invention also provides an electronic device, which includes a processor and a memory;

[0124] The memory stores the program;

[0125] The processor executes a program to perform the aforementioned geographic event knowledge graph alignment method; the electronic device has the function of carrying and running the software system for geographic event knowledge graph alignment provided in the embodiments of the present invention, such as a personal computer, minicomputer, mainframe, workstation, network or distributed computing environment, standalone or integrated computer platform, or communicating with charged particle tools or other imaging devices, etc.

[0126] This invention also provides a computer-readable storage medium storing a program that is executed by a processor to implement the geographic event knowledge graph alignment method described above.

[0127] In some alternative embodiments, the functions / operations mentioned in the block diagrams may not occur in the order shown in the operation diagrams. For example, depending on the functions / operations involved, two consecutively shown blocks may actually be executed substantially simultaneously, or the blocks may sometimes be executed in reverse order. Furthermore, the embodiments presented and described in the flowcharts of this invention are provided by way of example to provide a more comprehensive understanding of the technology. The disclosed methods are not limited to the operations and logic flows presented in the embodiments of this invention. Alternative embodiments are contemplated, in which the order of various operations is changed and sub-operations described as part of a larger operation are executed independently.

[0128] This invention also discloses a computer program product or computer program, which includes computer instructions stored in a computer-readable storage medium. A processor of a computer device can read the computer instructions from the computer-readable storage medium and execute the computer instructions, causing the computer device to perform the aforementioned geographic event knowledge graph alignment method.

[0129] Furthermore, although the invention has been described in the context of functional modules, it should be understood that, unless otherwise stated, one or more of the described functions and / or features may be integrated into a single physical device and / or software module, or one or more functions and / or features may be implemented in a separate physical device or software module. It is also understood that a detailed discussion of the actual implementation of each module is unnecessary for understanding the invention. Rather, considering the properties, functions, and internal relationships of the various functional modules in the apparatus disclosed in the embodiments of the invention, the actual implementation of the module will be understood within the scope of conventional skill of an engineer. Therefore, those skilled in the art can implement the invention as set forth in the claims using ordinary techniques without excessive experimentation. It is also understood that the specific concepts disclosed are merely illustrative and are not intended to limit the scope of the invention, which is determined by the full scope of the appended claims and their equivalents.

[0130] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, essentially, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0131] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can include, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0132] More specific examples of computer-readable media (a non-exhaustive list) include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which the program can be printed, because the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.

[0133] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0134] In the description of this specification, references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0135] Although embodiments of the invention have been shown and described, those skilled in the art will understand that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the claims and their equivalents.

[0136] The above is a detailed description of the preferred embodiments of the present invention, but the present invention is not limited to the embodiments described. Those skilled in the art can make various equivalent modifications or substitutions without departing from the spirit of the present invention, and these equivalent modifications or substitutions are all included within the scope defined by the claims of this application.

Claims

1. A method for aligning geographical event knowledge graphs, characterized in that, include: Based on the target geographic event, construct spatial neighborhood windows at multiple scales and acquire multi-source heterogeneous data from the spatial neighborhood windows; Based on the digital elevation data of the spatial neighborhood window, the spatial neighborhood window is divided into spatial semantic units; The multi-source attribute data of spatial semantic units are aggregated and classified to obtain the attribute vectors and structured labels of spatial semantic units; Using the center point of the target geographic event as the anchor node, a multi-scale scene semantic structure diagram is determined based on the anchor node, the spatial semantic units adjacent to the anchor node, and the attribute vectors of the spatial semantic units. The multi-scale scene semantic structure graph is encoded, and the encoding result is used to perform multi-level aggregation and update of the multi-scale scene semantic structure graph to obtain the scene semantic feature representation. Similarity clustering is performed on the semantic feature representation of the scene to obtain the knowledge graph alignment result; The process of constructing spatial neighborhood windows at multiple scales based on the target geographic event and acquiring multi-source heterogeneous data for these spatial neighborhood windows includes: Multiple spatial neighborhood windows of different scales are constructed with the target geographic event as the center point; Multi-source heterogeneous data for each scale of spatial neighborhood window are collected and integrated, including monitoring data and event factor data; The target geographic event is a landslide event; the monitoring data includes rainfall and displacement data, and the event factor data includes geomorphic zoning, earthquakes, DEM and land use types; the multi-source attribute data includes average slope, average aspect, mean NDVI value, land use type ratio, fault density and geological unit type. The process of determining a multi-scale scene semantic structure graph, using the center point of the target geographic event as the anchor node and based on the anchor node, the spatial semantic units adjacent to the anchor node, and the attribute vectors of the spatial semantic units, includes: The spatial semantic unit where the center point of the target geographic event is located is used as the anchor node; Using the spatial semantic units adjacent to the anchor node as unit nodes, the topological edges between the anchor node and the unit node are determined. The types of topological edges include grid adjacency, slope aspect continuity, and structural unit connectivity. The multi-scale scene semantic structure graph is determined based on the attribute vectors of anchor nodes, unit nodes, topological edges, and spatial semantic units. for: in, It is a set of nodes consisting of anchor nodes and element nodes. For the set of topological edges, A set of attribute vectors for spatial semantic units; The process of encoding the multi-scale scene semantic structure graph, and then performing multi-layer aggregation and updating of the multi-scale scene semantic structure graph based on the encoding results to obtain a scene semantic feature representation includes: Based on the attribute types of anchor nodes and unit nodes in the multi-scale scene semantic structure graph, encoding processing is performed to obtain the initial node representation, where the encoding includes: The numerical attributes of anchor nodes and unit nodes are normalized to obtain numerical vectors. The structured labels of anchor nodes and unit nodes are converted into semantic embedding vectors using a BERT pre-trained model; The initial node representation is obtained by concatenating the numerical vector and the semantic embedding vector. ,in It is a numerical vector. It is a semantic embedding vector; Based on the initial node representations, the multi-scale scene semantic structure graph is aggregated and updated in multiple layers to obtain the final node representations. The formula for multi-level aggregation and updating of the semantic structure graph of a multi-scale scene is as follows: in, For node vectors, To control the weights of a node's own features and its neighbors' features, Representative node The initial feature vector, Represents a node The set of neighboring nodes.

2. The geographical event knowledge graph alignment method according to claim 1, characterized in that, The step of dividing the spatial neighborhood window into spatial semantic units based on the digital elevation data of the spatial neighborhood window includes: Based on digital elevation data, ArcGIS is used to divide the spatial neighborhood window into multiple spatial semantic units, and the boundaries between adjacent spatial semantic units are defined by watersheds or terrain features.

3. The geographical event knowledge graph alignment method according to claim 1, characterized in that, The aggregation and classification of multi-source attribute data of spatial semantic units to obtain attribute vectors and structured labels for spatial semantic units includes: The multi-source attribute data of each spatial semantic unit are aggregated to obtain the attribute vector of the spatial semantic unit; Based on the attribute vectors of spatial semantic units, semantic classification is performed using event expertise and preset threshold rules to obtain structured labels for spatial semantic units.

4. The geographical event knowledge graph alignment method according to claim 1, characterized in that, The step of performing similarity clustering on the semantic feature representation of the scene to obtain the knowledge graph alignment result includes: The final representation of the node Perform weighted average aggregation to obtain graph-level vectors. for: in, As weight, The total number of node vectors; The semantic structure graph of a multi-scale scene is identified by a distance threshold and density clustering method to obtain spatially semantically similar events or event clusters, and the knowledge graph alignment results are obtained. Obtain event nodes of the same type of scenario from events or event clusters with similar spatial semantics, and then obtain an aligned list of event nodes.

5. A geographic event knowledge graph alignment device, characterized in that, include: The first module is used to construct spatial neighborhood windows of multiple scales based on the target geographic event and to acquire multi-source heterogeneous data of the spatial neighborhood windows; The second module is used to divide the spatial neighborhood window into spatial semantic units based on the digital elevation data of the spatial neighborhood window. The third module is used to aggregate and classify the multi-source attribute data of spatial semantic units to obtain the attribute vectors and structured labels of spatial semantic units. The fourth module is used to determine a multi-scale scene semantic structure diagram based on the center point of the target geographic event as the anchor node, the spatial semantic units adjacent to the anchor node, and the attribute vectors of the spatial semantic units. The fifth module is used to encode the multi-scale scene semantic structure graph, and to perform multi-level aggregation and update of the multi-scale scene semantic structure graph through the encoding results to obtain the scene semantic feature representation. The sixth module is used to perform similarity clustering on the semantic feature representation of the scene to obtain the knowledge graph alignment result; The process of constructing spatial neighborhood windows at multiple scales based on the target geographic event and acquiring multi-source heterogeneous data for these spatial neighborhood windows includes: Multiple spatial neighborhood windows of different scales are constructed with the target geographic event as the center point; Multi-source heterogeneous data for each scale of spatial neighborhood window are collected and integrated, including monitoring data and event factor data; The process of determining a multi-scale scene semantic structure graph, using the center point of the target geographic event as the anchor node and based on the anchor node, the spatial semantic units adjacent to the anchor node, and the attribute vectors of the spatial semantic units, includes: The spatial semantic unit where the center point of the target geographic event is located is used as the anchor node; Using the spatial semantic units adjacent to the anchor node as unit nodes, the topological edges between the anchor node and the unit node are determined. The types of topological edges include grid adjacency, slope aspect continuity, and structural unit connectivity. The target geographic event is a landslide event; the monitoring data includes rainfall and displacement data, and the event factor data includes geomorphic zoning, earthquakes, DEM and land use types; the multi-source attribute data includes average slope, average aspect, mean NDVI value, land use type ratio, fault density and geological unit type. The multi-scale scene semantic structure graph is determined based on the attribute vectors of anchor nodes, unit nodes, topological edges, and spatial semantic units. for: in, It is a set of nodes consisting of anchor nodes and element nodes. For the set of topological edges, A set of attribute vectors for spatial semantic units; The process of encoding the multi-scale scene semantic structure graph, and then performing multi-layer aggregation and updating of the multi-scale scene semantic structure graph based on the encoding results to obtain a scene semantic feature representation includes: Based on the attribute types of anchor nodes and unit nodes in the multi-scale scene semantic structure graph, encoding processing is performed to obtain the initial node representation, where the encoding includes: The numerical attributes of anchor nodes and unit nodes are normalized to obtain numerical vectors. The structured labels of anchor nodes and unit nodes are converted into semantic embedding vectors using a BERT pre-trained model; The initial node representation is obtained by concatenating the numerical vector and the semantic embedding vector. ,in It is a numerical vector. It is a semantic embedding vector; Based on the initial node representations, the multi-scale scene semantic structure graph is aggregated and updated in multiple layers to obtain the final node representations. The formula for multi-level aggregation and updating of the semantic structure graph of a multi-scale scene is as follows: in, For node vectors, To control the weights of a node's own features and its neighbors' features, Representative node The initial feature vector, Represents a node The set of neighboring nodes.

6. An electronic device, characterized in that, Including the processor and memory; The memory is used to store programs; The processor executes the program to implement the geographic event knowledge graph alignment method as described in any one of claims 1-4.

7. A computer-readable storage medium, characterized in that, The storage medium stores a program that is executed by a processor to implement the geographic event knowledge graph alignment method as described in any one of claims 1-4.

Images