CN-121999182-A - Urban digital twin scene static LOD processing method, equipment and storage medium

CN121999182ACN 121999182 ACN121999182 ACN 121999182ACN-121999182-A

Abstract

Aiming at the problems that in the prior art, a three-dimensional model has easily damaged semantic structures, distorted geometric features and difficult control of multi-level errors in the LOD simplification process, the application discloses a static LOD processing method, equipment and a storage medium for an urban digital twin scene, and belongs to the technical field of computer graphics and three-dimensional modeling. The method comprises the steps of carrying out semantic recognition and region division on a three-dimensional grid model of an urban building to construct a semantic labeling grid, establishing a simplified control model based on the semantic labeling grid to determine a simplified constraint, executing grid simplification processing under the constraint, carrying out geometric feature preservation processing on a simplified result, carrying out error evaluation on each LOD hierarchical model and self-adaptive correction, and carrying out substitution expression on semantic components which cannot be reserved to generate multi-level LOD model data. The method reduces the complexity of the model, realizes the cooperative maintenance of semantic consistency and geometric accuracy, and is suitable for efficient rendering and visualization application of urban three-dimensional scenes.

Inventors

SHEN XI
MA PIN
WANG HENGYU
BA TE
RAN JU
JIANG XUELIN
HUANG QIANYANG

Assignees

云南省地图院

Dates

Publication Date: 20260508
Application Date: 20260410

Claims (10)

1. A static LOD processing method for an urban digital twin scene is characterized by comprising the following steps: Carrying out semantic recognition and region division on building components in the urban building three-dimensional grid model to obtain a semantic annotation grid containing a plurality of semantic regions; Based on the semantic annotation grids, establishing a semantic-related simplified control model, and enabling each semantic region to correspond to a corresponding simplified constraint condition; Based on the simplified control model, performing semantic constraint grid simplification processing on the three-dimensional grid model of the urban building to obtain a simplified grid model; Executing geometric feature maintaining processing on the simplified grid model to maintain building contour features in the simplified grid model stable; And carrying out error evaluation on the simplified grid model of each LOD level based on the three-dimensional grid model of the urban building, and carrying out correction control on a local area with the error exceeding a preset threshold according to an evaluation result.
2. The method for processing the static LOD of the urban digital twin scene according to claim 1, wherein the semantic recognition and the region division are carried out on the building components in the urban building three-dimensional grid model to obtain a semantic annotation grid containing a plurality of semantic regions, and the method comprises the following steps: extracting local geometric descriptors including a surface normal vector, a Gaussian curvature, an average curvature and a surface area by taking a triangular surface patch as a basic unit; constructing adjacent patches into a surface adjacency graph, and expressing local geometric topological relations by taking the patches as nodes and shared edges as edges; performing message transmission and feature aggregation on the surface adjacency graph by using a multi-layer graph convolution network, and endowing each surface patch with a building element category label to obtain a surface patch semantic category label set; Performing connected domain analysis on the facial mask semantic category label set, and combining the facial mask with adjacent space and consistent categories to form a preliminary semantic region to obtain a preliminary semantic annotation grid; And extracting boundary line segments along shared edges of adjacent semantic regions aiming at boundary transition regions, judging a hard boundary and a soft boundary according to a normal vector included angle and a curvature gradient of the surface patch, executing sharp cutting and boundary alignment on the hard boundary, executing gradual transition distribution on the soft boundary, and distributing the surface patch to the adjacent regions to finally obtain the semantic annotation grid with definite space attribution and continuous boundary.
3. The method for processing the static LOD of the urban digital twin scene according to claim 2, wherein a simplified control model related to semantics is established based on the semantic annotation grid, so that each semantic region corresponds to a corresponding simplified constraint condition, comprising the following steps: Extracting characteristic parameters from four dimensions of the relative area ratio of the components, the average observation frequency, the structure identification information quantity and the outline saliency degree of each semantic region and the corresponding building components of each semantic region in the semantic annotation grid, determining the weight of each dimension by combining with the visual attention data of a large-scale urban scene user, and calculating to obtain the comprehensive visual importance score of each semantic region; dividing the semantic region into three simplification constraint levels of a high-fidelity level, a standard simplification level and an aggressive simplification level according to the comprehensive visual importance score; Setting corresponding simplification constraint for semantic areas of each simplification constraint level; Based on the simplified constraint level and the simplified constraint, a semantic related simplified control model is established.
4. A method for processing static LOD of urban digital twin scene according to claim 3, characterized in that based on the simplified control model, the grid simplifying process of semantic constraint is performed on the three-dimensional grid model of urban architecture, to obtain a simplified grid model, comprising the following steps: Under the guidance of the simplified control model, identifying candidate folding edges in the three-dimensional grid model of the urban building, and calculating the geometric secondary error cost value of each folding edge as a basic cost component; Checking semantic regions of vertexes at two ends of each candidate folding edge according to the simplified control model, and if the semantic regions belong to different regions, overlapping semantic cross-domain penalty coefficients on basic cost; Adding a sparse penalty term related to the residual vertex density of the region to the candidate folding edges in the high-fidelity semantic region based on the simplified control model, so that the folding rate of the high-fidelity region is reduced in a self-adaptive manner in the simplified process; after each folding operation is completed, updating the cost of all candidate folding edges in the vicinity of the affected vertex according to the simplified control model, and adjusting the sequence in a priority queue to ensure that the next folding operation is advanced according to the current optimal strategy; And circularly executing folding operation under the drive of the simplified control model and the priority queue, and realizing the simplification treatment of the three-dimensional grid model of the urban building under the condition of considering both geometric precision and semantic structure protection until the preset simplification target is reached, thereby obtaining the simplified grid model.
5. The method for processing static LOD of an urban digital twin scene according to claim 1, wherein the method for performing geometric feature preserving processing on the simplified grid model is as follows: extracting building outline feature lines from the three-dimensional grid model of the urban building; Carrying out sectional fold line fitting on the building outline characteristic line to obtain a fold line segment, and constructing a three-dimensional characteristic constraint frame by taking the end points of the fold line segment as anchor points; In the grid simplification process, applying displacement limitation to vertexes falling into the influence range of the three-dimensional feature constraint frame, and adaptively maintaining building volume sensing and contour continuity under different LOD levels; when the area where the broken line segment is located loses the vertex support due to simplification, reserved vertices are inserted into the broken line segment, so that the outline features in the simplified grid are ensured to exist continuously, and meanwhile, the vision and the structural continuity are maintained under different LOD levels.
6. The method for processing static LOD of urban digital twin scene according to claim 1, wherein the method for evaluating error of simplified grid model of each LOD level based on three-dimensional grid model of urban architecture is as follows: uniformly determining a plurality of sampling points on the surface of the simplified grid model of each LOD level; Calculating the shortest undirected point-to-plane distance from each sampling point to the surface of the three-dimensional grid model of the urban building, and representing geometric distance errors by using root mean square of absolute values of distances of all sampling points; comparing the unit normal vector included angles of the simplified grid model and the three-dimensional grid model of the urban building at the same sampling point, and taking the deviation mean value of the included angles to represent the normal deflection error; Calculating the closed volumes of all semantic areas in the three-dimensional grid model and the simplified model of the urban building respectively, and weighting and calculating the volume change rate according to the simplified constraint level of each semantic area to obtain a semantic volume error; and carrying out normalized weighting on the geometric distance error, the normal deflection error and the semantic volume error to obtain a comprehensive error score.
7. The method for processing static LOD of urban digital twin scene according to claim 1, wherein the error evaluation is performed on the simplified grid model of each LOD level based on the three-dimensional grid model of urban building, and the local area with the error exceeding the preset threshold is corrected and controlled according to the evaluation result, comprising the following steps: Identifying a local area with the error exceeding a preset threshold, marking all the illegal patches by taking triangular patches as units, expanding a preset buffer range based on a space bounding box of the illegal patches, and determining a correction scope; In the correction action domain, executing Loop subdivision for each violation triangular patch to divide the midpoint of each edge into four sub triangular patches uniformly so as to increase vertex density and improve local geometric adjustment freedom degree; The method comprises the steps of taking the surface of a local area corresponding to a three-dimensional grid model of the urban building as an optimization target, and adopting a Laplace surface deformation method to iteratively optimize the positions of vertexes in a correction action area so that the projection positions of the vertexes to the surface of the three-dimensional grid model of the urban building are gradually closed; and when the local region error still exceeds a preset threshold value, repeatedly executing illegal region positioning, self-adaptive grid refinement and local surface optimization until the local region error is converged below the preset threshold value.
8. The method for processing the static LOD of the urban digital twin scene according to claim 1, further comprising the steps of: when the LOD level reaches a preset simplification degree, judging whether semantic components which cannot retain geometric structures exist in the current simplified grid model or not; If a semantic component which cannot keep the geometric structure exists, representing the semantic component in an alternative expression mode, and finally obtaining multi-level LOD model data by combining the simplified grid models of all LOD levels; If the semantic components which cannot keep the geometric structure do not exist, the simplified grid models of all LOD levels are directly integrated, and multi-level LOD model data are obtained.
9. An urban digital twin scene static LOD processing device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the computer program being configured to implement the steps of the urban digital twin scene static LOD processing method according to any one of claims 1 to 8.
10. A storage medium, characterized in that the storage medium is a computer readable storage medium, on which a computer program is stored, which computer program, when being executed by a processor, implements the steps of the urban digital twinning scene static LOD processing method according to any one of claims 1 to 8.

Description

Urban digital twin scene static LOD processing method, equipment and storage medium Technical Field The application relates to the technical field of computer graphics and three-dimensional modeling, in particular to a method, equipment and a storage medium for processing static LOD of an urban digital twin scene. Background With the rapid development of urban information model (CIM) and digital twin technology, urban three-dimensional scenes are widely applied in the fields of planning and design, operation and maintenance management, emergency command, visual display and the like. Because the urban building model has huge data scale and complex geometric structure, the problems of large storage pressure, low loading efficiency, insufficient rendering performance and the like are often faced in the network transmission and real-time rendering process, and the model is usually required to be expressed in a grading manner through a multi-level detail (LOD) technology so as to realize the balance of precision and performance under different application scenes. The existing static LOD processing method is mostly based on a pure geometric simplification algorithm, and performs vertex folding or surface patch merging processing on the grid, so that the complexity of a model can be reduced, but the semantic attribute and the structural difference of building components are not fully considered, and key component distortion, fuzzy contour features and semantic information damage are easily caused. Meanwhile, in the multi-level generation process, a unified semantic constraint mechanism and an error control means are lacking, geometric continuity and visual consistency are difficult to maintain among different LOD levels, and the reality and usability of the digital twin scene are affected. In addition, aiming at the problems of local error accumulation and important component loss in the simplification process, the prior art lacks an effective closed-loop evaluation and self-adaptive correction mechanism, and is difficult to reasonably express and replace semantic components which cannot keep geometric structures, so that the integrity and expression capacity of the multi-level model data are affected. In summary, how to achieve high-quality, multi-level and sustainable optimization static LOD model generation by considering semantic structure protection, geometric feature maintenance and multi-dimensional error controllability in the three-dimensional grid simplification process of urban buildings has become a technical problem to be solved. Disclosure of Invention In order to overcome a series of defects existing in the prior art, the application aims to provide a static LOD processing method of an urban digital twin scene, which comprises the following steps: Carrying out semantic recognition and region division on building components in the urban building three-dimensional grid model to obtain a semantic annotation grid containing a plurality of semantic regions; Based on the semantic annotation grids, establishing a semantic-related simplified control model, and enabling each semantic region to correspond to a corresponding simplified constraint condition; Based on the simplified control model, performing semantic constraint grid simplification processing on the three-dimensional grid model of the urban building to obtain a simplified grid model; Executing geometric feature maintaining processing on the simplified grid model to maintain building contour features in the simplified grid model stable; And carrying out error evaluation on the simplified grid model of each LOD level based on the three-dimensional grid model of the urban building, and carrying out correction control on a local area with the error exceeding a preset threshold according to an evaluation result. In some embodiments, semantic recognition and region division are performed on building components in the urban building three-dimensional grid model to obtain a semantic annotation grid containing a plurality of semantic regions, and the method comprises the following steps: extracting local geometric descriptors including a surface normal vector, a Gaussian curvature, an average curvature and a surface area by taking a triangular surface patch as a basic unit; constructing adjacent patches into a surface adjacency graph, and expressing local geometric topological relations by taking the patches as nodes and shared edges as edges; performing message transmission and feature aggregation on the surface adjacency graph by using a multi-layer graph convolution network, and endowing each surface patch with a building element category label to obtain a surface patch semantic category label set; Performing connected domain analysis on the facial mask semantic category label set, and combining the facial mask with adjacent space and consistent categories to form a preliminary semantic region to obtain a preliminary semantic annotation grid; And extra