A multi-source three-dimensional geographic information data fusion method and system

Abstract

The present application relates to the cross technical field of computer graphics, geographic information technology, digital twinning and intelligent transportation, and discloses a kind of fusion method and system of multi-source three-dimensional geographic information data, including obtaining the oblique photography surface model to be fused, fine three-dimensional structure model and digital elevation model;Obtain spatially aligned real scene model and terrain model;Determine the first edge area and the second edge area based on the profile and spatial position of fine three-dimensional structure model respectively;The driving reconstruction related to fine three-dimensional structure model is carried out to terrain model;The transition deformation related to fine three-dimensional structure model is carried out to real scene model to make the deformed real scene model smoothly connect with fine three-dimensional structure model in the second edge area;Integrate fine three-dimensional structure model, the deformed real scene model and the reconstructed terrain model.The present application realizes the seamless in real geometric sense, and lays a reliable geometric foundation for high-precision engineering application.

Description

Technical Field

[0001] This invention relates to the interdisciplinary fields of computer graphics, geographic information technology, digital twins and intelligent transportation, and specifically to a method and system for fusing multi-source three-dimensional geographic information data. Background Technology

[0002] With the development of intelligent transportation and digital twin technologies, constructing high-precision, measurable, and analyzable 3D digital reality models throughout the entire lifecycle of highway engineering has become a core requirement. Currently, this field mainly relies on three types of data and technologies:

[0003] 1. Oblique photogrammetry: It can quickly and automatically generate large-scale, high-precision 3D models of real-world scenes (i.e., "oblique photogrammetric surface models"), which realistically reflect the appearance and spatial location of surface features. However, its models are continuous "skin" models, which cannot separate individual objects, and the bottom of the model (such as under a bridge or inside a tunnel) often has data holes and distortions.

[0004] 2. Refined 3D Modeling Technology: This involves manually constructing refined 3D structural models (BIM models) of key structures (such as bridges and tunnels) using specialized software (e.g., Bentley, Revit). These models are structurally complete, highly accurate, and can be supplemented with rich engineering attribute information.

[0005] 3. Digital Elevation Model (DEM): Represents the undulation of the land surface and is the base topographic data for engineering design.

[0006] However, directly combining data from different sources and with different characteristics results in a serious disconnect between the two sets of data:

[0007] Geometric conflicts: Inconsistencies such as intersections, floating, or penetrations occur between the BIM model and the oblique photogrammetry model and the terrain at the junctions.

[0008] Poor blending effect: Simple model overlay cannot handle geometric and texture cracks at the joints, resulting in an unrealistic visual appearance.

[0009] Low level of automation: Existing fusion methods mostly rely on manual editing and repair, which is labor-intensive, inefficient, costly, and difficult to maintain consistent quality.

[0010] Weak application support: Due to inaccurate integration, it is impossible to perform accurate earthwork volume calculation, sight distance analysis, conflict detection and other professional engineering applications on this basis.

[0011] Therefore, there is an urgent need for a method that can automatically and accurately solve the problem of seamless fusion of multi-source 3D geographic information data. Summary of the Invention

[0012] The present invention aims to overcome the above-mentioned defects of the prior art. The present invention provides a method and system for fusing multi-source three-dimensional geographic information data, which solves the problem of seamless fusion between three-dimensional models of different sources and different accuracies in highway engineering scenarios, and forms a complete, unified, visual, measurable and analyzable three-dimensional digital base.

[0013] This invention is achieved through the following technical solution:

[0014] A method for fusing multi-source 3D geographic information data includes:

[0015] Obtain the oblique photogrammetry surface model, the detailed 3D structural model, and the digital elevation model to be fused, and unify the oblique photogrammetry surface model, the detailed 3D structural model, and the digital elevation model into the same spatial reference system;

[0016] Global geometric matching is performed on the oblique photogrammetric surface model and the digital elevation model after coordinate unification to obtain a spatially aligned real-scene model and terrain model;

[0017] Based on the outline and spatial position of the refined 3D structural model, the first junction area between the refined 3D structural model and the terrain model, and the second junction area between the refined 3D structural model and the real scene model are determined respectively.

[0018] The terrain model is reconstructed in a driving manner related to the fine three-dimensional structure model, so that the reconstructed terrain model fits the fine three-dimensional structure model in the first edge region;

[0019] The real-world model is subjected to a transitional deformation related to the fine three-dimensional structure model, so that the deformed real-world model is smoothly connected to the fine three-dimensional structure model in the second edge region, and the deformed real-world model is naturally connected to the reconstructed terrain model.

[0020] The detailed 3D structural model, the deformed real-world model, and the reconstructed terrain model are integrated to generate a seamless, unified 3D model.

[0021] As an optimization, the specific process of unifying the oblique photogrammetry surface model, the detailed 3D structural model, and the digital elevation model into the same spatial reference system is as follows:

[0022] Determine the spatial reference frame corresponding to each of the oblique photogrammetry surface model, the detailed three-dimensional structural model, and the digital elevation model;

[0023] Based on the transformation relationship between their respective spatial reference systems, the oblique photogrammetry surface model, the detailed three-dimensional structural model, and the digital elevation model are uniformly transformed into the same target engineering coordinate system.

[0024] As an optimization, the specific process of performing global geometric matching between the coordinate-unified oblique photogrammetric surface model and the digital elevation model is as follows:

[0025] Extract the digital surface model from the oblique photogrammetry surface model;

[0026] Calculate the spatial transformation parameters between the digital surface model and the digital elevation model;

[0027] The oblique photography surface model is geometrically corrected based on the spatial transformation parameters.

[0028] As an optimization, based on the outline and spatial position of the refined 3D structural model, the specific process for determining the first boundary region between the refined 3D structural model and the terrain model, and the second boundary region between the refined 3D structural model and the real-world model, is as follows:

[0029] Establish a spatial index that includes the detailed 3D structural model, the terrain model, and the real-scene model;

[0030] Based on the spatial index, the spatial proximity region between the outer surface of the fine three-dimensional structural model and the terrain model and the real scene model is calculated through geometric buffer analysis or ray intersection analysis.

[0031] The calculated spatial neighbor regions are defined as the first border region and the second border region, respectively;

[0032] Based on the definitions of the first and second border regions, corresponding terrain data blocks and real-scene data blocks to be processed are segmented from the terrain model and the real-scene model, respectively.

[0033] As an optimization, the specific process of performing a driving reconstruction of the terrain model related to the refined three-dimensional structure model is as follows:

[0034] Obtain engineering semantic constraints related to the refined three-dimensional structural model, wherein the engineering semantic constraints include the slope design parameters of the refined three-dimensional structural model at the first junction area;

[0035] The base contour of the refined three-dimensional structural model in the first junction area is used as a geometric hard constraint;

[0036] Based on the aforementioned engineering semantic constraints, a set of elevation control points conforming to engineering design specifications is generated within the scope of the aforementioned geometric hard constraints.

[0037] Using the geometric hard constraints and the set of elevation control points as input, the terrain model is reconstructed using a constrained triangulation network to generate a reconstructed terrain that closely fits the base of the fine three-dimensional structural model and conforms to the engineering slope morphology.

[0038] As an optimization, the slope design parameters include at least one of the following: the design slope ratio of the slope, the outline dimensions of the structure, or the design elevation of the control points.

[0039] As an optimization, the specific process of performing transitional deformations on the real-world model related to the refined 3D structural model is as follows:

[0040] Based on the second edge region, the deformable region adjacent to the real scene model and the fine three-dimensional structure model is defined;

[0041] Obtain the boundary conditions of the region to be deformed, wherein the boundary conditions include the original geometric boundary on one side of the real scene model and the side surface geometric boundary of the fine three-dimensional structure model in the second junction region;

[0042] Based on the boundary conditions, the area to be deformed is smoothly deformed using a mesh deformation algorithm to generate a geometric transition surface; wherein, the geometric transition surface is smoothly connected to the side surface of the fine three-dimensional structure model; and the transitional deformation also allows the real scene model and the reconstructed terrain model to connect naturally at the boundary.

[0043] As an optimization, the fusion method also includes:

[0044] The textures at the junctions of the deformed real-world model, the refined 3D structural model, and the reconstructed terrain model are fused to achieve visual seamlessness.

[0045] As an optimization, the specific process of fusing the textures at the junctions of the deformed real-world model, the refined 3D structural model, and the reconstructed terrain model is as follows:

[0046] For the geometric transition surface and the boundary area between the real scene model and the reconstructed terrain model, the best texture image of the corresponding viewpoint is selected and extracted from the original image of the oblique photogrammetric surface model.

[0047] The extracted texture image is mapped onto the mesh corresponding to the geometric transition surface and the boundary region;

[0048] At the seams after texture mapping, an image fusion algorithm is used to progressively blend the colors and brightness of the textures of the real scene model, the fine three-dimensional structure model, and the reconstructed terrain model to eliminate visual seams.

[0049] For areas with missing textures caused by model occlusion, the surrounding texture information is used to repair or fill them.

[0050] This invention also discloses a multi-source three-dimensional geographic information data fusion system for performing the aforementioned multi-source three-dimensional geographic information data fusion method, comprising:

[0051] The coordinate unification module is used to acquire the oblique photogrammetry surface model, the fine three-dimensional structure model and the digital elevation model to be fused, and to unify the oblique photogrammetry surface model, the fine three-dimensional structure model and the digital elevation model into the same spatial reference system;

[0052] The spatial alignment module is used to perform global geometric matching on the coordinate-unified oblique photogrammetric surface model and the digital elevation model to obtain a spatially aligned real-world model and terrain model.

[0053] The fusion region determination module is used to determine, based on the outline and spatial position of the fine three-dimensional structure model, the first boundary region between the fine three-dimensional structure model and the terrain model, and the second boundary region between the fine three-dimensional structure model and the real scene model.

[0054] The terrain reconstruction module is used to perform a driving reconstruction of the terrain model in relation to the fine three-dimensional structure model, so that the reconstructed terrain model fits the fine three-dimensional structure model in the first edge region;

[0055] The real-scene deformation module is used to perform transitional deformation on the real-scene model related to the fine three-dimensional structure model, so that the deformed real-scene model can be smoothly connected with the fine three-dimensional structure model in the second edge area, and the deformed real-scene model can be naturally connected with the reconstructed terrain model.

[0056] The fusion module is used to integrate the detailed 3D structural model, the deformed real-world model, and the reconstructed terrain model to generate a seamless, integrated 3D model.

[0057] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0058] This invention eliminates systematic deviations between the oblique photogrammetry model and the terrain through global geometric matching, solving the problem of suspension or penetration at the macro scale. It also precisely handles local geometric cracks between structure-ground and structure-scene / scene-ground relationships through driven reconstruction and transitional deformation, respectively. This forms a two-layered guarantee mechanism of overall correction + local refinement, ensuring that the fused model has no geometric conflicts at any scale, achieving true geometric seamlessness and laying a reliable geometric foundation for high-precision engineering applications.

[0059] This invention automatically identifies and segments fusion regions based on spatial indexing and geometric analysis algorithms; it uses engineering semantic parameters (such as slope ratio) as input to drive terrain reconstruction; and it automatically generates geometric transition surfaces using a mesh deformation algorithm. This significantly reduces the time-consuming, labor-intensive, and subjective manual editing work required in traditional methods. This invention can automatically locate problems, understand design intent, and perform corrections, resulting in high processing efficiency and consistent output quality, providing a feasible technical path for model fusion in large-scale engineering scenarios.

[0060] This invention introduces engineering semantic constraints to drive terrain reconstruction, ensuring that the generated terrain (such as bridge abutment slopes) not only conforms to the structure but also strictly adheres to engineering design specifications (such as design slope ratio). This transforms the fused result from a simple geometric shape into a digital expression embodying engineering rules. Consequently, the fused 3D model can be directly used for professional engineering applications such as compliance checks, accurate earthwork calculations, and slope stability analysis, fundamentally enhancing the data value and application depth.

[0061] This invention, based on geometric seamlessness, performs texture fusion and reconstruction, including extracting textures from the optimal viewpoint, using the Poisson fusion algorithm to eliminate color differences at seams, and intelligently repairing occluded areas. This not only eliminates geometric cracks but also visual seams and texture abrupt changes. The texture transitions in the fused areas are natural, the colors are harmonious, and the overall model has a strong visual realism, meeting the needs of high-fidelity visualization, promotional displays, and immersive experiences.

[0062] This invention constructs a complete and standardized processing flow from data preparation → global matching → region identification → semantic-driven geometric correction → texture fusion → result generation. The flow is logically clear and each step is highly modular. It is not only applicable to the highway bridge in the embodiments, but its core idea (driven reconstruction and transitional deformation based on the structural model) can be easily extended to various highway structures such as tunnels, culverts, retaining walls, and toll stations, and even to multi-source 3D model fusion scenarios in similar fields such as railways, water conservancy, and municipal engineering, demonstrating strong versatility. Attached Figure Description

[0063] The accompanying drawings, which are included to provide a further understanding of embodiments of the invention and form part of this application, do not constitute a limitation thereof. In the drawings:

[0064] Figure 1 This is a flowchart of a method for fusing multi-source three-dimensional geographic information data according to the present invention;

[0065] Figure 2 This is a diagram showing the fusion effect achieved by the method of the present invention. Detailed Implementation

[0066] To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the embodiments and accompanying drawings. The illustrative embodiments and descriptions of the present invention are only used to explain the present invention and are not intended to limit the present invention.

[0067] This embodiment 1 uses the fusion of a detailed 3D structural model (BIM model) of a highway bridge with an oblique photogrammetric surface model of the bridge area and a high-precision digital elevation model (DEM) as an example to provide a method for fusing multi-source 3D geographic information data, and specifically and completely illustrates the implementation process of the present invention. (Refer to...) Figure 1 The method flowchart shown in this embodiment includes the following steps:

[0068] S1. Obtain the oblique photogrammetry surface model, the fine three-dimensional structure model, and the digital elevation model to be fused, and unify the oblique photogrammetry surface model, the fine three-dimensional structure model, and the digital elevation model into the same spatial reference system.

[0069] First, obtain the three core data sources to be fused: a detailed 3D structural model of the bridge (usually a BIM model, in a format such as .ifc), an oblique photogrammetric surface model of the bridge area (in a format such as .osgb), and a digital elevation model (DEM) covering the bridge area.

[0070] Before proceeding with the coordinate transformation process, the raw data needs to undergo necessary preprocessing to ensure that the data quality meets the requirements of the fusion algorithm.

[0071] For oblique photogrammetric surface models: check and fix any topological errors (such as non-manifold edges, self-intersecting patches), and remove redundant patches or islands that are far from the main model due to noise.

[0072] For detailed 3D structural models (BIM): Clean up redundant components or internal details in the model that are not related to geometric integration, and optimize the model structure.

[0073] For digital elevation models: check for data holes and fill them with appropriate interpolation; convert them to the specific format required by the algorithm (such as regular grid or triangular mesh TIN format).

[0074] After the above preprocessing, these data sources are in their optimal state. However, because these data originate from different acquisition methods and software platforms, their initial spatial reference systems differ. For example, oblique photogrammetry results may use the WGS-84 geographic coordinate system, BIM models may use a project-specific coordinate system, and DEMs may use the national geodetic coordinate system. Direct overlay will result in significant spatial location deviations.

[0075] Therefore, the core of this step is to establish a unified spatial benchmark. The specific process is as follows:

[0076] S1.1 Determine the spatial reference frame corresponding to each of the oblique photogrammetry surface model, the detailed three-dimensional structure model, and the digital elevation model;

[0077] S1.2 Based on the known measurement control points, feature points, or transformation parameters between coordinate systems, calculate the precise transformation relationship between the original coordinate system of each model and the selected target engineering coordinate system (e.g., CGCS2000 3° projection coordinate system).

[0078] S1.3. Based on this transformation relationship, the oblique photogrammetry surface model, the detailed 3D structural model, and the digital elevation model are all transformed to the unified target engineering coordinate system. This step provides a crucial geometric consistency foundation for all subsequent accurate spatial calculations and fusion operations.

[0079] S2. Perform global geometric matching on the oblique photogrammetric surface model and the digital elevation model after coordinate unification to obtain a spatially aligned real-world model and terrain model.

[0080] Based on coordinate unification, it is necessary to address potential systemic deviations between the oblique photogrammetry model and the terrain model (such as overall settlement or tilting caused by aerial triangulation accuracy). This step aims to perform macroscopic geometric correction, and the specific process is as follows:

[0081] S2.1 Automatically extract the digital surface model (DSM) from the coordinate-transformed oblique photogrammetric surface model. This DSM represents the elevation information of the top surfaces of all ground features, including the land surface, vegetation, and buildings.

[0082] S2.2. Perform a global comparison and matching analysis between the extracted DSM and the digital elevation model (DEM, representing bare land terrain) within the same range. By applying surface matching algorithms (such as the ICP algorithm or its variants), calculate the systematic deviation parameters between the two in three-dimensional space, which typically include translation, rotation, and scaling.

[0083] S2.3. Based on the calculated spatial transformation parameters, a comprehensive and rigid geometric correction is performed on the original oblique photogrammetry surface model. After this correction, the oblique photogrammetry model achieves a high degree of consistency with the terrain model on a macro scale, eliminating large-scale "floating" or "penetrating" phenomena. At this point, the corrected model is referred to as the reality model, and the terrain data serving as the base is called the terrain model. This step resolves systematic errors between different data sources, laying the foundation for subsequent local fine-grained fusion and preparing for the next step of local fine-grained fusion.

[0084] S3. Based on the outline and spatial position of the refined three-dimensional structural model, determine the first junction area between the refined three-dimensional structural model and the terrain model, and the second junction area between the refined three-dimensional structural model and the real scene model.

[0085] One of the core features of this invention is the automated location of areas requiring fine-grained processing. Driven by an immutable, high-resolution 3D structural model (bridge BIM), it automatically identifies the junctions between the bridge and its surrounding environment. The specific process is as follows:

[0086] S3.1 To improve the efficiency of spatial querying of massive data, a unified spatial index (such as an octree index) is established, which includes a detailed three-dimensional structural model, terrain model and real scene model.

[0087] S3.2 Based on this index, using algorithms such as geometric buffer analysis (identifying adjacent patches within a certain distance) and ray intersection analysis (accurately detecting contact or penetration), the spatial proximity areas between the outer surface of the BIM model (focusing on the bottom surface of the bridge pier, the bottom surface of the bridge abutment, and the side surface) and the terrain model and the reality model are automatically calculated.

[0088] The specific process is as follows:

[0089] The first stage (coarse screening) involves creating a 3D geometric buffer zone of a certain distance on the outer surface of the refined 3D structural model (focusing on the bottom surface of the bridge piers, the bottom surface of the abutments, and the sides). Using spatial indexing, triangular facets of the terrain model and the real-world model located within this buffer zone are quickly retrieved, and the areas containing these facets are preliminarily identified as potential neighboring regions. The second stage (refined assessment) involves emitting dense rays from feature points (such as vertices or facet centers) on the outer surface of the refined 3D structural model towards the terrain model and the real-world model, performing ray intersection analysis. By analyzing whether the rays intersect and the depth of the intersection, the spatial neighboring regions with actual contact or penetration are accurately determined and calculated.

[0090] S3.3 Logically divide the calculation results: Define the adjacent area between the BIM model and the terrain model (e.g., the bottom area of ​​the four bridge piers and the bottom area of ​​the two bridge abutments) as the first edge area; define the adjacent area between the BIM model and the real scene model (e.g., the contact zone between the side of the bridge abutment and the grassland and road surface model generated by the surrounding oblique photography) as the second edge area.

[0091] S3.4. Based on the definitions of the first and second border regions, corresponding, precisely defined terrain data blocks and real-scene data blocks to be processed are segmented from the complete terrain model and real-scene model, respectively. This local data block operation mode greatly improves the efficiency and targeting of subsequent processing algorithms.

[0092] S4. Perform a driving reconstruction of the terrain model related to the fine three-dimensional structure model, so that the reconstructed terrain model fits the fine three-dimensional structure model in the first edge region.

[0093] For the first junction area, this invention proposes an engineering semantic-driven approach to intelligently correct the terrain, ensuring it closely matches the structural model and conforms to engineering specifications. The specific process is as follows:

[0094] S4.1 Obtain engineering semantic constraints related to the current fusion scenario (the system reads or the user inputs engineering design parameters related to the current fusion scenario. These parameters constitute the engineering semantics driving the fusion). These conditions are the logical core driving terrain reconstruction and at least include the slope design parameters of the refined 3D structural model at the first junction area. For example, for the abutment backfill area, the design document clearly specifies the design slope ratio of its cone slope (also known as the slope ratio: used for abutment cone slopes, embankments, and cutting slopes (e.g., 1:1.5)). In addition, parameters may also include the precise dimensions of the structure outline (standardized design dimensions of retaining walls and drainage facilities), the design elevation of key control points (i.e., elevation control points, which can be understood as the design elevation of key points), etc.

[0095] S4.2 At the same time, the base profile of the BIM model in the first junction area (such as the precise boundary of the bridge abutment bottom surface) is used as an unchangeable geometric hard constraint.

[0096] S4.3. Based on the aforementioned engineering semantic constraints (such as the design slope ratio), within the range of the geometric hard constraints, automatically calculate and generate a set of elevation control points conforming to engineering design specifications. For example, based on a slope ratio of 1:1.5, the system automatically calculates the toe line from the abutment boundary to the original natural ground surface and generates a series of control elevation points along this slope. Specifically, this step includes: taking each side of the abutment base polygon as the starting line, along its outer normal direction, calculating the relationship between the horizontal displacement d and the elevation drop h (h=d*i) according to the slope ratio i. The system starts from the set of points on the starting line, moves gradually along the normal direction, and calculates the expected elevation. The expected elevation is compared with the elevation of the original terrain model. When the two are equal or the difference is within the tolerance range, the toe point of the slope direction is considered to have been found. Connecting the toe points in all directions forms a closed toe line. Subsequently, sampling is performed between the toe line and the abutment base boundary according to certain two-dimensional point layout rules (such as regular grids or triangular meshes), and the design elevation of each sampling point is interpolated according to the aforementioned slope formula, thereby forming a set of elevation control points used to constrain terrain reconstruction. For example, when generating the terrain of the abutment backfill area, the algorithm automatically calculates the toe line from the abutment boundary to the natural ground based on the input slope ratio, and generates a series of elevation control points that meet the slope requirements.

[0097] S4.4. Using the aforementioned geometric hard constraints (bridge abutment base contour line) and the set of elevation control points as input, perform constrained digital elevation model (DEM) interpolation to generate the reconstructed terrain. This process is specifically implemented using the constrained Delaunay triangulation algorithm:

[0098] (1) Constraint integration: The geometric hard constraints are used as forced edges, and each point in the elevation control point set is used as a forced point. Together with the terrain points in the original terrain data block to be processed, they form a new scattered dataset with constraints.

[0099] (2) Triangulation reconstruction: Based on the dataset, perform constrained Delaunay triangulation. This algorithm guarantees that in the generated triangulation, all forced edges are retained as edges of triangles, and all forced points are retained as vertices of triangles.

[0100] (3) Terrain Surface Generation: Based on this newly constructed constrained triangular mesh, a continuous terrain surface is generated. This surface strictly passes through all forced points and forced edges, thus achieving seamless geometric fit with the BIM model base and completely eliminating floating or penetration. At the same time, since the forced point set is calculated based on the engineering slope ratio, the shape of this surface at the connection points naturally presents a reasonable slope (such as a standard bridge abutment cone slope) that conforms to engineering design standards and soil mechanics principles, rather than a simple smooth curved surface. This achieves a leap from geometric fit to engineering rationality.

[0101] (4) Data replacement: Replace the original terrain data block to be processed with the newly generated, locally reconstructed terrain surface to complete the dynamic and accurate correction of the digital elevation model.

[0102] S5. Perform transitional deformation on the real-scene model related to the fine three-dimensional structure model, so that the deformed real-scene model can be smoothly connected with the fine three-dimensional structure model in the second edge area, and the deformed real-scene model can be naturally connected with the reconstructed terrain model.

[0103] For the second edge region, and the potential new gaps between the real-world model and the reconstructed terrain, local geometric deformation of the real-world model is required to achieve a smooth transition. The specific process is as follows:

[0104] S5.1. Based on the second edge region, define the deformation region adjacent to the real scene model and the fine three-dimensional structure model.

[0105] S5.2 Obtain the boundary conditions of the region to be deformed, wherein the boundary conditions include the original geometric boundary on one side of the real scene model and the side surface geometric boundary of the fine three-dimensional structure model in the second junction region.

[0106] S5.3 Based on this boundary condition, a mesh deformation algorithm (such as feature-based Morphing deformation algorithm or physical simulation-based elastic mesh deformation algorithm) is used to smooth the deformation of the region to be deformed. This process generates a gradual geometric transition surface. This geometric transition surface achieves a smooth geometric connection with the side surface of the BIM model, eliminating visually hard edges and cracks.

[0107] The specific process is as follows:

[0108] (1) Establishment of correspondence: A set of dense, spatially corresponding vertex pairs are automatically established on the original boundary of the real scene model and the side surface boundary of the fine three-dimensional structure model.

[0109] (2) Solving the deformation function: Treat the mesh of the region to be deformed as an elastic body. Using the corresponding vertex pairs mentioned above as positional constraints, and usually supplemented by smoothness constraints (such as minimizing the mesh deformation energy), construct a mathematical model that minimizes energy. By solving this model (for example, solving a large sparse linear system), the optimal displacement vector of each vertex in the region is obtained.

[0110] (3) Transition surface generation: Based on the calculated displacement vector, all vertices in the area to be deformed are driven to move, thereby generating a geometric transition surface that smoothly transitions from the original boundary of the real scene model to the side surface of the structural model. This surface is geometrically continuous (C0 continuous) and usually tangentially continuous (G1 continuous) with the side surface of the structural model, completely eliminating hard edges and cracks.

[0111] S5.4. When constructing the energy minimization model described above, the connection relationship between the outer edge of the real-world model and the boundary of the reconstructed terrain model generated in step S4 is added to the model as an additional positional constraint. That is, when solving the deformation function, both the corresponding point constraints on the side boundary of the structural model and the connection constraints on the boundary of the reconstructed terrain are satisfied. In this way, a single deformation calculation is configured to simultaneously drive the mesh vertices, so that the generated geometric transition surface can smoothly connect with the side surface of the structural model and its outer edge can naturally connect with the reconstructed terrain, thereby solving the two types of crack problems, "scenery-structure" and "scenery-land", in a closed loop in one go.

[0112] S6. The textures at the junctions of the deformed real-world model, the refined three-dimensional structural model, and the reconstructed terrain model are fused to achieve visual seamlessness.

[0113] Geometric seamlessness is fundamental; visual seamlessness is key. After completing geometric corrections, texture processing is applied to the blended area:

[0114] S6.1 Texture Mapping: For the geometric transition surface and the boundary area between the real scene model and the reconstructed terrain model, the system automatically backtracks from the oblique photogrammetry original image library and selects the source image with the most correct viewing angle, the highest resolution and the least occlusion as the new texture source for these areas, based on the UV coordinates and spatial position of its three-dimensional mesh.

[0115] S6.2 Texture Stitching: The selected new texture, as well as the textures inherent in the refined 3D structural model and the original textures of the reconstructed terrain model, are mapped onto their corresponding mesh surfaces. Then, at the seam boundaries of all different texture types, a Poisson blending algorithm is applied. This algorithm constructs a smooth gradient field at the seam by solving the Poisson equation, thereby driving seamless blending of pixel color and brightness, completely eliminating color differences and seam lines caused by different light sources and shooting times, achieving a visually continuous transition. Alternatively, image processing techniques such as gradient color blending can also be used at the seam boundaries to process the edges between the refined model texture and the oblique photographic texture, eliminating color and brightness differences and achieving a visually seamless transition.

[0116] S6.3 Texture Restoration: For texture-deficient areas caused by permanent occlusion by structures during the original data acquisition (such as the ground directly under a bridge), a sample-based image restoration algorithm is adopted. This algorithm uses the valid textures around the missing area as a sample library, and performs content filling and structural transmission into the missing area by matching and copying texture blocks to generate visually reasonable repair textures, ensuring the overall visual integrity of the model.

[0117] S7. Integrate the detailed 3D structural model, the deformed real-world model, and the reconstructed terrain model to generate a seamless integrated 3D model.

[0118] Finally, integrate all the processed components:

[0119] This system integrates the original detailed 3D structural model (BIM), the real-world model processed with global matching and local geometric textures, and the terrain model reconstructed through semantic drive. The output is a single, lightweight, seamless, integrated 3D model with unified texture coordinates. This model can be packaged and published using advanced streaming formats such as 3DTiles and S3M, and can be directly applied to web, mobile, or professional 3D platforms, supporting efficient visualization, spatial querying, profile analysis, earthwork calculation, view analysis, and other advanced engineering applications.

[0120] This embodiment forms a complete automated processing chain, from data preprocessing, global matching, and intelligent regional identification, to terrain-driven reconstruction based on engineering semantics and transitional deformation of the real-world model, and finally to fine texture fusion. This method not only successfully solves the "two-layer" problem of multi-source 3D model fusion in highway and bridge scenarios, but also ensures the geometric accuracy, visual realism, and engineering compliance of the fused results, providing a high-quality data foundation for the full lifecycle digital management of highway engineering.

[0121] In summary, the innovative points of this invention are as follows:

[0122] By directly embedding engineering design parameters (slope ratio, structural profile, etc.) as constraints into the geometry generation algorithm, the fusion process is dominated by engineering rules.

[0123] The system calls upon the design slope ratio defined in the project's design standards or soil parameters. For example, for backfill soil at a certain location, the specified slope ratio is 1:1.5. During terrain interpolation, the algorithm not only uses the abutment boundary as a constraint but also automatically calculates a series of contour lines or elevation points in the extended area behind the abutment according to the 1:1.5 slope requirement. Based on these scattered points that conform to both geometric constraints and engineering specifications, DEM interpolation and triangulation reconstruction are performed. The terrain automatically generated by this algorithm not only seamlessly connects to the abutment geometrically but also the shape of its connection part (i.e., the abutment cone slope) is a natural slope that is engineering-reasonable, stable, and meets design requirements.

[0124] This algorithm addresses geometric discontinuities at the seams between two independent 3D models (such as an oblique photogrammetry model and a BIM model, or oblique photogrammetry and terrain). It automatically creates and computes a gradual "Morphing Zone" between the boundaries of the two models, where the geometry smoothly transitions from one model to the other. Based on the feature-driven Morphing (morphing) deformation algorithm boundary, a smooth deformation function is calculated. This function drives each vertex within the transition zone, gradually transforming its position from the "oblique model boundary" to the "terrain model boundary." The transition zone mesh is treated as an elastic membrane, with one end "pulled" to the BIM model boundary and the other end fixed to the oblique model. Its smooth deformation morphology is calculated by simulating physical processes. Based on the terrain generated by this algorithm, a smoothly transitioning, crack-free, and overlapping geometric region is created between the oblique model and the original terrain, providing a perfect geometric foundation for subsequent texture blending.

[0125] The method of the present invention will be described next through specific examples.

[0126] Data: Bridge BIM model (.ifc format), bridge area oblique photogrammetry model (.osgb format), high-precision DEM.

[0127] Import the BIM model and DEM into the data source, store the oblique photogrammetry data, unify all data to the 2000 National Geodetic Coordinate System, and generate TIN terrain from the DEM data.

[0128] The algorithm automatically identifies the rectangular areas where the bottom surfaces of the four bridge piers are in contact with the ground, as well as the complex areas where the two abutments connect with the embankment.

[0129] Using the corner points of the bridge pier bottom as constraint points and the design slope ratio at that location, the slope toe line from the bridge abutment boundary to the natural ground is automatically calculated, and a series of elevation control points that meet the slope requirements are generated. The above geometric constraints and the control points generated by semantic driving are used together as hard constraints to perform triangular mesh interpolation on the DEM, generating terrain that fits perfectly with the bridge pier.

[0130] A 2-meter-wide transition zone is generated at the junction of the bridge abutment and the inclined model, and the mesh of the inclined model is smoothly deformed to fit the side of the bridge abutment.

[0131] The texture of the ground around the bridge piers is extracted from the tilted original image, and after Poisson fusion, it is mapped onto the corrected mesh to cover up processing traces.

[0132] The output is in 3DTiles format and loaded onto the web platform.

[0133] The diagram obtained through the above execution process is shown below. Figure 2 As shown.

[0134] Example 2 discloses a multi-source 3D geographic information data fusion system for executing the multi-source 3D geographic information data fusion method described in Example 1, including:

[0135] The coordinate unification module is used to acquire the oblique photogrammetry surface model, the fine three-dimensional structure model and the digital elevation model to be fused, and to unify the oblique photogrammetry surface model, the fine three-dimensional structure model and the digital elevation model into the same spatial reference system;

[0136] The spatial alignment module is used to perform global geometric matching on the coordinate-unified oblique photogrammetric surface model and the digital elevation model to obtain a spatially aligned real-world model and terrain model.

[0137] The fusion region determination module is used to determine, based on the outline and spatial position of the fine three-dimensional structure model, the first boundary region between the fine three-dimensional structure model and the terrain model, and the second boundary region between the fine three-dimensional structure model and the real scene model.

[0138] The terrain reconstruction module is used to perform a driving reconstruction of the terrain model in relation to the fine three-dimensional structure model, so that the reconstructed terrain model fits the fine three-dimensional structure model in the first edge region;

[0139] The real-scene deformation module is used to perform transitional deformation on the real-scene model related to the fine three-dimensional structure model, so that the deformed real-scene model can be smoothly connected with the fine three-dimensional structure model in the second edge area, and the deformed real-scene model can be naturally connected with the reconstructed terrain model.

[0140] The fusion module is used to integrate the detailed 3D structural model, the deformed real-world model, and the reconstructed terrain model to generate a seamless, integrated 3D model.

[0141] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for fusing multi-source three-dimensional geographic information data, characterized in that, include: Obtain the oblique photogrammetry surface model, the detailed 3D structural model, and the digital elevation model to be fused, and unify the oblique photogrammetry surface model, the detailed 3D structural model, and the digital elevation model into the same spatial reference system; Global geometric matching is performed on the oblique photogrammetric surface model and the digital elevation model after coordinate unification to obtain a spatially aligned real-scene model and terrain model; Based on the outline and spatial position of the refined 3D structural model, the first junction area between the refined 3D structural model and the terrain model, and the second junction area between the refined 3D structural model and the real scene model are determined respectively. The terrain model is reconstructed in a driving manner related to the fine three-dimensional structure model, so that the reconstructed terrain model fits the fine three-dimensional structure model in the first edge region; The specific process is as follows: Obtain engineering semantic constraints related to the refined three-dimensional structural model, wherein the engineering semantic constraints include the slope design parameters of the refined three-dimensional structural model at the first junction area; The base contour of the refined three-dimensional structural model in the first junction area is used as a geometric hard constraint; Based on the aforementioned engineering semantic constraints, a set of elevation control points conforming to engineering design specifications is generated within the scope of the aforementioned geometric hard constraints. Using the geometric hard constraints and the set of elevation control points as input, the terrain model is reconstructed by a constrained triangulation network to generate a reconstructed terrain that closely fits the base of the fine three-dimensional structural model and conforms to the shape of the engineering slope. The real-world model is subjected to a transitional deformation related to the fine three-dimensional structure model, so that the deformed real-world model is smoothly connected to the fine three-dimensional structure model in the second edge region, and the deformed real-world model is naturally connected to the reconstructed terrain model. The detailed 3D structural model, the deformed real-world model, and the reconstructed terrain model are integrated to generate a seamless, unified 3D model.

2. The method for fusing multi-source three-dimensional geographic information data according to claim 1, characterized in that, The specific process of unifying the oblique photogrammetry surface model, the detailed 3D structural model, and the digital elevation model into the same spatial reference system is as follows: Determine the spatial reference frame corresponding to each of the oblique photogrammetry surface model, the detailed three-dimensional structural model, and the digital elevation model; Based on the transformation relationship between their respective spatial reference systems, the oblique photogrammetry surface model, the detailed three-dimensional structural model, and the digital elevation model are uniformly transformed into the same target engineering coordinate system.

3. The method for fusing multi-source three-dimensional geographic information data according to claim 1, characterized in that, The specific process of performing global geometric matching between the coordinate-unified oblique photogrammetric surface model and the digital elevation model is as follows: Extract the digital surface model from the oblique photogrammetry surface model; Calculate the spatial transformation parameters between the digital surface model and the digital elevation model; The oblique photography surface model is geometrically corrected based on the spatial transformation parameters.

4. The method for fusing multi-source three-dimensional geographic information data according to claim 1, characterized in that, Based on the outline and spatial position of the refined 3D structural model, the specific process for determining the first boundary region between the refined 3D structural model and the terrain model, and the second boundary region between the refined 3D structural model and the real-world model, is as follows: Establish a spatial index that includes the detailed 3D structural model, the terrain model, and the real-scene model; Based on the spatial index, the spatial proximity region between the outer surface of the fine three-dimensional structural model and the terrain model and the real scene model is calculated through geometric buffer analysis or ray intersection analysis. The calculated spatial neighbor regions are defined as the first border region and the second border region, respectively; Based on the definitions of the first and second border regions, corresponding terrain data blocks and real-scene data blocks to be processed are segmented from the terrain model and the real-scene model, respectively.

5. The method for fusing multi-source three-dimensional geographic information data according to claim 1, characterized in that, The slope design parameters include at least one of the following: the design slope ratio of the slope, the outline dimensions of the structure, or the design elevation of the control points.

6. The method for fusing multi-source three-dimensional geographic information data according to claim 1, characterized in that, The specific process of performing transitional deformations on the real-world model related to the refined 3D structural model is as follows: Based on the second edge region, the deformable region adjacent to the real scene model and the fine three-dimensional structure model is defined; Obtain the boundary conditions of the region to be deformed, wherein the boundary conditions include the original geometric boundary on one side of the real scene model and the side surface geometric boundary of the fine three-dimensional structure model in the second junction region; Based on the boundary conditions, the area to be deformed is smoothly deformed using a mesh deformation algorithm to generate a geometric transition surface; wherein, the geometric transition surface is smoothly connected to the side surface of the fine three-dimensional structure model; and the transitional deformation also allows the real scene model and the reconstructed terrain model to connect naturally at the boundary.

7. The method for fusing multi-source three-dimensional geographic information data according to claim 6, characterized in that, Fusion methods also include: The textures at the junctions of the deformed real-world model, the refined 3D structural model, and the reconstructed terrain model are fused to achieve visual seamlessness.

8. The method for fusing multi-source three-dimensional geographic information data according to claim 7, characterized in that, The specific process of fusing the textures at the junctions of the deformed real-world model, the refined 3D structural model, and the reconstructed terrain model is as follows: For the geometric transition surface and the boundary area between the real scene model and the reconstructed terrain model, the best texture image of the corresponding viewpoint is selected and extracted from the original image of the oblique photogrammetric surface model. The extracted texture image is mapped onto the mesh corresponding to the geometric transition surface and the boundary region; At the seams after texture mapping, an image fusion algorithm is used to progressively blend the colors and brightness of the textures of the real scene model, the fine three-dimensional structure model, and the reconstructed terrain model to eliminate visual seams. For areas with missing textures caused by model occlusion, the surrounding texture information is used to repair or fill them.

9. A system for fusing multi-source three-dimensional geographic information data, used to execute the method for fusing multi-source three-dimensional geographic information data as described in any one of claims 1-8, characterized in that, include: The coordinate unification module is used to acquire the oblique photogrammetry surface model, the fine three-dimensional structure model and the digital elevation model to be fused, and to unify the oblique photogrammetry surface model, the fine three-dimensional structure model and the digital elevation model into the same spatial reference system; The spatial alignment module is used to perform global geometric matching on the coordinate-unified oblique photogrammetric surface model and the digital elevation model to obtain a spatially aligned real-world model and terrain model. The fusion region determination module is used to determine, based on the outline and spatial position of the fine three-dimensional structure model, the first boundary region between the fine three-dimensional structure model and the terrain model, and the second boundary region between the fine three-dimensional structure model and the real scene model. The terrain reconstruction module is used to perform a driving reconstruction of the terrain model in relation to the fine three-dimensional structure model, so that the reconstructed terrain model fits the fine three-dimensional structure model in the first edge region; The specific process is as follows: Obtain engineering semantic constraints related to the refined three-dimensional structural model, wherein the engineering semantic constraints include the slope design parameters of the refined three-dimensional structural model at the first junction area; The base contour of the refined three-dimensional structural model in the first junction area is used as a geometric hard constraint; Based on the aforementioned engineering semantic constraints, a set of elevation control points conforming to engineering design specifications is generated within the scope of the aforementioned geometric hard constraints. Using the geometric hard constraints and the set of elevation control points as input, the terrain model is reconstructed by a constrained triangulation network to generate a reconstructed terrain that closely fits the base of the fine three-dimensional structural model and conforms to the shape of the engineering slope. The real-scene deformation module is used to perform transitional deformation on the real-scene model related to the fine three-dimensional structure model, so that the deformed real-scene model can be smoothly connected with the fine three-dimensional structure model in the second edge area, and the deformed real-scene model can be naturally connected with the reconstructed terrain model. The fusion module is used to integrate the detailed 3D structural model, the deformed real-world model, and the reconstructed terrain model to generate a seamless, integrated 3D model.

Images