A real scene three-dimensional data space construction and knowledge expression method and related device

Abstract

The application discloses a real scene three-dimensional data space construction and knowledge expression method and a related device, relates to the technical fields of geographic information, three-dimensional reconstruction and knowledge engineering, and comprises the following steps: acquiring multi-source heterogeneous real scene three-dimensional data and uniformly processing coordinates to obtain a standardized geometric point set and a semantic label set; constructing a three-dimensional voxel grid, and calculating a fusion feature vector of each voxel fusion geometric feature and semantic feature; constructing a semantic enhanced index tree; constructing a real scene three-dimensional knowledge graph, and establishing a bidirectional mapping relationship between each knowledge graph entity and each index tree node; performing spatial semantic joint query based on a joint query request, the semantic enhanced index tree and the bidirectional mapping relationship, and outputting a spatial semantic joint query result. The application can realize the fusion of geometric data and semantic attributes of real scene three-dimensional data, and establish a bidirectional mapping between a spatial index node and a knowledge graph entity, thereby improving the efficiency of spatial semantic joint query.

Description

Technical Field

[0001] This application relates to the fields of geographic information, 3D reconstruction and knowledge engineering technology, and in particular to a method and related apparatus for constructing and representing real-scene 3D data space and knowledge representation. Background Technology

[0002] Most existing 3D reality technologies focus on high-precision geometric reconstruction, generating discrete meshes or point clouds, and generally lack semantic information. In applications such as digital twins and smart cities, geometric data and semantic attributes are usually stored separately, requiring external database queries, which is not only inefficient but also makes cross-dimensional reasoning difficult. Traditional spatial indexing methods are mostly based on geometric bounding boxes, and the index structure lacks semantics, failing to express the semantic relationships between entities, thus hindering the intelligent application of 3D spatial data.

[0003] Based on this, how to provide a method and related device that integrates geometric and semantic features, constructs semantically enhanced indexes, establishes bidirectional mapping of knowledge graphs, and supports efficient spatial semantic joint queries, in order to solve the current problems of geometric and semantic separation of real-world 3D data, lack of semantics in index structure, and low efficiency of spatial semantic queries, has become a technical problem that urgently needs to be solved in this field. Summary of the Invention

[0004] The purpose of this application is to provide a method and related apparatus for constructing and representing real-scene 3D data space, which can realize the fusion of geometric data and semantic attributes of real-scene 3D data, and establish a bidirectional mapping between spatial index nodes and knowledge graph entities, thereby achieving efficient spatial semantic joint query and improving the efficiency of spatial semantic joint query.

[0005] To achieve the above objectives, this application provides the following solution.

[0006] Firstly, this application provides a method for constructing and representing a real-scene 3D data space, including the following steps.

[0007] Acquire multi-source heterogeneous real-world 3D data.

[0008] The multi-source heterogeneous real-world 3D data is subjected to unified coordinate preprocessing to obtain a standardized geometric point set and semantic tag set.

[0009] Based on the standardized geometric point set and semantic tag set, a three-dimensional voxel mesh is constructed, and the fused feature vector of the fused geometric features and semantic features of each voxel in the three-dimensional voxel mesh is calculated.

[0010] Based on the fused feature vector, a semantically enhanced index tree is constructed; the semantically enhanced index tree includes several index tree nodes, each of which includes a spatial bounding box and a node feature center, and semantic information entropy is used as the basis for node splitting.

[0011] Based on the semantically enhanced index tree, a real-world 3D knowledge graph is constructed, and a bidirectional mapping relationship is established between each knowledge graph entity and each index tree node in the real-world 3D knowledge graph.

[0012] Obtain the joint query request, and perform a spatial semantic joint query based on the joint query request, the semantically enhanced index tree, and the bidirectional mapping relationship, and output the spatial semantic joint query result.

[0013] 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 steps of the real-scene three-dimensional data space construction and knowledge representation method described in any one of the above.

[0014] Thirdly, this application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the real-scene three-dimensional data space construction and knowledge representation method described above.

[0015] Fourthly, this application provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the real-scene three-dimensional data space construction and knowledge representation method described above.

[0016] According to the specific embodiments provided in this application, this application has the following technical effects.

[0017] This application provides a method and related apparatus for constructing and representing spatial knowledge from real-world 3D data. First, by performing unified coordinate preprocessing on multi-source heterogeneous real-world 3D data, the spatial reference of the multi-source data is made consistent, eliminating data deviations caused by coordinate differences, improving data compatibility and usability, and enhancing the accuracy of spatial semantic joint query results. Second, by constructing a 3D voxel mesh and calculating the fusion feature vector of the fused geometric and semantic features of each voxel, geometric and semantic information are expressed in an integrated manner at the voxel level, avoiding the separation of geometric and semantic data. This achieves the fusion of geometric data and semantic attributes of real-world 3D data, improving data integrity and expressive power, and facilitating more accurate and reliable spatial semantic joint query results. Third, by constructing a semantically enhanced index tree based on semantic information entropy and including spatial bounding boxes and node feature centers, the index partitioning is made more aligned with semantic distribution characteristics, improving the rationality and retrieval accuracy of the spatial index. Finally, by establishing a bidirectional mapping relationship between knowledge graph entities and index tree nodes, a deep association between the spatial index structure and semantic knowledge is achieved, opening up the interaction channel between spatial data and semantic knowledge. Finally, by performing spatial semantic joint queries based on semantically enhanced index trees and bidirectional mapping relationships, the target space can be quickly located and semantic conditions can be matched synchronously, achieving efficient spatial semantic joint queries and improving the accuracy and efficiency of spatial semantic joint queries. Attached Figure Description

[0018] 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.

[0019] Figure 1 An application environment diagram of a real-scene 3D data space construction and knowledge representation method provided in an embodiment of this application; Figure 2 A flowchart illustrating a method for constructing and representing a real-scene 3D data space according to an embodiment of this application; Figure 3 This is a schematic diagram of a semantically enhanced index tree structure provided in an embodiment of this application; Figure 4 This is a schematic diagram of the three-dimensional distribution of raw data provided in an embodiment of this application; Figure 5 This is a schematic diagram of the voxelization result provided in an embodiment of this application; Figure 6 This is a schematic diagram of an octree space partitioning structure provided in an embodiment of this application; Figure 7This is a schematic diagram of node entropy distribution provided in an embodiment of this application; Figure 8 This is a schematic diagram of the entity distribution of a knowledge graph provided in an embodiment of this application; Figure 9 This is a schematic diagram of the spatial semantic joint query results provided in an embodiment of this application; Figure 10 This is a schematic diagram illustrating the comparison of query performance according to an embodiment of this application; Figure 11 This is a schematic diagram of the semantic distribution statistics of voxels provided in an embodiment of this application; Figure 12 This is a schematic diagram of the octree depth distribution provided in an embodiment of this application; Figure 13 This is a schematic diagram of the structure of a computer device provided in an embodiment of this application. Detailed Implementation

[0020] 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.

[0021] The purpose of this application is to provide a method and related apparatus for constructing and representing real-scene 3D data space and knowledge representation. It aims to achieve efficient spatial semantic joint query by constructing a multi-scale data space that integrates geometric features and semantic vectors and establishing a bidirectional mapping between spatial index nodes and knowledge graph entities. This improves the efficiency of spatial semantic joint query and solves the current problems of geometric and semantic separation in real-scene 3D data, lack of semantics in index structure, and low efficiency of spatial semantic query.

[0022] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, this application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0023] The method for constructing and representing real-scene 3D data space and knowledge provided in this application can be applied to, for example... Figure 1In the application environment shown, terminal 102 communicates with server 104 via a network. A data storage system can store the data that server 104 needs to process. The data storage system can be set up independently, integrated into server 104, or placed in the cloud or on another server. Terminal 102 can send multi-source heterogeneous real-scene 3D data to server 104. After receiving the multi-source heterogeneous real-scene 3D data, server 104 performs unified coordinate preprocessing on the multi-source heterogeneous real-scene 3D data to obtain a standardized geometric point set and a semantic label set. Based on the standardized geometric point set and semantic label set, a 3D voxel mesh is constructed, and the fused feature vector of the fused geometric features and semantic features of each voxel in the 3D voxel mesh is calculated. Based on the fused feature vector, a semantically enhanced index tree is constructed. The semantically enhanced index tree includes several index tree nodes, each of which includes a spatial bounding box and a node feature center, and uses semantic information entropy as the basis for node splitting. Based on the semantically enhanced index tree, a real-scene 3D knowledge graph is constructed, and a bidirectional mapping relationship is established between each knowledge graph entity in the real-scene 3D knowledge graph and each index tree node. A joint query request is obtained, and a spatial semantic joint query is performed based on the joint query request, the semantically enhanced index tree, and the bidirectional mapping relationship, and the spatial semantic joint query result is output. Server 104 can feed back the spatial semantic joint query results to terminal 102. Furthermore, in some embodiments, the method for constructing and representing real-world 3D data space can also be implemented independently by server 104 or terminal 102. For example, terminal 102 can directly perform spatial semantic joint query processing on multi-source heterogeneous real-world 3D data, or server 104 can obtain multi-source heterogeneous real-world 3D data from the data storage system and perform spatial semantic joint query processing on the multi-source heterogeneous real-world 3D data.

[0024] The terminal 102 can be, but is not limited to, various desktop computers, laptops, smartphones, tablets, and IoT devices. The server 104 can be implemented using a standalone server or a server cluster consisting of multiple servers, or it can be a cloud server.

[0025] In one exemplary embodiment, such as Figure 2 As shown, a method for constructing and representing a real-scene 3D data space is provided. This method is executed by a computer device, specifically by a terminal or server alone, or by both a terminal and a server. In this embodiment, the method is applied to... Figure 1 Taking server 104 as an example, the following steps are included.

[0026] S1: Acquire multi-source heterogeneous real-world 3D data.

[0027] S2: Perform unified coordinate preprocessing on the multi-source heterogeneous real-world 3D data to obtain a standardized geometric point set and semantic tag set.

[0028] S3: Based on the standardized geometric point set and semantic label set, construct a three-dimensional voxel mesh, and calculate the fusion feature vector of the fused geometric features and semantic features of each voxel in the three-dimensional voxel mesh; wherein the fusion feature vector is obtained by weighted fusion of the geometric feature vector and the semantic feature vector, and the weights satisfy the sum of 1.

[0029] S4: Based on the fused feature vector, construct a semantically enhanced index tree; the semantically enhanced index tree includes several index tree nodes, each of which includes a spatial bounding box and a node feature center, and uses semantic information entropy as the basis for node splitting.

[0030] S5: Based on the semantically enhanced index tree, construct a real-world 3D knowledge graph and establish a bidirectional mapping relationship between each knowledge graph entity and each index tree node in the real-world 3D knowledge graph; wherein, the bidirectional mapping relationship is established by calculating the cosine similarity between the knowledge graph entity vector and the feature center of the index tree node, and a bidirectional link relationship is established when the cosine similarity is greater than a preset mapping threshold.

[0031] S6: Obtain the joint query request, and perform a spatial semantic joint query based on the joint query request, the semantically enhanced index tree, and the bidirectional mapping relationship, and output the spatial semantic joint query result; wherein, the spatial semantic joint query outputs a set of associated entities whose semantic matching score is higher than the semantic matching score threshold by calculating the weighted score of spatial matching degree and semantic matching degree.

[0032] By implementing steps S1 to S6 above, a multi-scale data space integrating geometric features and semantic vectors is constructed, and a bidirectional mapping between spatial index nodes and knowledge graph entities is established. This enables efficient spatial semantic joint queries, improves the efficiency of spatial semantic joint queries, and solves the current problems of geometric and semantic separation in real-world 3D data, lack of semantics in index structures, and low efficiency of spatial semantic queries.

[0033] As an optional implementation, in step S1, the multi-source heterogeneous real-scene 3D data includes at least one of oblique photogrammetry data, laser point cloud data, and vector semantic data.

[0034] As an optional implementation, in step S4, the semantically enhanced index tree is an octree index structure, and a split is performed when the semantic information entropy of the index tree node is greater than a preset threshold and has not reached the maximum depth.

[0035] As an optional implementation, in step S5, a real-world 3D knowledge graph is constructed based on the semantically enhanced index tree, and a bidirectional mapping relationship is established between each knowledge graph entity and each index tree node in the real-world 3D knowledge graph. This specifically includes the following steps.

[0036] S51: Based on the semantically enhanced index tree, construct a real-world 3D knowledge graph and define the mapping function between knowledge graph entities and index tree nodes in the real-world 3D knowledge graph.

[0037] S52: Calculate the cosine similarity value between each knowledge graph entity and the node feature center of each index tree node.

[0038] S53: Input the cosine similarity value into the mapping function to determine whether to establish a bidirectional mapping relationship between the knowledge graph entity and the index tree node based on the relationship between the cosine similarity value and the preset mapping threshold, specifically including the following situations.

[0039] (1) When the cosine similarity is greater than the preset mapping threshold, the mapping function outputs a valid association, and at this time a bidirectional mapping relationship is established between the knowledge graph entity and the index tree node.

[0040] (2) When the cosine similarity is less than or equal to the preset mapping threshold, the mapping function outputs an invalid association, and at this time, a bidirectional mapping relationship between the knowledge graph entity and the index tree node is not established.

[0041] As an optional implementation, step S6 involves obtaining a joint query request, performing a spatial semantic joint query based on the joint query request, the semantically enhanced index tree, and the bidirectional mapping relationship, and outputting the spatial semantic joint query result. This specifically includes the following steps.

[0042] S61: Receive a joint query request that includes spatial scope and semantic target.

[0043] S62: Based on the spatial range in the joint query request, locate the corresponding set of index tree nodes in the semantically enhanced index tree.

[0044] S63: For each index tree node in the set of index tree nodes, use the mapping set to find the corresponding associated entity; the mapping set is a set composed of bidirectional mapping relationships between each knowledge graph entity and each index tree node.

[0045] S64: Calculate the semantic matching score between each of the associated entities and the semantic target.

[0046] S65: Compare the semantic matching score with the semantic matching threshold to determine the associated entities corresponding to all semantic matching scores greater than the semantic matching threshold.

[0047] S66: Based on all the associated entities whose semantic matching scores are greater than the semantic matching threshold, construct an associated entity set as the spatial semantic joint query result.

[0048] In an exemplary embodiment, in order to explain in detail the technical solution provided by the embodiments of this application, a method for constructing and representing a real-scene three-dimensional data space is provided, which specifically includes the following steps.

[0049] Step 1: Acquisition and preprocessing of multi-source heterogeneous real-world 3D data.

[0050] In this embodiment of the application, the original multi-source heterogeneous real-scene 3D dataset is first defined as... Includes oblique photogrammetry data Laser point cloud data and vector semantic data Then, the original multi-source heterogeneous 3D reality data was subjected to coordinate system unification processing, and the target coordinate system was set as follows: The transformation matrix is For any spatial point coordinate in the original multi-source heterogeneous real-world 3D dataset... The new coordinates after the coordinate transformation The calculation formula is as follows.

[0051] = (1).

[0052] in, This represents the coordinate offset.

[0053] In this embodiment of the application, the preprocessed standard dataset is a standardized geometric point set and a semantic label set. , expressed as the following formula.

[0054] (2).

[0055] in, Represents a standardized set of geometric points and a set of semantic tags. For a geometric point set, Indicates the first Geometric points after coordinate transformation For semantic tag set, Indicates the first A semantic tag.

[0056] Step 2: Fusion of geometric and semantic features of real-world 3D entities.

[0057] In standardized geometric point sets and semantic tag sets Based on this, a three-dimensional voxel mesh is constructed. Set the voxel side length to For any voxel Its spatial location index is determined by three-dimensional integer coordinates (i, j, k). The fused feature vector of the voxel is calculated. The fused feature vector is composed of geometric feature vectors. and semantic feature vector Concatenated. Feature vector fusion. The formula for calculating is as follows.

[0058] (3).

[0059] in, Represents the fused feature vector. Represents geometric eigenvectors. Represents semantic feature vectors, and These are the weight coefficients of the geometric feature vector and the semantic feature vector, respectively. Geometric eigenvectors The semantic feature vector is composed of the statistical histogram of the normal directions of all points within the voxel. It consists of one-hot encoding of semantic labels, and the final output is a fused feature dataset. .

[0060] Step 3: Construct a semantically enhanced index tree.

[0061] Figure 3 The semantically enhanced index tree structure is illustrated, showing the structural details of storing node feature centers and knowledge graph link pointers (ptr_KG) within the index tree nodes. Specifically, the index tree nodes of the semantically enhanced index tree not only store the bounding box but also the node feature centers (i.e., node feature center vectors) for calculating cosine similarity. Furthermore, bidirectional link pointers directly point to knowledge graph entities. This achieves a deep integration of spatial indexing and semantic knowledge.

[0062] In this embodiment of the application, based on the fused feature dataset Construct an octree index structure, i.e., a semantically enhanced index tree. Define the index tree node as The spatial bounding box of each index tree node and node feature center Node feature center The update formula is as follows.

[0063] (4).

[0064] in, Indicates the first A fused feature vector, This represents the total number of voxels contained in the node.

[0065] Step 4: Construction and mapping of the real-world 3D knowledge graph.

[0066] Building knowledge graphs ,in For a knowledge graph entity set, Define a set of relations and define knowledge graph entities. With the 1 index tree node mapping function Calculate entity vectors in a knowledge graph. With node feature center The cosine similarity Sim is calculated using the following formula.

[0067] (5).

[0068] in, Representing entities in a knowledge graph With the 1 index tree node cosine similarity, This represents the magnitude of the vector. If the cosine similarity Sim is greater than a preset mapping threshold... λ Then, knowledge graph entities are established. With the 1 index tree node A bidirectional link is established; otherwise, no bidirectional link is established, ultimately forming a data space mapping set. .

[0069] In this embodiment of the application, the mapping function Using cosine similarity (Sim) as the core calculation basis, it is used to achieve quantitative matching and bidirectional association between knowledge graph entities and index tree nodes. Mapping function. The expression for is as follows.

[0070] (6).

[0071] in, Representing entities in a knowledge graph With the index tree nodes The mapping function, λ This is the preset mapping threshold.

[0072] In this embodiment, the cosine similarity Sim between the entity vector of the knowledge graph and the node feature center vector of the index tree node is first calculated to determine the degree of similarity between the two in the feature space; then, the cosine similarity Sim is input into the mapping function. , and preset mapping threshold λ The comparison includes two cases: (1) when the cosine similarity Sim is greater than the preset mapping threshold. λ At that time, the mapping function Output effective association, that is, the output of formula (6) is "1", at this time a bidirectional link is established between knowledge graph entities and index tree nodes; (2) when the cosine similarity Sim is less than or equal to the preset mapping threshold λ At that time, the mapping function The output is invalid, that is, the output of formula (6) is "0", and no bidirectional link is established at this time.

[0073] Step 5: Spatial semantic joint query based on data space.

[0074] Define a union query request ,in For spatial range, For semantic goals. First, utilize spatial scope. In semantically enhanced index trees Quickly locate the relevant set of index tree nodes. For each index tree node in the relevant set of index tree nodes, use the mapping set Find related entities The semantic matching score is calculated using a joint query formula. The formula for the joint query is as follows.

[0075] im (7).

[0076] in, Indicates the semantic matching score. I is the spatial weighting coefficient. For indicator functions, Indicates the first One index tree node, For spatial range, For semantic goals, Indicates related entities, im This represents the cosine similarity between the semantic target and the associated entity. The final output is the semantic matching score. The set of associated entities that are greater than the semantic matching degree threshold σ is taken as the result of the spatial semantic joint query.

[0077] The method for constructing and representing a real-scene 3D data space proposed in this application has the following advantages.

[0078] (1) Improve data integration: through the formula in step 2 This method deeply integrates geometric and semantic features at the voxel level, solving the problem of separating geometry and attributes in traditional methods and enabling each basic unit in the data space to have complete descriptive capabilities.

[0079] (2) Optimize the index structure: The semantically enhanced index tree constructed in step 3 can be used with the formula Semantic information entropy is introduced as the basis for splitting, whereby... The entropy value is the semantic information entropy. The total number of semantic categories. Assign semantic category number, For the node belonging to the first The proportion of semantically related voxels to the total number of voxels in a node changes the traditional splitting strategy that relies solely on geometric bounding boxes. This makes the hierarchical structure of the semantically enhanced index tree more consistent with human cognitive logic and improves retrieval accuracy.

[0080] (3) Improve query efficiency: Through the bidirectional mapping mechanism established in step 4 and the joint query formula in step 5, the parallel calculation of spatial range and semantic conditions is realized. Compared with the traditional serial method of first querying the space and then filtering the attributes, the computational complexity is greatly reduced and the response speed in large-scale scenarios is significantly improved.

[0081] This application takes the real-scene 3D data management of a smart park as an example to illustrate the specific implementation process of this application embodiment, which includes the following steps.

[0082] Step 1: Collect oblique photogrammetry data of the smart park and building outline vector semantic data The raw oblique photogrammetry data contains A spatial point, with coordinate system WGS84. Target coordinate system. The coordinate system is set to the CGCS2000 projection coordinate system. The transformation matrix is ​​obtained through parameter transformation calculation. and offset For any point Substitute into the formula = Calculate its new coordinates after coordinate transformation. For example, the original coordinates of a corner point of a building are (113.234, 22.567, 10.5). After calculation and transformation, the coordinates are... The result is (500034.2, 2445678.1, 10.5). After processing, a standardized geometric point set and a semantic tag set are generated. This includes a set of geometric points G and a set of semantic tags S for the corresponding buildings, roads, and vegetation.

[0083] Step 2: Set the voxel side length Meters, constructing a three-dimensional voxel mesh Traversing the standardized geometric point set and semantic tag set. The points in the data are assigned to the corresponding voxels. In the middle. For a certain building element It contains 150 laser points. Calculate the geometric eigenvectors of this voxel. and semantic feature vector (Architecture-related tag vector). This application embodiment sets weighting coefficients based on actual needs and experimental verification. , Substitute into the formula The fusion feature vector of the voxel is calculated. This generates a fusion feature dataset for the entire smart park. .

[0084] Step 3: Based on the fused feature dataset Constructing a semantically enhanced index tree based on an octree index structure Initialize the root node. Its spatial bounding box Covering the entire smart park. Using formulas Calculate the feature center of the root node ,in This represents the total number of voxels contained in a node. The percentages of buildings, roads, and vegetation within the root node are calculated, and the semantic information entropy is determined. Based on actual needs and experimental verification, this application sets an entropy threshold in its embodiments. =0.8, because the smart park contains various land features, the entropy value of semantic information entropy is... Therefore, the root node is split. This process is performed recursively until the node depth reaches [the specified value]. or entropy The final result is a semantically enhanced index tree containing 15,000 leaf nodes (i.e., index tree nodes).

[0085] Step 4: Construct a knowledge graph (KG) containing building entities within the smart park. Facility entity etc. For leaf nodes in a semantically enhanced index tree. (Representing an office building), its node feature center is Search for the corresponding building entity in the knowledge graph. Its vector is Substituting into the cosine similarity formula The calculated cosine similarity Sim = 0.95. Based on experimental verification and actual needs, this embodiment sets a preset mapping threshold. λ The preferred value is 0.85. Since 0.95 > Therefore, establish and The bidirectional mapping relationship is established and stored in the mapping set. .

[0086] Step 5: The user initiates a joint query request Q to find the spatial range of a rectangular area within the smart park. All commercial buildings within, i.e., semantic targets For "business". The system first uses a semantically enhanced index tree. Search and Spatial Range The corresponding set of index tree nodes intersecting .for Each node in the map set Find the corresponding related entities and use the union query formula. im Calculate the semantic matching score. This embodiment sets spatial weight coefficients based on experimental verification and practical needs. The preferred value is 0.5, and the preferred value for the semantic matching degree threshold σ is 0.6. If a node is located within the region and the associated entity is a commercial building, then the semantic matching degree score is calculated. =0.5×1+0.5×0.9=0.95. The final system output is all... A set of related entities with a value greater than 0.6 is used to form a list of related entities, enabling fast and accurate spatial semantic joint queries. Through the complete implementation of the above steps, this embodiment successfully constructs a real-world 3D data space for a smart park, realizing the entire process from raw data collection to knowledge graph construction.

[0087] The key statistical indicators of the embodiments of this application include the following.

[0088] Data scale: 44,634 spatial points were successfully processed, generating 3,986 voxel units.

[0089] Octree nodes: The constructed semantically enhanced index tree contains 400 leaf nodes, and the maximum depth of the semantically enhanced index tree reaches 8 levels.

[0090] Knowledge Graph Entities: Ten building entities were extracted and constructed from multi-source heterogeneous real-world 3D data, including five office buildings, four commercial buildings, and one residential building.

[0091] Two-way mapping relationship: A two-way mapping relationship between 790 entities and nodes was established, with an average similarity of 0.98.

[0092] Query performance: The average response time of spatial semantic joint query is 1.99 milliseconds, compared with the average response time of 4.00 milliseconds of the traditional serial query method. The response time of spatial semantic joint query in this application embodiment is significantly shortened, and the performance is improved by about 2 times.

[0093] Figure 4 The original data three-dimensional distribution is shown. This application embodiment is for the unified acquisition and preprocessing of multi-source heterogeneous real-scene three-dimensional data. The original data is unified to the CGCS2000 projection coordinate system through coordinate transformation. Figure 4 Red dots represent buildings, gray dots represent roads, and green dots represent vegetation, clearly demonstrating the spatial layout characteristics of the smart park. From Figure 4 It can be seen that the building point cloud is distributed at a higher spatial position (Z-axis direction), while the road point cloud is mainly distributed at the ground level (Z≈0), and the vegetation point cloud shows a medium height distribution. This spatial distribution characteristic verifies the completeness of data collection and the correctness of coordinate transformation.

[0094] Figure 5 The voxelization result is shown. This embodiment of the application converts discrete point cloud data into a regular three-dimensional voxel mesh structure by fusing the geometric and semantic features of real-world 3D entities. Each voxel unit in the 3D voxel mesh structure carries a fused feature vector. From Figure 5 It can be seen that the voxelized data retains the spatial distribution characteristics of the original point cloud while achieving structured organization of the data. The building area (red) has the highest voxel density, the road area (gray) is distributed in strips, and the vegetation area (green) is distributed in scattered spots, which is consistent with the actual distribution characteristics of ground features.

[0095] Figure 6 The spatial partitioning structure of an octree is shown. This embodiment of the application constructs a semantically enhanced index tree. The blue rectangles represent the spatial bounding boxes of the octree nodes, illustrating the spatial partitioning structure of the semantically enhanced index tree. From... Figure 6 It can be seen that the partition density of the octree is related to the spatial distribution characteristics of the data. In data-dense areas (such as building clusters), nodes are partitioned more finely; in data-sparse areas, nodes maintain a larger spatial range. This adaptive partitioning strategy effectively balances storage efficiency and query efficiency, allowing for further analysis of the construction quality of semantically enhanced index trees.

[0096] Figure 7 The distribution of node entropy values ​​is shown, illustrating the relationship between the semantic information entropy of each leaf node and the node depth. From Figure 7 As can be seen, the red dashed line represents the entropy threshold. =0.8. The semantic information entropy of most leaf nodes is between 0.4 and 0.8, indicating that the semantic categories within these nodes are relatively simple, and the splitting strategy effectively distinguishes different types of land features. A few high-entropy nodes (entropy value close to 1.0) still exist, which may be because these nodes are located in the boundary areas of different land features.

[0097] Figure 8 The diagram illustrates the distribution of entities within the knowledge graph, showcasing the spatial distribution of 10 building entities extracted from multi-source heterogeneous 3D real-world data. Figure 8 As can be seen, red markers represent commercial buildings, blue markers represent office buildings, and green markers represent residential buildings. Each entity is labeled with a name and type, realizing the transformation from geometric objects to semantic entities. This entity-level organization lays the foundation for subsequent knowledge retrieval and reasoning.

[0098] Figure 9 The results of the spatial semantic joint query are shown, where the yellow area represents the spatial scope of the query, and the red pentagram marks indicate the matched building entities. From Figure 9 As can be seen, the building entities within the query area were successfully identified and highlighted. The goal of this query was to find all commercial buildings within the specified area. The system matched 190 entities, with commercial buildings scoring 1.000 and non-commercial buildings (such as office buildings) scoring 0.650. This difference in scores effectively achieves semantic filtering, verifying the spatial semantic joint query capability of the method in this embodiment.

[0099] To verify the query efficiency of the method in the embodiments of this application, Figure 10 The performance differences between the traditional serial query method and the parallel query method of the embodiments of this application were compared. Figure 10 As can be seen, the query time of the traditional serial query method is 4.00 milliseconds, while the query time of the parallel query method in this embodiment is only 1.99 milliseconds, representing a performance improvement of approximately 2 times. This significant performance improvement stems from two key factors: first, by establishing a bidirectional mapping between the knowledge graph and the semantically enhanced index tree, parallel computation of spatial range query and semantic matching is achieved; second, based on the semantically enhanced index tree, relevant nodes can be quickly located, reducing unnecessary node traversal.

[0100] In order to gain a deeper understanding of the characteristics of multi-source heterogeneous real-world 3D data, this application embodiment statistically analyzed the voxel quantity distribution of each semantic type. Figure 11 The statistical distribution of voxel semantics is shown. From... Figure 11 As can be seen, there are 896 building voxels, 379 road voxels, and 2711 vegetation voxels. Building and vegetation voxels are the dominant types, which is consistent with the spatial characteristics of smart parks, which are dominated by buildings.

[0101] Figure 12 The depth distribution of the octree is shown, illustrating its depth distribution characteristics. From... Figure 12 It can be seen that leaf nodes are mainly concentrated between depths 3 and 6, with the largest number of leaf nodes at depth 5. This distribution characteristic indicates that the information entropy-based adaptive splitting strategy can automatically adjust the splitting depth according to the spatial and semantic features of the data, performing deeper splits in data-dense regions and maintaining a shallower depth in data-sparse regions.

[0102] Through the complete implementation process and visualization analysis of the embodiments described above, the feasibility and practicality of the methods in this application are fully verified. Specifically, regarding the improvement of data fusion, this application employs voxel-level feature fusion, where each voxel unit carries complete geometric and semantic information, enabling each basic unit in the data space to possess complete descriptive capabilities and successfully solving the problem of separating geometry and attributes in traditional methods. For the optimization of the index structure, this application utilizes an adaptive splitting strategy based on information entropy, making the semantically enhanced index tree structure more consistent with human cognitive logic. Deep splitting is performed in data-dense regions, while a shallow structure is maintained in data-sparse regions, achieving a balance between storage efficiency and query efficiency. To improve query efficiency, this application establishes a bidirectional mapping between a knowledge graph and a semantically enhanced index tree, enabling parallel computation of spatial range queries and semantic matching. Through the complete implementation process, detailed visualization analysis, and rigorous data verification, the effectiveness in constructing and representing real-world 3D data spaces is fully demonstrated, providing a solid technical foundation for the intelligent application of multi-source heterogeneous real-world 3D data.

[0103] 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 13As shown, this 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. The database stores multi-source heterogeneous real-world 3D data. The I / O interfaces are used for information exchange between the processor and external devices. The communication interface is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it implements a method for constructing and representing a real-world 3D data space.

[0104] Those skilled in the art will understand that Figure 13 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.

[0105] 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 steps in the above-described method embodiments.

[0106] In one exemplary embodiment, a computer-readable storage medium is provided storing a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0107] In one exemplary embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0108] 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. Moreover, the collection, use and processing of the relevant data are carried out in compliance with the relevant data protection laws and policies of the country where the location is located, and with the authorization granted by the owner of the corresponding device.

[0109] 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).

[0110] 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.

[0111] 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.

[0112] 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 and representing knowledge in a real-world 3D data space, characterized in that, include: Acquire multi-source heterogeneous real-world 3D data; The multi-source heterogeneous real-scene 3D data is subjected to unified coordinate preprocessing to obtain a standardized geometric point set and semantic tag set; Based on the standardized geometric point set and semantic tag set, a three-dimensional voxel mesh is constructed, and the fused geometric and semantic features of each voxel in the three-dimensional voxel mesh are calculated as a fused feature vector. Based on the fused feature vector, a semantically enhanced index tree is constructed; The semantically enhanced index tree includes several index tree nodes, each of which includes a spatial bounding box and a node feature center, and uses semantic information entropy as the basis for node splitting. Based on the semantically enhanced index tree, a real-world 3D knowledge graph is constructed, and a bidirectional mapping relationship is established between each knowledge graph entity and each index tree node in the real-world 3D knowledge graph. Obtain the joint query request, and perform a spatial semantic joint query based on the joint query request, the semantically enhanced index tree, and the bidirectional mapping relationship, and output the spatial semantic joint query result; Based on the semantically enhanced index tree, a real-world 3D knowledge graph is constructed, and a bidirectional mapping relationship is established between each knowledge graph entity and each index tree node in the real-world 3D knowledge graph, specifically including: Based on the semantically enhanced index tree, a real-world 3D knowledge graph is constructed, and a mapping function between knowledge graph entities and index tree nodes in the real-world 3D knowledge graph is defined. Calculate the cosine similarity value between the node feature centers of each knowledge graph entity and each index tree node; The cosine similarity value is input into the mapping function to determine whether to establish a bidirectional mapping relationship between the knowledge graph entity and the index tree node based on the relationship between the cosine similarity value and a preset mapping threshold. Specifically, this includes: When the cosine similarity is greater than the preset mapping threshold, the mapping function outputs a valid association, and at this time a bidirectional mapping relationship is established between the knowledge graph entity and the index tree node; When the cosine similarity is less than or equal to the preset mapping threshold, the mapping function outputs an invalid association, and at this time, a bidirectional mapping relationship between the knowledge graph entity and the index tree node is not established; Obtain the joint query request, and perform a spatial semantic joint query based on the joint query request, the semantically enhanced index tree, and the bidirectional mapping relationship, and output the spatial semantic joint query result, specifically including: Receive joint query requests that include spatial scope and semantic target; Based on the spatial range in the joint query request, locate the corresponding set of index tree nodes in the semantically enhanced index tree; For each index tree node in the set of index tree nodes, the corresponding associated entity is found using a mapping set; the mapping set is a set composed of bidirectional mapping relationships between each knowledge graph entity and each index tree node. Calculate the semantic matching score between each of the associated entities and the semantic target; The semantic matching scores of each entity are compared with the semantic matching threshold to determine the associated entities whose semantic matching scores are greater than the semantic matching threshold. Based on all the associated entities whose semantic matching scores are greater than the semantic matching threshold, a set of associated entities is constructed as the spatial semantic joint query result.

2. The method for constructing and representing real-scene 3D data space according to claim 1, characterized in that, The standardized geometric point set and semantic tag set The expression is: ； in, Represents a standardized set of geometric points and a set of semantic tags. For a geometric point set, Indicates the first Geometric points after coordinate transformation For semantic tag set, Indicates the first A semantic tag.

3. The method for constructing and representing real-scene 3D data space according to claim 1, characterized in that, The formula for calculating the fused feature vector is as follows: ； in, Represents the fused feature vector. Represents geometric eigenvectors. Represents semantic feature vectors, and These are the weight coefficients of the geometric feature vector and the semantic feature vector, respectively. .

4. The method for constructing and representing real-scene 3D data space according to claim 1, characterized in that, The semantically enhanced index tree is an octree index structure. When the semantic information entropy of the index tree node is greater than a preset threshold but has not reached the maximum depth, a split is performed.

5. The method for constructing and representing real-scene 3D data space according to claim 1, characterized in that, The semantic matching score between each of the associated entities and the semantic target is calculated using the following formula: im ； in, Indicates the semantic matching score. I is the spatial weighting coefficient. For indicator functions, Indicates the first One index tree node, For spatial range, For semantic goals, Indicates related entities, im It represents the cosine similarity between the semantic target and the associated entity.

6. A computer device, comprising: A memory, a processor, and a computer program stored in the memory and capable of running on the processor, characterized in that the processor executes the computer program to implement the method for constructing and representing a real-scene three-dimensional data space according to any one of claims 1-5.

7. 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 and representing real-scene three-dimensional data space according to any one of claims 1-5.

8. A computer program product, comprising a computer program, characterized in that, When executed by a processor, the computer program implements the method for constructing and representing real-scene three-dimensional data space according to any one of claims 1-5.

Images