Geographic knowledge graph-guided twin modeling method for complex three-dimensional scene

Abstract

The invention discloses a complex three-dimensional scene twinborn modeling method guided by a geographic knowledge map, belongs to the technical field of digital twinborn, and solves the problems that the existing twinborn scene modeling method mostly focuses on single space dimension fusion and does not consider the multi-dimension fusion requirement, such as the synergistic effect of objects, attributes and the whole scene is not considered, and the modeling effect is poor. The model precision is reduced, and the adaptability is insufficient. The method comprises the following steps: constructing a scene-feature-transition complex three-dimensional scene geographic knowledge graph based on obtained dynamic and static data of each geographic scene in a complex three-dimensional twinborn scene physical space, and updating in real time; and carrying out dynamic fusion modeling on the complex three-dimensional scene through progressive guidance layer by layer from an object level, an attribute level to a scene level by adopting a geographic knowledge graph of the complex three-dimensional scene. The method is used for complex three-dimensional scene twin modeling.

Description

Technical Field

[0001] A complex three-dimensional scene twin modeling method guided by a geographic knowledge graph is used for complex three-dimensional scene twin modeling and belongs to the field of digital twin technology. Background Art

[0002] Various three-dimensional scenes such as urban parks and mountain transportation span vast spaces such as underground and underground, indoor and outdoor, and include various entities. Their spatial structures are diverse and constantly reconstructed. At the same time, static geographical features are deeply integrated with dynamically changing human-machine objects to form a complex three-dimensional landscape. With the development of the global economy, various three-dimensional spaces are constantly being developed and expanded, and the management and planning of complex scenes face unprecedented challenges. Digital twins create high-precision virtual mappings of the physical world to achieve accurate characterization and real-time monitoring, providing decision support for the refined management of all elements of three-dimensional scene space. However, the extreme complexity of complex scenes makes the data information in key areas present massive, heterogeneous and dynamic characteristics, which is difficult to be efficiently integrated and expressed, restricting the in-depth application of digital twin technology. Knowledge graphs can formally describe three-dimensional scene entities, concepts and their relationships. They construct a systematic knowledge system from the perspective of graph theory, thereby guiding the fusion of information from different fields.

[0003] As an emerging technology, geographic knowledge graphs have both general knowledge connotations and spatiotemporal characteristics of geographic knowledge. They can integrate semantic and spatial information in a structured way, and have great application potential in the accurate expression, deep understanding, efficient acquisition and intelligent reasoning of three-dimensional scene knowledge. However, when describing complex scene information, existing geographic knowledge graphs often face the dilemma of information isolation and lack of systematicity, making it difficult to fully capture and describe the details of scene modeling. In addition, current graph construction methods often ignore the update mechanism, resulting in data lag, making it difficult to timely reflect the real-time status and changes of the scene, and unable to meet the stringent requirements of the digital twin era for the dynamic and real-time nature of modeling knowledge and data. Therefore, it is urgent to explore the construction of geographic knowledge graphs with adaptive update capabilities to achieve seamless integration and instant reflection of information.

[0004] Combining digital twins with geographic knowledge graphs and building highly realistic twin scenes through systematic knowledge guidance is the key to promoting the refined management of complex cities and other systems. With its powerful spatial relationship analysis and semantic information integration capabilities, geographic knowledge graphs can provide data foundation and knowledge support for the modeling process. However, existing twin scene modeling methods mostly focus on the fusion of a single spatial dimension, without considering multi-dimensional fusion needs, such as the synergy between objects, attributes and the scene as a whole, resulting in problems such as reduced model accuracy and insufficient adaptability. In addition, the traditional modeling method model is solidified, and the modeling process lacks dynamic response, making it difficult to adjust and dynamically build rapidly changing complex scenes in real time.

[0005] In summary, the existing twin scene modeling methods have the following technical problems:

[0006] 1. Existing twin scene modeling methods mostly focus on the fusion of a single spatial dimension, without considering the needs of multi-dimensional fusion, such as the synergy between objects, attributes and the entire scene, resulting in reduced model accuracy and insufficient adaptability;

[0007] 2. The existing twin scene modeling method has a rigid pattern, lacks flexibility in the modeling process, and cannot respond to dynamic changes in real time, resulting in poor results when quickly building complex scenes and difficulty in adapting to real-time modeling needs. Summary of the invention

[0008] The purpose of the present invention is to provide a complex three-dimensional scene twin modeling method guided by a geographic knowledge graph, so as to solve the problems that the existing twin scene modeling methods focus on the fusion of a single spatial dimension and do not consider the needs of multi-dimensional fusion, such as the synergy between objects, attributes and the scene as a whole, resulting in reduced model accuracy and insufficient adaptability.

[0009] In order to achieve the above object, the technical solution adopted by the present invention is:

[0010] A complex three-dimensional scene twin modeling method guided by a geographic knowledge graph includes the following steps:

[0011] Step 1: Based on the acquired dynamic and static data of each geographic scene in the physical space of the complex three-dimensional twin scene, a complex three-dimensional scene geographic knowledge graph of scene-feature-change is constructed and updated in real time;

[0012] Step 2: Use the geographic knowledge graph of complex three-dimensional scenes to guide the dynamic fusion modeling of complex three-dimensional scenes step by step from the object level, attribute level to the scene level.

[0013] Further, the specific steps of step 1 are:

[0014] Step 1.1, a "scene-feature-transition" spatiotemporal collaborative mapping model of a given complex three-dimensional scene, wherein the "scene-feature-transition" spatiotemporal collaborative mapping model includes scene space, feature space and transition space, wherein the scene space includes various geographic scenes, and each geographic scene object in each geographic scene shows different features, which are represented by feature sets in the feature space. The feature space not only describes the properties of each geographic scene object, but is also affected by the transition space. The transition space represents the evolution process of each geographic scene object in the scene space over time, that is, it directly affects the expression of the state and features of each geographic scene in the scene space;

[0015] Step 1.2: Based on the "scene-feature-transition" spatiotemporal collaborative mapping model, the domain ontology of each geographic scene is extracted from the dynamic and static data of the physical space of the complex three-dimensional twin scene obtained to obtain the scene ontology library, where the domain ontology is the pattern layer, including the transition domain corresponding to the transition space, the scene domain corresponding to the scene space and the feature domain corresponding to the feature space, and the semantic association relationship between the transition domain, the scene domain and the feature domain. The static data in the dynamic and static data include the location, shape and size of each geographic scene object contained in the geographic scene recorded in literature, expert knowledge, specifications and standards and spatial databases. The dynamic data includes the temporal evolution and the impact of human activities recorded in video surveillance, sensor data, social media and relational databases. The geographic scene refers to the three-dimensional geographical location or space range, including geographic scene objects and their associations. The geographic scene objects have physical objects and environmental elements. Physical objects include topography, buildings, transportation facilities, vegetation and water bodies. The formula of the domain ontology of each geographic scene is:

[0016] Ont=<CO,SO,FO|Relations|Properties>

[0017] Among them, Ont represents the domain ontology of geographic scene, CO represents the change domain, SO represents the scene domain, FO represents the feature domain, Relations represents the semantic relationship between the change domain, scene domain and feature domain, and Properties represents the attributes;

[0018] Step 1.3: Construct the data layer based on the obtained scene ontology library. The specific steps are as follows:

[0019] Parse structured and unstructured data based on the obtained scene ontology library, and extract knowledge based on given extraction rules after parsing, including entity extraction, attribute extraction, and extraction of association relationships between entities and between entities and attributes. Attributes include basic static attribute data, dynamic attribute data, and hierarchical partition attribute data. Basic static attribute data include geometric attributes, physical attributes, and spatiotemporal attributes. Geometric attributes include location, shape, and size. Physical attributes include material, mass, and motion. Spatiotemporal attributes include occurrence time and spatial distribution. Dynamic attribute data include traffic flow data, environmental monitoring data, and social media dynamics obtained by sensors. Traffic flow data includes the number of real-time vehicles and real-time changing locations. Environmental monitoring data includes air quality and temperature and humidity changes. Social media dynamics include geographic tags of real-time events or activities. Hierarchical partition attribute data includes the function, area, height, and material of each layer.

[0020] After knowledge extraction, the semantic, topological and spatial relationships between entities are sorted out, and the association relationship between entities is constructed. At the same time, the index value corresponding to each geographic scene object is given based on the importance of each geographic scene object and each constraint in the three-dimensional scene constraint, that is, the data layer is obtained, where the semantics is the extracted attribute, and the three-dimensional scene constraint includes scene interaction constraint, spatial distance constraint, indoor and outdoor scene constraint, above and below ground scene constraint and research area scene constraint;

[0021] Step 1.4: After mapping the pattern layer and the data layer, a complex three-dimensional scene geographic knowledge graph of scene-feature-transition is obtained, and the complex three-dimensional scene geographic knowledge graph is updated according to the dynamic data accessed in real time.

[0022] Further, the specific steps of step 2 are:

[0023] Step 2.1: Use the initial or updated complex three-dimensional scene geographic knowledge graph to perform object-level fusion of the dynamic and static data of each geographic scene in the physical space of the complex three-dimensional twin scene. The specific steps are as follows:

[0024] Spatial Matching:

[0025] The target position of the geographic scene object stored in the initial or updated complex three-dimensional scene geographic knowledge graph and the parameters of the three-dimensional rotation matrix are read to calculate the rotation matrix, and the posture data is calculated through the rotation matrix, wherein the posture parameters included in the posture data adjust the direction and angle of the coordinate position mapped to the geographic scene object in the virtual space, so that the posture in the virtual scene is consistent with the real world;

[0026] The formula for the rotation matrix is:

[0027]

[0028] Roll = arctan2(R 32 ,R 33 )

[0029] Yaw = arctan2(R 12 ,R 11 )

[0030] Among them, R represents the 3x3 rotation matrix, Pitch represents the yaw angle rotating around the Z axis, Roll represents the roll angle rotating around the X axis, and Yaw represents the pitch angle rotating around the Y axis. Represents the parameters of the rotation matrix. The first column represents the direction vector of the X axis after rotation, the second column represents the direction vector of the Y axis after rotation, and the third column represents the direction vector of the Z axis after rotation. 11 , R 21 and R 31Respectively represent the direction vector of the X-axis on the X-axis, Y-axis and Z-axis after rotation, R 12 , R 22 and R 32 Respectively represent the direction vector of the Y axis after rotation on the X axis, Y axis and Z axis, R 13 , R 23 and R 33 Respectively indicate that the direction vector of the Z axis after rotation is on the X axis, Y axis and Z axis;

[0031] Based on the target position and posture data of each geographic scene object, a spatial positioning algorithm is used to map the attributes of the geographic scene object to the corresponding coordinate position in the virtual space, thus achieving spatial matching;

[0032] Spatial topological constraints:

[0033] Read the relationship information between geographic scene objects recorded in the initial or updated complex three-dimensional scene geographic knowledge graph, and identify the geographic scene object pairs that need to establish topological constraints;

[0034] Then, the spatial topological relationship of the geographic scene object model is obtained after matching each pair of spaces according to the selected topological semantics, that is, the three-dimensional geographic scene model after object-level fusion is obtained, wherein the spatial topological relationship includes the adjacent spatial topological relationship, the separated spatial topological relationship and the included spatial topological relationship of the geographic scene object model;

[0035] The formula for the adjacent spatial topological relationship is:

[0036]

[0037] Among them, Adj(α,β) means that the geographic scene object model α is adjacent to the geographic scene object model β. represents the boundary of the geographic scene object model, α° and β° represent the interior of the geographic scene object model α and the geographic scene object model β respectively, represents the empty set;

[0038]

[0039] Where Dis(α,β) represents the separation of geographic scene object model α and geographic scene object model β;

[0040]

[0041] Among them, Con(α,β) means that the geographic scene object model α completely contains the geographic scene object model β;

[0042] T(α,β)=Adj(α,β)+Dis(α,β)+Con(α,β)

[0043] Wherein, T(α,β) represents the spatial topological relationship between geographic scene object model α and geographic scene object model β;

[0044] Step 2.2, using a dynamic and static attribute category constraint fusion method to fuse the three-dimensional geographic scene model obtained after object-level fusion with the attributes of dynamic and static data in the initial or updated complex three-dimensional scene geographic knowledge graph, wherein the dynamic and static attribute category constraint fusion method includes attribute matching, spatial feature expression, and symbol subordination binding of geographic scene objects;

[0045] The specific steps are:

[0046] First, the dynamic and static attribute data in the initial or updated complex three-dimensional scene geographic knowledge graph are preprocessed, including data cleaning, format conversion and attribute integration;

[0047] Secondly, according to the attribute categories in the complex three-dimensional scene geographic knowledge graph obtained by preprocessing, the dynamic and static attribute category constraint fusion method is adopted to perform different constraint modeling, that is, to obtain the attribute-level fused three-dimensional geographic scene model. The specific steps are as follows:

[0048] For basic static attribute data, a mapping structure is used to associate the basic static attribute data with each geographic scene object model in the three-dimensional geographic scene model obtained after object-level fusion, an attribute table is constructed according to the associated mapping relationship and stored in a structured form, an icon is created for the static attribute of each ID in the attribute table, each icon is associated with its corresponding attribute data, and the icon is visually displayed in the three-dimensional geographic scene model obtained after object-level fusion, wherein the attribute table contains the ID, attribute type, and attribute value of each geographic scene object in the three-dimensional geographic scene model after object-level fusion;

[0049] For the hierarchical partition attribute data, standardization is first performed, including unit and format unification. Secondly, the hierarchical structure of buildings in the three-dimensional geographic scene model obtained after object-level fusion is constructed based on the hierarchical partition attribute data obtained through standardization, and the corresponding spatial coordinates and geometric features are specified for each layer in the structural hierarchy. Then, based on the spatial coordinates and geometric features, the hierarchical partition attribute data obtained through standardization is bound to the corresponding hierarchical structure. Finally, the bound data is visualized in feature space according to the given display attribute fields of each layer, where the display attribute fields include display material, display color, display transparency, layer height or purpose of each layer.

[0050] For dynamic attribute data, first create a corresponding symbol identifier for each sensor object in the 3D geographic scene model obtained after object-level fusion based on the coordinate information extracted by the sensor, and locate it at the specified spatial coordinates in the 3D geographic scene model obtained after object-level fusion. Then, bind the dynamic attribute data to the corresponding symbol identifier to achieve dynamic display effects in the 3D engine.

[0051] Step 2.3: Based on the three-dimensional geographic scene model obtained after attribute-level fusion and the stereoscopic scene constraint conditions, a LOD model of the corresponding level is constructed for scene-level fusion. The specific implementation steps are as follows:

[0052] Based on the constraints of the stereo scene, the importance of the layers in the three-dimensional geographic scene model obtained after attribute-level fusion is evaluated and a comprehensive index is generated. The specific steps are as follows:

[0053] The index values ​​of each geographic scene object model obtained after attribute-level fusion are extracted from the initial or updated complex three-dimensional scene geographic knowledge graph, and different weights are assigned to each index value according to the determined three-dimensional scene constraints, and the weighted algorithm is used to calculate the comprehensive index of each geographic scene object model. The formula is:

[0054]

[0055] Among them, Imp metric represents the comprehensive index of each geographic scene object model calculated under the constraints of the stereo scene, ω j represents the weight of the jth index value in each geographic scene object model under the cth stereo scene constraint, m j represents the jth index value in each geographic scene object model, and n represents the total number of index values ​​in each geographic scene object model;

[0056] Based on the comprehensive indicators, the detail level of each geographic scene object in the three-dimensional geographic scene model obtained after attribute-level fusion is constructed to obtain the final LOD model, that is, the three-dimensional geographic scene model of scene-level fusion is obtained. The calculation formula of the detail level of the LOD model is:

[0057] LOD i =f(Imp metric )

[0058] Among them, LOD i It represents the detail level of the LOD model of the i-th geographic scene object model in the three-dimensional geographic scene model obtained after attribute level fusion. f is represented as a linear or nonlinear function, and the attribute level fusion is determined according to specific needs.

[0059] Compared with the prior art, the advantages of the present invention are:

[0060] The modeling process of existing fusion modeling technology lacks dynamism and flexibility. The constructed models are often of low precision and poor visualization efficiency. This paper proposes an iterative dynamic fusion modeling method, which realizes the rapid registration and unification of scene model space, the correct mapping of modeling relationships, the precise binding of attribute information and the efficient representation of scene models through the guidance of geographic knowledge graph and multi-dimensional fusion constraints. It is specifically embodied in:

[0061] 1. Based on the "scene-feature-change" spatiotemporal collaborative mapping model, the present invention accurately depicts the geographical elements of the three-dimensional scene and their interactive relationships, which not only promotes the consistent expression of scene knowledge, but also deepens the cognition and grasp of the dynamic characteristics of the geographical scene, and provides a more systematic and efficient information processing framework for traffic management, urban governance and other aspects;

[0062] 2. The present invention proposes a method for constructing a complex three-dimensional scene geographic knowledge graph driven by dynamic data. Based on the model layer obtained by the "scene-feature-change" spatiotemporal collaborative mapping model, the concepts and relationships of the three-dimensional geographic scene are mapped into the complex three-dimensional scene geographic knowledge graph, and the dynamic update of the edges and nodes of the complex three-dimensional scene geographic knowledge graph is realized in a data-driven manner in the data layer. The constructed complex three-dimensional scene geographic knowledge graph can be used to guide the subsequent dynamic modeling of three-dimensional twin scenes.

[0063] 3. The present invention constructs a multi-dimensional dynamic fusion mechanism from the perspective of object, attribute and scene modeling, combines the geographic knowledge graph of complex three-dimensional scenes, guides the high-precision fusion construction of geographic scenes from multiple dimensions, and can achieve rapid spatial registration and unification of geographic scene object models, correct mapping of modeling relationships, precise binding of attribute information and efficient representation of scene models;

[0064] 4. The dynamic fusion modeling of the present invention, supported by the dynamic and complex three-dimensional scene geographic knowledge graph, realizes the real-time fusion construction and dynamic adjustment and update of the twin scenes. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments will be briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0066] Figure 1 It is a schematic diagram of the overall framework of the present invention;

[0067] Figure 2 Schematic diagram of the “scene-feature-transition” spatiotemporal collaborative mapping model in the present invention;

[0068] Figure 3 This is a schematic diagram of the domain ontology of the complex three-dimensional scene twin geographic scene in the present invention;

[0069] Figure 4 A schematic diagram for constructing a geographic knowledge graph for complex three-dimensional scenes in the present invention;

[0070] Figure 5 Schematic diagram of dynamic fusion modeling of complex three-dimensional scenes in the present invention;

[0071] Figure 6 It is a schematic diagram of the spatial position and posture parameters of each geographical scene object in the present invention;

[0072] Figure 7 A schematic diagram of the spatial topological relationship of geographic scene objects in the present invention;

[0073] Figure 8 A schematic diagram of the dynamic and static attribute category constraints in the present invention;

[0074] Fig. 9 is a schematic diagram of information fusion based on attribute category constraints in the present invention;

[0075] Fig.10 is a schematic diagram of the stereoscopic scene constraint in the present invention;

[0076] Fig.11 This is an example of the hierarchical partitioning visualization of buildings in a three-dimensional geographic scene in the present invention. DETAILED DESCRIPTION

[0077] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0078] In order to solve the technical problems existing in the existing technology, it is necessary to innovate the twin scene fusion modeling method based on the needs of multi-dimensional geographic scene fusion. Specifically, we should explore cross-level fusion mechanisms and effectively integrate dynamic and static attribute data to ensure the real-time and accuracy of the data, while taking into account the balance between the realism and expression efficiency of the model, so as to achieve high-fidelity construction and dynamic optimization of the digital twin model.

[0079] The overall framework of the complex 3D scene twin modeling method guided by geographic knowledge graph is as follows: Figure 1As shown in the figure, first, the characteristics of objects in complex three-dimensional scenes and the correlation mechanism between them are explored, and a structured geographic knowledge graph of complex three-dimensional scenes is constructed to achieve accurate characterization and unified description of three-dimensional spatial objects and relationships; then, the guiding mechanism of the geographic knowledge graph of complex three-dimensional scenes is integrated into the fusion modeling process of geographic scenes, which progresses step by step from the object level, attribute level to the scene level. It not only focuses on the accurate modeling of a single object, but also emphasizes the fusion of attribute information between objects and the construction of the overall structure of the scene. By dynamically integrating domain ontology data, real-time sensor information and basic geographic scenes, the seamless connection and deep integration of the three are achieved, which can support the two-way mapping mechanism of digital twins, embodies the real-time mapping and dynamic adjustment between physical space and information space, thereby providing precise feedback control and promoting intelligent management and decision support in complex three-dimensional scenes. To elaborate, first, from physical space to information space, we construct a three-dimensional twin scene of information space through various data, information and knowledge of physical space. Secondly, from information space to physical space, through the constructed digital twin scene, we can further guide the management and decision-making in the real physical space. For example, managers can monitor the monitoring data through real-time sensors, judge the status of important places such as computer rooms and parking lots, and make some environmental control, resource allocation and other actions.

[0080] The geographic scene constructs a complex three-dimensional scene geographic knowledge graph of scene-feature-change based on the dynamic and static data of each geographic scene in the physical space of the complex three-dimensional twin scene, and updates it in real time; the specific steps are: In view of the complexity and dynamics of the geographic scene, it is difficult for traditional knowledge management methods to conduct systematic analysis and processing. Therefore, the present invention first deeply analyzes the concepts, attributes and internal connections of the multiple geographic elements involved in the geographic scene, focusing on the characteristics of the scene and scene objects and their change process. Based on this, a "scene-feature-change" spatiotemporal collaborative mapping model of complex three-dimensional scenes is given, such as Figure 2 As shown in the figure, the "scene-feature-transition" spatiotemporal collaborative mapping model includes scene space, feature space and transition space. The scene space includes various geographical scenes. Each geographical scene object in each geographical scene shows different features. The features are represented by feature sets in the feature space. The feature space not only describes the properties of each geographical scene object, but is also affected by the transition space. The transition space represents the evolution of each geographical scene object in the scene space over time, which directly affects the expression of the state and features of each geographical scene in the scene space. This model accurately depicts the geographical elements of the geographical scene and their interactive relationships, which not only promotes the consistent expression of scene knowledge, but also deepens the cognition and grasp of the dynamic characteristics of the scene, and provides a more systematic and efficient information processing framework for traffic management, urban governance and other aspects.

[0081] The scene space covers a wide range of entities in three-dimensional space, such as traffic networks, building communities, mobile populations, and environmental factors. The feature space accurately captures the characteristics of these objects, such as the smoothness and congestion of traffic, the freshness and pollution of the environment, the flow of time, and the layout of space. The change space records every moment of dynamics from facility updates to population flows, from environmental changes to traffic adjustments. Scene objects (i.e., geographic scene objects) in the scene space can interpret their feature changes through a series of changes, and the features represent the characteristics, forms, or situations of scene objects. For example, the roads of traffic facilities may be congested during peak hours, and this feature may be improved by changes in road widening projects. Similarly, the green coverage and air quality of urban parks will be updated with seasonal plant growth and maintenance activities. This close linkage of "scene-feature-change" constructs a dynamically updated geographic knowledge map to help managers and decision makers establish a real-time and comprehensive information perspective.

[0082] Under the guidance of the integrated "scene-feature-transition" spatiotemporal collaborative mapping model expression mode, the geographic scene deeply analyzes the domain knowledge of the three-dimensional scene (i.e., the geographic scene) to extract the correct concept attributes of geographic elements and their accurate associations, that is, based on the "scene-feature-transition" spatiotemporal collaborative mapping model, the domain ontology of each geographic scene is extracted from the dynamic and static data of the physical space of the complex three-dimensional twin scene to obtain the scene ontology library, where the domain ontology is the pattern layer, including the transition domain corresponding to the transition space, the scene domain corresponding to the scene space and the feature domain corresponding to the feature space, and the semantic association relationship between the transition domain, the scene domain and the feature domain. The geographic scene refers to the three-dimensional geographical location or space range, including geographic scene objects and their associations. The geographic scene objects have physical objects and environmental elements. Physical objects include topography, buildings, transportation facilities, vegetation and water bodies. The formula of the domain ontology of each geographic scene is:

[0083] Ont=<CO,SO,FO|Relations|Properties>

[0084] Among them, Ont represents the domain ontology of the geographic scene, CO represents the change domain, SO represents the scene domain, FO represents the feature domain, Relations represents the semantic relationship between the change domain, scene domain and feature domain, and Properties represents the attributed geographic scene. The domain ontology formula describes the geographic ontology concepts and semantic relationships, and thus constructs the domain geographic ontology model, such as Figure 3 The figure shows a schematic diagram of the geographic ontology constructed for complex geographic scenes. Nodes represent geographic concepts, entities and their attributes in the geographic scene, and edges represent semantic associations between them. This ontology structure allows for a more comprehensive expression of entities and their relationships in complex scenes, while also refining the characteristics and attributes of each entity, thus achieving the mapping and unified description of the integrated expression model to the scene's geographic elements.

[0085] The construction process of the geographic ontology model in the above-mentioned field forms the geographic knowledge graph model layer, which mainly involves the construction of related entity concepts and relationships. However, the core of the knowledge graph is to instantiate abstract concepts and build a knowledge system that can accurately associate and dynamically store data. Therefore, the present invention focuses on how to map the model layer to the data layer, that is, to effectively integrate multi-source knowledge data, fill it into the model framework, and then use real-time dynamic data to drive the update of knowledge nodes and edges to ensure the timeliness and accuracy of the graph. The construction and dynamic update process of the geographic knowledge graph for complex three-dimensional scenes is as follows: Figure 4 shown.

[0086] Knowledge mainly comes from literature, expert knowledge, norms and standards, static data such as the location, shape and size of each geographic scene object contained in the geographic scene recorded in the spatial database, and dynamic data such as the temporal evolution and the impact of human activities recorded in video surveillance, sensor networks, social media, and relational databases. First, by parsing structured and unstructured data, geographic scene knowledge extraction is implemented, including entity recognition, attribute extraction and association analysis, to achieve a unified semantic description and ensure the consistency of the knowledge structure. Subsequently, the semantic, topological, and spatial relationships between entities are sorted out, and the association relationships between entities are constructed. Among them, the attributes include basic static attribute data, dynamic attribute data, and hierarchical and partitioned attribute data. The basic static attribute data include geometric attributes, physical attributes, and spatiotemporal attributes. Geometric attributes include location, shape, and size. Physical attributes include material, mass, and motion. Spatiotemporal attributes include occurrence time and spatial distribution. Dynamic attribute data include traffic flow data acquired by sensors, environmental monitoring data, and social media dynamics. Traffic flow data includes the real-time number of vehicles and real-time changing locations. Environmental monitoring data includes air quality and temperature and humidity changes. Social media dynamics include geographic tags of real-time events or activities. Hierarchical and partitioned attribute data include the function, area, height, and material of each layer.

[0087] After knowledge extraction, the semantic, topological and spatial relationships between entities are sorted out, and the association relationship between entities is constructed. At the same time, the index value corresponding to each geographic scene object is given based on the importance of each geographic scene object and each constraint in the three-dimensional scene constraint, that is, the data layer is obtained, where the semantics is the extracted attribute, and the three-dimensional scene constraint includes scene interaction constraint, spatial distance constraint, indoor and outdoor scene constraint, above and below ground scene constraint and research area scene constraint;

[0088] Finally, after mapping the pattern layer and the data layer, a complex three-dimensional scene geographic knowledge graph of scene-feature-transition is obtained, and the complex three-dimensional scene geographic knowledge graph is updated according to the real-time access dynamic data. The present invention completes the integrated expression of geographic scene knowledge and realizes the mapping from the geographic ontology pattern layer to the data layer. The constructed complex three-dimensional scene geographic knowledge graph can be used to guide the subsequent three-dimensional twin scene modeling.

[0089] The complex three-dimensional scene geographic knowledge graph is used to guide the dynamic fusion modeling of complex three-dimensional scenes from the object level, attribute level to the scene level step by step. The specific steps are as follows: Fusion modeling is a key link in building complex three-dimensional digital twin scenes. It aims to integrate basic geographic scenes, facility equipment models and dynamic sensor data to achieve unified expression of complex three-dimensional scenes. The modeling process of existing fusion modeling technology lacks dynamism and is inflexible. The constructed models are often of low accuracy and poor visualization efficiency. The present invention proposes an iterative dynamic fusion modeling method. Through the guidance of geographic knowledge graph and multi-dimensional fusion constraints, the rapid registration and unification of scene model space, the correct mapping of modeling relationships, the precise binding of attribute information and the efficient representation of scene models are achieved.

[0090] At the object-level fusion level, we focus on the precise embedding and spatial relationship construction of geographic scene objects. For a single model, we use the spatial position and posture information in the complex three-dimensional scene geographic knowledge graph to ensure its precise positioning and orientation in the geographic scene. For multi-model fusion, we use the topological relationship rules defined in the complex three-dimensional scene geographic knowledge graph to maintain the spatial continuity and logic between models and achieve overall coordination of the scene. Attribute-level fusion focuses on the deep fusion of dynamic and static data. Based on the attribute classification and association in the complex three-dimensional scene geographic knowledge graph, we use the spatial subordinate binding mechanism to closely associate the attribute data with the corresponding entity. At the same time, we use the attribute space representation technology to intuitively display the spatial distribution and change rules behind the data, and enhance the perception and interaction capabilities of the scene. Scene-level fusion focuses on the fusion of three-dimensional scenes. Based on the LOD (Level Detail) model organization strategy, we dynamically adjust the fineness and complexity of the model according to the different spatial requirements of the geographic scene, such as above and below the ground, indoors and outdoors, to balance the sense of reality and computational efficiency.

[0091] In order to realize dynamic fusion modeling of complex geographical scenes, the present invention designs a set of iterative update processes, such as Figure 5As shown. First, dynamic and static data are extracted from the geographic knowledge graph of complex three-dimensional scenes, and a three-dimensional primary scene is constructed after preprocessing as the starting point for fusion modeling. Subsequently, under the guidance of multi-level fusion constraint rules, three-dimensional bases, building facilities and dynamic data are gradually connected, and scene details are gradually improved and optimized through operations such as rotation positioning, translation fitting, dynamic assignment and scaling matching. During this period, the scene is continuously checked to see if it meets the fusion rules to ensure the effective representation of dynamic data and the dynamic update of the knowledge graph. Finally, through the iterative update modeling of the scene, the dynamic construction and continuous optimization of the complex three-dimensional twin scene are realized.

[0092] Object-level fusion is specifically:

[0093] In the process of building a 3D digital twin scene, object-level fusion aims to achieve accurate positioning and posture adjustment of a single model object, as well as optimize the spatial relationship and dynamic adaptation between multiple model objects. This case will elaborate on the implementation method of object-level fusion from two aspects: spatial matching and spatial topological constraints.

[0094] (1) Spatial matching

[0095] Spatial fast matching aims to quickly and accurately embed geographic scene objects into three-dimensional virtual space, mainly involving the precise acquisition and dynamic application of the spatial position (i.e., target position) parameters (such as longitude and latitude, altitude, or three-dimensional Cartesian coordinates) and attitude parameters (such as roll angle, pitch angle, and yaw angle calculated by Euler angles, quaternions, or rotation matrices) of geographic scene objects. The spatial position and attitude parameters that spatial matching relies on are shown in the figure below: Figure 6 As shown, the target position of the geographic scene object stored in the initial or updated complex three-dimensional scene geographic knowledge graph and the parameters of the three-dimensional rotation matrix are read to calculate the rotation matrix, and the posture data is calculated through the rotation matrix, wherein the posture parameters included in the posture data adjust the direction and angle of the coordinate position mapped to the geographic scene object in the virtual space, so that the posture in the virtual scene is consistent with the real world.

[0096]

[0097] Roll = arctan2(R 32 ,R 33 )

[0098] Yaw = arctan2(R 12 ,R 11 )

[0099] Among them, R represents the 3x3 rotation matrix, Pitch represents the yaw angle rotating around the Z axis, Roll represents the roll angle rotating around the X axis, and Yaw represents the pitch angle rotating around the Y axis. Represents the parameters of the rotation matrix. The first column represents the direction vector of the X axis after rotation, the second column represents the direction vector of the Y axis after rotation, and the third column represents the direction vector of the Z axis after rotation. 11 , R 21 and R 31 Respectively represent the direction vector of the X-axis on the X-axis, Y-axis and Z-axis after rotation, R 12 , R 22 and R 32 Respectively represent the direction vector of the Y axis after rotation on the X axis, Y axis and Z axis, R 13 , R 23 and R 33 Respectively indicate that the direction vector of the Z axis after rotation is on the X axis, Y axis and Z axis;

[0100] Based on the target position and posture data of each geographic scene object, a spatial positioning algorithm is used to map the attributes of the geographic scene object to the corresponding coordinate position in the virtual space, thus achieving spatial matching; when the position or posture of the geographic scene object in the real world changes, the complex three-dimensional scene geographic knowledge graph will obtain and update this information in a timely manner, and trigger corresponding adjustments in the virtual scene to achieve dynamic matching.

[0101] (2) Spatial topological constraints

[0102] Spatial topological constraints ensure that the arrangement and combination of geographic scene objects in virtual space conforms to real logic and avoids spatial conflicts by defining clear topological rules such as proximity, separation, and inclusion. First, read the relationship information between geographic scene objects recorded in the initial or updated complex three-dimensional scene geographic knowledge graph, and identify the geographic scene object pairs that need to establish topological constraints. Then, the spatial topological relationship of the geographic scene object model obtained after each pair of spatial matching is obtained according to the selected topological semantics, that is, the three-dimensional geographic scene model after object-level fusion is obtained, where the spatial topological relationship includes the spatial topological relationship of the geographic scene object model pairs that are adjacent, separated, and included.

[0103] The formula for the adjacent spatial topological relationship is:

[0104]

[0105] Among them, Adj(α,β) means that the geographic scene object model α is adjacent to the geographic scene object model β. represents the boundary of the geographic scene object model, α° and β° represent the interior of the geographic scene object model α and the geographic scene object model β respectively, represents an empty set, for example, the adjacent floor models of a building fit closely together;

[0106]

[0107] Wherein, Dis(α,β) indicates that the geographic scene object model α is separated from the geographic scene object model β, for example, the server and other equipment are installed in the computer room;

[0108]

[0109] Among them, Con(α,β) means that the geographic scene object model α completely contains the geographic scene object model β. In the complex three-dimensional scene space structure, large buildings or complexes often contain multiple small space units;

[0110] T(α,β)=Adj(α,β)+Dis(α,β)+Con(α,β)

[0111] Among them, T(α,β) represents the spatial topological relationship between the geographic scene object model α and the geographic scene object model β. When using this formula to constrain the modeling of multiple geographic scene object models, it can ensure that the geographic scene objects have a correct spatial relationship. The spatial topological relationship is shown as follows: Figure 7 shown.

[0112] Attribute-level fusion is as follows:

[0113] Attribute data fusion is the key link between the physical world and the digital model. It realizes the integration and expression of multi-dimensional attribute data based on the spatial position and semantic subordination of each entity and its ancillary facilities in the three-dimensional geographic scene model obtained after object-level fusion. This process involves the attribute-level fusion and information association of the inherently dynamically changing ancillary facilities. On the basis of dynamic and static attribute analysis and data processing, we further explore the deep fusion mechanism of attribute data and scene models, aiming to achieve the precise binding of models and attribute information, as well as the intuitive display of attribute data in geographic scenes. The dynamic and static attribute category constraint fusion method includes three aspects: model attribute matching, spatial feature expression, and symbolic subordination binding. The main processes are as follows: Figure 8 shown.

[0114] Model attribute matching mainly extracts the subordinate relationship of the three-dimensional geographic scene model obtained after the fusion of attributes and object level, ensuring that the static attributes of geographic scene elements can be accurately associated with their virtual representation in the twin scene, laying the foundation for complex geographic scene information retrieval and spatial analysis; spatial representation modeling mainly targets complex information in static attributes (such as building stratification and household division, underground pipe network layout), and adopts modeling methods such as layered coloring and highlighting to intuitively display its spatial layout and ownership relationship, which helps to improve the cognition and understanding of complex spatial structures. Dynamic attribute symbolization and binding mainly involve dynamic data sources such as sensors and video surveillance. First, symbolic design is carried out to build the sensor equipment into easy-to-understand graphics or icons. Subsequently, these symbols are bound to the corresponding positions in the scene model to realize real-time update and display of dynamic data. The fusion constraint diagram is as follows Fig. 9 shown.

[0115] That is, the dynamic and static attribute category constraint fusion method is used to fuse the three-dimensional geographic scene model obtained after object-level fusion with the attributes of dynamic and static data in the initial or updated complex three-dimensional scene geographic knowledge graph, wherein the dynamic and static attribute category constraint fusion method includes attribute matching, spatial feature expression, and symbol subordinate binding of geographic scene objects; the specific steps are:

[0116] First, the dynamic and static attribute data in the initial or updated complex three-dimensional scene geographic knowledge graph are preprocessed, including data cleaning, format conversion and attribute integration;

[0117] Secondly, according to the attribute categories in the complex three-dimensional scene geographic knowledge graph obtained by preprocessing, the dynamic and static attribute category constraint fusion method is adopted to perform different constraint modeling, that is, to obtain the attribute-level fused three-dimensional geographic scene model. The specific steps are as follows:

[0118] For basic static attribute data, a mapping structure is used to associate the basic static attribute data with each geographic scene object model in the three-dimensional geographic scene model obtained after object-level fusion. An attribute table is constructed according to the associated mapping relationship and stored in a structured form. Icons are created for the static attributes of each ID in the attribute table. Each icon is associated with its corresponding attribute data, and the icon is visualized in the three-dimensional geographic scene model obtained after object-level fusion, wherein the attribute table contains the ID, attribute type, and attribute value of each geographic scene object in the three-dimensional geographic scene model after object-level fusion. The mapping structure refers to a system for associating attribute data with objects in the three-dimensional geographic scene model obtained after object-level fusion. It is realized by clarifying the corresponding relationship between different basic static attribute data and the three-dimensional model objects obtained after object-level fusion in an attribute table, in which each row represents the ID of a geographic scene object and its corresponding static attribute data field; secondly, a data model is designed, which may be in the form of a key-value pair, with the geographic scene object ID as the key and the basic attribute data as the value. Databases (such as SQL) or data structures (such as dictionaries and hash tables) can be used to store these mapping relationships; then, in the mapping structure, semantic identifiers are used to describe attribute types. For example, specific identifiers can be defined for each attribute (such as location, material, and purpose) to make it easier to identify and operate during data extraction and display; finally, the mapping relationship is stored in a database or memory for quick access. When acquiring a 3D model, the relevant static attribute data can be obtained by looking up the attribute table and dynamically updated.

[0119] For the hierarchical partition attribute data, standardization is first performed, including unit and format unification. Secondly, the hierarchical structure of buildings in the three-dimensional geographic scene model obtained after object-level fusion is constructed based on the hierarchical partition attribute data obtained through standardization, and the corresponding spatial coordinates and geometric features are specified for each layer in the structural hierarchy. Then, based on the spatial coordinates and geometric features, the hierarchical partition attribute data obtained through standardization is bound to the corresponding hierarchical structure. Finally, the bound data is visualized in feature space according to the given display attribute fields of each layer, where the display attribute fields include display material, display color, display transparency, layer height or purpose of each layer.

[0120] For dynamic attribute data, first create a corresponding symbol identifier for each sensor object in the three-dimensional geographic scene model obtained after object-level fusion based on the coordinate information extracted by the sensor, and position it at the specified spatial coordinates in the three-dimensional geographic scene model obtained after object-level fusion. Then, bind the dynamic attribute data to the corresponding symbol identifier to achieve dynamic display effects in the three-dimensional engine.

[0121] The specific scene-level fusion is as follows:

[0122] Scene-level fusion aims to optimize the efficiency of the overall scene. By comprehensively considering the optimization selection of the detail levels of multiple objects, the realism and rendering efficiency of the overall scene are ensured, thereby constructing a three-dimensional scene that can meet the visual experience requirements and effectively control resource consumption. The three-dimensional twin scene constructed above (the three-dimensional geographic scene model obtained after attribute-level fusion) focuses on the modeling effect, and its impact on subsequent visual expression is not foreseen. The loading requirements of a large number of refined models for complex three-dimensional scenes will increase the burden on server storage, network scheduling, and terminal rendering. Therefore, when constructing complex stereoscopic scenes, it is necessary to solve the problem of how to balance realism and rendering efficiency. While considering the multi-level LOD modeling strategy, the present invention designs a fusion mechanism for stereoscopic scene constraints. Mainly based on the stereoscopic scene constraint conditions, the importance of the target model in the scene is evaluated, and the LOD model of the corresponding level is constructed according to the degree of importance. The stereoscopic scene constraint is shown in the figure Fig.11 shown.

[0123] like Fig.10 As shown in the figure, the stereo scene constraint integrates multiple constraints, including the study area, underground scenes, indoor and outdoor scenes, etc., to form a systematic importance evaluation hierarchy. These constraints act together on the modeling of each geographic scene object model in the three-dimensional geographic scene model obtained after attribute-level fusion, and calculate and generate comprehensive indicators, so as to construct the detailed level of the target modeling object according to the importance. For example, for models in the core area of ​​the study and requiring high visualization, according to each stereo scene constraint, a higher weight is given to it and multiplied with the corresponding index value to obtain a weighted comprehensive index, and a high-precision LOD model is constructed according to the comprehensive index; for models located in non-critical areas such as underground and indoors or with low interaction frequency, their importance is reduced accordingly, and a low-precision model is constructed to optimize rendering performance. For example, scene interaction constraints, spatial distance constraints, indoor and outdoor scene constraints, underground and underground scene constraints, and study area scene constraints are assigned different index values ​​according to the importance of different geographic scene objects.

[0124] That is, based on the three-dimensional geographic scene model obtained after attribute-level fusion and the stereoscopic scene constraint conditions, the LOD model of the corresponding level is constructed for scene-level fusion. The specific implementation steps are as follows:

[0125] Based on the constraints of the stereo scene, the importance of the layers in the three-dimensional geographic scene model obtained after attribute-level fusion is evaluated and a comprehensive index is generated. The specific steps are as follows:

[0126] The index values ​​of each geographic scene object model obtained after attribute-level fusion are extracted from the initial or updated complex three-dimensional scene geographic knowledge graph, and different weights are assigned to each index value according to the determined three-dimensional scene constraints, and the weighted algorithm is used to calculate the comprehensive index of each geographic scene object model. The formula is:

[0127]

[0128] Among them, Imp metric represents the comprehensive index of each geographic scene object model calculated under the constraints of the stereo scene, ω j represents the weight of the jth index value in each geographic scene object model under the cth stereo scene constraint, m j represents the jth index value in each geographic scene object model, and n represents the total number of index values ​​in each geographic scene object model;

[0129] Based on the comprehensive indicators, the detail level of each geographic scene object in the three-dimensional geographic scene model obtained after attribute-level fusion is constructed to obtain the final LOD model, that is, the three-dimensional geographic scene model of scene-level fusion is obtained. The calculation formula of the detail level of the LOD model is:

[0130] LOD i =f(Imp metric )

[0131] Among them, LOD i The detail level of the LOD model of the i-th geographic scene object model in the three-dimensional geographic scene model obtained after attribute-level fusion. f is expressed as a linear or nonlinear function, and attribute-level fusion is determined according to specific needs. Based on the actual importance of the model, the modeling precision in the scene is dynamically adjusted. This ensures that key information is not lost and reduces the overall visualization burden.

[0132] In summary, the present invention first explores the characteristics and associations between three-dimensional scenes and modeling objects, designs a dynamic data-driven three-dimensional scene geographic knowledge graph construction algorithm; then, analyzes the associations between complex scene models and their dynamic and static attributes in terms of geometry, attributes, scale, and semantics, constructs a multi-dimensional dynamic fusion mechanism from the perspective of object, attribute, and scene modeling, and proposes a complex three-dimensional scene dynamic fusion modeling method guided by geographic knowledge graphs. This enables efficient fusion construction of twin scenes and promotes the transformation of operation management in various three-dimensional scenes to refinement and informatization.

Claims

1. A complex three-dimensional scene twin modeling method guided by geographic knowledge graph, characterized in that: The steps include: Step 1: Based on the acquired dynamic and static data of each geographic scene in the physical space of the complex three-dimensional twin scene, a complex three-dimensional scene geographic knowledge graph of scene-feature-change is constructed and updated in real time; Step 2: Use the geographic knowledge graph of complex three-dimensional scenes to guide the dynamic fusion modeling of complex three-dimensional scenes step by step from the object level, attribute level to the scene level.

2. According to claim 1, a complex three-dimensional scene twin modeling method guided by a geographic knowledge graph is characterized in that: The specific steps of step 1 are: Step 1.1, a "scene-feature-transition" spatiotemporal collaborative mapping model of a given complex three-dimensional scene, wherein the "scene-feature-transition" spatiotemporal collaborative mapping model includes scene space, feature space and transition space, wherein the scene space includes various geographic scenes, and each geographic scene object in each geographic scene shows different features, which are represented by feature sets in the feature space. The feature space not only describes the properties of each geographic scene object, but is also affected by the transition space. The transition space represents the evolution of each geographic scene object in the scene space over time, that is, it directly affects the expression of the state and features of each geographic scene in the scene space; Step 1.2: Based on the "scene-feature-transition" spatiotemporal collaborative mapping model, the domain ontology of each geographic scene is extracted from the dynamic and static data of the physical space of the complex three-dimensional twin scene obtained to obtain the scene ontology library, where the domain ontology is the pattern layer, including the transition domain corresponding to the transition space, the scene domain corresponding to the scene space and the feature domain corresponding to the feature space, and the semantic association relationship between the transition domain, the scene domain and the feature domain. The static data in the dynamic and static data include the location, shape and size of each geographic scene object contained in the geographic scene recorded in literature, expert knowledge, specifications and standards and spatial databases. The dynamic data includes the temporal evolution and the impact of human activities recorded in video surveillance, sensor data, social media and relational databases. The geographic scene refers to the three-dimensional geographical location or space range, including geographic scene objects and their associations. The geographic scene objects have physical objects and environmental elements. Physical objects include topography, buildings, transportation facilities, vegetation and water bodies. The formula of the domain ontology of each geographic scene is: Ont=<CO,SO,FO|Relations|Properties> Among them, Ont represents the domain ontology of geographic scene, CO represents the change domain, SO represents the scene domain, FO represents the feature domain, Relations represents the semantic relationship between the change domain, scene domain and feature domain, and Properties represents the attributes; Step 1.3: Construct the data layer based on the obtained scene ontology library. The specific steps are as follows: Parse structured and unstructured data based on the obtained scene ontology library, and extract knowledge based on given extraction rules after parsing, including entity extraction, attribute extraction, and extraction of association relationships between entities and between entities and attributes. Attributes include basic static attribute data, dynamic attribute data, and hierarchical partition attribute data. Basic static attribute data include geometric attributes, physical attributes, and spatiotemporal attributes. Geometric attributes include location, shape, and size. Physical attributes include material, mass, and motion. Spatiotemporal attributes include occurrence time and spatial distribution. Dynamic attribute data include traffic flow data, environmental monitoring data, and social media dynamics obtained by sensors. Traffic flow data includes the number of real-time vehicles and real-time changing locations. Environmental monitoring data includes air quality and temperature and humidity changes. Social media dynamics include geographic tags of real-time events or activities. Hierarchical partition attribute data includes the function, area, height, and material of each layer. After knowledge extraction, the semantic, topological and spatial relationships between entities are sorted out, and the association relationship between entities is constructed. At the same time, the index value corresponding to each geographic scene object is given based on the importance of each geographic scene object and each constraint in the three-dimensional scene constraint, that is, the data layer is obtained, where the semantics is the extracted attribute, and the three-dimensional scene constraint includes scene interaction constraint, spatial distance constraint, indoor and outdoor scene constraint, above and below ground scene constraint and research area scene constraint; Step 1.4: After mapping the pattern layer and the data layer, a complex three-dimensional scene geographic knowledge graph of scene-feature-transition is obtained, and the complex three-dimensional scene geographic knowledge graph is updated according to the dynamic data accessed in real time.

3. According to claim 2, a complex three-dimensional scene twin modeling method guided by a geographic knowledge graph is characterized in that: The specific steps of step 2 are: Step 2.1: Use the initial or updated complex three-dimensional scene geographic knowledge graph to perform object-level fusion of the dynamic and static data of each geographic scene in the physical space of the complex three-dimensional twin scene. The specific steps are as follows: Spatial Matching: The target position of the geographic scene object stored in the initial or updated complex three-dimensional scene geographic knowledge graph and the parameters of the three-dimensional rotation matrix are read to calculate the rotation matrix, and the posture data is calculated through the rotation matrix, wherein the posture parameters included in the posture data adjust the direction and angle of the coordinate position mapped to the geographic scene object in the virtual space, so that the posture in the virtual scene is consistent with the real world; The formula for the rotation matrix is: Roll=arctan2(R 32 ,R 33 ) Yaw=arctan2(R 12 ,R 11 ) Among them, R represents the 3x3 rotation matrix, Pitch represents the yaw angle rotating around the Z axis, Roll represents the roll angle rotating around the X axis, and Yaw represents the pitch angle rotating around the Y axis. Represents the parameters of the rotation matrix. The first column represents the direction vector of the X axis after rotation, the second column represents the direction vector of the Y axis after rotation, and the third column represents the direction vector of the Z axis after rotation. 11 , R 21 and R 31 Respectively represent the direction vector of the X-axis on the X-axis, Y-axis and Z-axis after rotation, R 12 , R 22 and R 32 Respectively represent the direction vector of the Y axis after rotation on the X axis, Y axis and Z axis, R 13 , R 23 and R 33 Respectively indicate that the direction vector of the Z axis after rotation is on the X axis, Y axis and Z axis; Based on the target position and posture data of each geographic scene object, a spatial positioning algorithm is used to map the attributes of the geographic scene object to the corresponding coordinate position in the virtual space, thus achieving spatial matching; Spatial topological constraints: Read the relationship information between geographic scene objects recorded in the initial or updated complex three-dimensional scene geographic knowledge graph, and identify the geographic scene object pairs that need to establish topological constraints; Then, the spatial topological relationship of the geographic scene object model is obtained after matching each pair of spaces according to the selected topological semantics, that is, the three-dimensional geographic scene model after object-level fusion is obtained, wherein the spatial topological relationship includes the adjacent spatial topological relationship, the separated spatial topological relationship and the included spatial topological relationship of the geographic scene object model; The formula for the adjacent spatial topological relationship is: Among them, Adj(α,β) means that the geographic scene object model α is adjacent to the geographic scene object model β. represents the boundary of the geographic scene object model, α° and β° represent the interior of the geographic scene object model α and the geographic scene object model β respectively, represents the empty set; Where Dis(α,β) represents the separation of geographic scene object model α and geographic scene object model β; Among them, Con(α,β) means that the geographic scene object model α completely contains the geographic scene object model β; T(α,β)=Adj(α,β)+Dis(α,β)+Con(α,β) Wherein, T(α,β) represents the spatial topological relationship between geographic scene object model α and geographic scene object model β; Step 2.2, using a dynamic and static attribute category constraint fusion method to fuse the three-dimensional geographic scene model obtained after object-level fusion with the attributes of dynamic and static data in the initial or updated complex three-dimensional scene geographic knowledge graph, wherein the dynamic and static attribute category constraint fusion method includes attribute matching, spatial feature expression, and symbol subordination binding of geographic scene objects; The specific steps are: First, the dynamic and static attribute data in the initial or updated complex three-dimensional scene geographic knowledge graph are preprocessed, including data cleaning, format conversion and attribute integration; Secondly, according to the attribute categories in the complex three-dimensional scene geographic knowledge graph obtained by preprocessing, the dynamic and static attribute category constraint fusion method is adopted to perform different constraint modeling, that is, to obtain the attribute-level fused three-dimensional geographic scene model. The specific steps are as follows: For basic static attribute data, a mapping structure is used to associate the basic static attribute data with each geographic scene object model in the three-dimensional geographic scene model obtained after object-level fusion, an attribute table is constructed according to the associated mapping relationship and stored in a structured form, an icon is created for the static attribute of each ID in the attribute table, each icon is associated with its corresponding attribute data, and the icon is visually displayed in the three-dimensional geographic scene model obtained after object-level fusion, wherein the attribute table contains the ID, attribute type, and attribute value of each geographic scene object in the three-dimensional geographic scene model after object-level fusion; For the hierarchical partition attribute data, standardization is first performed, including unit and format unification. Secondly, the hierarchical structure of buildings in the three-dimensional geographic scene model obtained after object-level fusion is constructed based on the hierarchical partition attribute data obtained through standardization, and the corresponding spatial coordinates and geometric features are specified for each layer in the structural hierarchy. Then, based on the spatial coordinates and geometric features, the hierarchical partition attribute data obtained through standardization is bound to the corresponding hierarchical structure. Finally, the bound data is visualized in feature space according to the given display attribute fields of each layer, where the display attribute fields include display material, display color, display transparency, layer height or purpose of each layer. For dynamic attribute data, first create a corresponding symbol identifier for each sensor object in the 3D geographic scene model obtained after object-level fusion based on the coordinate information extracted by the sensor, and locate it at the specified spatial coordinates in the 3D geographic scene model obtained after object-level fusion. Then, bind the dynamic attribute data to the corresponding symbol identifier to achieve dynamic display effects in the 3D engine. Step 2.3: Based on the three-dimensional geographic scene model obtained after attribute-level fusion and the stereoscopic scene constraint conditions, a LOD model of the corresponding level is constructed for scene-level fusion. The specific implementation steps are as follows: Based on the constraints of the stereo scene, the importance of the layers in the three-dimensional geographic scene model obtained after attribute-level fusion is evaluated and a comprehensive index is generated. The specific steps are as follows: The index values ​​of each geographic scene object model obtained after attribute-level fusion are extracted from the initial or updated complex three-dimensional scene geographic knowledge graph, and different weights are assigned to each index value according to the determined three-dimensional scene constraints, and the weighted algorithm is used to calculate the comprehensive index of each geographic scene object model. The formula is: Among them, Imp metric represents the comprehensive index of each geographic scene object model calculated under the constraints of the stereo scene, ω j represents the weight of the jth index value in each geographic scene object model under the cth stereo scene constraint, m j represents the jth index value in each geographic scene object model, and n represents the total number of index values ​​in each geographic scene object model; Based on the comprehensive indicators, the detail level of each geographic scene object in the three-dimensional geographic scene model obtained after attribute-level fusion is constructed to obtain the final LOD model, that is, the three-dimensional geographic scene model of scene-level fusion is obtained. The calculation formula of the detail level of the LOD model is: LOD i =f(Imp metric ) Among them, LOD i It represents the detail level of the LOD model of the i-th geographic scene object model in the three-dimensional geographic scene model obtained after attribute level fusion. f is represented as a linear or nonlinear function, and the attribute level fusion is determined according to specific needs.

Images