MIC prefabricated component rollover method based on BIM and CAD fusion

Abstract

The invention relates to the technical field of building information models and digital construction, and discloses an MIC prefabricated component rollover method based on BIM and CAD fusion, which comprises the following steps of: firstly, performing geometric cleaning and semantic feature extraction on two-dimensional CAD data of an MIC component, and determining a semantic type and an initial geometric parameter of a primitive; according to the semantic type and the initial geometric parameters, calculating the connection probability between the components, and constructing an MIC component space topological correlation graph for describing the internal logic relation of the MIC components; and calculating visual saliency feature points of the component as invisible anchor points. According to the method, multi-dimensional geometric feature vectors are constructed, a weighted scoring function is used for semantic judgment, the strong dependence of a traditional rollover technology on a CAD graph layer name is eliminated, the parallelism, the closure and the relative inclusion relation of graphic elements are deeply analyzed, a system can accurately distinguish structural parts and electromechanical pipelines according to the geometric nature of graphs, and the reliability of the structural parts and the electromechanical pipelines is improved. And the robustness and accuracy of semantic recognition of the MIC component are remarkably improved.

Description

Technical Field

[0001] This invention relates to the fields of building information modeling and digital construction technology, specifically to a method for reverse molding of prefabricated components based on the integration of BIM and CAD. Background Technology

[0002] Modular integrated building (MIC), as a highly integrated form of building industrialization, requires high-precision prefabrication and assembly of structure, decoration and electromechanical pipelines in the factory. This makes the accuracy and efficiency of the conversion from two-dimensional design drawings to three-dimensional information models (BIM) a key factor in determining the success or failure of a project. Although automatic model conversion technology based on CAD drawings has been applied in the general building field, existing technologies still have significant limitations when facing special scenarios such as MIC with high density and high integration.

[0003] In existing CAD modeling technologies, mainstream solutions rely heavily on layer naming or simple geometric rules for element recognition and 3D conversion. This approach is overly dependent on the rigor of upstream design-end drawing standards. Once CAD drawings contain disordered layers, non-standard line types, or non-standard elements, modeling software often fails to correctly identify the semantics of components, resulting in a large number of erroneous geometric entities in the generated BIM model. More importantly, existing modeling technologies typically treat each component as an independent geometric object and discretize it, lacking in-depth analysis of the physical connection logic and spatial topological relationships between components. This results in the generated model being merely a collection of geometric shapes, unable to reflect the logical connections between the complex electromechanical pipelines and structural skeleton within the MIC module. This directly limits the application value of the model in subsequent detailed design and construction guidance.

[0004] Furthermore, in the on-site comparison of virtual and real data and the application of drawings, the existing technical processes are often disconnected. The generation of BIM models and the positioning data of on-site augmented reality (AR) applications are often out of sync. On-site AR positioning mainly relies on QR code markers or GPS signals, but these traditional positioning methods are prone to failure or insufficient accuracy due to the complex electromagnetic environment and line-of-sight obstruction inside the MIC module. At the same time, due to the lack of preprocessing of the model's logical relationships, mobile devices often have to render the full amount of data when loading high-precision MIC models, making it difficult to dynamically load local data according to the connection logic of components. This results in severe rendering lag and fails to meet the requirements for transparent acceptance of concealed works. In addition, the current model-changing process exhibits a unidirectional data flow characteristic. Model deviations found by on-site measurements can usually only be corrected manually. There is a lack of a closed-loop mechanism to feed on-site measurement data to the front-end model-changing algorithm, which prevents the model-changing algorithm from automatically optimizing recognition parameters by learning historical deviations. This makes it difficult to fundamentally solve the systemic problem of inaccurate identification of non-standard components. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a method for reverse molding of prefabricated MIC components based on the integration of BIM and CAD. This method solves the problems of existing MIC component reverse molding technologies, which rely heavily on CAD layer specifications, resulting in low semantic recognition rates, lack of internal logical topological relationships in the generated models, inability of model data to support high-precision AR positioning and dynamic rendering of unmarked points on site, and lack of a reverse self-correction mechanism for reverse molding parameters based on on-site measured data.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a MIC prefabricated component formwork transformation method based on the integration of BIM and CAD, comprising the following steps: S1. First, perform geometric cleaning and semantic feature extraction on the two-dimensional CAD data of the MIC component to determine the semantic type and initial geometric parameters of the primitives. S2. Based on the semantic type and initial geometric parameters, calculate the connection probability between components and construct a spatial topology association map of MIC components that describes the internal logical relationship of MIC components. S3. Calculate the visually salient feature points of the component as hidden anchor points, and encapsulate the data of the hidden anchor points and the spatial topological association map of the MIC component into a parametric BIM model to generate a BIM model that integrates topological semantics. S4. In the on-site application phase, spatial positioning is performed using the invisible anchor points in the BIM model, and augmented reality (AR) dynamic rendering is performed on the BIM model based on the spatial topology association map of the MIC components. S5. Collect on-site measured point cloud data, calculate the geometric deviation between the on-site measured point cloud data and the BIM model, correct the key parameters in the semantic feature extraction stage in reverse based on the geometric deviation, and regenerate the BIM model using the corrected key parameters.

[0007] Preferably, the geometric cleaning of the two-dimensional CAD data of the MIC component includes: Traverse all geometric primitives in the two-dimensional CAD data and set noise threshold and closure determination threshold; If the geometric length of a geometric primitive is less than the noise threshold, the geometric primitive is considered noise and is removed. If the Euclidean distance between the endpoints of two geometric primitives is less than the closure threshold, then a vertex merging operation is performed to construct a closed geometric path.

[0008] Preferably, determining the semantic type of a primitive includes: Construct a multidimensional geometric feature vector for primitives, wherein the components of the multidimensional geometric feature vector include at least the parallelism, closure, aspect ratio, and relative inclusion relationship of the primitives; The semantic score of a primitive is calculated using a weighted scoring function. The semantic score is the sum of the normalized values ​​of each feature component and the corresponding weight coefficients. A preset set of classification thresholds is used to compare the semantic score with the set of classification thresholds to determine whether the semantic type of the graphic element belongs to a structural component, electromechanical pipeline, or other component.

[0009] Preferably, the construction of the MIC component spatial topology association map describing the internal logical relationships of MIC components includes: Select nodes identified as structural components and nodes of electromechanical pipelines, and calculate the connection probability that the two are logically related; the connection probability is jointly determined by the minimum Euclidean distance of the three-dimensional bounding box and the overlap state of the two-dimensional projection. A topological threshold is set. If the calculated connection probability is greater than the topological threshold, a connection edge is established in the graph to form the spatial topological association graph of the MIC component.

[0010] Preferably, the visually salient feature points of the computational component are used as implicit anchor points, including: For the identified rigid structural primitives, geometric corner points are extracted as candidate points; The visual saliency score of the candidate point is calculated. The visual saliency score is determined by the local curvature modulus, the line density gradient, and the construction occlusion factor, wherein the construction occlusion factor is based on the installation procedure preset of the MIC component. Select several points that rank highly in the visual saliency score as the hidden anchor points, and record their absolute coordinates and visual descriptors.

[0011] Preferably, the step of encapsulating the data of the hidden anchor points and the spatial topology association map of the MIC components into a parametric BIM model includes: Based on the semantic type, the preset MIC parameterized family library is invoked to map the extracted initial geometric parameters to family parameters to generate a 3D solid. The topology ID of the generated MIC component spatial topology association map and the calculated visual descriptor of the hidden anchor point are written into the extended attribute field of the corresponding BIM component to realize data ID association.

[0012] Preferably, the spatial positioning using the hidden anchor points in the BIM model includes: AR devices are used to collect images of the scene and extract feature points. The on-site feature points are matched with the visual descriptors of the hidden anchor points pre-embedded in the BIM model; Based on the matching results, establish the transformation matrix between the world coordinate system and the model coordinate system.

[0013] Preferably, the step of performing augmented reality (AR) dynamic rendering on the BIM model based on the spatial topology association map of the MIC components includes: When the current component node is selected, the remaining component nodes are traversed in the MIC component space topology association graph. Calculate the shortest path hop count from the remaining component nodes to the current component node; Set a rendering level depth threshold, and only render components whose shortest path hop count is less than or equal to the rendering level depth threshold, while hiding components that have no logical connection.

[0014] Preferably, calculating the geometric deviation between the measured point cloud data and the BIM model includes: The local sparse point cloud on site is obtained as the measured point cloud data on site, and the optimal rigid body transformation matrix of the measured point cloud data on site and the BIM mesh model is calculated and registered using the iterative nearest point algorithm. Calculate the average deviation vector of the key feature surface. If the magnitude of the average deviation vector exceeds the preset tolerance, it is determined that the geometric deviation exists.

[0015] Preferably, the step of reversing the key parameters of the semantic feature extraction stage based on the geometric deviation includes: The parameters that affect the geometry of the component are identified as the key parameters. Calculate the component of the average deviation vector in the direction of the normal vector on the component surface; The key parameters are updated using the components and a preset learning rate, and the updated key parameters are fed back to the semantic feature extraction step for re-determination of semantics.

[0016] This invention provides a method for MIC prefabricated component formwork transformation based on the integration of BIM and CAD. It has the following beneficial effects: 1. This invention constructs multidimensional geometric feature vectors and uses a weighted scoring function for semantic determination, thus eliminating the strong dependence of traditional model-making technology on CAD layer names. This method deeply analyzes the parallelism, closure, and relative inclusion relationships of graphic elements, enabling the system to accurately distinguish between structural components and electromechanical pipelines based on the geometric essence of the graphics even when the original CAD drawings are poorly managed or the drawings are not standardized. This significantly improves the robustness and accuracy of semantic recognition of MIC components and avoids model generation errors caused by layer errors.

[0017] 2. This invention constructs a spatial topological relationship map inside the MIC component by calculating the three-dimensional spatial distance and two-dimensional projection overlap between components, realizing a leap from simple geometric modeling to logical modeling. This feature enables the generated BIM model to not only contain appearance data, but also store the physical connection and assembly logic between components using this topological map as a carrier. This solves the problem of isolated components and inability to perform systematic logical tracing in traditional mold-making models, and provides a solid data foundation for subsequent intelligent drawing review.

[0018] 3. This invention pre-calculates invisible anchor points by combining geometric curvature, line density gradient and construction occlusion factor during the molding stage, and encapsulates the anchor point data into the BIM model, thereby achieving high-precision AR positioning of markerless points on the construction site. This effectively overcomes the shortcomings of traditional QR codes being easily damaged or lost and GPS signals being unstable in indoor environments such as substations, ensuring the stability of virtual-real integration.

[0019] 4. This invention solves the problem of mobile rendering lag and visual occlusion caused by dense internal pipelines in MIC prefabricated components by making AR dynamic rendering decisions based on the shortest path hop count of the spatial topology association map. This enables the system to intelligently display only the pipeline system that is logically connected to the currently interested component, and automatically hide unrelated decorative layers or remote components. While reducing the computing load on the equipment, it greatly improves the intuitiveness and efficiency of on-site hidden works acceptance.

[0020] 5. This invention calculates the geometric deviation between the measured point cloud data on site and the BIM model, and constructs a reverse correction mechanism to update the key parameters in the semantic feature extraction stage, forming a technical closed loop of design-manufacturing-feedback-optimization. This enables the system to adaptively eliminate systematic deviations caused by differences in CAD drawing habits or manufacturing errors of non-standard components, and continuously improve the accuracy of the model generated by the mold and the matching degree with the actual object. Attached Figure Description

[0021] Figure 1 This is one of the schematic diagrams of the method flow of the present invention; Figure 2 This is a second schematic diagram of the method flow of the present invention; Figure 3 This is the third schematic diagram of the method flow of the present invention; Figure 4 This is the fourth schematic diagram of the method flow of the present invention. Detailed Implementation

[0022] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] Please see the appendix Figure 1 -Appendix Figure 4 This invention provides a method for reverse molding of prefabricated MIC components based on the integration of BIM and CAD, including the following steps: S1. First, perform geometric cleaning and semantic feature extraction on the 2D CAD data of the MIC component to determine the semantic type and initial geometric parameters of the primitives. Determining the semantic type of the primitives includes: constructing a multi-dimensional geometric feature vector of the primitives. The components of the multi-dimensional geometric feature vector include at least the parallelism, closure, aspect ratio, and relative inclusion relationship of the primitives; calculating the semantic score of the primitives using a weighted scoring function. The semantic score is the sum of the products of the normalized values ​​of each feature component and the corresponding weight coefficients; setting a preset classification threshold set, comparing the semantic score with the classification threshold set, and determining whether the semantic type of the primitive belongs to a structural component, electromechanical pipeline, or other component. Perform geometric cleaning on the 2D CAD data of the MIC component, including: traversing all geometric primitives in the 2D CAD data, setting a noise threshold and a closure judgment threshold; if the geometric length of a geometric primitive is less than the noise threshold, the geometric primitive is regarded as noise and removed; if the Euclidean distance between the endpoints of two geometric primitives is less than the closure judgment threshold, a vertex merging operation is performed to construct a closed geometric path. Specifically, in the MIC project, the raw data input is usually two-dimensional CAD drawings. Due to differences in design habits, the drawings often contain a large number of unstructured redundant lines and non-standard layer names. This step aims to convert these unstructured data into structured semantic data that computers can understand. First, the system reads all geometric metadata sets from the CAD file. To eliminate plotting errors and redundant data, perform the following operations: Noise reduction: Set noise threshold (e.g., 2mm), traverse all primitives and calculate their geometric lengths. .like If the element is determined to be a noise point or a broken line caused by accidental touch, it will be removed from the set. Topology closure processing: This addresses the common "visual closure but data disconnection" scenario in CAD by setting a closure determination threshold. (e.g., 1mm), calculate the endpoints of any two primitives. and Euclidean distance between ) If the distance is less than Then, the vertex snapping and merging operation will be automatically performed to connect broken lines into continuous geometric paths, so as to ensure that the subsequently generated BIM components have closed sections. To eliminate the dependence on CAD layer names, this invention employs a geometric feature analysis method to analyze each cleaned closed primitive. Construct a multidimensional geometric feature vector The vector contains the following key components: Parallelism feature: Determines whether a graphic element is composed of two parallel lines (commonly seen in walls and pipes); Closure characteristic: Boolean value, indicating whether the elements are connected end to end; Aspect Ratio Characteristics: The aspect ratio of the smallest bounding rectangle of a graphic element (long strips are usually walls or beams, while squares or circles are usually columns or manholes). Containment relationship feature: Determine whether the graphic element is completely contained by other large-scale graphic elements (such as an opening in a wall); Based on the aforementioned feature vectors, a linear weighted scoring model is used to calculate the semantic score of the primitives.

[0024] in, For the first The weight coefficients of each feature (set based on experience or training data), The system uses a pre-defined set of classification thresholds as the normalization function. The semantic determination logic is as follows: like These are classified as structural components (such as walls, steel frames, and beams); like It was determined to be a mechanical and electrical pipeline (MEP). Otherwise, it is determined to be other decorative components. This step outputs a geometric dataset with clear semantic labels, providing attribute basis for subsequent 3D modeling.

[0025] S2. Based on the semantic type and initial geometric parameters, calculate the connection probability between components and construct a spatial topological association graph of MIC components describing the internal logical relationships of MIC components. This includes: selecting nodes identified as structural components and nodes of electromechanical pipelines, and calculating the connection probability that the two have a logical relationship; the connection probability is jointly determined by the minimum Euclidean distance of the three-dimensional bounding box and the overlap state of the two-dimensional projection; setting a topological threshold, if the calculated connection probability is greater than the topological threshold, then establishing connection edges in the graph to form a spatial topological association graph of MIC components. Specifically, MIC components are highly integrated with intricate internal pipelines. Traditional molding only generates independent geometric models and lacks logical relationships between components. This step aims to construct a diagram describing the physical connection logic. The system traverses all structural nodes identified in step S1. and electromechanical pipeline nodes Calculate the probability that there is a logical relationship between them (such as pipelines passing through the keel or pipelines being fixed to the bracket). This probabilistic model integrates the relationship between three-dimensional spatial distance and two-dimensional projection:

[0026] The specific parameters have the following meanings: : Calculation component and The minimum Euclidean distance between bounding boxes in three-dimensional space; the closer the distance, the higher the probability of association. : Smoothing constant to prevent the denominator from being zero; : Two-dimensional projection overlap indicator function, if and If there is an overlapping area in the projection of the CAD plane, the value is 1; otherwise, it is 0. This reflects that the overlapping of lines in CAD drawings usually implies an assembly relationship.

[0027] These are distance weights and overlap weights, used to balance the effects of spatial proximity and visual overlap. Set a topology decision threshold The system iterates through all node pairs: If the calculated result In the topology graph In the node and Establish an undirected edge between them The final generated topological map It not only records the geometric position of the components, but also the logical assembly relationship of A supporting B and C passing through A.

[0028] S3. Calculate the visual saliency feature points of the components as hidden anchor points, and encapsulate the data of the hidden anchor points and the spatial topology association map of the MIC components into a parametric BIM model to generate a BIM model that integrates topological semantics. Encapsulating the data of the hidden anchor points and the spatial topology association map of the MIC components into the parametric BIM model includes: calling the preset MIC parametric family library according to the semantic type, mapping the extracted initial geometric parameters to the family parameters to generate a three-dimensional entity; writing the topology ID of the generated spatial topology association map of the MIC components and the visual descriptors of the calculated hidden anchor points into the extended attribute fields of the corresponding BIM components to achieve data ID association; calculating the visual saliency feature points of the components as hidden anchor points includes: extracting geometric corner points as candidate points for the identified rigid structural elements; calculating the visual saliency score of the candidate points, which is jointly determined by the local curvature modulus, line density gradient, and construction occlusion factor, where the construction occlusion factor is based on the installation procedure preset of the MIC components; selecting several points with the highest visual saliency scores as hidden anchor points, and recording their absolute coordinates and visual descriptors. Specifically, the preliminary data preparation work for AR applications was first integrated into the mold-making stage. In order to achieve high-precision AR positioning on-site without QR codes, the system analyzed the rigid structural primitives extracted in step S1 (such as the steel structural corner pieces and embedded parts edges of the MIC module), and searched for feature points suitable for computer vision recognition. For candidate feature points... Calculate its visual saliency score :

[0029] in (Geometric curvature): Prioritize points with obvious geometric features such as corners and apex; (Contrast Gradient): Prioritize areas with dense lines and high contrast with the background; (Construction Obscuration Factor): This is a crucial prior parameter. Based on the MIC assembly sequence (sequence attributes in BIM), if a point will be covered by a decorative panel in a subsequent sequence, then... This avoids selecting invalid anchor points and selects... highest Each point serves as an implicit anchor point, recording its absolute coordinates in the model coordinate system. and the corresponding visual descriptor; When generating the final BIM model (such as Revit or IFC format), perform the following data encapsulation operations: Geometric Instantiation: Using the parameters (length, width, height, and positioning points) extracted by S1, a preset parametric family is used to generate a 3D solid. Information embedding: The topology ID generated by S2 and the anchor point descriptor generated by S3 are written into the extended attributes of the BIM component. The resulting BIM model is no longer a simple geometric model, but an intelligent model with its own positioning navigation map and logical relationship network.

[0030] S4. In the on-site application phase, spatial positioning is achieved using invisible anchor points in the BIM model, and augmented reality (AR) dynamic rendering is performed on the BIM model based on the MIC component spatial topology association map. This includes: when the current component node is selected, traversing the remaining component nodes in the MIC component spatial topology association map; calculating the shortest path hop count from the remaining component nodes to the current component node; setting a rendering layer depth threshold, rendering only components with a shortest path hop count less than or equal to the rendering layer depth threshold, and hiding components with no logical connection. Spatial positioning is achieved using invisible anchor points in the BIM model, including: acquiring on-site images through AR devices and extracting on-site feature points; matching the on-site feature points with the visual descriptors of the pre-embedded invisible anchor points in the BIM model; and establishing a transformation matrix between the world coordinate system and the model coordinate system based on the matching results. Specifically, on the construction site, AR devices (such as tablets or AR glasses) load the generated BIM model. The AR device then captures images of the site via a camera, extracts image feature points in real time, and performs feature matching with pre-stored hidden anchor point descriptors in the BIM model. The PnP algorithm is used to calculate the camera pose, thereby establishing a precise transformation matrix between the world coordinate system and the model coordinate system. Since the anchor points are selected based on construction visibility, they have extremely high robustness. To address the issue of complex internal structures in the MIC module and the potential for obstruction of view when fully displayed, the system employs a logical perspective strategy. When the user selects (or identifies) the current component in the AR screen... When rendering a main keel (e.g.), the rendering engine reads the topology map from its extended properties and calculates the remaining components. arrive Shortest path hop count Set the rendering layer depth threshold (For example, setting it to 1 means only directly connected components are displayed), render collection Defined as:

[0031] The system only renders collections The system automatically hides irrelevant decorative layers or remote pipelines in the components, which can greatly improve the efficiency of drawing review.

[0032] S5. Collect on-site measured point cloud data, calculate the geometric deviation between the on-site measured point cloud data and the BIM model, and reverse-correct the key parameters in the semantic feature extraction stage based on the geometric deviation. Then, regenerate the BIM model using the corrected key parameters. The calculation of the geometric deviation between the on-site measured point cloud data and the BIM model includes: acquiring local sparse point cloud data as on-site measured point cloud data; using the iterative nearest point algorithm to calculate the optimal rigid body transformation matrix between the on-site measured point cloud data and the BIM mesh model for registration; calculating the average deviation vector of key feature surfaces; if the magnitude of the average deviation vector exceeds the preset tolerance, a geometric deviation is determined to exist; and reversing the key parameters in the semantic feature extraction stage based on the geometric deviation, including: determining the parameters affecting the geometric shape of the component as key parameters; calculating the component of the average deviation vector in the direction of the normal vector on the component surface; updating the key parameters using the component and the preset learning rate, and feeding the updated key parameters back to the semantic feature extraction step for re-semantic judgment. Specifically, firstly, a mobile device with LiDAR capability is used to acquire local sparse point clouds of the on-site components. The ICP algorithm is used to combine point clouds with BIM mesh models. Perform registration and calculate the optimal transformation matrix. :

[0033] After registration, calculate the average normal deviation vector of key feature surfaces (such as walls). ; If the deviation modulus Exceeding the preset tolerance (This indicates that the BIM model dimensions do not match the actual dimensions, and construction errors have been ruled out), thus triggering a correction mechanism. Assuming the wall thickness parameter identified in stage S1 is... (This value is derived from CAD line spacing), the correction formula is:

[0034] in For learning rate, For the wall normal vector, the system will correct the parameters. Feedback is sent back to stage S1 to update the judgment threshold or geometric parameter extraction rules in semantic feature extraction. When it is found that all the walls reproduced are thinner than the actual objects, the system will automatically increase the thickness compensation value in the recognition algorithm and automatically regenerate the BIM model. As the number of times it is used increases, the accuracy of the reproduction becomes higher and higher.

[0035] In summary, this invention provides a method for reverse engineering of prefabricated MIC components based on the integration of BIM and CAD. By constructing multi-dimensional geometric feature vectors and using a weighted scoring function for semantic determination, it eliminates the strong dependence of traditional reverse engineering techniques on CAD layer names. This method deeply analyzes the parallelism, closure, and relative inclusion relationships of graphic elements, enabling the system to accurately distinguish between structural components and electromechanical pipelines based on the geometric essence of the graphics, even when the original CAD drawings have chaotic layer management or are not drawn in a standardized manner. This significantly improves the robustness and accuracy of semantic recognition of MIC components, avoids model generation errors caused by layer errors, and calculates the geometric deviation between the BIM model and the actual measured point cloud data. A reverse correction mechanism is then constructed to update the key parameters in the semantic feature extraction stage, forming a technical closed loop of design-manufacturing-feedback-optimization. This mechanism can use actual on-site data to reverse-calibrate the recognition threshold of the reverse engineering algorithm, enabling the system to adaptively eliminate systematic deviations caused by differences in CAD drawing habits or manufacturing errors of non-standard components. With the increase in the number of applications, the accuracy of the reverse-engineered model and the matching degree with the actual object are continuously improved.

[0036] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for reverse molding of prefabricated MIC components based on the integration of BIM and CAD, characterized in that, Includes the following steps: S1. First, perform geometric cleaning and semantic feature extraction on the two-dimensional CAD data of the MIC component to determine the semantic type and initial geometric parameters of the primitives. S2. Based on the semantic type and initial geometric parameters, calculate the connection probability between components and construct a spatial topology association map of MIC components that describes the internal logical relationship of MIC components. S3. Calculate the visually salient feature points of the component as hidden anchor points, and encapsulate the data of the hidden anchor points and the spatial topological association map of the MIC component into a parametric BIM model to generate a BIM model that integrates topological semantics. S4. In the on-site application phase, spatial positioning is performed using the invisible anchor points in the BIM model, and augmented reality (AR) dynamic rendering is performed on the BIM model based on the spatial topology association map of the MIC components. S5. Collect on-site measured point cloud data, calculate the geometric deviation between the on-site measured point cloud data and the BIM model, correct the key parameters in the semantic feature extraction stage in reverse based on the geometric deviation, and regenerate the BIM model using the corrected key parameters.

2. The method for MIC prefabricated component formwork transformation based on BIM and CAD integration according to claim 1, characterized in that, The geometric cleaning of the 2D CAD data of the MIC component includes: Traverse all geometric primitives in the two-dimensional CAD data and set noise threshold and closure determination threshold; If the geometric length of a geometric primitive is less than the noise threshold, the geometric primitive is considered noise and is removed. If the Euclidean distance between the endpoints of two geometric primitives is less than the closure threshold, then a vertex merging operation is performed to construct a closed geometric path.

3. The method for reverse molding of MIC prefabricated components based on the integration of BIM and CAD as described in claim 1, characterized in that, The determination of the semantic type of the primitive includes: Construct a multidimensional geometric feature vector for primitives, wherein the components of the multidimensional geometric feature vector include at least the parallelism, closure, aspect ratio, and relative inclusion relationship of the primitives; The semantic score of a primitive is calculated using a weighted scoring function. The semantic score is the sum of the normalized values ​​of each feature component and the corresponding weight coefficients. A preset set of classification thresholds is used to compare the semantic score with the set of classification thresholds to determine whether the semantic type of the graphic element belongs to a structural component, electromechanical pipeline, or other component.

4. The MIC prefabricated component formwork method based on BIM and CAD integration according to claim 1, characterized in that, The construction of the spatial topological association map of MIC components, which describes the internal logical relationships of MIC components, includes: Select nodes identified as structural components and nodes of electromechanical pipelines, and calculate the connection probability that the two are logically related; the connection probability is jointly determined by the minimum Euclidean distance of the three-dimensional bounding box and the overlap state of the two-dimensional projection. A topological threshold is set. If the calculated connection probability is greater than the topological threshold, a connection edge is established in the graph to form the spatial topological association graph of the MIC component.

5. The method for reverse molding of MIC prefabricated components based on the integration of BIM and CAD according to claim 1, characterized in that, The visually salient feature points of the computational component serve as implicit anchor points, including: For the identified rigid structural primitives, geometric corner points are extracted as candidate points; The visual saliency score of the candidate point is calculated. The visual saliency score is determined by the local curvature modulus, the line density gradient, and the construction occlusion factor, wherein the construction occlusion factor is based on the installation procedure preset of the MIC component. Select several points that rank highly in the visual saliency score as the hidden anchor points, and record their absolute coordinates and visual descriptors.

6. The method for reverse molding of MIC prefabricated components based on the integration of BIM and CAD according to claim 1, characterized in that, The step of encapsulating the data of the hidden anchor points and the spatial topology association map of the MIC components into a parametric BIM model includes: Based on the semantic type, the preset MIC parameterized family library is invoked to map the extracted initial geometric parameters to family parameters to generate a 3D solid. The topology ID of the generated MIC component spatial topology association map and the calculated visual descriptor of the hidden anchor point are written into the extended attribute field of the corresponding BIM component to realize data ID association.

7. The method for reverse molding of MIC prefabricated components based on the integration of BIM and CAD according to claim 1, characterized in that, The spatial positioning using the hidden anchor points in the BIM model includes: AR devices are used to collect images of the scene and extract feature points. The on-site feature points are matched with the visual descriptors of the hidden anchor points pre-embedded in the BIM model; Based on the matching results, establish the transformation matrix between the world coordinate system and the model coordinate system.

8. The method for reverse molding of MIC prefabricated components based on the integration of BIM and CAD according to claim 1, characterized in that, The step of performing augmented reality (AR) dynamic rendering on the BIM model based on the spatial topology association map of the MIC components includes: When the current component node is selected, the remaining component nodes are traversed in the MIC component space topology association graph. Calculate the shortest path hop count from the remaining component nodes to the current component node; Set a rendering level depth threshold, and only render components whose shortest path hop count is less than or equal to the rendering level depth threshold, while hiding components that have no logical connection.

9. The method for reverse molding of MIC prefabricated components based on the integration of BIM and CAD according to claim 1, characterized in that, The calculation of the geometric deviation between the measured point cloud data and the BIM model includes: The local sparse point cloud on site is obtained as the measured point cloud data on site, and the optimal rigid body transformation matrix of the measured point cloud data on site and the BIM mesh model is calculated and registered using the iterative nearest point algorithm. Calculate the average deviation vector of the key feature surface. If the magnitude of the average deviation vector exceeds the preset tolerance, it is determined that the geometric deviation exists.

10. The MIC prefabricated component formwork method based on BIM and CAD integration according to claim 9, characterized in that, The key parameters for the semantic feature extraction stage, which are corrected in reverse based on the geometric deviation, include: The parameters that affect the geometry of the component are identified as the key parameters. Calculate the component of the average deviation vector in the direction of the normal vector on the component surface; The key parameters are updated using the components and a preset learning rate, and the updated key parameters are fed back to the semantic feature extraction step for re-determination of semantics.

Images