Multi-source heterogeneous live-action three-dimensional database construction method, equipment and medium

Abstract

The invention discloses a multi-source heterogeneous live-action three-dimensional database construction method and device and a medium, and relates to the field of database construction. The method comprises the steps of obtaining multi-source heterogeneous data; performing standardization processing on the multi-source heterogeneous data to obtain standardized multi-source heterogeneous data; the method comprises the following steps: constructing an entity classification set, an entity attribute set and an entity relationship set based on a real scene three-dimensional data classification system, and constructing an RDF triple by using a graph convolutional neural network; and constructing a multi-source heterogeneous live-action three-dimensional database based on the standardized multi-source heterogeneous data and the RDF triple. According to the method, the complexity of data conversion is reduced through the preset data classification system and the standardized naming rule, and the automation level of data matching and storage is improved in combination with the knowledge graph driven by priori cognition. Meanwhile, a mixed framework of a relational database, a distributed file system and a graph database is adopted, a multi-stage index structure is combined, efficient storage and rapid retrieval of multi-modal data are supported, and the query time is remarkably shortened.

Description

Technical Field

[0001] This application relates to the field of database construction, and in particular to a method, device and medium for constructing a multi-source heterogeneous real-scene 3D database. Background Technology

[0002] The multi-source nature (such as oblique photogrammetry models, laser point clouds, BIM (Building Information Modeling), remote sensing images, etc.) and heterogeneity (differences in data format, coordinate system, and semantic standards) of real-world 3D data pose serious challenges to data integration, mainly including: (1) Inefficient traditional data management: Existing technologies mostly rely on a single database (such as relational or file systems) to store 3D data, lacking a multimodal hybrid storage architecture, resulting in low efficiency in the storage of massive amounts of data and difficulty in meeting real-time update requirements. (2) Insufficient data quality and consistency: Differences in coordinate systems and inconsistent semantic descriptions of multi-source data can easily lead to data redundancy and logical conflicts, while existing methods rely on manual intervention or simple rule cleaning, making it difficult to achieve accurate coordinate transformation and semantic fusion. (3) Lack of a standardized system: Although some work attempts to standardize the 3D data acquisition process, there is a lack of a standardized framework covering the entire chain of data classification, naming, and storage, creating barriers to cross-regional and cross-industry data sharing and collaborative applications. Summary of the Invention

[0003] The purpose of this application is to provide a method, device and medium for constructing a multi-source heterogeneous real-scene 3D database, so as to realize the rapid parsing and batch storage of multi-source heterogeneous data through an automated process, thereby shortening the processing cycle.

[0004] To achieve the above objectives, this application provides the following solution: Firstly, this application provides a method for constructing a multi-source heterogeneous real-scene 3D database, including: Acquire multi-source heterogeneous data; the multi-source heterogeneous data includes digital elevation models, digital surface models, satellite orthophotos, UAV orthophotos, 3D mesh models, laser point clouds, basic geographic entities, thematic geographic entities, and corresponding metadata, wherein basic geographic entities and thematic geographic entities include entity map metadata, entity attribute tables, and entity relationship tables; The multi-source heterogeneous data is standardized to obtain standardized multi-source heterogeneous data; the standardization process includes coordinate transformation, GeoSOT multi-level partitioning coding, and standardized data naming. Based on the real-scene 3D data classification system, an entity classification set, an entity attribute set, and an entity relationship set are constructed, and an RDF triple is constructed using a graph convolutional neural network; Based on the standardized multi-source heterogeneous data and the RDF triples, a multi-source heterogeneous real-scene 3D database is constructed; the multi-source heterogeneous real-scene 3D database includes a geographic scene database, a geographic entity database, and a metadata database; the geographic scene database includes a digital elevation model library and a digital orthophoto model library; the geographic entity database includes a basic geographic entity database and a thematic geographic entity database; the metadata database includes a geographic scene metadata database and a geographic entity metadata database.

[0005] In one embodiment, the multi-source heterogeneous data is standardized to obtain standardized multi-source heterogeneous data, specifically including: The coordinate system of the multi-source heterogeneous data is checked, and the coordinate transformation is performed on the multi-source heterogeneous data that is inconsistent with the target coordinate system to obtain the transformed multi-source heterogeneous data. The transformed multi-source heterogeneous data is subjected to GeoSOT multi-level partitioning and encoding to obtain encoded multi-source heterogeneous data; The encoded multi-source heterogeneous data is then standardized and named to obtain standardized multi-source heterogeneous data.

[0006] In one embodiment, the coordinate system of the multi-source heterogeneous data is checked, and the coordinate system of the multi-source heterogeneous data that is inconsistent with the target coordinate system is transformed to obtain the transformed multi-source heterogeneous data, specifically including: Determine whether the coordinate system of the multi-source heterogeneous data is consistent with the target coordinate system to obtain a first determination result; If the first judgment result is negative, then it is determined whether the transformation parameters of the multi-source heterogeneous data that are inconsistent with the target coordinate system are known, and a second judgment result is obtained. If the second judgment result is yes, then the transformation parameters are used to perform coordinate transformation on the multi-source heterogeneous data that is inconsistent with the target coordinate system to obtain the transformed multi-source heterogeneous data; If the second judgment result is negative, then extract the coordinates of the common points of the multi-source heterogeneous data with unknown transformation parameters, solve the corresponding transformation parameters using the global least squares method, and use the transformation parameters to perform coordinate transformation on the multi-source heterogeneous data that is inconsistent with the target coordinate system to obtain the transformed multi-source heterogeneous data.

[0007] In one embodiment, a coordinate transformation is performed on multi-source heterogeneous data that is inconsistent with the target coordinate system using transformation parameters to obtain transformed multi-source heterogeneous data, specifically including: When the coordinate system of multi-source heterogeneous data that is inconsistent with the target coordinate system is the geographic coordinate system, the coordinate transformation of the multi-source heterogeneous data that is inconsistent with the target coordinate system is performed using a three-dimensional 7-parameter transformation model according to the transformation parameters, and the transformed multi-source heterogeneous data is obtained. When the coordinate system of multi-source heterogeneous data that is inconsistent with the target coordinate system is the projected coordinate system, a two-dimensional four-parameter transformation model is used to transform the coordinates of the multi-source heterogeneous data that is inconsistent with the target coordinate system according to the transformation parameters, so as to obtain the transformed multi-source heterogeneous data.

[0008] In one embodiment, the real-scene 3D data classification system includes three levels of classification; the first level is divided into geographic scenes and geographic entities; the second level is divided into terrain models, orthophoto models, and 3D models for geographic scenes, and basic geographic entities for geographic entities; the third level is divided into digital elevation models and digital surface models for terrain models, satellite orthophoto models and UAV orthophoto models for orthophoto models, 3D models for 3D mesh models and laser point clouds, basic geographic entities for natural geographic entities, artificial geographic entities, and administrative geographic entities; and thematic geographic entities for geological disaster themes, ecological environment themes, and water conservancy and river and lake themes.

[0009] In one implementation, based on a real-scene 3D data classification system, an entity classification set, an entity attribute set, and an entity relation set are constructed. Then, using a graph convolutional neural network, RDF triples are constructed, specifically including: The entity data in the three-level categories of the real-scene 3D data classification system is refined in multiple levels to obtain several subclasses for each entity. Ontology objects are extracted from the subclasses to construct an entity classification set. The entity data includes natural geographic entities, artificial geographic entities, administrative geographic entities, geological disaster topics, ecological environment topics, and water conservancy and river and lake topics. Based on the entity attribute table, all attributes of each subclass entity are obtained and divided into basic attributes and special attributes to construct an entity attribute set; among them, basic attributes are attributes common to all entities, and special attributes are attributes unique to each entity. Retrieve the names of all relationships from the entity relationship table and construct an entity relationship set; Based on the entity classification set, the entity attribute set, and the entity relation set, RDF triples are determined using a knowledge triple inference model; wherein the knowledge triple inference model is constructed based on a graph convolutional neural network.

[0010] In one embodiment, a multi-source heterogeneous real-scene 3D database is constructed based on the standardized multi-source heterogeneous data and the RDF triples, specifically including: The standardized multi-source heterogeneous data is organized according to the real-scene 3D data classification system to form the pre-database results data; A multi-engine hybrid database architecture is adopted to construct a multi-source heterogeneous real-scene 3D database; the multi-engine hybrid database architecture includes relational database, document database, distributed file system and graph database; The geographic scene data in the pre-entry result data is stored in the distributed file system, and the metadata corresponding to the geographic scene data is stored in the relational database; The map data, attribute data, and metadata corresponding to the basic geographic entities and thematic geographic entities in the pre-database results data are stored in the relational database. The RDF triples are stored in the graph database.

[0011] In one embodiment, a multi-source heterogeneous real-scene 3D database is constructed based on the standardized multi-source heterogeneous data and the RDF triples, and then the process further includes: The newly collected multi-source heterogeneous data is standardized to obtain standardized latest multi-source heterogeneous data. Based on the latest standardized multi-source heterogeneous data, deep learning technology is used for change detection and analysis, incremental updates are performed on the changed areas, and spatial replacement, attribute update and relation update are used for data entry processing to complete the dynamic update of the multi-source heterogeneous real-scene 3D database; among them, the change detection and analysis methods include topological relation comparison method, change extraction method and feature point difference matching method.

[0012] Secondly, this application provides a computer device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the multi-source heterogeneous real-scene 3D database construction method described in any one of the above.

[0013] Thirdly, this application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the multi-source heterogeneous real-scene 3D database construction method described above.

[0014] According to the specific embodiments provided in this application, this application has the following technical effects: This application provides a method, device, and medium for constructing a multi-source heterogeneous real-scene 3D database. The method includes acquiring multi-source heterogeneous data; standardizing the multi-source heterogeneous data to obtain standardized multi-source heterogeneous data; the standardization process includes coordinate transformation, GeoSOT multi-level subdivision coding, and standardized data naming; constructing entity classification sets, entity attribute sets, and entity relation sets based on a real-scene 3D data classification system, and constructing RDF triples using a graph convolutional neural network; and constructing a multi-source heterogeneous real-scene 3D database based on the standardized multi-source heterogeneous data and RDF triples. This application reduces the complexity of data transformation through a pre-defined data classification system and standardized naming rules, and improves the automation level of data matching and database entry by combining prior knowledge-driven knowledge graphs. Simultaneously, it adopts a hybrid architecture of relational databases, distributed file systems, and graph databases, combined with a multi-level index structure, to support efficient storage and fast retrieval of multimodal data, significantly reducing query time. This application achieves rapid parsing and batch database entry of multi-source heterogeneous data through automated processes, shortening the processing cycle. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the 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.

[0016] Figure 1 A flowchart illustrating a method for constructing a multi-source heterogeneous real-scene 3D database according to an embodiment of this application; Figure 2 This is a schematic diagram of the data coordinate transformation process; Figure 3 This is a schematic diagram of the grid coding calculation process; Figure 4 A schematic diagram illustrating the specific process of constructing a knowledge triplet reasoning model based on a graph convolutional neural network model; Figure 5 The intent of the process for creating the mapping relationship table; Figure 6 This is a diagram illustrating the data organization structure. Figure 7 A schematic diagram of the organization of a multi-source heterogeneous real-scene 3D database; Figure 8 Flowchart for updating a multi-source heterogeneous real-scene 3D database; Figure 9 This is a schematic diagram of the structure of a computer device provided in an embodiment of this application. Detailed Implementation

[0017] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0018] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0019] Current construction of real-scene 3D databases faces several technical challenges due to the diverse data sources, complex information, and varied structures: First, the fusion of multi-source heterogeneous data is difficult. Data formats such as tilt models, DEMs (Digital Elevation Models), and DOMs (Digital Orthophoto Maps) differ significantly, with inconsistent coordinate systems and resolutions, leading to low fusion efficiency. Second, the representation of geographic entity relationships is insufficient. Existing databases lack semantic associations and topological descriptions of industry entities (such as roads, buildings, and pipelines), making it difficult to meet the needs of rapid analysis applications. Third, dynamic update capabilities are weak. Traditional construction methods rely on manual intervention, making incremental data updates and dynamic management difficult. Fourth, computational resource consumption is high. During large-scale data storage and retrieval, the lack of optimized multi-source data indexing structures results in low I / O efficiency. Therefore, how to achieve efficient fusion and unified storage of tilt models, DEMs, DOMs, and industry geographic entity relationships; how to reduce the human cost of multi-source data conversion and standardization; and how to improve the dynamic scalability and cross-industry compatibility of the database have become core issues that urgently need to be addressed. This application proposes to integrate multi-source data processing, heterogeneous data fusion, entity relationship mapping, distributed storage and index optimization into a real-scene 3D database system, supporting the efficient and rapid utilization of real-scene 3D data.

[0020] This application proposes a method for constructing a multi-source heterogeneous real-scene 3D database. It aims to address core issues such as low efficiency in integrating multi-source heterogeneous data, poor quality control, and inconsistent management standards. By integrating geographic information science, knowledge graphs, and big data storage technologies, it eliminates heterogeneity interference through standardized data classification and naming rules, enhances data semantic consistency using knowledge graphs, and improves the dynamic management capabilities of multimodal data based on a hybrid storage architecture. Ultimately, it constructs an efficient, high-precision, and scalable real-scene 3D database construction system, solving the shortcomings of traditional methods that prioritize storage over association and format over semantics.

[0021] This application optimizes data entry efficiency through a hybrid storage architecture, ensures data accuracy through knowledge graphs and quality models, and enhances management standardization through a standardized design across the entire process. It achieves efficient integration and in-depth application of multi-source heterogeneous data in the field of 3D geospatial data. Its technical framework not only responds to the Ministry of Natural Resources' policy requirements for real-scene 3D construction but also integrates cutting-edge methods for multi-source data processing, possessing strong industry applicability and practical guidance value.

[0022] The proposed method for constructing a multi-source heterogeneous real-scene 3D database first involves classifying and acquiring multi-source heterogeneous real-scene 3D data, including designing a data classification system and collecting multi-source heterogeneous data. Second, it performs standardization processing of the real-scene 3D data, including coordinate transformation, GeoSOT (Geographic Coordinate Subdividing Grid with One-Dimension Integral Coding on 2n-Tree) multi-level subdivision coding, and standardized data naming. Third, it constructs geographic entities and knowledge graphs based on prior knowledge, including knowledge ontology construction driven by prior knowledge and knowledge extraction based on a graph neural network model. Finally, it adopts a distributed hybrid storage strategy, designs the database structure, and stores spatial data and relational data in a NoSQL database (Not Only SQL) and a graph database, respectively. Simultaneously, it uses deep learning technology to perform change detection analysis, enabling incremental dynamic updates of the real-scene 3D data.

[0023] The method for constructing a multi-source heterogeneous real-scene 3D database in this application includes multi-source data classification, heterogeneous data standardization processing, entity relationship knowledge graph construction, data storage and updating. By establishing a multi-source heterogeneous database construction system, the management and updating effect of real-scene 3D data can be achieved through "standardized database construction and dynamic updating".

[0024] The following explains the concepts of real-world 3D, real-world 3D database, knowledge graph, geographic entities, entity relationships, and resource description framework involved in this application.

[0025] Real-world 3D: A digital space that realistically, three-dimensionally, and temporally reflects and expresses human production, living, and ecological spaces within a certain range; it is a standardized product of new basic surveying and mapping.

[0026] Real-scene 3D database: A warehouse that organizes and stores hierarchical, classified, and time-series real-scene 3D data according to a certain data structure. It is a collection of real-scene 3D data that is stored in a computer for a long time, is organized, can be shared and utilized, and is uniformly managed.

[0027] Knowledge graph: A collection of knowledge elements and their relationships described in a structured form.

[0028] Geographic entity: A geographic object in the real world that occupies a certain and continuous spatial location and range and has the same attribute or complete function.

[0029] Entity relationships: Describes the spatial relationships, attribute relationships, and temporal relationships between entities.

[0030] Resource Description Framework (RDF) is a resource description language.

[0031] In one exemplary embodiment, such as Figure 1 As shown, a method for constructing a multi-source heterogeneous real-scene 3D database is provided, including the following steps: S1: Acquire multi-source heterogeneous data; the multi-source heterogeneous data includes digital elevation models, digital surface models, satellite orthophotos, UAV orthophotos, 3D mesh models, laser point clouds, basic geographic entities, thematic geographic entities, and corresponding metadata, wherein basic geographic entities and thematic geographic entities include entity map metadata, entity attribute tables, and entity relationship tables.

[0032] In practical applications, the first step is to design a data classification system (a real-scene 3D data classification system).

[0033] First, the heterogeneous data from multiple sources, including real-world 3D raster imagery, vector entity data, 3D point cloud data, and OSGB models, are classified into three levels. The first level is divided into geographic scenes and geographic entities. The second level is divided into geographic scenes into terrain models, orthophoto models, and 3D models, and geographic entities into basic geographic entities and thematic geographic entities. The third level is divided into terrain models into digital elevation models and digital surface models, orthophoto models into satellite orthophotos and UAV orthophotos, 3D models into 3D mesh models and laser point clouds, basic geographic entities into mountains, water bodies, snow and ice areas, agricultural and forestry land and other land, water conservancy, transportation, buildings (structures), pipelines, courtyards, administrative division units, place names, territorial spatial planning units, and other management areas. Thematic geographic entities are classified into three levels according to the specific thematic industry characteristics (for example, geological disaster entities can be divided into landslides, collapses, debris flows, ground subsidence, ground collapse, etc.), as shown in Table 1.

[0034] Table 1. Classification System of Real-Scene 3D Data

[0035] Then collect heterogeneous data from multiple sources.

[0036] According to the data classification system, we collected multi-source heterogeneous 3D reality data, mainly including digital elevation models (DEM), digital surface models (DSM), satellite orthophotos (RS-DOM), UAV orthophotos (UVA-DOM), 3D mesh models, laser point clouds, basic geographic entities, thematic geographic entities, and their corresponding metadata. Among them, the basic geographic entity and thematic geographic entity data include entity map metadata, entity attribute tables, and entity relationship tables.

[0037] S2: Standardize the multi-source heterogeneous data to obtain standardized multi-source heterogeneous data; the standardization process includes coordinate transformation, GeoSOT multi-level partitioning coding, and standardized data naming.

[0038] In one embodiment, S2 specifically includes: S21: Check the coordinate system of the multi-source heterogeneous data, and perform coordinate transformation on the multi-source heterogeneous data that is inconsistent with the target coordinate system to obtain the transformed multi-source heterogeneous data.

[0039] In one implementation, such as Figure 2 As shown, S21 specifically includes: S211: Determine whether the coordinate system of the multi-source heterogeneous data is consistent with the target coordinate system, and obtain the first determination result.

[0040] S212: If the first judgment result is negative, then determine whether the transformation parameters of the multi-source heterogeneous data that are inconsistent with the target coordinate system are known, and obtain the second judgment result.

[0041] S213: If the second judgment result is yes, then the coordinate transformation of the multi-source heterogeneous data that is inconsistent with the target coordinate system is performed using the transformation parameters to obtain the transformed multi-source heterogeneous data.

[0042] S214: If the second judgment result is negative, extract the coordinates of the common points of the multi-source heterogeneous data with unknown transformation parameters, solve the corresponding transformation parameters using the overall least squares method, and use the transformation parameters to perform coordinate transformation on the multi-source heterogeneous data that is inconsistent with the target coordinate system to obtain the transformed multi-source heterogeneous data.

[0043] In one embodiment, a coordinate transformation is performed on multi-source heterogeneous data that is inconsistent with the target coordinate system using transformation parameters to obtain transformed multi-source heterogeneous data, specifically including: When the coordinate system of multi-source heterogeneous data that is inconsistent with the target coordinate system is the geographic coordinate system, the coordinate transformation of the multi-source heterogeneous data that is inconsistent with the target coordinate system is carried out using a three-dimensional 7-parameter transformation model according to the transformation parameters, so as to obtain the transformed multi-source heterogeneous data.

[0044] When the coordinate system of multi-source heterogeneous data that is inconsistent with the target coordinate system is the projected coordinate system, a two-dimensional four-parameter transformation model is used to transform the coordinates of the multi-source heterogeneous data that is inconsistent with the target coordinate system according to the transformation parameters, so as to obtain the transformed multi-source heterogeneous data.

[0045] In practical applications, coordinate system checks are performed on the collected multi-source heterogeneous data, and coordinate transformation is performed on data that is inconsistent with the target coordinate system.

[0046] The geographic coordinate system adopts a three-dimensional 7-parameter transformation model, as shown in formula (1), and the projected coordinate system adopts a two-dimensional 4-parameter transformation model, as shown in formula (2). For multi-source heterogeneous data with unknown coordinate transformation parameters, common point coordinates are extracted, and the coordinate transformation parameters are solved using the overall least squares method; at least 3 common point coordinates are extracted for the geographic coordinate system, and at least 2 common point coordinates are extracted for the projected coordinate system.

[0047] (1) In the formula, , and Here, k is the translation parameter, and k is the scale factor. , For Euler rotation angle parameters.

[0048]

[0049] S22: Perform GeoSOT multi-level partitioning and encoding on the transformed multi-source heterogeneous data to obtain encoded multi-source heterogeneous data.

[0050] Based on the transformed multi-source heterogeneous data and GeoSOT global subdivision rules, a hierarchical grid code is generated through recursive subdivision using an octree. This establishes a four-dimensional spatiotemporal index structure for the real-scene 3D data (3D space + UTC timestamp), generating a spatial grid code for the data block index. The specific operation process of GeoSOT multi-level subdivision coding is as follows: Figure 3 As shown.

[0051] First, read the real-world 3D data and calculate the center coordinates; then convert the center coordinates into binary numbers in degrees, minutes, seconds, and fractional seconds, and determine the grid level based on the real-world 3D data; determine the number of bits in the binary number based on the grid level; use Morton cross-coding to generate binary codes, convert them into quaternary codes, add the hemisphere number, and finally output the grid code.

[0052] S23: Perform data standardization naming on the encoded multi-source heterogeneous data to obtain standardized multi-source heterogeneous data.

[0053] In practical applications, the naming rules are extended according to the ISO 19115 metadata standard to generate a composite naming structure containing data type identifier, spatial grid code, spatial resolution, and sensor ID. Standardized data from S21 and spatial grid codes from S22 are obtained. Geographic scenes are named using the format "Data Type + Grid Size / Resolution_Data Current Year_Spatial Grid Code"; geographic entities are named using the format "Entity Type Abbreviation_Geometric Type_Data Production Year_Spatial Grid Code". The specific abbreviations for entity types are shown in Table 2.

[0054] Table 2. List of English abbreviations for entity types

[0055] S3: Based on the real-scene 3D data classification system, construct entity classification set, entity attribute set and entity relationship set, and use graph convolutional neural network to construct RDF triples.

[0056] In one embodiment, S3 specifically includes: S31: The entity data in the three-level categories of the real-scene 3D data classification system is refined in multiple levels to obtain several subclasses of each entity. Ontology objects are extracted from the subclasses to construct an entity classification set. The entity data includes natural geographic entities, artificial geographic entities, administrative geographic entities, geological disaster topics, ecological environment topics, and water conservancy and river and lake topics.

[0057] S32: Obtain all attributes of each subclass entity according to the entity attribute table, and divide them into basic attributes and special attributes to construct an entity attribute set; wherein, basic attributes are attributes common to all entities, and special attributes are attributes unique to each entity.

[0058] S33: Obtain the names of all relationships from the entity relationship table and construct the entity relationship set.

[0059] S34: Based on the entity classification set, the entity attribute set, and the entity relation set, determine the RDF triples using the knowledge triple inference model; wherein the knowledge triple inference model is constructed based on a graph convolutional neural network.

[0060] In practical applications, the first step is to construct a knowledge ontology driven by prior knowledge.

[0061] The knowledge ontology construction comprises three parts: entity classification construction, entity attribute set construction, and entity relation set construction. ① Entity refinement: The entity data in the three-level categories of the real-world 3D data classification system is refined at multiple levels to obtain several subclasses for each entity class, from which ontology objects are extracted. ② Entity attribute set construction: All attributes of each subclass entity are obtained from the geographic entity attribute table, and these are divided into basic attributes and specific attributes. Basic attributes are those common to all entities, while specific attributes are those unique to each entity. ③ Entity relation set construction: The names of all relations are obtained from the geographic entity relation table.

[0062] Then, knowledge extraction is performed based on a graph neural network model.

[0063] A knowledge triplet reasoning model is constructed based on graph convolutional neural networks to achieve joint attribute-relation extraction. The process of constructing the knowledge triplet reasoning model based on graph convolutional neural networks is as follows: Figure 4 As shown, the graph convolutional neural network includes an input layer, a graph convolutional layer, and an output layer. The input is an entity, the graph convolutional layer initializes node features, and the output is attribute triples and relation triples.

[0064] The knowledge extraction process is as follows: ① Define RDF triple description rules: The triple description rule for entity attributes is "<entity, entity attribute item, attribute value>", and the triple description rule for entity relations is "<entity, entity relation, entity>".

[0065] ② Based on the constructed knowledge ontology and the defined RDF triple description rules, a mapping table between various entities and RDF elements is formulated, such as... Figure 5 As shown, knowledge extraction of geographic entities is achieved based on the mapping relationship table.

[0066] S4: Based on the standardized multi-source heterogeneous data and the RDF triples, construct a multi-source heterogeneous real-scene 3D database.

[0067] In one embodiment, S4 specifically includes: S41: Organize the standardized multi-source heterogeneous data according to the real-scene 3D data classification system to form the pre-entry result data.

[0068] S42: A multi-engine hybrid database architecture is adopted to construct a multi-source heterogeneous real-scene 3D database; the multi-engine hybrid database architecture includes relational database, document database, distributed file system and graph database.

[0069] S43: Store the geographic scene data in the pre-entry result data in the distributed file system, and store the metadata corresponding to the geographic scene data in the relational database.

[0070] S44: Store the map data, attribute data, and metadata corresponding to the basic geographic entities and thematic geographic entities in the pre-entry result data in the relational database.

[0071] S45: The RDF triples are stored in the graph database.

[0072] In practical applications, S2's standardized multi-source heterogeneous data is organized according to "first-level category, second-level category, and third-level category" data types to form the final data before formal data entry into the database, such as... Figure 6 As shown.

[0073] A multi-engine hybrid database architecture is adopted, including relational databases (PostgreSQL), document databases (MongoDB), distributed file systems (HDFS), and graph databases (Neo4j), to build a heterogeneous storage cluster. This ensures cross-engine data consistency through a multi-database transaction atomicity guarantee mechanism. The resulting multi-source heterogeneous real-scene 3D database (i.e., a geological disaster real-scene 3D database) is as follows: Figure 7 As shown, the multi-source heterogeneous real-scene 3D database includes a geographic scene database, a geographic entity database, and a metadata database; the geographic scene database includes a digital elevation model library and a digital orthophoto model library; the geographic entity database includes a basic geographic entity database and a thematic geographic entity database; and the metadata database includes a geographic scene metadata database and a geographic entity metadata database.

[0074] In one embodiment, a multi-source heterogeneous real-scene 3D database is constructed based on the standardized multi-source heterogeneous data and the RDF triples, and then the process further includes: The newly collected multi-source heterogeneous data is standardized to obtain standardized latest multi-source heterogeneous data.

[0075] Based on the latest standardized multi-source heterogeneous data, deep learning technology is used for change detection and analysis, incremental updates are performed on the changed areas, and spatial replacement, attribute update and relation update are used for data entry processing to complete the dynamic update of the multi-source heterogeneous real-scene 3D database; among them, the change detection and analysis methods include topological relation comparison method, change extraction method and feature point difference matching method.

[0076] In practical applications, different databases are used for storage based on different data types: ① Geographic scene data is stored in the database. Geographic scene data is organized according to S41, such as DOM image files of different resolutions, DEM / DSM data of different grid sizes, and their corresponding mosaic datasets, which are stored in the file system, while the corresponding metadata is stored in the relational database. ② Geographic entity data is stored in the database. Data on basic geographic entities and thematic geographic entities is organized according to S41, and their corresponding graph metadata, attribute data, and metadata are stored in the relational database. ③ Knowledge graph data is stored in the database. The RDF triple data extracted in step three is stored in the graph database.

[0077] The update process of a multi-source heterogeneous real-scene 3D database is as follows: Figure 8 As shown, based on the latest collected real-scene 3D data, after data standardization processing, deep learning technology is used to carry out change detection and analysis, including using topological relationship comparison method for vector data, change extraction method for raster data, and feature point difference matching method for 3D models. Incremental updates are performed on the discovered change areas, and spatial replacement, attribute update, and relationship update are used for database processing to achieve dynamic updates of the multi-source heterogeneous real-scene 3D database.

[0078] The experimental area covers 4,086 square kilometers. The terrain is characterized by high mountains and deep valleys, and the region is prone to earthquakes and geological disasters. Experimental data was collected using satellite remote sensing orthophotos and high-resolution sub-meter imagery, resulting in real-world 3D data. Based on disaster entities such as landslides, mudslides, and debris flows, a multi-source heterogeneous real-world 3D data database was established.

[0079] Following steps S1-S2 above, satellite remote sensing results and basic geographic entities for the experimental area are acquired. After data cleaning and analysis, coordinate unification and naming standardization are performed on the multi-source heterogeneous data. Then, following step S3, geological hazard entity data such as collapses, landslides, and debris flows are obtained based on prior knowledge of geological hazards. On this basis, following step S4, a multi-source heterogeneous real-scene 3D database of geological hazards in the experimental area is constructed.

[0080] Traditional techniques for constructing real-scene 3D databases typically rely on manual processing and human experience, resulting in low efficiency and poor quality. Compared to traditional methods, a new method for constructing multi-source heterogeneous real-scene 3D databases offers the following significant advantages: (1) In terms of data quality, this application uses S2 and S3 to ensure the accuracy of spatial data through accurate coordinate analysis and transformation techniques, and combines semantic fuzzy matching of knowledge graphs to solve the problem of geometric and attribute mismatch in multi-source data. Through prior knowledge-driven quality control, a geographic entity knowledge graph is constructed, knowledge ontology is introduced, and the semantic consistency and logical relevance of data are enhanced, which is superior to the simple standardization preprocessing of traditional real-scene 3D data.

[0081] (2) Regarding data entry efficiency, this application utilizes a pre-defined data classification system and standardized naming rules in S1-S4 to reduce the complexity of data conversion. Combined with a priori knowledge-driven knowledge graph, it enhances the automation level of data matching and entry. Simultaneously, it employs a hybrid architecture of relational databases, distributed file systems, and graph databases, along with a multi-level index structure, to support efficient storage and rapid retrieval of multimodal data, significantly reducing query time. Compared to traditional manual coding or middleware development, this patent achieves rapid parsing and batch entry of multi-source heterogeneous data through automated processes, shortening the processing cycle.

[0082] (3) In terms of management standardization, unified standards have been established for data classification, naming and storage rules, supporting integrated management of multi-level data at the provincial, municipal and county levels, and meeting the requirements of the Ministry of Natural Resources for the construction of "Real Scene 3D China". This application is based on S4's hybrid database architecture framework, which is compatible with multi-modal data such as vector, raster and point cloud, and changes the limitations of relying on SQL filtering, thus meeting the dynamic management needs of real scene 3D data.

[0083] In one exemplary embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the above-described method for constructing a multi-source heterogeneous real-scene 3D database.

[0084] In one exemplary embodiment, a computer-readable storage medium is provided storing a computer program that, when executed by a processor, implements the above-described method for constructing a multi-source heterogeneous real-scene 3D database.

[0085] In one exemplary embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the above-described method for constructing a multi-source heterogeneous real-scene 3D database.

[0086] In one exemplary embodiment, a computer device is provided, which may be a server or a terminal, and its internal structure diagram may be as follows. Figure 9As shown, the computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operating system and computer programs stored in the non-volatile storage media to run. The I / O interfaces are used for exchanging information between the processor and external devices. The communication interface is used for communicating with external terminals via a network connection. When the computer program is executed by the processor, it implements a method for constructing a multi-source heterogeneous real-scene 3D database.

[0087] Those skilled in the art will understand that Figure 9 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0088] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0089] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM).

[0090] The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0091] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0092] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A method for constructing a multi-source heterogeneous real-scene 3D database, characterized in that, include: Acquire multi-source heterogeneous data; the multi-source heterogeneous data includes digital elevation models, digital surface models, satellite orthophotos, UAV orthophotos, 3D mesh models, laser point clouds, basic geographic entities, thematic geographic entities, and corresponding metadata, wherein basic geographic entities and thematic geographic entities include entity map metadata, entity attribute tables, and entity relationship tables; The multi-source heterogeneous data is standardized to obtain standardized multi-source heterogeneous data; the standardization process includes coordinate transformation, GeoSOT multi-level partitioning coding, and standardized data naming. Based on the real-scene 3D data classification system, an entity classification set, an entity attribute set, and an entity relationship set are constructed, and an RDF triple is constructed using a graph convolutional neural network; Based on the standardized multi-source heterogeneous data and the RDF triples, a multi-source heterogeneous real-scene 3D database is constructed; the multi-source heterogeneous real-scene 3D database includes a geographic scene database, a geographic entity database, and a metadata database; the geographic scene database includes a digital elevation model library and a digital orthophoto model library; the geographic entity database includes a basic geographic entity database and a thematic geographic entity database; the metadata database includes a geographic scene metadata database and a geographic entity metadata database.

2. The method for constructing a multi-source heterogeneous real-scene 3D database according to claim 1, characterized in that, The multi-source heterogeneous data is standardized to obtain standardized multi-source heterogeneous data, specifically including: The coordinate system of the multi-source heterogeneous data is checked, and the coordinate transformation is performed on the multi-source heterogeneous data that is inconsistent with the target coordinate system to obtain the transformed multi-source heterogeneous data. The transformed multi-source heterogeneous data is subjected to GeoSOT multi-level partitioning and encoding to obtain encoded multi-source heterogeneous data; The encoded multi-source heterogeneous data is then standardized and named to obtain standardized multi-source heterogeneous data.

3. The method for constructing a multi-source heterogeneous real-scene 3D database according to claim 2, characterized in that, The coordinate system of the multi-source heterogeneous data is checked, and the coordinate system of the multi-source heterogeneous data that is inconsistent with the target coordinate system is transformed to obtain the transformed multi-source heterogeneous data, specifically including: Determine whether the coordinate system of the multi-source heterogeneous data is consistent with the target coordinate system to obtain a first determination result; If the first judgment result is negative, then it is determined whether the transformation parameters of the multi-source heterogeneous data that are inconsistent with the target coordinate system are known, and a second judgment result is obtained. If the second judgment result is yes, then the transformation parameters are used to perform coordinate transformation on the multi-source heterogeneous data that is inconsistent with the target coordinate system to obtain the transformed multi-source heterogeneous data; If the second judgment result is negative, then extract the coordinates of the common points of the multi-source heterogeneous data with unknown transformation parameters, solve the corresponding transformation parameters using the global least squares method, and use the transformation parameters to perform coordinate transformation on the multi-source heterogeneous data that is inconsistent with the target coordinate system to obtain the transformed multi-source heterogeneous data.

4. The method for constructing a multi-source heterogeneous real-scene 3D database according to claim 3, characterized in that, Using transformation parameters, coordinate transformation is performed on multi-source heterogeneous data that is inconsistent with the target coordinate system to obtain transformed multi-source heterogeneous data, specifically including: When the coordinate system of multi-source heterogeneous data that is inconsistent with the target coordinate system is the geographic coordinate system, the coordinate transformation of the multi-source heterogeneous data that is inconsistent with the target coordinate system is performed using a three-dimensional 7-parameter transformation model according to the transformation parameters, and the transformed multi-source heterogeneous data is obtained. When the coordinate system of multi-source heterogeneous data that is inconsistent with the target coordinate system is the projected coordinate system, a two-dimensional four-parameter transformation model is used to transform the coordinates of the multi-source heterogeneous data that is inconsistent with the target coordinate system according to the transformation parameters, so as to obtain the transformed multi-source heterogeneous data.

5. The method for constructing a multi-source heterogeneous real-scene 3D database according to claim 1, characterized in that, The real-scene 3D data classification system includes three levels of classification: the first level is divided into geographic scenes and geographic entities; the second level is divided into geographic scenes into terrain models, orthophoto models and 3D models, and geographic entities into basic geographic entities and thematic geographic entities. In the three-level categories, terrain models are divided into digital elevation models and digital surface models; orthophoto models are divided into satellite orthophotos and UAV orthophotos; 3D models are divided into 3D mesh models and laser point clouds; basic geographic entities are divided into natural geographic entities, artificial geographic entities, and administrative geographic entities; and thematic geographic entities are divided into geological disaster themes, ecological environment themes, and water conservancy and river and lake themes.

6. The method for constructing a multi-source heterogeneous real-scene 3D database according to claim 5, characterized in that, Based on a real-scene 3D data classification system, an entity classification set, an entity attribute set, and an entity relation set are constructed. Furthermore, an RDF triple is built using a graph convolutional neural network, specifically including: The entity data in the three-level categories of the real-scene 3D data classification system is refined in multiple levels to obtain several subclasses for each entity. Ontology objects are extracted from the subclasses to construct an entity classification set. The entity data includes natural geographic entities, artificial geographic entities, administrative geographic entities, geological disaster topics, ecological environment topics, and water conservancy and river and lake topics. Based on the entity attribute table, all attributes of each subclass entity are obtained and divided into basic attributes and special attributes to construct an entity attribute set; among them, basic attributes are attributes common to all entities, and special attributes are attributes unique to each entity. Retrieve the names of all relationships from the entity relationship table and construct an entity relationship set; Based on the entity classification set, the entity attribute set, and the entity relation set, RDF triples are determined using a knowledge triple inference model; wherein the knowledge triple inference model is constructed based on a graph convolutional neural network.

7. The method for constructing a multi-source heterogeneous real-scene 3D database according to claim 1, characterized in that, Based on the standardized multi-source heterogeneous data and the RDF triples, a multi-source heterogeneous real-scene 3D database is constructed, specifically including: The standardized multi-source heterogeneous data is organized according to the real-scene 3D data classification system to form the pre-database results data; A multi-engine hybrid database architecture is adopted to construct a multi-source heterogeneous real-scene 3D database; the multi-engine hybrid database architecture includes relational database, document database, distributed file system and graph database; The geographic scene data in the pre-entry result data is stored in the distributed file system, and the metadata corresponding to the geographic scene data is stored in the relational database; The map data, attribute data, and metadata corresponding to the basic geographic entities and thematic geographic entities in the pre-database results data are stored in the relational database. The RDF triples are stored in the graph database.

8. The method for constructing a multi-source heterogeneous real-scene 3D database according to claim 1, characterized in that, Based on the standardized multi-source heterogeneous data and the RDF triples, a multi-source heterogeneous real-scene 3D database is constructed, which then includes: The newly collected multi-source heterogeneous data is standardized to obtain standardized latest multi-source heterogeneous data. Based on the latest standardized multi-source heterogeneous data, deep learning technology is used for change detection and analysis, incremental updates are performed on the changed areas, and spatial replacement, attribute update and relation update are used for data entry processing to complete the dynamic update of the multi-source heterogeneous real-scene 3D database; among them, the change detection and analysis methods include topological relation comparison method, change extraction method and feature point difference matching method.

9. A computer device, comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that the processor executes the computer program to implement the method for constructing a multi-source heterogeneous real-scene 3D database according to any one of claims 1-8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the computer program implements the method for constructing a multi-source heterogeneous real-scene 3D database as described in any one of claims 1-8.

Images