Multi-modal matching method and system for identifying repeated components in BIM model

Abstract

The invention provides a multi-modal matching method and system for identifying repeated components in a BIM model, and relates to the technical field of building information model lightweight, and the method comprises the steps: obtaining BIM format data of to-be-matched component pairs, and separating geometric attributes, semantic attributes and topological attributes of each component from the BIM format data; on the basis of geometric attributes, Hash bucket dividing is carried out on each component, geometric similarity calculation is carried out on component pairs in the same bucket, and a rough matching result is obtained according to the geometric similarity; and generating a multi-dimensional Hash signature of the component by using the geometric attribute, the semantic attribute and the topological attribute, and performing fine registration verification on the rough matching result through the multi-dimensional Hash signature to obtain a final multi-modal matching result which is used for identifying the repeated component in the BIM model. According to the method, geometric attributes, semantic attributes and topological attributes of the components are combined, and the detection speed and accuracy of the repeated components in the BIM model are improved by means of parallel computing, hash acceleration and the like.

Description

Technical Field

[0001] This invention relates to the field of lightweight building information modeling (BIM) technology, specifically to a multimodal matching method and system for identifying repetitive components in a BIM model. Background Technology

[0002] In complex building projects, BIM models contain a massive number of components, often with duplicate components that are structurally identical or similar, posing challenges to model management, data exchange, and component reuse. Traditional component matching often relies on component IDs (such as IDs automatically assigned by Revit), but ID conflicts and a lack of semantic information can occur across models or projects.

[0003] Existing patent CN114461600A proposes a data reuse method based on component identity tags (CITs), which generates unique tags through component classification codes and location codes. This method assigns category and location information to component tags, enabling multi-dimensional reuse of BIM data across platforms and stages. However, this patent focuses on data identification and centralized management, and has limitations in component matching application scenarios.

[0004] Another patent, CN110991553B, discloses a BIM model comparison method. This method constructs a shape distribution based on the geometric information of components and calculates the geometric similarity between components in different versions of the model. It also calculates the positional similarity of components by registering a coordinate system. Then, it performs a weighted superposition of geometric and positional similarities for one-to-one matching. Although this method automates the comparison of two versions of BIM models, improving the efficiency and accuracy of the comparison, it is mainly aimed at comparing the differences between different versions of the model and has shortcomings in the accuracy, performance, and stability of the matching.

[0005] In summary, existing technologies for identifying duplicate components in BIM rely on partial information for matching, which cannot balance high speed and high accuracy, and are insufficient in terms of robustness and resistance to transformation. Summary of the Invention

[0006] To address the aforementioned issues, this invention proposes a multimodal matching method and system for identifying duplicate components in BIM models. By combining the geometric, semantic, and topological attributes of components and employing parallel computing and hash acceleration techniques, the detection speed and accuracy of duplicate components in BIM models are improved. This enables the merging and deduplication of geometric components in the model, thereby reducing the size of the model to be displayed and ultimately achieving a lightweight BIM engine with a superior interactive experience.

[0007] According to some embodiments, the present invention adopts the following technical solution: Multimodal matching methods for identifying repetitive components in a BIM model include: Obtain the BIM format data of the component pairs to be matched, and extract the geometric, semantic, and topological attributes of each component. Based on geometric attributes, each component is hash-bucketed, and geometric similarity is calculated for component pairs in the same bucket. A rough matching result is obtained based on the geometric similarity. By utilizing geometric, semantic, and topological attributes, a multidimensional hash signature of the component is generated. The coarse matching result is then finely registered and verified using the multidimensional hash signature to obtain the final multimodal matching result, which is used to identify duplicate components in the BIM model.

[0008] According to some embodiments, the present invention adopts the following technical solution: A multimodal matching system for identifying repetitive components in a BIM model includes: The data acquisition module is configured to: acquire BIM format data of the component pairs to be matched, and extract the geometric, semantic and topological attributes of each component from them; The coarse matching module is configured to: perform hash binning on each component based on geometric attributes, calculate the geometric similarity of component pairs in the same bin, and obtain the coarse matching result based on the geometric similarity; The fine registration module is configured to generate a multidimensional hash signature for a component using geometric, semantic, and topological attributes, and then verify the coarse matching result using the multidimensional hash signature to obtain the final multimodal matching result, which is used to identify duplicate components in the BIM model.

[0009] According to some embodiments, the present invention adopts the following technical solution: A computer program product includes a computer program that, when executed by a processor, implements the multimodal matching method for identifying repetitive components in a BIM model.

[0010] According to some embodiments, the present invention adopts the following technical solution: A non-transitory computer-readable storage medium is provided for storing computer instructions, which, when executed by a processor, implement the multimodal matching method for identifying repetitive components in a BIM model.

[0011] According to some embodiments, the present invention adopts the following technical solution: An electronic device includes a processor, a memory, and a computer program; wherein the processor is connected to the memory, the computer program is stored in the memory, and when the electronic device is running, the processor executes the computer program stored in the memory to enable the electronic device to perform the multimodal matching method for identifying repetitive components in a BIM model.

[0012] Compared with the prior art, the beneficial effects of the present invention are as follows: First, by using pre-hashing filtering and GPU parallel computing, the number of matching pairs is significantly reduced and similarity calculation is accelerated, enabling the rapid location of candidate duplicate components in large-scale models. Secondly, multidimensional hash signatures greatly improve matching accuracy and enhance robustness to position and orientation changes by integrating geometric, semantic and topological attributes, and will not easily mismatch components of different categories or connection methods together.

[0013] By comprehensively utilizing the above methods, this invention can significantly improve the efficiency and accuracy of duplicate component detection. For example, the hash acceleration scheme reported by Lin et al. can reduce BIM change detection time by 98%, and this invention can adopt similar strategies to achieve the effect of accelerating detection. At the same time, semantic classification reduces false alarms, making the detection accuracy close to 100%. Compared with the prior art, this invention significantly reduces computational overhead while maintaining high-precision matching, and solves problems such as volatile component IDs, low matching efficiency, and sensitivity to model transformations. Attached Figure Description

[0014] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0015] Figure 1 This is a flowchart of the method in Example 1. Detailed Implementation The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0016] It should be noted that the following detailed descriptions are exemplary and intended to provide further illustration of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0017] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of exemplary embodiments according to the invention. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0018] Example 1 One embodiment of the present invention provides a multimodal matching method for identifying repetitive components in a BIM model, comprising: Step 1: Obtain the BIM format data of the component pairs to be matched, and extract the geometric attributes, semantic attributes and topological attributes of each component; Step 2: Based on geometric attributes, hash bucket each component, calculate the geometric similarity of component pairs in the same bucket, and obtain a rough matching result based on the geometric similarity. Step 3: Using geometric, semantic, and topological attributes, generate multidimensional hash signatures for components. Use the multidimensional hash signatures to perform fine registration verification on the coarse matching results to obtain the final multimodal matching results, which are used to identify duplicate components in the BIM model.

[0019] As one embodiment, the multimodal matching method of the present invention is used to identify duplicate components in a BIM model, thereby reducing the size of the model to be displayed and ultimately realizing the interactive experience of the lightweight BIM engine. The following describes the specific implementation process using the example of how to deduplicate and render the target BIM model.

[0020] To demonstrate the role of this invention in the entire BIM lightweighting process, the overall implementation process of the BIM model lightweighting project is described below. The focus is on how to deduplicate and render components in the target BIM model. Deduplication involves using a multimodal matching method to match component pairs. Matched component pairs are considered identical components, and these identical components are merged, retaining only one, thus achieving the effect of deduplication and lightweighting. Other lightweighting steps are only briefly described, such as... Figure 1 As shown, the specific implementation steps are as follows: 1. Format conversion: Obtain the original BIM format of the target BIM model, such as RVT, IFC, etc.

[0021] The original BIM format is parsed to extract the geometric information (triangular mesh, curved surface, etc.) and semantic attributes (component attributes, material information, component category, material, floor number, etc.) of each component.

[0022] 2. Separation of digital and analog models: Separate geometric information from semantic attributes, strip the semantic attributes from the original file, store them in an independent database (such as NoSQL), and call them on demand when identifying duplicate components and front-end display.

[0023] 3. Triangulation and parameterization: The curved surfaces in the geometric information are converted into triangular networks. Combined with the triangulation algorithm, the network topology is optimized and the number of triangular faces is reduced.

[0024] The triangulated geometric information is parameterized to obtain geometric attributes, including basic geometric primitives, parameterized geometric bodies, spatial positions, rotation angles, and other information. Connectivity information, including connections, adjacency, and containment relationships between geometric elements, is extracted from the triangulated geometric information and used as topological attributes.

[0025] 4. Identification of duplicate components: Duplicate components are identified using an algorithm, and only one merged component is stored, reducing storage redundancy. The detailed steps are as follows: (1) Coarse matching By combining the geometric properties of the components, a rough matching of repeating components is achieved, which is specifically divided into two sub-steps: 1) Pre-hash filtering: Based on the geometric properties of each component, a coarse hash is generated to pre-hash the components, such as quantizing the bounding box or voxel histogram of the component and building hash buckets to quickly filter component pairs of different categories or with large size differences.

[0026] Specifically, the bounding box size and voxelized histogram are calculated for each component. This information is then discretized and bucketed using a hash function to quickly eliminate component pairs that are significantly different in category or size. After pre-hashing filtering, only component combinations that may be duplicated in the hash buckets are retained as candidate sets for subsequent matching.

[0027] 2) Fast geometric matching: For the pre-screened candidate component pairs, the geometric similarity is calculated in parallel using GPU (such as calculating the overlap rate of point cloud voxels or the similarity of shape distribution).

[0028] Specifically, the distance or correlation between two components in a component pair is computed in parallel in the GPU to quickly estimate the geometric similarity of their shape histograms. This step utilizes the parallel computing capabilities of the GPU, which significantly improves the processing speed of batch matching and reduces the total computation time.

[0029] (2) Fine-grained registration verification: For component pairs with high geometric similarity, a simplified ICP registration algorithm is used for fine-grained comparison and verification based on multidimensional hash signatures, which consists of two sub-steps: 1) Generate multidimensional hash signatures: The geometric attributes, semantic categories, and floor information of components obtained through rough matching, as well as the topological attributes such as the connection features of adjacent components (such as the number of neighbors and the category of neighbors), are combined (i.e., spliced) into a feature vector. Then, a binary signature is generated through a multiple random projection function. This signature is a multidimensional hash signature, which contains compressed geometric features, semantic labels, and topological coding information, and ensures that the signatures of similar components are close in the Hamming space.

[0030] 2) Fine-grained registration verification: For candidate pairs (A, B), further precise alignment verification is performed based on geometry, semantics, and topology.

[0031] Specifically, the Hamming spatial network is extracted from the multidimensional hash signature of the component, the Iterative Closest Point (ICP) algorithm is run to calculate the optimal rigid transformation, and A is registered to B. The residual after registration is calculated using the mean squared error of the distance between points. If the residual is lower than the threshold and the transformation matrix meets the expectations, it is confirmed to be the same component; otherwise, it is considered a mismatch and excluded.

[0032] The transformation matrix here is the spatial transformation matrix between components. For the identified duplicate components, since the main structure is the same, the only difference is the rotation angle of the position, i.e. the spatial transformation matrix. This is achieved through matrix calculation during the later rendering process to restore N components.

[0033] By combining coarse matching and fine registration, this process enables rapid and accurate identification of repetitive components in BIM models. Compared to traditional methods, this process does not rely on component IDs and is unaffected by the component's position or orientation in the model, enabling efficient identification of similar components and supporting their reuse.

[0034] 5. Rendering process For repetitive components (such as reinforcing bars), GPU instantiation is used for rendering, and rendering instructions for the same component are merged into a single rendering, reducing the computational requirements of the graphics card. The model space is divided into spatial blocks, and an octree data structure is constructed to render only the blocks within the view frustum. As the user's viewpoint moves, the view frustum pruning algorithm dynamically calculates the visible area and schedules only the LOD data of the blocks within the view frustum. This achieves the following: at long distances, only the wireframe outline of the parametric geometry is rendered (LOD0); at mid-range distances, a simplified mesh that preserves the topology is switched to (LOD2); and at close distances, a high-precision mesh with millions of triangular faces is loaded (LOD4).

[0035] Specifically, the LOD data of blocks within the view frustum is implemented using LOD layering, as follows: Constructing multiple levels of detail (such as LOD0 with 100 polygons, LOD1 with 1000 polygons, LOD2 with 10000 polygons, LOD3 with 100000 polygons, and LOD4 with 1000000 polygons) and dynamically loading them according to the viewing distance, i.e., using a low-poly model (LOD0) for distant scenes and a high-poly model (LOD4) for close-up scenes.

[0036] Example 2 One embodiment of the present invention provides a multimodal matching system for identifying repetitive components in a BIM model, comprising: The data acquisition module is configured to: acquire BIM format data of the component pairs to be matched, and extract the geometric, semantic and topological attributes of each component from them; The coarse matching module is configured to: perform hash binning on each component based on geometric attributes, calculate the geometric similarity of component pairs in the same bin, and obtain the coarse matching result based on the geometric similarity; The fine registration module is configured to generate a multidimensional hash signature for a component using geometric, semantic, and topological attributes, and then verify the coarse matching result using the multidimensional hash signature to obtain the final multimodal matching result, which is used to identify duplicate components in the BIM model.

[0037] Example 3 One embodiment of the present invention provides a computer program product, including a computer program that, when executed by a processor, implements the multimodal matching method for identifying repetitive components in a BIM model.

[0038] Example 4 In one embodiment of the present invention, a non-transitory computer-readable storage medium is provided for storing computer instructions. When the computer instructions are executed by a processor, they implement the multimodal matching method for identifying repetitive components in a BIM model.

[0039] Example 5 One embodiment of the present invention provides an electronic device, including: a processor, a memory, and a computer program; wherein the processor is connected to the memory, the computer program is stored in the memory, and when the electronic device is running, the processor executes the computer program stored in the memory to enable the electronic device to implement the multimodal matching method for identifying repetitive components in a BIM model.

[0040] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0041] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0042] While the specific embodiments of the present invention have been described above in conjunction with the accompanying drawings, this is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art without creative effort based on the technical solutions of the present invention are still within the scope of protection of the present invention.

Claims

1. A multimodal matching method for identifying repetitive components in a BIM model, characterized in that, include: Obtain the BIM format data of the component pairs to be matched, and extract the geometric, semantic, and topological attributes of each component. Based on geometric attributes, each component is hash-bucketed, and geometric similarity is calculated for component pairs in the same bucket. A rough matching result is obtained based on the geometric similarity. By utilizing geometric, semantic, and topological attributes, a multidimensional hash signature of the component is generated. The coarse matching result is then finely registered and verified using the multidimensional hash signature to obtain the final multimodal matching result, which is used to identify duplicate components in the BIM model.

2. The multimodal matching method for identifying duplicate components in a BIM model as described in claim 1, characterized in that, The geometric attributes include basic geometric primitives, parametric geometry, spatial position, and rotation angle; The semantic attributes include component attributes, material information, component category, material, and floor number; The topological attributes include the connections, adjacencies, and containments between geometric elements.

3. The multimodal matching method for identifying duplicate components in a BIM model as described in claim 1, characterized in that, The separation specifically refers to: Parse BIM format data to extract geometric information and semantic attributes; Geometric information is triangulated and parameterized to obtain geometric properties; Connectivity information is extracted from the triangulated geometric information and used as topological attributes.

4. The multimodal matching method for identifying duplicate components in a BIM model as described in claim 1, characterized in that, The coarse matching includes two sub-steps: pre-hash filtering and fast geometric matching; The pre-hashing filtering is based on geometric attributes, and each component is hashed and bucketed. Component pairs in the same bucket are used as candidate component pairs for subsequent fast geometric matching. The fast geometric matching method involves computing the distance or correlation between two distributions in parallel on a GPU for candidate component pairs after pre-hashing screening, thereby quickly estimating the geometric similarity between them.

5. The multimodal matching method for identifying duplicate components in a BIM model as described in claim 4, characterized in that, The hash bucketing is specifically as follows: Based on geometric properties, the bounding box size and voxelized histogram of the component are calculated. After discretizing the bounding box size and voxelized histogram, the component is bucketed using a hash function to quickly exclude component pairs that are significantly different in category or size, and only retain component pairs that may be duplicated in the hash bucket.

6. The multimodal matching method for identifying repetitive components in a BIM model as described in claim 1, characterized in that, The generation of the multidimensional hash signature is specifically as follows: Geometric, semantic, and topological attributes are concatenated into a feature vector, and a binary signature is generated through a multiple random projection function to obtain a multidimensional hash signature.

7. The multimodal matching method for identifying duplicate components in a BIM model as described in claim 1, characterized in that, The detailed verification of the coarse matching result through multidimensional hash signature is as follows: Hamming spatial networks are extracted from the multidimensional hash signatures of the components. The Hamming spatial networks of the two components are registered using the ICP algorithm. Fine-grained registration verification is then performed based on the mean square error of the distance between the registered points.

8. A multimodal matching system for identifying repetitive components in a BIM model, characterized in that, include: The data acquisition module is configured to: acquire BIM format data of the component pairs to be matched, and extract the geometric, semantic and topological attributes of each component from them; The coarse matching module is configured to: perform hash binning on each component based on geometric attributes, calculate the geometric similarity of component pairs in the same bin, and obtain the coarse matching result based on the geometric similarity; The fine registration module is configured to generate a multidimensional hash signature for a component using geometric, semantic, and topological attributes, and then verify the coarse matching result using the multidimensional hash signature to obtain the final multimodal matching result, which is used to identify duplicate components in the BIM model.

9. A non-transitory computer-readable storage medium, characterized in that, The non-transitory computer-readable storage medium is used to store computer instructions, which, when executed by a processor, implement the multimodal matching method for identifying repetitive components in a BIM model as described in any one of claims 1-7.

10. An electronic device, characterized in that, include: The device includes a processor, a memory, and a computer program; wherein the processor is connected to the memory, the computer program is stored in the memory, and when the electronic device is running, the processor executes the computer program stored in the memory to enable the electronic device to perform the multimodal matching method for identifying repetitive components in a BIM model as described in any one of claims 1-7.

Images