Railway BIM model lightweight processing method and system based on multi-source data processing

Abstract

The present application relates to the technical field of data processing, in particular to a railway BIM model lightweight processing method and system based on multi-source data processing, comprising: obtaining a design BIM model, laser point cloud, oblique photography model and GIS terrain data, and performing unified coordinate conversion, component ID labeling and triangulation to obtain a unified triangular mesh; obtaining the geometric consistency coefficient, professional category coefficient and continuity coefficient of each component in the unified triangular mesh along the railway mileage direction, and calculating the component engineering importance according to the preset weight; in the edge folding iteration process, the mean value and mean absolute deviation of all candidate edge folding costs are counted, the candidate edge folding cost is semantically weighted combined with the component engineering importance, and an adaptive stop threshold is constructed based on the mean value and mean absolute deviation, and the semantically weighted edge folding is iteratively executed to generate a lightweight mesh. The present application solves the problem of low compression effect caused by repeated model loading.

Description

Technical Field

[0001] This invention relates to the field of data processing technology. More specifically, this invention relates to a method and system for lightweight processing of railway BIM models based on multi-source data processing. Background Technology

[0002] Railway engineering is characterized by its long track spans, diverse professional types, massive number of components, long construction periods, and extended operation and maintenance phases. This generates a large amount of heterogeneous data during the surveying and design, construction, operation and maintenance, and digital twin management processes. With the application of Building Information Modeling (BIM) technology in railway engineering, design firms typically create railway BIM models that include professional components such as tracks, roadbeds, bridges, tunnels, tracks, stations, overhead contact lines, communications, signaling, and power systems to achieve information management throughout the entire project lifecycle.

[0003] With the increasing digitalization of railway engineering, the scale of railway BIM models after multi-source data fusion is rapidly expanding. These models not only contain a vast amount of component geometric information but also integrate a large amount of attribute data, topological relationships, and multi-source spatial data, resulting in millions or even tens of millions of triangular facets. When such models are viewed in real-time on web platforms, mobile devices, or digital twin platforms, they typically consume significant amounts of network bandwidth, storage space, and graphics rendering resources, easily leading to problems such as excessively long model loading times, high memory usage, slow interactive responses, and decreased rendering frame rates. Existing technologies typically employ lightweight methods such as edge collapsing, mesh simplification, and hierarchical detail model construction to compress the model, thereby reducing model complexity and improving rendering efficiency.

[0004] Most existing lightweighting methods use geometric or topological errors as the sole optimization objective, applying a uniform simplification strategy to all components and determining the edge folding order solely based on geometric features. This fails to adequately consider the functional differences between components from different disciplines in railway engineering applications, as well as the consistency between components and their actual on-site state. Furthermore, existing lightweighting technologies for railway BIM models generally use fixed error thresholds or preset patch compression ratios as the termination condition for edge folding, making it difficult to dynamically adjust the simplification process according to the model's complexity and the geometric features of different regions. When local areas of the model have complex structures or dense components, fixed thresholds can easily lead to oversimplification in important areas, damaging geometric details and engineering semantic information; conversely, for structurally simple areas, insufficient simplification may occur, resulting in poor compression effects from repeated model loading. Summary of the Invention

[0005] To address the problem of low compression efficiency in the repetitive loading model mentioned in the background section, this invention provides solutions in the following aspects.

[0006] In a first aspect, the present invention provides a lightweight processing method for railway BIM models based on multi-source data processing, comprising: acquiring design BIM models, laser point clouds, oblique photogrammetry models, and GIS terrain data, and performing unified coordinate transformation, component ID labeling, and triangulation to obtain a unified triangular mesh; acquiring the geometric consistency coefficient, professional category coefficient, and continuity coefficient along the railway mileage direction of any component in the unified triangular mesh and the laser point cloud, and then weighting them according to preset weights to obtain the engineering importance of any component; during the edge folding iteration process, extracting the mean and mean absolute deviation of the folding costs of all candidate edges, wherein the candidate edge is any edge in the unified triangular mesh; performing semantically weighted edge folding on the folding costs of all candidate edges using the engineering importance, constructing an adaptive stopping threshold using the mean and mean absolute deviation, iteratively performing semantically weighted edge folding based on the adaptive stopping threshold to obtain a lightweight mesh, and outputting three-dimensional tile data by segmentation along the railway mileage direction.

[0007] The aforementioned technical solution improves the spatial consistency and geometric accuracy of railway BIM models after multi-source data fusion. It ensures that the model simplification process balances the importance and geometric characteristics of components, preventing oversimplification of critical components and important areas during the lightweighting process, thereby enhancing the model's structural integrity and engineering representation capabilities. Simultaneously, it can dynamically adjust the lightweighting process based on model complexity, improving model compression efficiency while maintaining geometric accuracy, and reducing data storage, transmission, and rendering overhead. This enhances the loading speed, visualization efficiency, and application performance of railway BIM models in large-scale scenarios.

[0008] Furthermore, the unified coordinate transformation specifically involves: converting the design BIM model, laser point cloud, oblique photogrammetry model, and GIS terrain data to a planar projection coordinate system, and then merging the design BIM model, laser point cloud, oblique photogrammetry model, and GIS terrain data according to their spatial positions within the planar projection coordinate system to complete the unified coordinate transformation.

[0009] The above technical solution eliminates coordinate differences between different data sources, improves the spatial consistency and registration accuracy of multi-source data, and enables various types of data to correspond accurately under the same spatial reference. This provides a unified data foundation for subsequent model construction, component analysis and lightweight processing, avoids positional offset and fusion errors caused by inconsistent coordinates, and improves the accuracy and reliability of railway BIM model processing results.

[0010] Furthermore, the unified triangular mesh specifically involves: extracting the globally unique identifier of each component as the component ID; generating a first triangular mesh by Delaunay triangulation of the design BIM model; generating a second triangular mesh by rolling sphere triangulation of the laser point cloud with a preset neighborhood radius; generating a third triangular mesh by merging triangular faces extracted from the oblique photogrammetry model according to tiles; generating a fourth triangular mesh by triangulating the GIS terrain data according to a regular raster; and associating each face in the first, second, third, and fourth triangular meshes with its corresponding component ID to obtain the unified triangular mesh.

[0011] The aforementioned technical solution employs appropriate mesh construction methods tailored to the characteristics of data from different sources, improving the geometric representation accuracy and modeling quality of various data types. It also establishes a unified mesh data structure, enabling data from different sources to be organized and managed uniformly within the same model. Furthermore, by associating each triangular facet with a corresponding component identifier, the correspondence between geometric information and component attribute information is maintained, providing a reliable data foundation for subsequent component analysis, model lightweighting, and data management. This enhances the integrity, consistency, and traceability of multi-source data fusion.

[0012] Furthermore, the project importance is specifically defined as follows: pre-setting professional category weight, geometric consistency weight, and continuity weight; for any component, multiplying the professional category coefficient of the component by the professional category weight to obtain a first component, multiplying the geometric consistency coefficient by the geometric consistency weight to obtain a second component, and multiplying the mileage continuity coefficient by the mileage continuity weight to obtain a third component; summing the first component, the second component, and the third component to obtain the project importance of the component.

[0013] The above technical solution comprehensively considers the professional attributes, geometric matching degree, and spatial continuity of components to make a more comprehensive and objective evaluation of the importance of components, improves the accuracy and rationality of importance determination, and enables different components to reflect differentiated processing strategies in subsequent processing. This is conducive to preserving the geometric features and attribute information of key components and important areas, and improving the structural integrity, geometric accuracy and engineering application value of the model after lightweighting.

[0014] Furthermore, the process of obtaining the professional category coefficient includes: establishing a mapping table from IFC entity type to professional category coefficient; reading the IFC entity type of any component and looking up the value in the mapping table to obtain the professional category coefficient of any component.

[0015] Furthermore, the process of obtaining the geometric consistency coefficient includes: for any triangular facet on the surface of each component, calculating the distance from the centroid of any triangular facet to the nearest point in the laser point cloud; taking the arithmetic mean of the distances corresponding to all facets of each component to obtain the component-level average distance; and using the difference between 1 and the ratio of the component-level average distance to the distance normalization upper limit as the geometric consistency coefficient of the component.

[0016] The above technical solution can objectively evaluate the degree of geometric fit between the design model and the real-world data from the overall component level, reduce the impact of local errors on the evaluation results, improve the accuracy and stability of geometric consistency analysis, and make the correspondence between the components and the actual engineering state more realistic. This provides a reliable basis for subsequent lightweight model processing and helps to improve the geometric fidelity, engineering realism and overall processing quality of the model.

[0017] Further, the mean and average absolute deviation of the folding costs of all candidate edges are extracted. Specifically, during the edge folding iteration, the folding cost of each candidate edge in the unified triangular mesh is calculated using a quadratic error metric and added to a priority queue. In the current iteration, the arithmetic mean of the folding costs of all candidate edges in the priority queue is taken as the current mean. The arithmetic mean of the absolute values ​​of the differences between the folding costs of each candidate edge in the priority queue and the current mean is taken as the current average absolute deviation.

[0018] The above technical solution can reflect the overall level and distribution characteristics of candidate edge folding cost in real time, accurately characterize the complexity and geometric changes of the current mesh, provide reliable statistical basis for the subsequent edge folding process, make the model simplification process more in line with the actual characteristics of the mesh, improve the stability and rationality of lightweight processing, and help to balance model compression efficiency and geometric fidelity.

[0019] Furthermore, the process of semantically weighted edge folding and adaptive stopping threshold includes: for each candidate edge in the priority queue, dividing the folding cost of the quadratic error metric of each candidate edge by the component engineering importance of its component to obtain the semantically weighted cost of each candidate edge; adding the current mean and the product of the preset distribution expansion coefficient and the current mean absolute deviation to obtain the current adaptive stopping threshold; performing edge folding in ascending order of the semantically weighted cost; when the semantically weighted cost of the head of the priority queue exceeds the current adaptive stopping threshold, stopping the edge folding iteration of the current level, using the current mesh as the input of the next level and continuing to perform edge folding until all preset levels of lightweight mesh are obtained, and outputting three-dimensional tile data by dividing along the railway mileage direction according to the preset unit length.

[0020] The aforementioned technical solution enables the model simplification process to consider both the importance and geometric features of components, prioritizing the preservation of geometric information for key components and important regions, thus avoiding the loss of important details during model lightweighting. Simultaneously, it can dynamically adjust the iteration termination condition based on the distribution of candidate edge folding costs, making the lightweighting process more consistent with the actual complexity of the model, improving model compression efficiency and geometric fidelity, and generating lightweight models at different accuracy levels. Furthermore, organizing and outputting data in blocks according to railway mileage direction facilitates on-demand loading and rapid scheduling of model data, reducing data transmission and rendering overhead, and improving the visualization loading efficiency and interactive performance of railway BIM models in large-scale scenarios.

[0021] In a second aspect, the present invention provides a lightweight processing system for railway BIM models based on multi-source data processing, including a memory and a processor. The memory stores computer program instructions, which, when executed by the processor, implement any of the above-described lightweight processing methods for railway BIM models based on multi-source data processing.

[0022] The beneficial effects of this invention are as follows: This invention enables efficient fusion of railway BIM models and multi-source spatial data. While ensuring the geometric accuracy, structural integrity, and consistency of attribute information of the model, it implements differentiated lightweighting processing for components of varying importance, improving the rationality and accuracy of model simplification. Simultaneously, it can dynamically control the lightweighting process according to the actual complexity of the model, balancing model compression efficiency and geometric fidelity, reducing data storage, transmission, and rendering resource consumption. Furthermore, by organizing model data along the railway line direction, it improves the loading efficiency, display smoothness, and interactive performance of large-scale railway scenes, providing efficient and reliable technical support for railway engineering design, construction, operation, and digital management. Attached Figure Description

[0023] Figure 1 This is a flowchart illustrating a lightweight processing method for railway BIM models based on multi-source data processing according to an embodiment of the present invention. Detailed Implementation

[0024] Example of a lightweight processing method for railway BIM models based on multi-source data processing.

[0025] like Figure 1 The flowchart shown below illustrates a lightweight processing method for railway BIM models based on multi-source data processing, according to an embodiment of the present invention. The method includes the following steps: S1: Obtain the design BIM model, laser point cloud, oblique photogrammetry model and GIS terrain data, and perform unified coordinate transformation, component ID labeling and triangulation to obtain a unified triangular mesh.

[0026] In a preferred embodiment, the spatial reference information of the design BIM model, laser point cloud, oblique photogrammetry model, and GIS terrain data is first read. Based on the corresponding coordinate reference and projection parameters, all types of data are uniformly converted to a preset planar projection coordinate system, so that data from different sources and in different formats are under the same spatial reference frame. Then, the overall registration and merging are completed according to the spatial positional relationship of each data, so that the design BIM model, laser point cloud, oblique photogrammetry model, and GIS terrain data form a unified spatial expression.

[0027] By employing the above methods, positional offsets caused by differences in coordinate references, projection methods, and spatial scales between different data sources are eliminated, improving the spatial consistency between multi-source heterogeneous data. This provides a reliable data foundation for subsequent geometric matching, attribute association, and unified calculation between different data sources. At the same time, it avoids component misalignment and topological distortion caused by the accumulation of coordinate errors, thereby improving the spatial accuracy of the overall model.

[0028] Furthermore, the component ID labeling and unified triangular mesh construction specifically involve: extracting a globally unique identifier corresponding to each component in the design BIM model as the component ID, and establishing a mapping relationship between the component ID and the corresponding geometric entity; subsequently, processing different types of data using a meshing method adapted to their data structure, wherein: for the design BIM model, Delaunay triangulation is used to generate a first triangular mesh based on the component boundary geometric information, converting each component entity into a continuous set of triangular facets; for laser point clouds, a preset neighborhood radius is set according to the spatial distribution of the point cloud, and the rolling sphere triangulation algorithm is used to reconstruct the point cloud surface to generate a second triangular mesh, in order to maintain the continuity of the point cloud surface and geometric details; for oblique photogrammetry models, triangular facets in each tile are extracted using tiles as the basic processing unit, and adjacent tiles are spliced ​​and merged according to spatial topological relationships to generate a third triangular mesh, thereby maintaining a large-scale continuous expression of the photogrammetry model; For GIS terrain data, elevation nodes are established based on regular rasters, and regular triangulation is performed on adjacent raster cells to generate a fourth triangular grid to accurately represent the changes in terrain undulation. Finally, each triangular facet in the first, second, third, and fourth triangular meshes is associated with its corresponding component ID, and a unified facet index relationship is established to obtain a unified triangular mesh.

[0029] By employing the above methods, design models, measured point clouds, oblique photogrammetry models, and GIS topographic data are all converted into a unified triangular mesh representation, eliminating the expression differences between different data structures and achieving a unified geometric description of multi-source heterogeneous data. Simultaneously, each triangular facet carries a corresponding component ID, establishing a stable association between geometric information and component semantic information. This preserves component-level semantic attributes while maintaining spatial geometric continuity, supporting subsequent rapid retrieval, spatial positioning, and attribute queries by component. It also facilitates facet-level registration, difference analysis, collision detection, and topological relationship calculations between different data sources, thereby significantly improving the computational efficiency, positioning accuracy, and reliability of subsequent multi-source data fusion analysis.

[0030] S2: After obtaining the geometric consistency coefficient, professional category coefficient, and continuity coefficient along the railway mileage direction of any component in the unified triangular mesh and the laser point cloud, the engineering importance of any component is obtained by weighting it according to the preset weight.

[0031] In a preferred embodiment, the project importance is specifically determined by: pre-setting a professional category weight, a geometric consistency weight, and a continuity weight; for any component, multiplying the professional category coefficient of the component by the professional category weight to obtain a first component, multiplying the geometric consistency coefficient by the geometric consistency weight to obtain a second component, and multiplying the mileage continuity coefficient by the mileage continuity weight to obtain a third component; and summing the first, second, and third components to obtain the project importance of the component.

[0032] This approach uses the professional semantic attributes, geometric realism, and spatial continuity of railway lines as the basis for evaluating engineering value. First, it assigns pre-defined weights to the professional category coefficient, geometric consistency coefficient, and mileage continuity coefficient, allowing different evaluation factors to participate in the comprehensive calculation according to their importance to the railway project. Then, it calculates the contribution value of each evaluation factor and merges these contributions to obtain the engineering importance of the component. This constructs a multi-dimensional comprehensive evaluation mechanism that considers semantic information, geometric accuracy, and railway topology. Compared to judging importance based solely on a single geometric feature or model structure, this scheme more comprehensively reflects the actual value of railway BIM components in the project. Components undertaking key functions, exhibiting high consistency with the site, and strong continuity along the railway line receive higher engineering importance, while ordinary or auxiliary components receive relatively lower importance. This provides a scientific, stable, and engineering-significant decision-making basis for subsequent semantic weighted edge folding, effectively preventing the oversimplification of key components during the lightweighting process. This improves the structural integrity, engineering readability, geometric fidelity, and accuracy and reliability of railway scene visualization applications after model lightweighting.

[0033] The process of obtaining the professional category coefficient includes: establishing a mapping table from IFC entity types to professional category coefficients; reading the IFC entity type of any component and looking it up in the mapping table to obtain the professional category coefficient of any component. The process of obtaining the geometric consistency coefficient includes: for any triangular facet on the surface of each component, calculating the distance from the centroid of any triangular facet to the nearest point in the laser point cloud; taking the arithmetic mean of the distances corresponding to all facets of each component to obtain the component-level average distance; and using the difference between 1 and the ratio of the component-level average distance to the distance normalization upper limit as the geometric consistency coefficient of the component.

[0034] Based on the IFC entity type information inherent in the components of the railway BIM model, a correspondence between entity types and professional category coefficients is pre-established. Professional category coefficients for each component are quickly obtained through table lookup, enabling important components of different specialties and functions to have differentiated semantic weights in subsequent processing. Simultaneously, using laser point clouds as real-world data, the spatial distance between each triangular facet on the component surface and its nearest point in the point cloud is calculated. All distances are statistically analyzed to obtain an average distance reflecting the overall geometric fit of the component. Then, a geometric consistency coefficient is calculated based on the deviation of the average distance from the upper limit of the normalized distance, objectively characterizing the geometric consistency between the design model and the actual site. By quantifying professional semantic information and geometric realism information separately and forming a unified evaluation index, the functional differences of different professional components in railway engineering can be fully reflected, and the consistency between the components and the actual site conditions can be accurately reflected. This provides an evaluation basis with both semantic attributes and geometric credibility for subsequent engineering importance calculations, thereby avoiding excessive distortion of important components due to geometric simplification during the lightweighting process and improving the engineering expression capability, spatial accuracy, and practical application reliability of the lightweight model.

[0035] S3: Adaptive semantic weighted lightweight processing based on engineering importance.

[0036] In a preferred embodiment, during the edge folding iteration process, the mean and mean absolute deviation of the folding costs of all candidate edges are extracted, where the candidate edge is any edge in the unified triangular mesh; the folding costs of all candidate edges are semantically weighted using the engineering importance, and an adaptive stopping threshold is constructed using the mean and mean absolute deviation; and a lightweight mesh is obtained by iteratively performing semantically weighted edge folding based on the adaptive stopping threshold.

[0037] The extraction of the mean and average absolute deviation of the folding costs of all candidate edges is specifically as follows: during the edge folding iteration process, the folding cost of each candidate edge in the unified triangular mesh is calculated according to the quadratic error metric and added to the priority queue; in the current iteration, the arithmetic mean of the folding costs of all candidate edges in the priority queue is taken as the current mean; the arithmetic mean of the absolute values ​​of the differences between the folding costs of each candidate edge in the priority queue and the current mean is taken as the current average absolute deviation.

[0038] The above technical solution first uses a quadratic error metric to calculate the folding cost of each candidate edge in the unified triangular mesh during the edge folding iteration process, and organizes all candidate edges into a priority queue according to their folding costs to establish a candidate set that can dynamically reflect the difficulty of the current mesh geometric simplification. Subsequently, the folding costs of all candidate edges in the priority queue are statistically analyzed, and their mean and mean absolute deviation are calculated respectively. The mean can reflect the overall folding cost level of the current mesh, and the mean absolute deviation can reflect the dispersion and fluctuation of the folding costs of each candidate edge. On this basis, the folding costs of candidate edges are semantically weighted in combination with the importance of the component engineering, so that candidate edges belonging to components with higher engineering importance have higher retention priority, while candidate edges belonging to components with lower engineering importance participate in simplification first. Furthermore, an adaptive stopping threshold is dynamically constructed based on the overall distribution characteristics of the folding costs to control the termination time of the edge folding iteration. By incorporating statistical distribution characteristics and engineering semantic information into the edge folding optimization process, this scheme avoids the oversimplification or undersimplification problems caused by the fixed stopping conditions in traditional lightweighting methods. It enables the lightweighting process to automatically adjust the iteration depth according to the current mesh complexity, ensuring that the geometric features, engineering semantics, and overall topology of key components are fully preserved while improving the simplification efficiency of non-critical areas. This achieves a dynamic balance between model compression rate, geometric fidelity, and computational efficiency, further enhancing the stability, adaptability, and engineering application value of lightweighting of railway BIM models.

[0039] The semantically weighted edge folding and adaptive stopping threshold process includes: for each candidate edge in the priority queue, dividing the folding cost of the quadratic error metric of each candidate edge by the component engineering importance of its component to obtain the semantically weighted cost of each candidate edge; adding the current mean and the product of the preset distribution expansion coefficient and the current mean absolute deviation to obtain the current adaptive stopping threshold; performing edge folding in ascending order of the semantically weighted cost; when the semantically weighted cost at the head of the priority queue exceeds the current adaptive stopping threshold, stopping the edge folding iteration of the current level, using the current mesh as the input of the next level and continuing to perform edge folding until all preset levels of lightweight meshes are obtained.

[0040] By integrating the geometric folding cost of each candidate edge with the engineering importance of its corresponding component, the final folding priority of candidate edges is not only affected by geometric errors but also comprehensively reflects the engineering value of the component. This results in candidate edges corresponding to components with higher engineering importance having a higher retention priority, while candidate edges corresponding to components with lower engineering importance are given priority in geometric simplification. Simultaneously, an adaptive stopping threshold is dynamically generated by combining the overall mean and dispersion of the current candidate edge folding costs. This allows the stopping condition to be adjusted in real-time according to changes in mesh complexity and cost distribution after each round of edge folding, rather than relying on a fixed threshold for uniform control. Subsequently, edge folding is performed sequentially from low to high semantic weighted cost. When the overall semantic weighted cost of the remaining candidate edges exceeds the current adaptive stopping threshold, it is considered that continued folding may lead to significant loss of important geometric features or engineering semantic information. Therefore, the simplification at the current level is terminated, and the current mesh is used as the input for the next level of lightweight processing, generating lightweight meshes of different accuracy levels layer by layer. By organically combining engineering semantic constraints, geometric error evaluation, and statistical distribution characteristics, this scheme enables the edge folding process to be dynamically adjusted according to the local features and overall complexity of the model. This not only effectively avoids the problems of oversimplification of key components or retention of redundant details in ordinary areas in traditional methods, but also enables the progressive construction of multi-level lightweight meshes. While ensuring the structural integrity, geometric fidelity, and engineering expression capabilities of the railway BIM model, it improves the model compression efficiency, the smoothness of layer switching, and the overall performance of real-time loading and visualization of 3D scenes.

[0041] S4: Lightweight model segmentation and 3D tile output.

[0042] In a preferred embodiment, the lightweight grid is segmented according to the railway mileage direction, and corresponding three-dimensional tile data is generated to achieve lightweight organization, storage and efficient visualization loading of the railway BIM model.

[0043] Specifically, using railway line mileage as the dividing benchmark, multiple continuous spatial segments are sequentially divided according to preset mileage lengths. Based on the spatial position of each component along the railway mileage direction, components in the lightweight mesh are assigned to their corresponding segments. For components spanning multiple segments, cross-tile relationships are established based on the component's spatial coverage to ensure the integrity of component display. Subsequently, local vertex data, triangular face indexes, component attribute information, and spatial bounding boxes are established for the lightweight mesh within each segment. Corresponding tile model data and hierarchical index data are generated according to the 3D tile data format. The hierarchical index records the spatial position, hierarchical relationship, and access path between tiles, while the tile model data stores the lightweight geometric model and component attribute information of the corresponding segment. During model browsing, based on the current viewpoint position, observation range, and display precision, and combined with the railway mileage direction, the 3D tile data corresponding to the target segment is dynamically loaded, and tile data that has left the field of view is unloaded, achieving continuous loading and progressive display of the model along the railway line.

[0044] By employing the above methods, the railway BIM model is organized into blocks according to the spatial distribution characteristics of the railway line. This allows model data to be loaded and dynamically scheduled on demand according to the railway mileage direction, avoiding the problems of excessive memory consumption and reduced rendering efficiency caused by loading the entire model at once. Simultaneously, by combining lightweight meshes with 3D tile data, while ensuring the integrity of component geometric features and attribute information, the amount of data transmitted per transaction and the graphics rendering overhead are effectively reduced. This improves the loading speed, interactive smoothness, and visualization application efficiency of ultra-long railway line BIM models in large-scale scenarios, and provides an efficient data organization method for digital twin display, operation and maintenance management, and full lifecycle applications of railway engineering.

[0045] The solution of this invention enables the effective fusion of railway BIM models with multi-source spatial data, improving the consistency and data integrity between the model and the actual engineering scenario. It also implements differentiated lightweighting based on the importance of components, ensuring the integrity of the geometric features and attribute information of key components while improving model compression efficiency and geometric fidelity. Furthermore, it can dynamically control the lightweighting process according to the model's complexity, avoiding over-simplification or under-simplification. Combined with the characteristics of railway lines, it enables the block organization and on-demand loading of model data, reducing data storage, transmission, and rendering overhead, and improving the visualization loading efficiency, interactive performance, and engineering application value of railway BIM models in large-scale scenarios. Example of a lightweight processing system for railway BIM models based on multi-source data processing: The lightweight railway BIM model processing system based on multi-source data processing according to an embodiment of the present invention includes a processor and a memory. The memory stores computer program instructions, which, when executed by the processor, implement the lightweight railway BIM model processing method based on multi-source data processing described above according to the present invention.

[0046] The lightweight processing system for railway BIM models based on multi-source data processing also includes other components well-known to those skilled in the art, such as communication interfaces. Their settings and functions are known in the art and will not be described in detail here.

[0047] While this specification has shown and described numerous embodiments of the invention, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Many modifications, alterations, and alternatives will occur to those skilled in the art without departing from the spirit and essence of the invention. It should be understood that various alternatives to the embodiments of the invention described herein may be employed in the practice of this invention.

Claims

1. A lightweight processing method for railway BIM models based on multi-source data processing, characterized in that, include: After acquiring the design BIM model, laser point cloud, oblique photogrammetry model and GIS terrain data, and performing unified coordinate transformation, component ID labeling and triangulation to obtain a unified triangular mesh; After obtaining the geometric consistency coefficient, professional category coefficient, and continuity coefficient along the railway mileage direction of any component in the unified triangular mesh and the laser point cloud, the engineering importance of any component is obtained by weighting it according to the preset weight. During the edge folding iteration process, the mean and mean absolute deviation of the folding costs of all candidate edges are extracted. The candidate edge is any edge in the unified triangular mesh. The folding costs of all candidate edges are semantically weighted and folded using the engineering importance. An adaptive stopping threshold is constructed using the mean and mean absolute deviation. Based on the adaptive stopping threshold, semantically weighted edge folding is iteratively performed to obtain a lightweight mesh. The three-dimensional tile data is then segmented and output along the railway mileage direction.

2. The lightweight processing method for railway BIM models based on multi-source data processing according to claim 1, characterized in that, The unified coordinate transformation is specifically as follows: The design BIM model, laser point cloud, oblique photogrammetry model, and GIS terrain data are converted to a plane projection coordinate system, and then merged according to spatial location in the plane projection coordinate system to complete the unified coordinate transformation.

3. The lightweight processing method for railway BIM models based on multi-source data processing according to claim 1, characterized in that, The unified triangular mesh is specifically as follows: Extract the globally unique identifier of each component as the component ID, use Delaunay triangulation to generate the first triangular mesh for the design BIM model; use rolling sphere triangulation with a preset neighborhood radius to generate the second triangular mesh for the laser point cloud; extract triangular facets from the oblique photogrammetry model according to the tiles and merge them to generate the third triangular mesh; use regular raster triangulation to generate the fourth triangular mesh for the GIS terrain data. Each facet in the first, second, third, and fourth triangular meshes is associated with its corresponding component ID to obtain a unified triangular mesh.

4. The lightweight processing method for railway BIM models based on multi-source data processing according to claim 1, characterized in that, The importance of the project is specifically as follows: Preset weights for professional categories, geometric consistency, and continuity; For any component, the professional category coefficient of the component is multiplied by the professional category weight to obtain the first component, the geometric consistency coefficient is multiplied by the geometric consistency weight to obtain the second component, and the mileage continuity coefficient is multiplied by the mileage continuity weight to obtain the third component. The first component, the second component, and the third component are summed to obtain the engineering importance of any component.

5. The lightweight processing method for railway BIM models based on multi-source data processing according to claim 3, characterized in that, The process of obtaining the professional category coefficient includes: Establish a mapping table between IFC entity types and professional category coefficients; read the IFC entity type of any component and look it up in the mapping table to obtain the professional category coefficient of any component.

6. The lightweight processing method for railway BIM models based on multi-source data processing according to claim 3, characterized in that, The process of obtaining the geometric consistency coefficient includes: For any triangular facet on the surface of each component, calculate the distance from the centroid of any triangular facet to the nearest point in the laser point cloud; The arithmetic mean of the distances corresponding to all the faces of each component is taken to obtain the component-level average distance; the difference between 1 and the ratio of the component-level average distance to the upper limit of distance normalization is taken as the geometric consistency coefficient of the component.

7. The lightweight processing method for railway BIM models based on multi-source data processing according to claim 1, characterized in that, Extract the mean and average absolute deviation of the folding costs of all candidate edges. Specifically, during the edge folding iteration process, calculate the folding cost of each candidate edge in the unified triangular mesh according to the quadratic error metric and add it to the priority queue. In the current iteration, the arithmetic mean of the folding costs of all candidate edges in the priority queue is taken as the current mean; the arithmetic mean of the absolute values ​​of the differences between the folding costs of each candidate edge in the priority queue and the current mean is taken as the current average absolute deviation.

8. The lightweight processing method for railway BIM models based on multi-source data processing according to claim 1, characterized in that, The process of semantically weighted edge folding and adaptive stopping thresholding includes: For each candidate edge in the priority queue, the semantic weighted cost of each candidate edge is obtained by dividing the quadratic error metric folding cost of each candidate edge by the component engineering importance of its component. The current adaptive stopping threshold is obtained by adding the product of the current mean and the preset distribution expansion coefficient with the current mean absolute deviation. Edge folding is performed in ascending order of semantic weighted cost. When the semantic weighted cost of the head of the priority queue exceeds the current adaptive stopping threshold, the edge folding iteration of the current level is stopped. The current grid is used as the input of the next level and edge folding is continued until all preset levels of lightweight grids are obtained. The three-dimensional tile data is then output by dividing the data along the railway mileage direction according to the preset unit length.

9. A lightweight processing system for railway BIM models based on multi-source data processing, characterized in that, It includes a memory and a processor, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the railway BIM model lightweight processing method based on multi-source data processing as described in any one of claims 1 to 8 is implemented.

Images